Surface Evolver Manual

Surface Evolver Manual
Version 2.70
August 25, 2013
Kenneth A. Brakke
Mathematics Department
Susquehanna University
Selinsgrove, PA 17870
[email protected]
http://www.susqu.edu/brakke
Contents
1
Introduction.
1.1 General description
1.2 Portability . . . . .
1.3 Bug reports . . . .
1.4 Web home page . .
1.5 Newsletter . . . . .
1.6 Acknowledgements
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
9
9
9
10
10
10
10
Installation.
2.1 Microsoft Windows . . . . . . . . . . . .
2.1.1 Modern GUI installation . . . . .
2.1.2 Old-fashioned installation by hand
2.2 Macintosh . . . . . . . . . . . . . . . . .
2.3 Unix/Linux . . . . . . . . . . . . . . . .
2.3.1 Compiling . . . . . . . . . . . .
2.4 Geomview graphics . . . . . . . . . . . .
2.5 X-Windows graphics . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
11
11
11
12
12
12
13
14
14
Tutorial
3.1 Basic Concepts . . . . . . . . . . . . . . .
3.2 Example 1. Cube evolving into a sphere . .
3.3 Example 2. Mound with gravity . . . . . .
3.4 Example 3. Catenoid . . . . . . . . . . . .
3.5 Example 4. Torus partitioned into two cells
3.6 Example 5. Ring around rotating rod . . . .
3.7 Example 6. Column of liquid solder . . . .
3.8 Example 7. Rocket fuel tank . . . . . . . .
3.8.1 Surface energy . . . . . . . . . . .
3.8.2 Volume . . . . . . . . . . . . . . .
3.8.3 Gravity . . . . . . . . . . . . . . .
3.8.4 Running . . . . . . . . . . . . . . .
3.9 Example 8. Spherical tank . . . . . . . . .
3.9.1 Surface energy . . . . . . . . . . .
3.9.2 Volume . . . . . . . . . . . . . . .
3.9.3 Gravity . . . . . . . . . . . . . . .
3.9.4 Running . . . . . . . . . . . . . . .
3.10 Example 9. Crystalline integrand . . . . . .
3.11 Tutorial in Advanced Calculus . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
15
15
16
19
21
23
25
30
32
33
34
35
36
38
39
40
41
42
44
45
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

4
Geometric elements
4.1 Vertices . . . . . . . . . .
4.2 Edges . . . . . . . . . . .
4.3 Facets . . . . . . . . . . .
4.4 Bodies . . . . . . . . . .
4.5 Facetedges . . . . . . . .
4.6 Common element attributes
4.7 Vertex attributes . . . . . .
4.8 Edge attributes . . . . . .
4.9 Facet attributes . . . . . .
4.10 Body attributes . . . . . .
4.11 Facetedge attributes . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
49
49
50
50
51
51
51
54
60
65
70
72
The Model
5.1 Dimension of surface . . . . . . . . . . . . . . . . . . . . .
5.2 Quadratic model . . . . . . . . . . . . . . . . . . . . . . . .
5.3 Lagrange model . . . . . . . . . . . . . . . . . . . . . . . .
5.4 Simplex model . . . . . . . . . . . . . . . . . . . . . . . .
5.5 Dimension of ambient space . . . . . . . . . . . . . . . . .
5.6 Riemannian metric . . . . . . . . . . . . . . . . . . . . . .
5.7 Torus domain. . . . . . . . . . . . . . . . . . . . . . . . .
5.8 Quotient spaces and general symmetry . . . . . . . . . . . .
5.8.1 TORUS symmetry group . . . . . . . . . . . . . . .
5.8.2 ROTATE symmetry group . . . . . . . . . . . . . .
5.8.3 FLIP ROTATE symmetry group . . . . . . . . . . .
5.8.4 CUBOCTA symmetry group . . . . . . . . . . . . .
5.8.5 CUBELAT symmetry group . . . . . . . . . . . . .
5.8.6 XYZ symmetry group . . . . . . . . . . . . . . . .
5.8.7 GENUS2 symmetry group . . . . . . . . . . . . . .
5.8.8 DODECAHEDRON symmetry group . . . . . . . .
5.8.9 CENTRAL SYMMETRY symmetry group . . . . .
5.8.10 SCREW SYMMETRY symmetry group . . . . . . .
5.8.11 QUARTER_TURN symmetry group . . . . . . . . .
5.9 Symmetric surfaces . . . . . . . . . . . . . . . . . . . . . .
5.10 Level set constraints . . . . . . . . . . . . . . . . . . . . . .
5.11 Boundaries . . . . . . . . . . . . . . . . . . . . . . . . . .
5.12 Energy. . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.13 Named quantities and methods . . . . . . . . . . . . . . . .
5.14 Pressure . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.15 Volume or content . . . . . . . . . . . . . . . . . . . . . . .
5.16 Diffusion . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.17 Motion . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.18 Hessian . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.19 Eigenvalues and eigenvectors . . . . . . . . . . . . . . . . .
5.20 Mobility . . . . . . . . . . . . . . . . . . . . . . . . . . . .
5.20.1 Vertex mobility . . . . . . . . . . . . . . . . . . . .
5.20.2 Area normalization . . . . . . . . . . . . . . . . . .
5.20.3 Area normalization with effective area . . . . . . . .
5.20.4 Approximate polyhedral curvature . . . . . . . . . .
5.20.5 Approximate polyhedral curvature with effective area
5.20.6 User-defined mobility . . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
74
74
74
74
75
75
75
75
77
77
77
78
78
78
78
79
79
79
79
79
80
80
81
82
83
84
84
85
85
86
86
87
88
88
88
88
88
89
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
5.21 Stability . . . . . . . . . . . . . . . . . . . . . . . .
5.21.1 Zigzag string . . . . . . . . . . . . . . . . .
5.21.2 Perturbed sheet with equilateral triangulation
5.22 Topology changes . . . . . . . . . . . . . . . . . . .
5.23 Refinement . . . . . . . . . . . . . . . . . . . . . .
5.24 Adjustable parameters and variables . . . . . . . . .
5.25 The String Model . . . . . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
89
89
89
90
90
90
90
The Datafile
6.1 Datafile organization . . . . . . . . . . . . . .
6.2 Lexical format . . . . . . . . . . . . . . . . . .
6.2.1 Comments . . . . . . . . . . . . . . .
6.2.2 Lines and line splicing . . . . . . . . .
6.2.3 Including files . . . . . . . . . . . . .
6.2.4 Macros . . . . . . . . . . . . . . . . .
6.2.5 Case . . . . . . . . . . . . . . . . . . .
6.2.6 Whitespace . . . . . . . . . . . . . . .
6.2.7 Identifiers . . . . . . . . . . . . . . . .
6.2.8 Strings . . . . . . . . . . . . . . . . .
6.2.9 Numbers . . . . . . . . . . . . . . . .
6.2.10 Keywords . . . . . . . . . . . . . . . .
6.2.11 Colors . . . . . . . . . . . . . . . . . .
6.2.12 Expressions . . . . . . . . . . . . . . .
6.3 Datafile top section: definitions and options . .
6.3.1 Macros . . . . . . . . . . . . . . . . .
6.3.2 Version check . . . . . . . . . . . . . .
6.3.3 Element id numbers . . . . . . . . . .
6.3.4 Variables . . . . . . . . . . . . . . . .
6.3.5 Arrays . . . . . . . . . . . . . . . . . .
6.3.6 Dimensionality . . . . . . . . . . . . .
6.3.7 Domain . . . . . . . . . . . . . . . . .
6.3.8 Length method . . . . . . . . . . . . .
6.3.9 Area method . . . . . . . . . . . . . .
6.3.10 Volume method . . . . . . . . . . . . .
6.3.11 Representation . . . . . . . . . . . . .
6.3.12 Hessian special normal vector . . . . .
6.3.13 Dynamic load libraries . . . . . . . . .
6.3.14 Extra attributes . . . . . . . . . . . . .
6.3.15 Surface tension energy . . . . . . . . .
6.3.16 Squared mean curvature . . . . . . . .
6.3.17 Integrated mean curvature . . . . . . .
6.3.18 Gaussian curvature . . . . . . . . . . .
6.3.19 Squared Gaussian curvature . . . . . .
6.3.20 Ideal gas model . . . . . . . . . . . . .
6.3.21 Gravity . . . . . . . . . . . . . . . . .
6.3.22 Gap energy . . . . . . . . . . . . . . .
6.3.23 Knot energy . . . . . . . . . . . . . . .
6.3.24 Mobility and motion by mean curvature
6.3.25 Annealing . . . . . . . . . . . . . . . .
6.3.26 Diffusion . . . . . . . . . . . . . . . .
6.3.27 Named method instances . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
91
91
91
91
91
92
92
92
92
92
92
92
93
93
93
94
94
94
94
95
95
96
96
97
97
98
98
98
99
99
100
100
100
100
101
101
101
101
101
101
102
102
102
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
4
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
6.4
6.5
6.6
6.7
6.8
6.9
7
6.3.28 Named quantities . . . . . . . . . .

6.3.29 Level set constraints . . . . . . . .
6.3.30 Constraint tolerance . . . . . . . .
6.3.31 Boundaries . . . . . . . . . . . . .
6.3.32 Numerical integration precision . .
6.3.33 Scale factor . . . . . . . . . . . . .
6.3.34 Mobility . . . . . . . . . . . . . .
6.3.35 Metric . . . . . . . . . . . . . . . .
6.3.36 Autochopping . . . . . . . . . . . .
6.3.37 Autopopping . . . . . . . . . . . .
6.3.38 Total time . . . . . . . . . . . . . .
6.3.39 Runge-Kutta . . . . . . . . . . . .
6.3.40 Homothety scaling . . . . . . . . .
6.3.41 Viewing matrix . . . . . . . . . . .
6.3.42 View transforms . . . . . . . . . .
6.3.43 View transform generators . . . . .
6.3.44 Zoom parameter . . . . . . . . . .
6.3.45 Alternate volume method . . . . . .
6.3.46 Fixed area constraint . . . . . . . .
6.3.47 Merit factor . . . . . . . . . . . . .
6.3.48 Parameter files . . . . . . . . . . .
6.3.49 Suppressing warnings . . . . . . .
6.3.50 Giving hints on memory allocation .
Element lists . . . . . . . . . . . . . . . . .
Vertex list . . . . . . . . . . . . . . . . . .
Edge list . . . . . . . . . . . . . . . . . . .
Face list . . . . . . . . . . . . . . . . . . .
Bodies . . . . . . . . . . . . . . . . . . . .
Commands . . . . . . . . . . . . . . . . .
Language syntax
7.1 General language syntax . . . . . .
7.1.1 Command separator . . . .
7.1.2 Compound commands . . .
7.1.3 Command repetition . . . .
7.1.4 Piping command output . .
7.1.5 Redirecting command output
7.1.6 Flow of control . . . . . . .
7.1.7 User-defined procedures . .
7.1.8 User-defined functions . . .
7.2 Expressions . . . . . . . . . . . . .
7.3 Aggregate expressions . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
103
104
105
106
106
106
107
107
107
107
107
108
108
108
108
109
109
109
109
109
109
110
110
110
110
111
111
112
112
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
113
113
113
113
114
114
114
115
116
117
117
119
Internal variables
120
8.1 Utility functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
Single-letter commands
128
9.1 Single-letter command summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
9.2 Alphabetical single-letter command reference . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
10 Toggles
135
11 General commands
11.1 Variable assignment . . . . . . . . . . . . . . . . . . . . . . .
11.2 Element generators and SQL-type queries on sets of elements .
11.3 Array operations. . . . . . . . . . . . . . . . . . . . . . . . .
11.4 Keyword commands . . . . . . . . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
12 Script examples
13 Operation
13.1 System command line
13.2 Initialization . . . . .
13.3 Error handling . . . .
13.4 Interrupts . . . . . . .
145
145
147
148
148
173
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
176
176
177
177
178
14 Graphics
14.1 Graphics commands . . . .
14.2 Graphics output file formats
14.2.1 Pixar . . . . . . . .
14.2.2 Geomview . . . . .
14.2.3 PostScript . . . . .
14.2.4 Triangle file . . . .
14.2.5 SoftImage file . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
179
179
184
184
184
185
185
185
15 Named Methods and Quantities

15.1 Introduction . . . . . . . . . . . . . . . .
15.2 Named methods . . . . . . . . . . . . . .
15.3 Method instances . . . . . . . . . . . . .
15.4 Named quantities . . . . . . . . . . . . .
15.5 Implemented methods . . . . . . . . . . .
15.6 Method descriptions . . . . . . . . . . . .
15.6.1 0-dimensional methods . . . . .
15.6.2 1-dimensional methods . . . . . .
15.6.3 2-dimensional methods . . . . . .
15.6.4 Surface curvature methods . . . .
15.6.5 General dimension methods . . .
15.6.6 Knot energy methods . . . . . . .
15.6.7 Elastic stretching energy methods
15.6.8 Weird and miscellaneous methods
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
186
186
186
186
187
188
192
192
192
198
202
207
208
214
217
16 Technical Reference
16.1 Notation . . . . . . . . . . . . . . .
16.2 Surface representation . . . . . . .
16.3 Energies and forces . . . . . . . . .
16.3.1 Surface tension . . . . . . .
16.3.2 Crystalline integrand . . . .
16.3.3 Gravity . . . . . . . . . . .
16.3.4 Level set constraint integrals
16.3.5 Gap areas . . . . . . . . . .
16.3.6 Ideal gas compressibility . .
16.3.7 Prescribed pressure . . . . .
16.3.8 Squared mean curvature . .
16.3.9 Squared Gaussian curvature
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
219
219
219
220
220
220
220
221
221
221
222
222
223
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
16.4 Named quantities and methods . . . . . . . . . . . . . . . . . . . . . .

16.4.1 Vertex value . . . . . . . . . . . . . . . . . . . . . . . . . . . .
16.4.2 Edge length . . . . . . . . . . . . . . . . . . . . . . . . . . . .
16.4.3 Facet area . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
16.4.4 Path integrals . . . . . . . . . . . . . . . . . . . . . . . . . . .
16.4.5 Line integrals . . . . . . . . . . . . . . . . . . . . . . . . . . .
16.4.6 Scalar surface integral . . . . . . . . . . . . . . . . . . . . . .
16.4.7 Vector surface integral . . . . . . . . . . . . . . . . . . . . . .
16.4.8 2-form surface integral . . . . . . . . . . . . . . . . . . . . . .
16.4.9 General edge integral . . . . . . . . . . . . . . . . . . . . . . .
16.4.10 General facet integral . . . . . . . . . . . . . . . . . . . . . . .
16.4.11 String area integral . . . . . . . . . . . . . . . . . . . . . . . .
16.4.12 Volume integral . . . . . . . . . . . . . . . . . . . . . . . . . .
16.4.13 Gravity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
16.4.14 Hooke energy . . . . . . . . . . . . . . . . . . . . . . . . . . .
16.4.15 Local Hooke energy . . . . . . . . . . . . . . . . . . . . . . .
16.4.16 Integral of mean curvature . . . . . . . . . . . . . . . . . . . .
16.4.17 Integral of squared mean curvature . . . . . . . . . . . . . . . .
16.4.18 Integral of Gaussian curvature . . . . . . . . . . . . . . . . . .
16.4.19 Average crossing number . . . . . . . . . . . . . . . . . . . . .
16.4.20 Linear elastic energy . . . . . . . . . . . . . . . . . . . . . . .
16.4.21 Knot energies . . . . . . . . . . . . . . . . . . . . . . . . . . .
16.5 Volumes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
16.5.1 Default facet integral . . . . . . . . . . . . . . . . . . . . . . .
16.5.2 Symmetric content facet integral . . . . . . . . . . . . . . . . .
16.5.3 Edge content integrals . . . . . . . . . . . . . . . . . . . . . .
16.5.4 Volume in torus domain . . . . . . . . . . . . . . . . . . . . .
16.6 Constraint projection . . . . . . . . . . . . . . . . . . . . . . . . . . .
16.6.1 Projection of vertex to constraints . . . . . . . . . . . . . . . .
16.6.2 Projection of vector onto constraint tangent space . . . . . . . .
16.7 Volume and quantity constraints . . . . . . . . . . . . . . . . . . . . .
16.7.1 Volume restoring motion . . . . . . . . . . . . . . . . . . . . .
16.7.2 Motion projection in gradient mode . . . . . . . . . . . . . . .
16.7.3 Force projection in mean curvature mode . . . . . . . . . . . .
16.7.4 Pressure at z = 0 . . . . . . . . . . . . . . . . . . . . . . . . .
16.8 Iteration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
16.8.1 Fixed scale motion . . . . . . . . . . . . . . . . . . . . . . . .
16.8.2 Optimizing scale motion . . . . . . . . . . . . . . . . . . . . .
16.8.3 Conjugate gradient mode . . . . . . . . . . . . . . . . . . . . .
16.9 Hessian iteration . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
16.10Dirichlet and Sobolev approximate Hessians . . . . . . . . . . . . . . .
16.11Calculating Forces and Torques on Rigid Bodies . . . . . . . . . . . . .
16.11.1 Method 1. Finite differences . . . . . . . . . . . . . . . . . . .
16.11.2 Method 2. Principle of Virtual Work by Finite Differences . . .
16.11.3 Method 3. Principle of Virtual Work using Lagrange Multipliers
16.11.4 Method 4. Explicit forces . . . . . . . . . . . . . . . . . . . .
16.11.5 Method 5. Variational formulation . . . . . . . . . . . . . . . .
16.11.6 Example of variational integrals . . . . . . . . . . . . . . . . .
16.12Equiangulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
16.13Dihedral angle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
16.14Area normalization . . . . . . . . . . . . . . . . . . . . . . . . . . . .
7
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
223
224
224
224
225
225
225
225
225
225
225
226
226
226
227
227
227
227
228
228
228
228
231
231
231
232
232
233
233
233
234
234
234
234
235
235
235
235
236
236
239
241
241
241
242
243
244
245
250
251
251
16.15Hidden surfaces . . . . . . . . . . . . . . . .
16.16Extrapolation . . . . . . . . . . . . . . . . .
16.17Curvature test . . . . . . . . . . . . . . . . .
16.18Annealing (jiggling) . . . . . . . . . . . . . .
16.19Long wavelength perturbations (long jiggling)
16.20Homothety . . . . . . . . . . . . . . . . . .
16.21Popping non-minimal edges . . . . . . . . .
16.22Popping non-minimal vertex cones . . . . . .
16.23Refining . . . . . . . . . . . . . . . . . . . .
16.24Refining in the simplex model . . . . . . . .
16.25Removing tiny edges . . . . . . . . . . . . .
16.26Weeding small triangles . . . . . . . . . . . .
16.27Vertex averaging . . . . . . . . . . . . . . .
16.28Zooming in on vertex . . . . . . . . . . . . .
16.29Mobility and approximate curvature . . . . .
17 Miscellaneous
17.1 Customizing graphics . . . . . .
17.1.1 Random-order interface
17.1.2 Painter interface . . . .
17.2 Dynamic load libraries . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
251
252
252
252
252
252
253
253
253
253
254
254
254
255
255
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
257
257
257
258
258
18 Helpful hints and notes

260
18.1 Hints . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 260
18.2 Checking your datafile . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 262
18.3 Reasonable scale factors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 263
19 Bugs
264
20 Version history
265
21 Bibliography
276
Chapter 1
Introduction.
1.1 General description
The Surface Evolver is an interactive program for the study of surfaces shaped by surface tension and other energies.
A surface is implemented as a simplicial complex, that is, a union of triangles. The user defines an initial surface in a
datafile. The Evolver evolves the surface toward minimal energy by a gradient descent method. The evolution is meant
to be a computer model of the process of evolution by mean curvature, which was studied in [B1] for surface tension
energy in the context of varifolds and geometric measure theory. The energy in the Evolver can be a combination of
surface tension, gravitational energy, squared mean curvature, user-defined surface integrals, or knot energies. The
Evolver can handle arbitrary topology (as seen in real soap bubble clusters), volume constraints, boundary constraints,
boundary contact angles, prescribed mean curvature, crystalline integrands, gravity, and constraints expressed as surface integrals. The surface can be in an ambient space of arbitrary dimension, which can have a Riemannian metric,
and the ambient space can be a quotient space under a group action. The user can interactively modify the surface to
change its properties or to keep the evolution well-behaved. The Evolver was written for one and two dimensional
surfaces, but it can do higher dimensional surfaces with some restrictions on the features available. Graphical output
is available as screen graphics and in several file formats, including PostScript.
The Surface Evolver program is freely available (see chapter 2) and is in use by a number of researchers. Some of
the applications of the Evolver so far include modelling the shape of fuel in rocket tanks in low gravity [Te], calculating
areas for the Opaque Cube Problem [B4],[B6], computing capillary surfaces in cubes [MH] and in exotic containers
[C], simulating grain growth [FT][WM], studying grain boundaries pinned by inclusions, finding partitions of space
more efficient than Kelvins tetrakaidecahedra [WP][KS1], calculating the rheology of foams [KR1][KR2], modelling
the shape of molten solder on microcircuits [RSB], studying polymer chain packing, modelling cell membranes [MB],
knot energies [KS2], sphere eversion [FS], and classifying minimal surface singularities.
The strength of the Surface Evolver program is in the breadth of problems it handles, rather than optimal treatment
of some specific problem. It is under continuing development, and users are invited to suggest new features.
This manual contains operational details and mathematical descriptions (please excuse the inconsistent fonts and
formatting; the manual grows by bits and pieces). Much of the manual (the stuff without a lot of mathematical
formulas) is also in HTML format, included in the distribution. A journal article description of the Evolver appeared
in [B2].
Previous users of the Evolver should consult the History chapter for the new features.
1.2 Portability
The Evolver is written in portable C and has been run on several systems: Microsoft Windows, Linux, Unix, and
Macintosh. It is meant to be easily portable to any system that has C. It compiles in either 32 or 64 bits.
10
1.3 Bug reports

Bug reports should be submitted by email to [email protected]. Please include the Evolver version number, a description of the problem, the initial datafile, and the sequence of commands necessary to reproduce the problem.
1.4 Web home page

My Web home page is http://www.susqu.edu/brakke and my Evolver page is http://www.susqu.edu/brakke/evolver.
Evolver-related material and links will be posted there. In particular, there are many examples of surfaces.
1.5 Newsletter
The group of Surface Evolver users has grown large enough that I have started a newsletter. Mostly it consists of
announcements of new Evolver versions. If you would like to be on the mailing list, send your email address to
[email protected]. Back issues are included in the HTML part of the distribution.
1.6 Acknowledgements
The Evolver was originally written as part of the Minimal Surfaces Group of the Geometry Supercomputing Project
(now The Geometry Center), sponsored by the National Science Foundation, the Department of Energy, Minnesota
Technology, Inc., and the University of Minnesota. The program is available free of charge.
10
Chapter 2
Installation.
This chapter explains how to get and install the Evolver. Evolver is written to be portable between systems. There are
pre-compiled versions for Windows and Macintosh; source files and a Makefile are provided for unix/Linux systems.
The distribution packages for various systems are available from
http://www.susqu.edu/brakke/evolver/evolver.html
Each package also contains documentation and sample datafiles and scripts. The documentation subdirectory is named
doc, and contains the manual in PDF format, an HTML version of the documentation (except for the mathematical
parts), a text file used by the Evolvers help command, and a brief The samples are in the subdirectory fe (which is the
file extension I use for datafiles; it stands for facet-edge, referring to the internal structure of surfaces in the Evolver).
This manual is separately downloadable in PostScript format from
http://www.susqu.edu/brakke/evolver/manual250.ps
or PDF format from

http://www.susqu.edu/brakke/evolver/manual250.pdf
The PDF version is included in the standard Evolver distributions.

The HTML manual can be read on-line at
http://www.susqu.edu/brakke/evolver/html/default.htm.
2.1 Microsoft Windows

2.1.1
Modern GUI installation
The distributions available here include both the 32-bit version of Evolver (evolver.exe) and the 64-bit version
(evolver64.exe). The 64-bit version is meant for very large models (hundreds of thousands of facets). Unless you
have such a large surface, use the 32-bit version since it is 10 to 20 percent faster, having less data to move around.
The file
http://www.susqu.edu/brakke/aux/downloads/Evolver-2.50-Win32.msi
is a fancy graphical installation programs that will lead you through installation. They will automatically set up the
environment variables described below.
11
2.1.2
12
Old-fashioned installation by hand
The file http://www.susqu.edu/brakke/aux/downloads/evolver250-Win32.zip has the executable files evolver.exe,

evolver64,exe along with the documentation and sample datafile subdirectories. Create a directory (such as C:\evolver),
and unzip the distribution package there. You can leave evolver.exe there and add C:\evolver to your PATH, or you can
copy evolver.exe to someplace in your PATH, such as C:\windows\system32.
You should also create an environment variable EVOLVERPATH telling Evolver where to search for various files.
Do this by opening Control Panel/System/Advanced/Environment Variables, clicking New under System Variables,
entering EVOLVERPATH for the Variable name, and c:\evolver\fe;c:\evolver\doc for the Variable value. You may add
further paths of your own to this list if you wish.
To make Evolver start automatically when you click on a *.fe file, you can associate Evolver with the file extension .fe by opening My Computer/Tools/Folder Options/File Types/New, entering the File Extension fe, clicking OK,
clicking Change, and browsing for the evolver.exe program. (This sequence of actions may vary on different Windows
versions.)
The Windows version uses OpenGL/GLUT graphics. OpenGL is standard in Windows, and all the necessary
GLUT components are included in the executable, so you dont have to install anything.
2.2 Macintosh
I am not a Mac person, and have only learned enough Mac to get Evolver minimally ported, so there are no Mac bells
and whistles.
There is a Mac OSX version at http://www.susqu.edu/brakke/aux/downloads/Evolver250-OSX.tar.gz. After downloading and unpacking it, you probably get a folder Evolver250-OSX, which has the executable file Evolver, the
samples folder fe, and the documentation folder doc. You can move the executable to some place on your PATH, or
add the folder to your PATH. You should also create an environment variable EVOLVERPATH containing paths to the
fe and doc folders by placing the following line in your .tcshrc file, with appropriate modifications:
setenv EVOLVERPATH /User/yourname/Evolver250-OSX/fe:/User/yourname/Evolver250-OSX/doc
This descends from the unix version, not the Mac OS 9 version, so you must run it from a terminal window. It uses
OpenGL GLUT graphics, which are standard with OSX.
2.3 Unix/Linux
The program is distributed in a compressed tar file evolver-2.50.tar.gz (for unix systems; see below for others). Get
this file into a working directory. Uncompress it with
gunzip evolver-2.50.tar.gz
Extract the files with

tar xvf evolver.tar
This will unpack into three subdirectories: src (source code), doc (the manual), and fe (sample datafiles). The packed
archive is about 2MB, unpacks to about 5MB. You will probably need another 3 or 4 MB to compile. See below for
compilation instructions.
Evolver needs to find the initial datafile and sometimes other files (e.g. command files for the read command,
or the HTML version of the manual). If the file is not in the current directory, then an environment variable called
EVOLVERPATH will be consulted for a directory search list. The datafile directory and the documentation directory
with the HTML files should definitely be included. The format is the same as the usual PATH variable. Set it up as
usual in your system:
Unix C shell:
12
13
setenv EVOLVERPATH /usr/you/evolver/fe:/usr/you/evolver/doc
Bourne shell:
EVOLVERPATH=/usr/you/evolver/fe:/usr/you/evolver/doc
export EVOLVERPATH
2.3.1
Compiling
First, you need to modify Makefile for your system. Makefile begins with sets of macro definitions for various systems.
If your system is listed, remove the comment symbols # from start of your definitions. If your system is not there,
use the GENERIC defines, or set up your own, and leave the graphics file as glutgraph.o if you have OpenGL/GLUT,
else xgraph.o if you have X-windows, and nulgraph.o otherwise. If you do define your own system, be sure to put a
corresponding section in include.h. After you get a working program you can write a screen graphics interface if you
cant use one of those provided. Edit CFLAGS to have the proper options (optimization, floating point option, etc.).
GRAPH should be the name of a screen graphics interface file. GRAPHLIB should be the appropriate graphics library
plus any other libraries needed. Check all paths and change if necessary for your system.
GLUT graphics uses a separate thread to display graphics, so if you use GLUT, you must compile with
-DPTHREADS and put -lpthread in GRAPHLIB.
If you want Evolver to be able to use geomview, include -DOOGL in CFLAGS.
If you wish to use the commands based on the METIS partitioning software (metis, kmetis, and metis_factor), then
you should download the METIS package from
http://www-users.cs.umn.edu/ karypis/metis/
and "make" the library libmetis.a (on some systems, make complains it cannot find ranlib, but the resulting libmetis.a
still works). In Evolvers Makefile, add -DMETIS to CFLAGS, and add -lmetis to GRAPHLIB. You will probably also
have to add -Lpath to GRAPHLIB to tell the linker where to find libmetis.a. Note that METIS is incorporated in the
Windows executable. If you are using hessian commands on very large surfaces, then metis_factor can be much faster
than the other sparse matrix factoring schemes in Evolver, and I highly recommend it.
If you want Evolver to operate in a higher space dimension n than the default maximum of 4, include
-DMAXCOORD=n in CFLAGS. This sets the upper limit of dimensionality, and is used for allocating space in data
structures. You can use -DMAXCOORD=3 to save a little memory.
If your system supports the long double data type, you can compute in higher precision by compiling with
-DLONGDOUBLE in CFLAGS. This should only be used by the precision-obsessed. And on Intel chips, it only does 80
bit precision, even if it allocates 128 bits of storage, since that is the hardware precision of the 8087 math coprocessor.
For 128 bit precision (in software, so much slower), compile with cf -DFLOAT128 and link with the quadmath library
(which you have to find on your own).
You can let the compiler optimize better if you hard-wire the space dimension into the code with the compiler
option -DSDIM=n, where n is the desired dimension of space. But such an Evolver can handle only the given space
dimension, no more, no less.
Silicon Graphics systems with multiple processors may compile a version that will use all processors for some
calculations by including -DSGI_MULTI in CFLAGS. This version will run fine with one processor, also. Currently, the
only calculations done in parallel are the named quantities. The number of processes actually done in parallel can
be controlled with the -pn command line option.
The file include.h lists the include files for various systems. If your system isnt listed, you will have to put in a list
of your own. Try copying the generic list (or one of the others) and compiling. Your compiler will no doubt be very
friendly and helpful in pointing out unfound header files.
include.h also has various other system-specific defines. See the GENERIC section of it for comments on these,
and include the appropriate ones for your system.
Now try compiling by running make. Hopefully, you will only have to change the system-specific parts of Makefile
and include.h to get things to work. If significant changes to other files are needed, let me know. For example, the
return value of sprintf() varies among systems, and that caused problems once upon a time.
Test by running on the cube.fe sample file as described in the Tutorial chapter.
13
14
2.4 Geomview graphics

Evolver has built-in OpenGL/GLUT graphics that will work on most all current systems, but it is also possible to use
the external Geomview program on unix systems. geomview is available from http://www.geomview.org For those with
surfaces already defined for Geomview, files in Geomviews OFF format may be converted to the Evolvers datafile
format by the program Polymerge, which is included in the Geomview distribution. Use the -b option. Files in other
Geomview formats may be converted to OFF with anytooff, also included with Geomview.
2.5 X-Windows graphics

There is a very primitive X-Windows graphics interface, which can be used on unix systems if for some reason OpenGL
doesnt work. To have Evolver use X-Windows graphics for its native graphics, edit Makefile to have xgraph.o as the
graphics module and have all the proper libraries linked in. The X window will appear when you do the s command.
Evolver does not continually poll for window redraw events, so if you move or resize the window you will have to do
s again to redraw the surface.
14
Chapter 3
Tutorial
This chapter introduces some basic concepts and then guides the user through a series of examples. Not all the sample
datafiles that come with Evolver are discussed; users may browse them at their liesure. This chapter can be read before
the following chapters. The following chapters assume you have read the Basic Concepts section of this chapter. If
you are a little rusty on Advanced Calculus, there is a short tutorial on it at the end of this chapter.
3.1 Basic Concepts

The basic geometric elements used to represent a surface are vertices, edges, facets, and bodies. Vertices are points in
Euclidean 3-space. Edges are straight line segments joining pairs of vertices. Facets are flat triangles bounded by three
edges. A surface is a union of facets. (Actually, there is no separate surface entity in the program; all facets essentially
belong to one logical surface.) A body is defined by giving its bounding facets.
The term surface", when used to refer to the entity upon which the Evolver operates, refers to all the geometric
elements plus auxiliary data such as constraints, boundaries, and forces.
There are no limitations on how many edges may share a vertex nor on how many facets may share an edge. Thus
arbitrary topologies are possible, including the triple junctions of surfaces characteristic of soap films.
Edges and facets are oriented for bookkeeping purposes, but there are no restrictions on the orientation of neighboring facets. Unoriented surfaces are thus possible.
A surface is deemed to have a total energy, arising from surface tension, gravitational energy, and possibly other
sources. It is this energy which the Evolver minimizes.
No particular units of measurement are used. The program only deals with numerical values. If you wish to relate
the program values to the real world, then all values should be within one consistent system, such as cgs or MKS.
The initial surface is specified in a text file (hereafter referred to as the datafile) that may be created with any
standard text editor. (The .fe extension I always use for datafiles stands for facet-edge, which refers to the internal data
structure used to represent the surface. You may use any name you wish for a datafile.)
The basic operation of the Evolver is to read in a datafile and take commands from the user. The main command
prompt is
Enter command:
Basic commands are one letter (case is significant), sometimes with a numerical parameter. The most frequently used
commands are:
gn
s
r
P
q
do n iterations
show surface on screen (or P option 8 for geomview)
refine triangulation of surface
graphics output (option 3 for PostScript)
quit
15
16
8
7
5
6
12
11
10
3
1
2
Figure 3.1: The cube skeleton.
There is also a more elaborate command language (in which case is not significant). Commands must be followed
with the ENTER key; Evolver only reads complete lines.
An iteration is one evolution step. The motion for the step is calculated as follows: First, the force on each vertex
is calculated from the gradient of the total energy of the surface as a function of the position of that vertex. The force
gives the direction of motion. Second, the force is made to conform to whatever constraints are applicable. Third, the
actual motion is found by multiplying the force by a global scale factor.
3.2 Example 1. Cube evolving into a sphere

A sample datafile cube.fe comes with Evolver. The initial surface is a unit cube. The surface bounds one body, and
the body is constrained to have volume 1. There is no gravity or any other force besides surface tension. Hence the
minimal energy surface will turn out to be a sphere. This example illustrates the basic datafile format and some basic
commands.
Lets look at the datafile that specifies the initial unit cube:
// cube.fe
// Evolver data for cube of prescribed volume.
vertices /* given by coordinates */
1 0.0 0.0 0.0
2 1.0 0.0 0.0
3 1.0 1.0 0.0
4 0.0 1.0 0.0
5 0.0 0.0 1.0
6 1.0 0.0 1.0
7 1.0 1.0 1.0
8 0.0 1.0 1.0
16
17
edges /* given by endpoints */

1 12
2 23
3 34
4 41
5 56
6 67
7 78
8 85
9 15
10 2 6
11 3 7
12 4 8
faces /* given by oriented edge loop */
1 1 10 -5 -9
2 2 11 -6 -10
3 3 12 -7 -11
4 4 9 -8 -12
5 5 6 7 8
6 -4 -3 -2 -1
bodies /* one body, defined by its oriented faces */
1 1 2 3 4 5 6 volume 1
The datafile is organized in lines, with one geometric element defined per line. Vertices must be defined first, then
edges, then faces, then bodies. Each element is numbered for later reference in the datafile.
Comments are delimited by /* */ as in C, or follow // until the end of the line as in C++. Case is not significant, and
all input is lower-cased immediately. Hence error messages about your datafiles will refer to items in lower case, even
when you typed them in upper case.
The datafile syntax is based on keywords. The keywords VERTICES, EDGES, FACES, and BODIES signal the start
of the respective sections. Note that the faces are not necessarily triangles (which is why they are called FACES and
not FACETS). Any non-triangular face will be automatically triangulated by putting a vertex at its center and putting
in edges to each of the original vertices. Faces dont have to be planar. Note that a minus sign on an edge means that
the edge is traversed in the opposite direction from that defined for it in the EDGES section. The face oriented normal
is defined by the usual right hand rule. The cube faces all have outward normals, so they all are positive in the body
list. In defining a body, the boundary faces must have outward normals. If a face as defined has an inward normal, it
must be listed with a minus sign.
That the body is constrained to have a volume of 1 is indicated by the keyword VOLUME after the body definition,
with the value of the volume following. Any attributes or properties an element has are given on the same line after its
definition.
Start Evolver and load the datafile with the command line
evolver cube.fe
You should get a prompt

Enter command:
Give the command s to show the surface. You should see a square divided into four triangles by diagonals. This is
the front side of the cube; you are looking in along the positive x-axis, with the z axis vertical and the positive y axis
to the right. On most systems, you can manipulate the displayed surface with the mouse: dragging the mouse over
17
18
the surface with the left button down rotates the surface; you can change to "zoom" mode by hitting the z key, to
"translate" by hitting t, to "spin" by hitting c, and back to "rotate" by hitting r. Hit the h key with the mouse focus
in the graphics window to get a summary of the possibilities. You can also give graphics commands at the graphics
command prompt; this is good for precise control. The graphics command prompt is
Graphics command:
It takes strings of letters, each letter making a viewing transformation on the surface: The most used ones are
r
l
u
d
R
q
rotate right by 6 degrees

rotate left by 6 degrees
rotate up by 6 degrees
rotate down by 6 degrees
reset to original position
quit back to main command prompt
Try rrdd to get an oblique view of the cube. Any transformations you make will remain in effect the next time you
show the surface. Now do q to get back to the main prompt.
If you are using geomview for graphics, do command P option 8 to get a display. It takes a couple of seconds to
initialize. You can manipulate the geomview display as usual independently of the Evolver. Evolver will automatically
update the image whenever the surface changes.
Now do some iterations. Give the command g 5 to do 5 iterations. You should get this:
5. area:
4. area:
3. area:
2. area:
1. area:
5.11442065156005 energy:
5.11237323810972 energy:
5.11249312304592 energy:
5.11249312772740 energy:
5.11249312772740 energy:
5.11442065156005
5.11237323810972
5.11249312304592
5.11249312772740
5.11249312772740
scale: 0.186828
scale: 0.21885
scale: 0.204012
scale: 0.20398
scale: 0.554771
Note that after each iteration a line is printed with the iterations countdown, area, energy, and current scale factor. By
default, the Evolver seeks the optimal scale factor to minimize energy. At first, there are large motions, and the volume
constraint may not be exactly satisfied. The energy may increase due to the volume constraint taking hold. Volume
constraints are not exactly enforced, but each iteration tries to bring the volume closer to the target. Here that results
in increases in area. You can find the current volumes with the v command:
Body
target volume
1 1.000000000000000
actual volume
pressure
0.999999779366360 3.408026016427987
The pressure in the last column is actually the Lagrange multiplier for the volume constraint. Now lets refine the
triangulation with the r command. This subdivides each facet into four smaller similar facets. The printout here gives
the counts of the geometric elements and the memory they take:
Vertices: 50 Edges: 144 Facets: 96 Facetedges: 288 Memory: 27554
Iterate another 10 times:

10. area: 4.908899804670224 energy: 4.908899804670224 scale: 0.268161
18
19
You can continue iterating and refining as long as you have time and memory.
Eventually, you will want to quit. So give the q command. You get
Enter new datafile name (none to continue, q to quit):
You can start a new surface by entering a datafile name (it can be the same one you just did, to start over), or continue
with the present surface by hitting ENTER with no name (in case you pressed q by accident, or suddenly you remember
something you didnt do), or you can really quit with another q.
3.3 Example 2. Mound with gravity

8
7
9
11
10
12
12
15
11
14
16
1
2
9
13
10
Figure 3.2: The mound skeleton.

This example is a mound of liquid sitting on a tabletop with gravity acting on it. The contact angle between the
drop surface and the tabletop is adjustable, to simulate the different degrees to which the liquid wets the table. This
example illustrates macros, variables, constraints with energy, and omitting faces from body surfaces.
The drop starts as a cube with one face (face 6 of example 1) on the tabletop (the z = 0 plane). The most straightforward way to specify a contact angle is to declare face 6 to be constrained to stay on the tabletop and give it a surface
tension different than the default of 1. But this leads to problems described below. The way the contact angle is
handled instead is to omit face 6 and give the edges around face 6 an energy integrand that results in the same energy
we would get if we did include face 6. If we let the interface energy density for face 6 be T , then we want a vectorfield
w such that
=
w dl.
Tk dS
( f ace 6)
f ace 6
So by Greens Theorem, all we need is curl w = Tk, and I have used w = Tyi. In practice, I dont think about Greens
Theorem as such; I just write down a line integral that sums up strips of surface.
I have chosen to parameterize the contact angle as the angle in degrees between the table and the surface on the
interior of the drop. This angle can be adjusted by assigning a value to the variable angle at run time. I could have
made WALLT the parameter directly, but then I wouldnt have had an excuse to show a macro.
Here is the datafile mound.fe:
19
20
// mound.fe
// Evolver data for drop of prescribed volume sitting on plane with gravity.
// Contact angle with plane can be varied.
PARAMETER angle = 90
// interior angle between plane and surface, degrees
#define T (-cos(angle*pi/180)) // virtual tension of facet on plane

constraint 1 /* the table top */
formula: x3 = 0
energy: // for contact angle
e1: -T*y
e2: 0
e3: 0
vertices
1 0.0 0.0 0.0 constraint 1 /* 4 vertices on plane */
2 1.0 0.0 0.0 constraint 1
3 1.0 1.0 0.0 constraint 1
4 0.0 1.0 0.0 constraint 1
5 0.0 0.0 1.0
6 1.0 0.0 1.0
7 1.0 1.0 1.0
8 0.0 1.0 1.0
9 2.0 2.0 0.0 fixed /* for table top */
10 2.0 -1.0 0.0 fixed
11 -1.0 -1.0 0.0 fixed
12 -1.0 2.0 0.0 fixed
edges /* given by endpoints and attribute */
1 1 2 constraint 1 /* 4 edges on plane */
2 2 3 constraint 1
3 3 4 constraint 1
4 4 1 constraint 1
5 56
6 67
7 78
8 85
9 15
10 2 6
11 3 7
12 4 8
13 9 10 fixed /* for table top */
14 10 11 fixed
15 11 12 fixed
16 12 9 fixed
1 1 10 -5 -9
2 2 11 -6 -10
3 3 12 -7 -11
20
21
4 4 9 -8 -12
5 5 6 7 8
7 13 14 15 16 density 0 fixed /* table top for display */
1 1 2 3 4 5 volume 1 density 1
The mound itself was basically copied from cube.fe, but with face 6 deleted. The reason for this is that face 6 is
not needed, and would actually get in the way. It is not needed for the volume calculation since it would always be at
z = 0 and thus not contribute to the surface integral for volume. The bottom edges of the side faces are constrained to
lie in the plane z = 0, so face 6 is not needed to keep them from catastrophically shrivelling up. We could have handled
the contact angle by including face 6 with a surface tension equal to the interface energy density between the liquid
and surface, but that can cause problems if the edges around face 6 try to migrate inward. After refinement a couple
of times, interior vertices of the original face 6 have no forces acting on them, so they dont move. Hence it would be
tough for face 6 to shrink when its outer vertices ran up against its inner vertices. The tabletop face, face 7, is entirely
extraneous to the calculations. Its only purpose is to make a nice display. You could remove it and all its vertices and
edges without affecting the shape of the mound. Its constraint 1 that is the tabletop as far as the mound is concerned.
To see what happens with the bottom face present, load moundB.fe and do "run".
Now run Evolver on mound.fe. Refine and iterate a while. You should get a nice mound. Its not a hemisphere,
since gravity is on by default with G = 1. If you use the G command to set G 0 and iterate a while, you get a
hemisphere. Try changing the contact angle, to 45 degrees (with the command "angle := 45" or 135 degrees for
example. You can also play with gravity. Do "G 10" to get a flattened drop, or "G -5" to get a drop hanging from the
ceiling. "G -10" will cause the drop to try to break loose, but it cant, since its vertices are still constrained.
3.4 Example 3. Catenoid

The catenoid is the minimal surface formed between two rings not too far apart. In cylindrical coordinates, its equation
is r = (1/a) cosh(az). In cat.fe, both the upper and lower rings are given as one-parameter boundary wires. The
separation and radius are parameters, so you can change them during a run with the A command. The initial radius
given is the minimum for which a catenoid can exist for the given separation of the rings. To get a stable catenoid, you
will have to increase this value. However, if you do run with the original value, you can watch the neck pinch out.
The initial surface consists of six rectangles forming a cylinder between the two circles. The vertices on the
boundaries are fixed, elsewise they would slide along the boundary to short-cut the curvature; two diameters is shorter
than one circumference. The boundary edges are fixed so that vertices arising from subdividing the edges are likewise
fixed.
Here is the catenoid datafile:
// cat.fe
// Evolver data for catenoid.
PARAMETER RMAX = 1.5088795 // minimum radius for height
PARAMETER ZMAX = 1.0
boundary 1 parameters 1
x1: RMAX * cos(p1)
x2: RMAX * sin(p1)
x3: ZMAX
// upper ring
x1: RMAX * cos(p1)
// lower ring
21
22
4
3
4
5
3
2
5
6
16
6
1
17
15
10
18
11
10
14
9
8
13
11
12
8
7
12
7
Figure 3.3: The catenoid skeleton. Vertices and edges 1-6 are on circular boundary 1, and vertices and edges 7-12 are
on circular boundary 2.
x2: RMAX * sin(p1)

x3: -ZMAX
vertices // given in terms of boundary parameter
1 0.00 boundary 1 fixed
2 pi/3 boundary 1 fixed
3 2*pi/3 boundary 1 fixed
4 pi boundary 1 fixed
edges
1 1 2 boundary 1 fixed
22

13
14
15
16
17
18
1
2
3
4
5
6
23
7
8
9
10
11
12
faces
1 1 14 -7 -13
2 2 15 -8 -14
3 3 16 -9 -15
4 4 17 -10 -16
5 5 18 -11 -17
6 6 13 -12 -18
The parameter in a boundary definition is always P1 (and P2 in a two-parameter boundary). The Evolver can
handle periodic parameterizations, as is done in this example.
Try this sequence of commands (displaying at your convenience):
r
u
g 120
t
0.05
o
g
(refine to get a crude, but workable, triangulation)

(equiangulation makes much better triangulation)
(takes this many iterations for neck to collapse)
(collapse neck to single vertex by eliminating all edges shorter than 0.05)
(split neck vertex to separate top and bottom surfaces)
(spikes collapse)
The catenoid shows some of the subtleties of evolution. Suppose the initial radius is set to RMAX = 1.0 and the
initial height to ZMAX = 0.55. Fifty iterations with optimizing scale factor result in an area of 6.458483. At this
point, each iteration is reducing the area by only .0000001, the triangles are all nearly equilateral, everything looks
nice, and the innocent user might conclude the surface is very near its minimum. But this is really a saddle point
of energy. Further iteration shows that the area change per iteration bottoms out about iteration 70, and by iteration
300 the area is down to 6.4336. The triangulation really wants to twist around so that there are edges following the
lines of curvature, which are vertical meridians and horizontal circles. Hence the optimum triangulation appears to be
rectangles with diagonals.
The evolution can be speeded up by turning on the conjugate gradient method with the U command. For conjugate
gradient cognoscenti, the saddle point demonstrates the difference between the Fletcher-Reeves and Polak-Ribiere
versions of conjugate gradient (7.8.3). The saddle point seems to confuse the Fletcher-Reeves version (which used to
be the default). However, the Polak-Ribiere version (the current default) has little problem. The U toggles conjugate
gradient on and off, and ribiere toggles the Polak-Ribiere version. With Fletcher-Reeves conjugate gradient in effect,
the saddle point is reached at iteration 17 and area starts decreasing again until iteration 30, when it reaches 6.4486.
But then iteration stalls out, and the conjugate gradient mode has to be turned off and on to erase the history vector.
Once restarted, another 20 iterations will get the area down to 6.4334. In Polak-Ribiere mode, no restart is necessary.
Exercise for the reader: Get the Surface Evolver to display an unstable catenoid by declaring the catenoid facets to
be the boundary of a body, and adjusting the body volume with the b command to get zero pressure.
3.5 Example 4. Torus partitioned into two cells

This example has a flat 3-torus (i.e. periodic boundary conditions) divided into two bodies. The unit cell is a unit cube,
and the surface has the topology of Kelvins partitioning of space into tetrakaidecahedra [TW], which was the least
area partitioning of space into equal volumes known until recently [WP]. The datafile handles the wrapping of edges
23
24
Figure 3.4: Pair of Kelvin tetrakaidecahedra.
around the torus by specifying for each direction whether an edge wraps positively (+), negatively (-), or not at all (*).
The display of a surface in a torus can be done several ways. The connected command (my favorite) shows each
body as a single unit. The clipped command shows the surface clipped to the fundamental parallelpiped. The raw_cells
command shows the unedited surface.
The Weaire-Phelan structure [WP] is in the datafile phelanc.fe. It has area 0.3% less than Kelvins.
// twointor.fe
// Two Kelvin tetrakaidecahedra in a torus.
TORUS_FILLED
periods
1.000000 0.000000 0.000000
0.000000 1.000000 0.000000
0.000000 0.000000 1.000000
vertices
1 0.499733 0.015302 0.792314
2 0.270081 0.015548 0.500199
3 0.026251 0.264043 0.500458
4 0.755123 0.015258 0.499302
5 0.026509 0.499036 0.794636
6 0.500631 0.015486 0.293622
7 0.025918 0.750639 0.499952
8 0.499627 0.251759 0.087858
9 0.256701 0.499113 0.087842
10 0.026281 0.500286 0.292918
11 0.500693 0.765009 0.086526
12 0.770240 0.499837 0.087382
edges
1 12***
2 23***
24
25
3 14***
4 35***
5 26***
6 27*-*
7 18**+
8 46***
9 59**+
10 3 10 * * *
11 3 4 - * *
12 6 8 * * *
13 6 11 * - *
14 7 4 - + *
15 8 12 * * *
16 9 8 * * *
17 9 11 * * *
18 10 7 * * *
19 11 1 * + 20 12 5 + * 21 5 7 * * *
22 11 12 * * *
23 10 12 - * *
24 9 10 * * *
faces
1 1 2 4 9 16 -7
2 -2 5 12 -16 24 -10
3 -4 10 18 -21
4 7 15 20 -4 11 -3
5 -1 3 8 -5
6 6 14 -11 -2
7 5 13 -17 24 18 -6
8 -12 13 19 7
9 -16 17 22 -15
10 -10 11 8 12 15 -23
11 -21 9 17 19 1 6
12 -14 -18 23 -22 -13 -8
13 -24 -9 -20 -23
14 -19 22 20 21 14 -3
bodies
1 -1 -2 -3 -4 -5 9 7 11 -9 10 12 5 14 3 volume 0.500
2 2 -6 -7 8 -10 -12 -11 -13 1 13 -14 6 4 -8 volume 0.500
Doing some refining and iterating will show that the optimal shape is curved a bit.
3.6 Example 5. Ring around rotating rod

This example consists of a ring of liquid forming a torus around a rod rotating along its long axis (z axis) in weightlessness. The liquid has controllable contact angle with the rod. The interesting question is the stability of the ring as
the spin increases.
25
26
8
8
11
9
13
10
10
12
12
11
16
7
6
6
15
4
14
4
3
1
2
Figure 3.5: The ringblob.fe skeleton. This takes advantage of symmetry to do just the upper half. The rod passes
vertically through the hole in the middle. Only the free liquid surface need be done, since the contact angle on the rod
can specified by a constraint energy integral on edges 9-12, and the liquid/rod surface contributes zero to the volume
calculation.
The effect of the rotation is incorporated in the energy through an integral using the divergence theorem:
E =
1 2 2
r dV =
B2
1 2 2
(x + y2 )zk dA
B 2
where B is the region of the liquid, is the fluid density and is the angular velocity. Note the energy is negative,
because spin makes the liquid want to move outward. This has to be countered by surface tension forces holding the
liquid on the rod. If is negative, then one has a toroidal bubble in a rotating liquid, and high spin stabilizes the
torus. The spin energy is put in the datafile using the named quantity syntax (see below). centrip is a user-chosen
name for the quantity, energy declares that this quantity is part of the total energy, global_method says that the
following method is to be applied to the whole surface, facet_vector_integral is the pre-defined name of the method
that integrates vector fields over facets, and vector_integrand introduces the components of the vectorfield.
The rod surface is defined to be constraint 1 with equation x2 + y2 = R2 , where R is the radius of the rod. The
contact energy of the liquid with the rod is taken care of with an edge integral over the edges where the liquid surface
meets the rod:

z
zds = T cos()
T cos()dA = T cos()
E=
(yi x j) ds
R
S
S
S
Here S is the rod surface not included as facets in B, T is the surface tension of the free surface, and is the internal
contact angle.
Constraint 2 is a horizontal symmetry plane. By assuming symmetry, we only have to do half the work.
Constraint 3 is a one-sided constraint that keeps the liquid outside the rod. Merely having boundary edges on the
rod with constraint 1 is not enough in case the contact angle is near 180 and the liquid volume is large. Constraint
3 may be put on any vertices, edges, or faces likely to try to invade the rod. However, it should be noted that if you
26
27
put constraint 3 on only some vertices and edges, equiangulation will be prevented between facets having different
constraints.
Constraint 4 is a device to keep the vertices on the rod surface evenly spaced. Edges on curved constraints often
tend to become very uneven, since long edges short-cutting the curve can save energy. Hence the need for a way
to keep the vertices evenly spread circumferentially, but free to move vertically. One way to do that is with another
constraint with level sets being vertical planes through the z axis at evenly spaced angles. Constraint 4 uses the real
modulus function with arctangent to create a periodic constraint. Each refinement, the parameters need to be halved
to cut the period in half. This is done with the special rr refinement command at the end of the datafile. The regular
r refinement command should never be used on this file. Note that autorecalc is temporarily turned off to prevent
projecting vertices to the constraint when it is in an invalid state. Also note the /6 offset to avoid the discontinuity in
the modulus function. /6 was cleverly chosen so that all refinements would also avoid the discontinuity.
For detecting stability, one can perturb the torus by defining a command
perturb := set vertex y y + .01 where not on_constraint 1
This sets the y coordinate of each vertex to y + .01. For convenience, this command is defined in the read section at
the end of the datafile, where you can put whatever commands you want to execute immediately after the datafile is
loaded. To detect small perturbations, and get numerical values for the size of perturbations, the y moment of the liquid
is calculated in the named quantity ymoment. It is not part of the energy, as indicated by the info_only keyword.
You can see the value with the v command.
// ringblob.fe
// Toroidal liquid ring on a rotating rod in weightlessness.
// Half of full torus
// Using second periodic constraint surface intersecting rod to
// confine vertices on rod to vertical motion.
// Important note to user: Use only the rr command defined at
// the end of this file to do refinement. This is due to the
// nature of constraint 4 below.
// This permits drawing both halves of the ring
view_transforms 1
1000
0100
0 0 -1 0
0001
// Basic parameters. These may be adjusted at runtime with the
// A command. Only spin is being adjusted in these experiments.
parameter rodr = 1 // rod radius
parameter spin = 0.0 // angular velocity
parameter angle = 30 // internal contact angle with rod
parameter tens = 1 // surface tension of free surface
#define rode (-tens*cos(angle*pi/180)) // liquid-rod contact energy
parameter dens = 1 // density of liquid, negative for bubble
// spin centripetal energy
quantity centrip energy global_method facet_vector_integral
vector_integrand:
q1: 0
27
28
q2: 0
q3: -0.5*dens*spin*spin*(x^2+y^2)*z
// y moment, for detecting instability
quantity ymoment info_only global_method facet_vector_integral
vector_integrand:
q1: 0
q2: 0
q3: y*z
// Constraint for vertices and edges confined to rod surface,
// with integral for blob area on rod
constraint 1
formula: x^2 + y^2 = rodr^2
energy:
e1: -rode*z*y/rodr
e2: rode*z*x/rodr
e3: 0
// Horizontal symmetry plane
constraint 2
formula: z = 0
// Rod surface as one-sided constraint, to keep stuff from caving in
// Can be added to vertices, edges, facets that try to cave in
constraint 3 nonnegative
formula: x^2 + y^2 = rodr^2
// Constraint to force vertices on rod to move only vertically.
// Expressed in periodic form, so one constraint fits arbitrarily
// many vertices. Note offset to pi/6 to avoid difficulties with
// modulus discontinuity at 0.
parameter pp = pi/2 /* to be halved each refinement */
parameter qq = pi/6 /* to be halved each refinement */
constraint 4
formula: (atan2(y,x)+pi/6) % pp = qq
//initial dimensions
#define ht 2
#define wd 3
vertices
1 0 -wd 0 constraints 2 // equatorial vertices
2 wd 0 0 constraints 2
3 0 wd 0 constraints 2
4 -wd 0 0 constraint 2
5 0 -wd ht
// upper outer corners
6 wd 0 ht
7 0 wd ht
8 -wd 0 ht
9 0 -rodr ht constraints 1,4 // vertices on rod
28
29
10 rodr 0 ht constraints 1,4

11 0 rodr ht constraints 1,4
12 -rodr 0 ht constraints 1,4
edges
1 1 2 constraint 2 // equatorial edges
2 2 3 constraint 2
3 3 4 constraint 2
4 4 1 constraint 2
5 56
// upper outer edges
6 67
7 78
8 85
9 9 10 constraint 1,4 // edges on rod
10 10 11 constraint 1,4
13 1 5
// vertical outer edges
14 2 6
15 3 7
16 4 8
17 5 9
// cutting up top face
18 6 10
19 7 11
20 8 12
1 1 14 -5 -13 tension tens // side faces
2 2 15 -6 -14 tension tens // Remember you cant change facet tension
3 3 16 -7 -15 tension tens // dynamically just by changing tens; you have
4 4 13 -8 -16 tension tens // to do "tens := 2; set facet tension tens"
5 5 18 -9 -17 tension tens // top faces
6 6 19 -10 -18 tension tens
1 1 2 3 4 5 6 7 8 volume 25.28
read // some initializations
transforms off // just show fundamental region to start with
// special refinement command redefinition
r :::= { autorecalc off; pp := pp/2; qq := qq % pp; r; autorecalc on; }
// a slight perturbation, to check stability
perturb := set vertex y y+.01 where not on_constraint 1
hessian_normal // to make Hessian well-behaved
linear_metric // to normalize eigenvalues
29
30
3.7 Example 6. Column of liquid solder

4
6
2
6
2
16
17
15
18
14
13
10
11
10
9
11
12
8
12
7
Figure 3.6: The column skeleton.

Here we have a tiny drop of liquid solder that bridges between two parallel, horizontal planes, the lower at z = 0
and the upper at z = ZH. On each plane there is a circular pad that the solder perfectly wets, and the solder is perfectly
nonwetting off the pads. This would be just a catenoid problem with fixed volume, except that the pads are offset, and
it is desired to find out what horizontal and vertical forces the solder exerts. The surface is defined the same way as
in the catenoid example, except the upper boundary ring has a shift variable SHIFT in it to provide an offset in the
y direction. This makes the shift adjustable at run time. Since the top and bottom facets of the body are not included,
the constant volume they account for is provided by content integrals around the upper boundary, and the gravitational
energy is provided by an energy integral. One could use the volconst attribute of the body instead for the volume, but
then one would have to reset that every time ZH changed.
The interesting part of this example is the calculation of the forces. One could incrementally shift the pad, minimize
the energy at each shift, and numerically differentiate the energy to get the force. Or one could set up integrals to
calculate the force directly. But the simplest method is to use the Principle of Virtual Work by shifting the pad,
recalculating the energy without re-evolving, and correcting for the volume change. Re-evolution is not necessary
because when a surface is at an equilibrium, then by definition any perturbation that respects constraints does not
change the energy to first order. To adjust for changes in constraints such as volume, the Lagrange multipliers (pressure
for the volume constraint) tell how much the energy changes for given change in the constraints:
DE = LT DC
where DE is the energy change, L is the vector of Lagrange multipliers and DC is the vector of constraint value
changes. Therefore, the adjusted energy after a change in a parameter is
Ead j = Eraw LT DC
where Eraw is the actual energy and DC is the vector of differences of constraint values from target values. The
commands do_yforce and do_zforce in the datafile do central difference calculations of the forces on the top pad, and
put the surface back to where it was originally. Note that the perturbations are made smoothly, i.e. the shear varies
linearly from bottom to top. This is not absolutely necessary, but it gives a smoother perturbation and hence a bit more
accuracy.
30
31
// column.fe
// example of using facet_general_integral to measure restoring force
// of tiny column of liquid solder in shape of skewed catenoid.
// All units cgs
parameter RAD = 0.05 // ring radius
parameter ZH = 0.04
// height
parameter SHI = 0.025 // shift
#define SG 8
// specific gravity of solder
#define TENS 460 // surface tension of solder
#define GR 980 // acceleration of gravity
quantity grav_energy energy modulus GR*SG global_method gravity_method
method_instance dvoly method facet_vector_integral modulus -1
vector_integrand:
q1: 0
q2: (z+ZH)/2/ZH
q3: 0
// following method instances and quantity for aligning force
method_instance dareay method facet_general_integral
scalar_integrand: TENS*x6*x5/2/ZH/sqrt(x4^2+x5^2+x6^2)
method_instance dgrav method facet_vector_integral
vector_integrand:
q1: 0
q2: -0.25*GR*SG*z^2/ZH
q3: 0
quantity forcey info_only global_method dareay global_method dvoly
global_method dgrav
BOUNDARY 1 PARAMETERS 1
X1: RAD*cos(P1)
X2: RAD*sin(P1)
X3: ZH
BOUNDARY 2 PARAMETERS 1
X1: RAD*cos(P1)
X2: RAD*sin(P1) + SHI
X3: -ZH
vertices // given in terms of boundary parameter

31
32

edges
13 1 7
14 2 8
15 3 9
16 4 10
17 5 11
18 6 12
faces
1 1 14 -7 -13 density TENS
bodies
1 -1 -2 -3 -4 -5 -6 volume 0.000275 volconst 2*pi*RAD^2*ZH
read
hessian_normal
dodiff := { olde := total_energy; shi := shi + .0001 ; olddarea := dareay.value;
conj_grad; ribiere; g 40;
printf "Actual force: %2.15g\\n", (total_energy-olde)/.0001;
printf "darea force: %2.15g\\n", (dareay.value/2+olddarea/2)/2;
}
3.8 Example 7. Rocket fuel tank

This example discusses a cylindrical flat-ended tank in low gravity with some liquid inside. Complete understanding
of this example qualifies you as an Evolver Power User.
32
33
8
16
7
4
2
8
15
10
1
5
3
14
11
6
2
4
5
12
13
6
Figure 3.7: The tank skeleton, left, and surface, right. Only the free surface needs to be explicitly represented.
The initial configuration is shown in the figure. The axis is horizontal with the base at the left. We will assume
the fuel is in a ring around the base of the tank. The fuel comprises one body, and the boundary of the body has four
components:
1. The covered part of the flat circular base at y = 0, implemented as constraint 1..
2. The cylindrical wall at x2 + z2 = R2 , implemented as constraint 2..
3. The triangular facets (8 of them in the figure).
4. The gaps between the edges on the wall (the gap edges, edges 5,6,7,8 in the figure) and the wall. We will
assume that the gap surfaces are generated by lines radially outward from the central axis (the y axis) through
the points of the gap edges.
To specify contact angles, let B be the interior contact angle with the base, and let W be the interior contact angle
with the cylindrical wall.
3.8.1
Surface energy
Base
The surface energy here comes from the contact angle. The angle is represented by pretending the uncovered base has
surface tension TB = cos(B ). That we handle with an energy integral on constraint 1. We want

E=
base
TBj dA.
(3.1)
By Stokes theorem, this is the integral around the edges in the base (edges 1,2,3,4)
E=
edges
TB zi ds,
(3.2)
where the edges are oriented oppositely from the figure. Hence the energy integrand for constraint 1 should be TB zi.
33
34
Wall
The contact angle is represented by pretending the uncovered wall has surface tension TW = cos(W ). The uncovered
wall in infinite, so we will assume equivalently that the covered wall has tension TW . The energy is

E=
wall
TW dA =
TW (xi + zk) dA.

R
wall
(3.3)
We want to turn this into an integral over the boundary of the wall, which is made of two paths, the base circumference
and the gap trace where the gap surfaces meet the wall. This calls for a vector potential, but our integrand above
does not have zero divergence. But we could choose a divergenceless integrand that is equal on the wall:

E=
wall
where
F = TW
F dA.
(3.4)
R
(xi + zk).
(x2 + z2 )
(3.5)
Now we can get a vector potential w

w = TW
Ry
(x2 + z2 )
Hence
E=
basecircum
(zi + xk),
w ds +
w = F.
(3.6)
w ds.
(3.7)
gap trace
The first of these is zero since y = 0 there. The second is equal to the integral over the gap edges, since w is
parallel to the gap surfaces! So
E=
gap edges
w ds,
(3.8)
where the gap edges are oppositely oriented from the figure. Hence the constraint 2 energy integrand includes w.
The facets
The surface tension energy here is automatically taken care of by the Evolver. We will leave the surface tension as the
default value of 1.
The gaps
We will declare constraint 2 to be CONVEX so that Evolver will include a gap surface energy for all the gap edges. We
put the spring constant to be 1, as that value give the best approximation to the true area for small gaps.
3.8.2
Volume
where F = 1. The integral

F dA
The body volume V =
dV can be calculated as a surface integral V =
over the facets is handled automatically, but there are two choices for F. We will take F = (xi + yj + zk)/3 by putting
SYMMETRIC_CONTENT in the top of the datafile. (This choice works better for the tank on its side; for an upright
tank, the default F = zk is easier.) The surface integral has four components:
Base
This part of the integral is zero, since y = 0 and the normal is j.
34
35
The wall
Here the j component of F can be dropped, and the rest put in the equivalent divergenceless form
FW =
R2
(xi + zk).
3(x2 + z2 )
(3.9)
This has vector potential

R2 y
(zi + xk),
w = FW .
(3.10)
3(x2 + z2 )
By the same reasoning as for wall area, this reduces to the integral of w over the gap edges. Thus we include w in
the content integrand of constraint 2.
w =
Facets
Evolver takes care of these automatically.
Gaps
We subtract some radial component from F to get a divergenceless vector FG with the same surface integral over the
gap surfaces:
FG = 1 xi + 1 yj 1 zk.
(3.11)
6
3
6
This has vector potential
1
1
w = zyi xyk.
(3.12)
6
6
Hence

F dA =
w ds +
w ds.
(3.13)
gaps
gap trace
gap edges
The second part we can put directly into the constraint 2 content integrand (with opposite orientation of the edges).
The gap trace part we convert to an equivalent integrand v:
v =
R2 y
(zi xk).
6(x2 + z2 )
which has radial curl. Hence the gap trace integral is the same as
orientation as the figure.
3.8.3
(3.14)
v ds,
gap edges
this time with the edges in the same
Gravity
Just for fun, were going to put in gravity in such a way that the user can make it act in any direction. Let the
= gyj + gzk. gy and gz will be adjustable constants. Let the fuel density be . Then the
acceleration of gravity be G
gravitational potential energy is
E =
body
First we change this to a surface integral

E =
It turns out convenient to take
xdV =
G

body
F dA,
body
(gy y + gz z)dV.
F = gy y + gz z.
(3.15)
(3.16)
F = (gy y + gz yz)j.
2
The surface integral has four components:
35
(3.17)
36
Base
This part is zero since y = 0.
Wall
This part is also zero since F is parallel to the wall.
Facets
This part is taken care of by installing F as the integrand for named quantity arb_grav for the facets.
Gaps
We get an equivalent divergence-free integrand FG by adding a radial component to F:
)
( 2
(
)
FG = gy y j y (xi + zk) + gz yzj z (xi + zk) .
2
2
3
This has vector potential w
( 2
)
y
yz
w = gy + gz
(zi xk).
4
3
(3.18)
(3.19)
This is integrated over the gap edges with negative orientation, so w goes into the energy integrand of constraint 2,
and over the gap trace. For the gap trace part, we construct an equivalent wG that has radial curl:
( 2
)
y
R2
yz
R3
wG = gy
(zi xk),
(3.20)
+
g
z
4 (x2 + y2 )
3 (x2 + z2 )(3/2)
so
gap trace
=
w ds
gap edges
wG ds.
(3.21)
Here the integral is over the same orientation as in the figure, so wG also goes into the energy integrand of constaint
2.
3.8.4
Running
Refine once before iterating. Iterate and refine to your hearts content. You might try command u once in a while to
regularize the triangulation. Play around with command A to change gravity and wall contact angles. Keep an eye on
the scale factor. If it gets towards zero, check out the t and w histograms.
// tankex.fe
// Equilibrium shape of liquid in flat-ended cylindrical tank.
// Tank has axis along y-axis and flat bottom in x-z plane. This
// is so gravity acting vertically draws liquid toward wall.
// Straight edges cannot conform exactly to curved walls.
// We need to give them an area so that area cannot shrink by straght edges
// pulling away from the walls. The gaps are also accounted for
// in volume and gravitational energy.
SYMMETRIC_CONTENT
// for volume calculations
36
37
// Contact angles, initially for 45 degrees.

PARAMETER ENDT = 0.707 /* surface tension of uncovered base */
PARAMETER WALLT = 0.707 /* surface tension of uncovered side wall */
// Gravity components
PARAMETER GY = 0
PARAMETER GZ = -1
SPRING_CONSTANT 1 // for most accurate gap areas for constraint 2
#define TR 1.00
#define RHO 1.00
/* tank radius */
/* fuel density */
constraint 1 // flat base

function: y = 0
energy:
// for contact energy line integral
e1: -ENDT*z
e2: 0
e3: 0
#define wstuff (WALLT*TR*y/(x^2 + z^2)) // common wall area term
#define vstuff (TR^2/3*y/(x^2 + z^2)) // common wall volume term
#define gstuff (GY*TR^2*y^2/4/(x^2 + z^2) + GZ*TR^3*y*z/3/(x^2+z^2)**1.5)
// common gap gravity term
constraint 2 CONVEX
// cylindrical wall
function: x^2 + z^2 = TR^2
energy:
// for contact energy and gravity
e1: -wstuff*z + RHO*GY*y^2*z/4 + RHO*GZ*y*z^2/3 - RHO*gstuff*z
e2: 0
e3: wstuff*x - RHO*GY*y^2*x/4 - RHO*GZ*y*z*x/3 + RHO*gstuff*x
content:
// so volumes calculated correctly
c1: vstuff*z - z*y/6 + vstuff*z/2
c2: 0
c3: -vstuff*x + x*y/6 - vstuff*x/2
// named quantity for arbitrary direction gravity on facets
quantity arb_grav energy method facet_vector_integral global
vector_integrand:
q1: 0
q2: -RHO*GY*y^2/2 - RHO*GZ*y*z
q3: 0
// Now the specification of the initial shape

vertices
1 0.5 0.0 0.5 constraint 1
2 0.5 0.0 -0.5 constraint 1
3 -0.5 0.0 -0.5 constraint 1
37
38
4 -0.5 0.0 0.5 constraint 1

5 1.0 0.5 0.0 constraint 2
6 0.0 0.5 -1.0 constraint 2
7 -1.0 0.5 0.0 constraint 2
8 0.0 0.5 1.0 constraint 2
edges
1 2 1 constraint 1
2 1 4 constraint 1
3 4 3 constraint 1
4 3 2 constraint 1
5 5 6 constraint 2
6 6 7 constraint 2
7 7 8 constraint 2
8 8 5 constraint 2
9 1 8
10 1 5
11 2 5
12 2 6
13 3 6
14 3 7
15 4 7
16 4 8
faces
1 13 6 -14
2 3 14 -15
3 15 7 -16
4 2 16 -9
5 9 8 -10
6 1 10 -11
7 11 5 -12
8 4 12 -13
bodies
1 1 2 3 4 5 6 7 8 volume 0.6 density 0 /* no default gravity */
3.9 Example 8. Spherical tank

This example discusses a sperical tank in low gravity with some liquid inside. Complete understanding of this example
qualifies you as an Evolver Power User.
The initial configuration is shown above. The liquid comprises one body, and the boundary of the body has three
components:
1. The sphere at x2 + y2 + z2 = R2 , implemented as constraint 1. ( Actually, implemented as x2 + y2 + z2 ) = R

so that projection to the constraint is linear in the radius, which gives much faster convergence.)
2. The faceted surface (initially, the single face).
3. The gaps between the edges on the sphere (the gap edges, edges 1,2,3,4 in the figure) and the sphere. We will
38
39
5
8
9
3
8
5
12
2
10
15
7
2
3
9
4
14
10
16
11
1
13
7
1
Figure 3.8: The liquid surface (equatorial plane) and the rear hemisphere.
assume that the gap surfaces are generated by lines radially outward from the center of the sphere through the
points of the gap edges.
We will often be looking for vectorfields that have radial curl, since their integral over the gap edges will be the
same as over the gap trace. As a useful fact for all the curls below, note that if
w =
zn
(yi xj)
(x2 + y2 )(x2 + y2 + z2 )n/2
then
w =
nzn1
(xi + yj + zk).
(x2 + y2 + z2 )1+n/2
(3.22)
(3.23)
To specify contact angles, let W be the interior contact angle with the spherical wall.
3.9.1
Surface energy
Wall
The contact angle is represented by pretending the uncovered wall has surface tension TW = cos(W ). We will assume
equivalently that the covered wall has tension TW . The energy is

E=
wall
TW dA =

wall
TW (xi + yj + zk) dA.

R
(3.24)
We want to turn this into an integral over the boundary of the covered wall, the gap trace where the gap surfaces
meet the wall. This calls for a vector potential, but our integrand above does not have zero divergence. But we could
choose a divergenceless integrand that is equal on the wall:

E=
wall
39
F dA,
(3.25)
40
where
F = TW
R2
(xi + yj + zk).
(x2 + y2 + z2 )3/2
(3.26)
Now we can get a vector potential w (see useful fact above with n = 1)
w = TW
R2 z
(x2 + y2 )(x2 + y2 + z2 )1/2
Hence
E=
gap trace
(yi xj),
w = F.
w ds.
(3.27)
(3.28)
This is equal to the integral over the gap edges, since w is parallel to the gap surfaces. So
E=
gap edges
w ds,
(3.29)
where the gap edges are oppositely oriented from the figure. Hence the constraint energy integrand includes w.
Note the potential w is zero at z = 0, so it is actually measuring area above the equator. So one would need to add a
hemispheres worth of energy 2TW R2 /3.
The facets
The surface tension energy here is automatically taken care of by the Evolver. We will leave the surface tension as the
default value of 1.
The gaps
We will declare constraint 1 to be CONVEX so that Evolver will include a gap surface energy for all the gap edges. We
put the spring constant to be 1, as that value give the best approximation to the true area for small gaps.
3.9.2
Volume
where F = 1. The integral

F dA
The body volume V =
dV can be calculated as a surface integral V =
over the facets is handled automatically, but there are two choices for F. We will take F = (xi + yj + zk)/3 by putting
SYMMETRIC_CONTENT in the datafile. Thus
V=
1
3

body sur f ace
(xi + yj + zk) dA.
(3.30)
The surface integral has three components:

The wall
Like wall energy, this is first put in the equivalent divergenceless form
R3
F = 1
(xi + yj + zk).
3 (x2 + y2 + z2 )3/2
(3.31)
This has vector potential (again useful fact above with n = 1)

w =
R3 z
1
(yi xj),
2
2
3 (x + y )(x2 + y2 + z2 )1/2
w = F.
(3.32)
By the same reasoning as for wall area, this reduces to the integral of w over the gap edges. Thus we include w
in the content integrand of constraint 1. However, we also note a potential problem at z = R where x2 + y2 = 0.
40
41
We must be sure that this point always remains within the covered surface. As long as it does, it will have a constant
effect. In fact, we see that at z = 0 the integrand w is zero, and the other volume contributions are zero at z = 0, so the
singularity at z = R must have a contribution to the volume of half the volume of the sphere. We can account for this
by using VOLCONST 2*pi*R^3/3 with the body.
Facets
Evolver takes care of these automatically.
Gaps
The integrand is parallel to the gap surfaces, hence this contribution is zero.
3.9.3
Gravity
The sphere being very symmetric, all directions of gravity are the same. Therefore we will use Evolvers default
gravity. Let the magnitude of gravity be G (downwards). Let the liquid density be . Then the gravitational potential
energy is

E =G
zdV.
(3.33)
body
First we change this to a surface integral

E = G
body
F dA,
F = z.
(3.34)
The default gravity mechanism uses

2
F = G z k.
2
(3.35)
The surface integral has three components:

Wall
First, we convert to the equivalent divergence-free integrand
F = G
R4 z3
(xi + yj + zk).
2(x2 + y2 + z2 )3
(3.36)
Note this has the same dot product with xi + yj + zk on the surface of the sphere. The vector potential is
w = G
R4 z4
8(x2 + y2 )(x2 + y2 + z2 )2
So
E=
(yi xj),
w = F.
(3.37)
GR .
w ds
4
gap trace
(3.38)
The constant term here is the gravitational energy of the lower hemisphere, since w = 0 when z = 0. It will not show
up in the Evolver energy; you must remember to adjust for it yourself. Due to the radial curl of w, the integral is also
the integral over the gap edges. The orientation is opposite that given for the gap edges, so w goes in the energy for
constraint 1.
Facets
The integrals over facets are automatically taken care of by the Evolver.
41
42
Gaps
We get an equivalent divergence-free integrand FG by adding a radial component (z/4)(xi + yj + zk) to F:
FG = G (xzi + yzj + z2k).
4
(3.39)
This has vector potential w

Gz2
(yi x j).
(3.40)
8
This is integrated over the gap edges with opposite orientation, so w goes into the energy integrand of constraint 1,
and over the gap trace. For the gap trace part, we construct an equivalent wG that has radial curl (using two versions
of useful fact with n = 2 and n = 4 to show radial curl):
w =
wG
=
=
=
G z2 (R2 z2 )
(yi x j)
8 (x2 + y2 )
(
)
G
z2 R4
z4 R4
(yi xj)
8 (x2 + y2 )(x2 + y2 + z2 ) (x2 + y2 )(x2 + y2 + z2 )2
G
z2 R4
(yi xj)
2
8 (x + y2 + z2 )2
so
gap trace
=
w ds
(3.41)
gap edges
wG ds.
(3.42)
Here the integral is over the same orientation as in the figure, so wG also goes into the energy integrand of constaint 1.
3.9.4
Running
Refine once before iterating. Iterate and refine to your hearts content. You might try command u once in a while to
regularize the triangulation. Play around with command A to change gravity and wall contact angles. Keep an eye on
the scale factor. If it gets towards zero, check out the t and w histograms.
// sphere.fe
// Spherical tank partially filled with liquid with contact angle.
SYMMETRIC_CONTENT // natural for a sphere
parameter rad = 1.00 // sphere radius
parameter angle = 45 // contact angle, in degrees
// cosine of contact angle
#define WALLT cos(angle*pi/180)
// density of body
#define RHO 1.0
// Various expressions used in constraints
#define wstuff (rad^2*z/(x^2 + y^2)/sqrt(x^2+y^2+z^2))
#define gstuff (G*RHO/8*rad^4*z^4/(x^2 + y^2)*(x^2+y^2+z^2)^2)
#define gapstuff (G*RHO*rad^4/8*z^2/(x^2+y^2+z^2)^2)
42
43
constraint 1 convex // the sphere, as wall for liquid

formula: sqrt(x^2 + y^2 + z^2) = rad
energy:
e1: WALLT*wstuff*y - gstuff*y + G*RHO/8*y*z^2 - gapstuff*y
e2: -WALLT*wstuff*x + gstuff*x - G*RHO/8*x*z^2 + gapstuff*x
e3: 0
content:
c1: -rad*wstuff*y/3
c2: rad*wstuff*x/3
c3: 0
constraint 2 // the sphere, for display only
formula: sqrt(x^2 + y^2 + z^2) = rad
vertices:
1 rad 0 0
2 0 rad 0
3 -rad 0 0
4 0 -rad 0
5 0 0 rad
6 0 0 -rad
7 0 rad 0
8 -rad 0 0
9 0 -rad 0
edges:
1 1 2
2 2 3
3 3 4
4 4 1
5 5 9
6 9 6
7 6 7
8 7 5
9 5 8
10 9 8
11 6 8
12 7 8
constraint 1
constraint 1
constraint 1
constraint 1
constraint 2 FIXED // to show back hemisphere
constraint 2 FIXED
constraint 2 FIXED
constraint 2 FIXED
constraint 2 FIXED
constraint 1
constraint 1
constraint 1
constraint 1
constraint 2 FIXED // to show back hemisphere
constraint 2 FIXED
constraint 2 FIXED
constraint 2 FIXED
constraint 2 FIXED
constraint 2 FIXED
constraint 2 FIXED
constraint 2 FIXED
faces:
1 1234
2 5 10 -9 constraint 2 density 0 FIXED // to show back hemisphere
3 -10 6 11 constraint 2 density 0 FIXED
4 -11 7 12 constraint 2 density 0 FIXED
5 8 9 -12 constraint 2 density 0 FIXED
bodies: // start with sphere half full
1 1 volume 2*pi*rad^3/3 volconst 2*pi*rad^3/3 density RHO
read
43
44
// Typical short evolution

gogo := { g 5; r; g 5; r; g 5; r; g 10; conj_grad; g 20; }
3.10 Example 9. Crystalline integrand

This example shows how to use Wulff vectors for a crystalline integrand. In general, the Wulff vector at a point on
a surface is a vector that is dotted with the surface normal to give the energy density of the surface. For ordinary
area, it is the unit normal itself. But if a finite set of Wulff vectors is given, and the one used at a point is the one
with the maximum dot product, then the minimal surface has only flat faces, i.e. it is crystalline. The crystal shape in
this example is an octahedron. The Wulff vectors are the vertices of the octahedron, and are in octa.wlf. The datafile
crystal.fe describes a cube with extra edges around the equator. These edges provide places for octahedron edges to
form. Things work a lot better if the facet edges are where the crystal edges want to be.
To see the cube turn into an octahedron, refine once, use the m command to set the scale to 0.05, and iterate about
50 times. Fixed scale works much better than optimizing scale for crystalline integrands, since crystalline energy is
not a smooth function of the scale factor. Sometimes you have to put in jiggling also to get over local obstacles to
minimizing energy.
Here is the Wulff vector file octa.wlf:
0
0
1
0
0 -1
0.707 0.707
-0.707 0.707
0.707 -0.707
-0.707 -0.707
0
0
0
0
Here is the datafile:

// crystal.fe
// Evolver data for cube of prescribed volume with
// crystalline surface energy. Evolves into an octahedron.
Wulff "octa.wlf"
// Wulff vectors for octahedral crystal
vertices
1 0.0 0.0 0.0
2 1.0 0.0 0.0
3 1.0 1.0 0.0
4 0.0 1.0 0.0
5 0.0 0.0 1.0
6 1.0 0.0 1.0
7 1.0 1.0 1.0
8 0.0 1.0 1.0
9 0.0 0.0 0.5
10 1.0 0.0 0.5
11 1.0 1.0 0.5
12 0.0 1.0 0.5
edges /* given by endpoints and attribute */
1 12
2 23
44
45
3 34
4 41
5 56
6 67
7 78
8 85
9 19
10 2 10
11 3 11
12 4 12
13 9 5
14 10 6
15 11 7
16 12 8
17 9 10
18 10 11
19 11 12
20 12 9
1 1 10 -17 -9
2 2 11 -18 -10
3 3 12 -19 -11
4 4 9 -20 -12
5 5 6 7 8
6 -4 -3 -2 -1
7 17 14 -5 -13
8 18 15 -6 -14
9 19 16 -7 -15
10 20 13 -8 -16
1 1 2 3 4 5 6 7 8 9 10 volume 1
3.11 Tutorial in Advanced Calculus

Some Evolver users may be a bit rusty in advanced calculus, so this section states the basic theorems and shows how
to apply them to find appropriate edge integrals for area and volume.
Notation: If S is an oriented surface or plane region, then S is its oriented boundary, oriented counterclockwise
with respect to the unit normal vector N.
Greens Theorem: If S is a plane region and w(x, y) = M(x, y)i + N(x, y)j is a vectorfield, then
=
w dl
N M
dx dy.
y
S x
Greens Theorem is actually a special case of Stokes Theorem:

Stokes Theorem: If S is an oriented surface in R3 and w(x, y, z) = wx (x, y, z)i+wy (x, y, z)j +wz (x, y, z)k is a vectorfield,
then

=
curl w N dA
w dl
S
45
46
([
where
curl w = w =
] [
] [
] )
wz wy wx wz wy wx
i,
j,
k .
y
z
z
z
x
y
Notation: If B is a region in R3 , then B is its boundary surface, oriented with outward unit normal N.
Divergence Theorem: If B is a region in R3 and w is a vectorfield, then

divw dx dy dz =
where
divw = w =
w N dA,
wx wy wz
+
+
.
x
y
z
Theorem: div curl w = 0.

Application to volume integrals.
Suppose B is a body we wish to calculate the volume of. The volume can be defined as the triple integral

V=
1 dx dy dz.
B
Evolver cant do volume integrals, but it can do surface integrals, so Evolver uses the Divergence Theorem with w = zk,
since div zk = 1:

V=
zk N dA.
B
Taking an upward pointing normal to the surface, this can be written as

V=
z dx dy.
It is often inconvenient and unnecessary to integrate over all of B. For example, any portion lying in the plane z = 0
may obviously be omitted, along with any portion that is vertical, i.e. has a horizontal tangent N. Other parts of B
may be omitted if we specifically include compensating integrals. Suppose a portion S of B is part of a constraint
level set F(x, y, z) = 0. Then we want to use Stokes Theorem to rewrite

S
zk N dA =
w dl
for some vectorfield w with zk = curl w. Unfortunately, div zk = 0, so we cant do that. We have to get an equivalent
form of zk valid on S. Clearly the equivalent form cannot contain z as such, so we have to solve the implicit function
F(x, y, z) = 0 explicitly for z = f (x, y). Then it remains to solve f (x, y)k = curl w for w. There are many possible
solutions to this (all differing by the gradient of some scalar function), but one convenient solution is to take wx =
M(x, y), wy = N(x, y), and wz = 0, with
N M
= f (x, y).
x
y
There are still many solutions of this. One could take partial antiderivatives
M(x, y) = 0,
or
M(x, y) =
N(x, y) =
f (x, y)dx,
f (x, y)dy,
N(x, y) = 0,
or some combination thereof. The choice is guided by the principle that Mi + Nj should be perpendicular to the
portions of S that are not included in edge constraint integrals.
46
47
Often, the problem has rotational symmetry around the z axis. Then it is usually convenient to choose a vectorfield
w with rotational symmetry, of the form
w = g(t)(yi + xj),
where t = x2 + y2 . Using the Chain Rule on t, we get
N M
= g (t) 2x x + g(t) + g (t) 2y y + g(t) = 2g (t)t + 2g(t) = 2(tg(t)) .

x
y
Writing f (x, y) = h(t) , the solution of
2(tg(t)) = h(t)
is
1
h(t)dt.
2t
Symbolic integration programs like Mathematica or Maple can be useful in doing this indefinite integral. The result
needs to have t replaced with x2 + y2 , and the relation F(x, y, z) = 0 is often useful in simplifying the result. Great care
may be needed to get the result into a form that is valid everywhere, particularly if there are square roots of ambiguous
sign. Always check your results for reasonableness, and check any datafiles to make sure the initial values for volumes
are what you expect.
Example: Hyperboloid constraint x2 + y2 z2 = R2 . Here the surface is a disk spanning the hyperboloid, and
the body is the region between the z = 0 plane and the disk. The edge integral must compensate for the volume below
the surface but outside the hyperboloid. We have z = f (x, y) = (x2 + y2 R2 )1/2 , so h(t) = (t R2 )1/2 , and integrating
gives
1 2
g(t) =
(t R2 )3/2 .
2t 3
Hence
(x2 + y2 R2 )3/2
w =
(yi + xj).
3(x2 + y2 )
g(t) =
Using the hyperboloid equation, this simplifies to

w =
z3
3(x2 + y2 )
(yi + xj).
Notice this form is smooth everywhere on the surface and zero when z = 0. When integrated over the disk boundary,
it properly gives negative volumes for curves with negative z. In the datafile, this vector would go in the constraint
integral with reversed signs, since we want to subtract the volume under the hyperboloid from the volume under the
disk.
Application to area. Here we want to calculate the area on a constraint surface S bounded by a curve S for
the purpose of prescribing the contact angle of the curve. We want to use Stokes Theorem to rewrite the area as a
boundary line integral:
curl w dl.
v N dA =
1 dA =
A=
S
To this end, we need to find a vectorfield v with v N = 1 and divv = 0. Again, one way to do this is to solve for the
surface as the graph of z = f (x, y) and use v = (1 + fx2 + fy2 )1/2k to write the area as

S
(1 + fx2 + fy2 )1/2 dx dy,
and use Greens Theorem. We need to solve

N M
= (1 + fx2 + fy2 )1/2 .

x
y
47
48
The same remarks as in the volume case apply.

Example: Hyperboloid constraint x2 + y2 z2 = R2 . We have z = f (x, y) = (x2 + y2 R2 )1/2 , so
[
(1 + fx2 + fy2 )1/2 = 1 +
x 2 + y2
2
x + y2 R2
]1/2
.
Using the rotationally symmetric approach as in the volume example, we have

h(t) = [1 + t/(t R2 )]1/2 ,
and
1
2t
Mathematica gives the integral in the horrible form
g(t) =
[1 + t/(t R2 )]1/2 dt.
R2 2t
R2 2t
2 log(3 R2 + 2 2 R2
+
4t
2
R
2 R2 t t)
( 2 )
R2 t
R
+
t
.
R2 t
2 2
R2 2t
Taking advantage of the fact that t R2 , this becomes

R2 log(3 R2 + 2 2 (2t R2 )(t R2 ) + 4t)

2
2
(2t R )(t R )
.
2 2
And using t R2 = z2 ,
R2 log(3 R2 + 2 2 z 2t R2 ) + 4t)
2
z 2t R
.
2 2
Adding a constant of integration to make this 0 at z = 0 gives
z
)
)

R2 ( (
2t R2 log 4t 3 R2 + 2 2 z 2t R2 log(R2 ) .
2 2
This is in fact an odd function of z. The final vectorfield is

( (
)
)

2 2z 2(x2 + y2 ) R2 R2 log 4(x2 + y2 ) 3 R2 + 2 2 z 2(x2 + y2 ) R2 log(R2 )
(yi + xj).
w =
4 2(x2 + y2 )
48
Chapter 4
Geometric elements
The surface is defined in terms of its geometric elements of each dimension. Each element has its own set of attributes.
Some may be set by the user; others are set internally but may be queried by the user. It is also possible to dynamically
define extra attributes for any type of element, which may be single values or vectors of values. Attribute values can
be specified in the datafile, and queried and sometimes set with commands. The types of elements are:
Vertices
Edges
Facets
Bodies
Facetedges
This chapter first defines each type and summarizes their attributes, then has a detailed list of the attributes of each.
4.1 Vertices
A vertex is a point in space. The coordinates of the vertices are the parameters that determine the location of the
surface. It is the coordinates that are changed when the surface evolves. A vertex carries no default energy, but may
have energy by being on a level set constraint in the string model, or by having a named quantity energy applied to it.
The vertices of the original surface are defined in the vertices section of the datafile.
Vertex attributes (grouped somewhat relatedly):
id
original
boundary
mean_curvature
coordinates, x,y,z, __x[] parameter values, p1,p2, p[] square mean curvature
vertex edges
on_boundary
mid_edge
vertex facets
v_boundary
mid_facet
valence
extra attributes
axial_point
bare
named quantities
triple_point
fixed
on_quantity
tetra_point
constraints
on_method_instance
v_force
on_constraint
v_method_list
v_velocity
hit_constraint
vertex_normal
raw_velocity
value_of_constraint
dihedral
v_oldx
v_constraint_list
constraint normal
49
50
4.2 Edges
An edge is a one-dimensional geometric element. In the linear model, an edge is an oriented line segment between a
tail vertex and a head vertex. In the quadratic model, an edge is defined by quadratic intepolation of two endpoints and
a midpoint. In the lagrange model, an edge is defined by the appropriate order interpolation with the edge vertices. In
thestring model, edges carry a default surface tension energy proportional to their length. Edges may also carry energy
by being on level set constraints in the soapfilm model, or by having named quantity energies applied to them. The
edges of the original surface are defined in the edges section of the datafile.
Edge attributes (grouped somewhat relatedly>:
on_constraint
orientation
id
e_constraint_list
frontbody
oid
boundary
backbody
original
on_boundary
dihedral
length
e_boundary
noncontent
density or tension
color
quantities
fixed
edge_vector, x,y,z
on_quantity
vertices
no_refine
method instances
midv
no_transform
on_method_instance
facets
wrap
e_method_list
valence
wrap_list
extra attributes
bare
show
constraints
4.3 Facets
In thesoapfilm model, a facet is an oriented triangle defined by a cycle of three edges. In thelinear model, a facet is
a flat triangle. In thequadratic model, the facet is a curved surface defined by quadratic interpolation among the three
facet corner vertices and the three edge midpoints. In the Lagrange model, lagrange_order interpolation is done among
(lagrange_order+1)(lagrange_order+2)/2 vertices. Although individual facets are oriented, there are no restrictions on
the orientations of adjacent facets. By default, a facet carries a surface tension energy equal to its area.
In thestring model, a facet is a chain of an arbitrary number of edges. The chain need not be closed. Usually a
facet is defined in the string model in order to define a body, so the space dimension is 2 and the facet is planar, one
facet corresponding to a body. Facets carry no energy by themselves.
In thesimplex model, a facet is a simplex of dimension surface_dimension defined by surface_dimension+1 vertices. The surface_dimension may be any dimension less than or equal to the space_dimension. The simplex is oriented
according to the order of the vertices. By default, a simplex carries a surface tension energy proportional to its volume.
Facets may carry additional energy by having named quantity energies applied to them.
The facets of the original surface are defined in the faces section of the datafile.
Facet attributes (grouped somewhat relatedly):
50
id
oid
original
density or tension
area
fixed
vertices
edges
bodies
frontbody
backbody
f_body_list
valence
51
constraints
on_constraint
f_constraint_list
boundary
on_boundary
f_boundary
color
frontcolor
backcolor
opacity
facet_normal, x,y,z
no_display
no_refine
no_transform
orientation
noncontent
phase
quantities
on_quantity
method instances
on_method_instance
f_method_list
extra attributes
f_next_vfacet
f_next_bfacet
4.4 Bodies
A body is a full-dimensional region of space. Bodies are not triangulated. Rather, they are determined by their
boundary facets (oredges in 2D). These facets are used for calculating body volume and gravitational energy. Only
those facets needed for correct calculation need be given. In the string model, usually a body corresponds to one facet.
Bodies of the original surface are defined in the bodies section of the datafile.
Body attributes:
volconst
id
actual_volume
original
pressure
facets
phase
density
centerofmass
volume
extra attributes
target
volfixed
4.5 Facetedges
A facetedge is a pairing of a facet and one of its edges, with orientation such that the edge orientation is consistent
with the facet orientation. Facetedges are used internally by Evolver, and are seldom of interest to the user. They
carry no energy. The C integrity-checking command will sometimes refer to facetedges if the surface is inconsistent.
Facetedge can be used as an element generator.
Facetedge attributes:
id
prevedge
edge
nextfacet
facet
prevfacet
nextedge extra attributes
4.6 Common element attributes

These are attributes that apply to all types of elements (or almost all).
id
Geometric element read-only integer attribute. The id of an element is a positive integer uniquely associated with
that element. The Evolver will assign ids to elements read from the datafile in the order they are read, unless the -i
51
52
command line option was used or keep_originals is in the top of the datafile, in which case the datafile element number
is the id. In either case, you can access the datafile id with the original attribute. Examples:
list vertex where id < 10
set edge color red where id == 4 or id == 6 or id == 9
foreach facet ff do
{ printf "%d %d %d %d\n",ff.id,ff.edge[1].id,ff.edge[2].id,ff.edge[3].id }
oid
Geometric element read-only integer attribute. The oid of an element is the "oriented id" of an element as used in
an expression. It is theid number signed according to whether the use of the element is with the same or opposite
orientation as the way it is stored. Example: to get an edge list for a facet as in the datafile, use oid instead of id:
foreach facet ff do { printf "%d %d %d %d\n",ff.id,ff.edge[1].oid,
ff.edge[2].oid,ff.edge[3].oid }
fixed
Geometric element read-write boolean attribute. For vertices, fixed means they dont move during various evolution
and triangulation grooming commands. For edges and facets, fixed means vertices generated from them by refinement
are fixed, although declaring a facet or edge fixed does not automatically make its vertices fixed.
For a body, fixed means its volume is constrained to be its target value. Likewise, fixed as an attribute of anamed
quantity means the quantity value is constrained.
Fixedness can be changed with the fix andunfix commands.
Examples:
fix vertex where z = 0
unfix vertex where on_constraint 1
list edge where fixed
print body[1] fixed
on_constraint
Vertex, edge, or facet read-only attribute. Boolean attribute for whether an element is on a given constraint. The full
syntax of the attribute is "on_constraint n" where n is the number or name of the constraint. Examples:
list edge where on_constraint 3
print vertex[3].on_constraint floorcon
on_boundary
Vertex, edge, or facet read-only boolean attribute.

The status of whether an element is on a parametric boundary The full syntax of the attribute is "on_boundary n"
where n is the number or name of the boundary. Examples:
list vertex where on_boundary 1
list edge where on_boundary topbdry
52
53
on_quantity
Vertex, edge, or facet read-only attribute. Boolean attribute for whether an element contributes to a given named
quantity. Actually, it tests whether the element is on any of the method instances comprising a quantity. The full
syntax of the attribute is "on_quantity quantityname". Examples:
list facet where on_quantity center_of_mass_x
print vertex[3].on_quantity blue_area
on_method_instance
Vertex, edge, or facet read-only attribute. Boolean attribute for whether an element contributes to a given named
method instance. The full syntax of the attribute is "on_method_instance instancename". Examples:
list facet where on_method_instance center_of_mass_x_edges
print vertex[3].on_method_instance blue_area_1
original
Geometric element read-write integer attribute. For elements read from the datafile, this is the number given to the
element in the datafile, which may be overridden by an explicit original attribute value in the datafile line defining the
element. The value is inherited by all elements of the same type that result from subdivision. For elements otherwise
generated at run time, the original attribute value is -1. Example: to show which facets descended from face 1 in the
datafile:
set facet color red where original == 1
Named quantity values as attributes
Named quantities and method instances can be applied to geomtric elements either in thedatafile (by adding the quantity or method name to the line defining an element) or with theset command. Nonglobal quantities or methods can
beunset for individual elements. The values for individual elements can be accessed using attribute syntax. Examples:
Suppose there is a named quantity "xmoment" that can be evaluated for facets. Then one could give commands
foreach facet do printf "%d %f\n",id,xmoment
list facet where xmoment > 4
Extra attributes
Geometric element read-write attributes. Ifextra attributes have been defined in the datafile or with a define command,
they can be accessed with attribute syntax. Extra attribute values in the datafile can be initialized for an element by
adding the attribute name and value to the line defining the element. Extra attributes may also be arrays of numeric
values (but not string values, even though global arrays can be strings), initialized with standard nested bracket syntax.
Example:
define vertex attribute newx real
define vertex attribute vmat real[3][2]
vertices
1 2 0 0 newx 3 vmat {{1,2},{3,4},{5,6}}
53
54
The command language can use the name with the same syntax as built-in attributes, and can define extra attributes at
run time:
set vertex newx x
define edge attribute vibel real[2]
set edge[2] vibel[1] 3; set edge[2] vibel[2] 4
print vertex[3].newx
Attribute array sizes may be changed at run time by executing another definition of the attribute, but the number of
dimensions must be the same. Array entry values are preserved as far as possible when sizes are changed.
The value of an extra attribute can also be calculated by user-supplied code. The attribute definition is followed by the
keyword "function" and then the code in brackets. In the code, the keyword "self" is used to refer to the element the
attribute is being calculated for. Example: To implement the lowest z value of a facet as an attribute:
define facet attribute minz real function
{self.minz := min(self.vertex,z);}
4.7 Vertex attributes

Vertex id
See id for general elements.

Vertex original
See original for general elements.

Vertex coordinates: x,y,z, x[], __x[]
Vertex read-write real attribute. The coordinates of a vertex are its location in space. By default, these are Euclidean
coordinates, but they may represent any coordinate system if the user defines appropriate length, area, volume, etc.
integrals. But graphics always treat the coordinates as Euclidean. The individual coordinates may be referred to as
x,y,z, w or x1,x2,x3,.., or x[1], x[2], x[3], ... or __x[1], __x[2], __x[3], .... In thevertices section of the datafile, vertices
of the original surface have their coordinates given unless they are on a parametric boundary. Vertices on parametric
boundaries have their coordinates calculated from their parameter values. Coordinates may be read or modified with
the command language. The form __x is useful to refer to the coordinates as a vector, for example in dot products.
Examples:
foreach vertex do printf "%d %f %f %f\n",id,x,y,z
set vertex z z+.1*x
print vertex[3].x[2]
dotprod := vertex[1].__x * vertex[2].__x
Vertex edges
Vertex read-only generator attribute. Generates edges attached to a vertex, oriented so vertex is the edge tail. The
edges are in no particular order. Examples:
54
55
list vertex[3].edges
foreach vertex vv do { foreach vv.edge do print id }
Always use ".edges" to generate vertex edges; using "edges" with an implicit element, as in "foreach vertex do list
edges" will list all edges in the surface over and over again.
Vertex facets
Vertex read-only generator attribute. Generates facets attached to a vertex, with positive facet orientation. The facets
are in no particular order. Examples:
list vertex[3].facets
foreach vertex vv do { foreach vv.facet do print id }
Always use ".facets" to generate vertex facets; using "facets" with an implicit element, as in "foreach vertex do list
facets" will list all facets in the surface over and over again.
Vertex valence
Vertex read-only integer attribute. The valence of a vertex is defined to be the number of edges it is a member of.
Example:
list vertices where valence == 6
histogram(vertex,valence)
Vertex bare
Vertex read-write boolean attribute. Declaring a vertex bare says that a vertex is not supposed to have any adjacent
edges. Useful in avoiding warning messages. A vertex may be declared bare in thevertices section of the datafile by
adding the keyword bare to the line defining the vertex. Bare is not simply a synonym for zero valence; it is a separate
attribute you set to say you intend for it to have zero valence. Examples:
set vertex bare where valence==0
list vertex where bare
Vertex fixed
Vertex read-write boolean attribute. A fixed vertex will not move during iteration (except to satisfy level set constraints)
or other operations, except if coordinates are explicitly changed by a ">set vertices ..." command. A vertex may be
declared fixed in the datafile by putting fixed on the line defining the vertex, after the coordinates. From the command
prompt, one can fix or unfix vertices with the fix and unfix commands. Examples: formula
list vertex where fixed
fix vertex where on_constraint 1
unfix vertices where on_boundary 1
Vertex constraints
Vertex read-write attribute. A level-set constraint is a restriction of vertices to lie on the zero level-set of a function.
A constraint declared nonnegative in the datafile forces a vertex to have a nonnegative value of the function. A
nonpositive constraint forces a vertex to have a nonpositive value of the function. A constraint may be declared global,
in which case it applies to all vertices. A vertex may be put on a constraint in the vertices section of the datafile by
listing the constraint numbers after the keyword constraint. See mound.fe for an example. In commands, the status
of a vertex can be read with the on_constraint and hit_constraint attributes. The status can be changed with the set or
unset commands. Examples: Datafile:
55
56
constraint 1
formula: z = 0
constraint wallcon nonpositive
formula: x = 4
Runtime:
list vertex where on_constraint wallcon
set vertex constraint 1 where id == 4 or id == 6
unset vertex constraint 1
Vertex on_constraint
See the on_constraint attribute for general elements.

Vertex hit_constraint
Vertex read-only attribute. Boolean attribute for whether a vertex exactly satisfies a given constraint. Particularly
meant for vertices on one-sided constraints. The full syntax of the attribute is "hit_constraint n" where n is the number
or name of the constraint. Examples:
list vertex where hit_constraint 3
print vertex[3].hit_constraint 1
Vertex value_of_constraint
Vertex read-only attribute. Built-in vertex attribute for finding the value of a level-set constraint for the current vertex
position. The vertex need not officially be on the constraint. Examples:
print vertex[4] value_of_constraint 1
list vertex where on_constraint topcon and value_of_constraint topcon > 1e-5
It is particularly meant for vertices on one-sided constraints.

Vertex v_constraint_list
This read-only integer array attribute gives access to the list of constraints a vertex is on. v_constraint_list[1] is the
number of constraints in the list, followed by the numbers of the constraints. Note that for named constraints, the
internally assigned numbers are used. Because this is the actual internal datastructure, the entries may have some high
bits used as flags, so to get the plain constraint numbers you should mask out the high bits with "imod 0x100000".
Example:
foreach vertex vv do
{ for ( spot := 2 ; spot <= vv.v_constraint_list[1]+1 ; spot++ )
{ connum := vv.v_constraint_list[spot] imod 0x100000;
if vv hit_constraint connum then
printf "Vertex %d hits constraint %d.\n",vv.id,connum;
}
};
Vertex constraint normal The unit normal vector of a level-set constraint at a vertex may be found with the vertex
vector attribute
56
or
57
constraint[number].normal
constraint[name].normal
number" may be an expression; name" is the unquoted name of the constraint, if it has one. Example:
print vertex[1].constraint[floorcon].normal
would print the unit normal of constraint floorcon at vertex 1. And you can put on a subscript to get individual
components. For example, the y component:
print vertex[1].constraint[floorcon].normal[2]
There is no necessity for the constraint to be applied to the vertex; the vertex is just used as a source of coordinates for
evaluating the gradient of the constraint formula.
Vertex boundary
Vertex read-write attribute. A vertex may be on aparameterized boundary, in which case its position is specified by
parameter values; i.e. the space coordinates are functions of the parameters. The keyword boundary is used as an
attribute only in the datafile declaration of elements being on boundaries. At runtime, one uses the attributes p1, p2, ..
for the parameter values, on_boundary to see if a vertex is on a particular boundary, and v_boundary to get the number
of the boundary. If you want to set a vertex on a boundary at runtime (tricky, since you have to set the parameters
yourself), the use the "set vertex boundary ..." command.
Vertex parameters p1, p2, p[]
Vertex read-write real attribute. Vertices on parametric boundaries are located according to the parameter values.
Parameters are referred to as p1,p2,... Usually only p1 is used, since one-parameter curves used as boundary wires are
most common. There is also an array form of the name, p, which is useful in array computations such as dot product in
the case of multiple parameters. Such vertices in the original surface have their parameter values given in the vertices
section of the datafile instead of their coordinates. Vertex parameters may be read or modified with the command
language. Example:
foreach vertex do printf "%d %f\n",id,p1
set vertex[1] p1 1.2
dotprod := vertex[1].p * vertex[2].p
Vertex on_boundary
See on_boundary for general elements.

Vertex v_boundary
Vertex read-only integer attribute. Internal attribute containing the number of any parameterized boundary the vertex
is on. Recall that named boundaries have internal id numbers, which are used here.
Vertex extra attributes
Seeextra attributes for general elements.

Vertex named quantity
See named quantities for general elements.

57
58
Vertex on_quantity
See on_quantity for general elements.

Vertex named method instances
Seenamed quantities for general elements.

Vertex on_method_instance
See on_method_instance for general elements.

Vertex v_method_list
Vertex read-only integer array attribute. Internal name for the array holding the id numbers of the method instances
to which this vertex contributes. Vertices do not directly record which quantities they are on, they only record which
method instances.
Vertex vertex_normal
Vertex read-only real array attribute. One-dimensional, size is space dimension. This is an indexed attribute consisting
of the components of a normal to the surface at a vertex, normalized to unit length. This is the same normal as used in
hessian_normal mode. For most vertices in the soapfilm model, the normal is the number average of the unit normals
of the surrounding facets. Along triple edges and such where hessian_normal has a multi-dimensional normal plane,
the vertex_normal is the first basis vector of the normal plane. Example: To print the normal components of vertex 3:
print vertex[3].vertex_normal[1];
print vertex[3].vertex_normal[2];
The vertex_normal can also be printed as an array:

print vertex[3].vertex_normal
Vertexnormal is an old synonym for vertex_normal.
Vertex dihedral
Vertex read-only real attribute in the string model. This is the angle from straightness of two edges at a vertex. If there
are less than two edges, the value is 0. If two or more edges, the value is 2*asin(F/2), where F is the magnitude of the
net force on the vertex, assuming each edge has tension 1. Upper limit clamped to pi.
Vertex mid_edge
Vertex read-only boolean attribute. True (1) if the vertex is on an edge but not an endpoint. Relevant in the quadratic
model or Lagrange model. Example:
list edge[23].vertex vv where vv.mid_edge
Vertex mid_facet
Vertex read-only boolean attribute. True (1) if the vertex is an interior control point of a facet in the Lagrange
model. Example:
58
59
list facet[23].vertex vv where vv.mid_facet
Vertex mean_curvature
Vertex read-only ral attribute, available in the string and soapfilm model. The mean curvature is calculated as the
magnitude of the gradient of area (or length in the string model) divided by the area (or length) associated with the
vertex, which is one-third the area of the facets adjacent to the vertex (or one-half of the length of adjacent edges). It
is divided by 2 in the soapfilm model to account for the "mean" part of the definition. The sign of the mean curvature
is relative to the orientation of the first adjacent facet (or edge) Evolver finds. This calculation can be done even if the
vertex is on a triple junction or other non-planar topology, even if it doesnt interpret well as mean curvature there.
Vertex sqcurve
Geometric element read-only real attribute. Sqcurv is the squared mean curvature at a vertex. This is only a discrete
approximation, of course, The method used to calculate it is the same as the sq_mean_curvature named method,
except if the normal_curvature toggle is on, in which case the calculation is as in the normal_sq_mean_curvature
named method. Does not require any other square mean curvature features to be active.
Vertex axial_point
Vertex read-write boolean attribute. Certainsymmetry groups (e.g.cubocta or rotate) have axes of rotation that are
invariant under some non-identity group element. A vertex on such an axis must be labeled in the datafile with the
attribute axial_point, since these vertices pose special problems for the wrap algorithms. If you are only using a
subgroup of the full group, then you only need to label vertices on the axes of the subgroup. The net wrap around a
facet containing an axial point need not be the identity. Edges out of an axial point must have the axial point at their
tail, and must have zero wrap. Facets including an axial point must have the axial point at the tail of the first edge in
the facet. It is your responsibility to use constraints to guarantee the vertex remains on the axis.
Vertex triple_point
Vertex read-write boolean attribute. For telling Evolver three films meet at this vertex. Used when effective_area is on
to adjust motion of vertex by making the effective area around the vertex 1/sqrt(3) of actual.
Vertex tetra_point
Vertex read-write booloean attribute. For telling Evolver six films meet at this vertex. Used when effective_area is on
to adjust motion of vertex by making the effective area around the vertex 1/sqrt(6) of actual.
Vertex v_force
Vertex read-only real array attribute. This is an indexed attribute giving the components of the force (negative energy
gradient as projected to constraints). One-dimensional, size is space dimension. Meant for debugging use. This is not
directly used for the motion; see v_velocity.
Vertex v_velocity
Vertex read-only real array attribute. This is an indexed attribute giving the components of the vector used for vertex
motion in the >g command. The motion of a vertex is thescale factor times this vector. The velocity vector is
calculated from the force vector by applying area normalization, mobilty, etc. Also, if a vertex is on a boundary, the
velocity is projected back to parameters. One-dimensional, size is space dimension.
59
60
Vertex raw_velocity
Vertex read-only real array attribute. Internal vertex attribute used when one-sided level-set constraints are present, so
the Lagrange multipliers for said constraints can be calculated. This is the velocity before any projection to volume or
level-set constraints. One-dimensional, size is space dimension. Not of interest to the ordinary user.
Vertex v_oldx
Vertex read-only real array attribute. Internal vertex array attribute used to store old coordinates when doing an
optimizing move. One-dimensional, size is space dimension.
4.8 Edge attributes

Edge id

Edge oid
See oid for general elements.

Edge original

Edge length
Edge read-only real attribute. Length of the edge. Examples:

histogram(edge where on_constraint 1, length)
print edge[3].length
Edge density or tension
Edge read-write real attribute. "Density" and "tension" are synonyms. Energy per unit length of edge. Default 1 in
string model, 0 in soapfilm model. The tension may be modified in the datafile edges section by adding "tension value"
to the line defining the edge. The tension may be modified with theset command. Examples:
set edge tension .5 where id < 10
loghistogram(edge,density)
Fixed edge
Edge read-write attribute. For an edge to be "fixed" means that any vertex or edge created by refining the edge will
inherit the "fixed" attribute. Declaring an edge fixed in the datafile will not fix vertices on the edge, and fixing an edge
from the command prompt will not fix any vertices. An edge may be declared fixed in the datafile edges section by
adding fixed to the line defining the edge. From the command prompt, one can fix or unfix edges with thefix and unfix
commands. Examples:
60
61
fix edge where on_constraint 1

list edges where fixed
set edge color red where fixed
unfix edge[3]
Edge vertices inxedge vertices
Edge read-only attribute. Acts as a generator for the vertices on an edge. In thelinear model, this means the tail and
head vertices. In the quadratic model, it means the tail, head, and middle vertices. For theLagrange model, it means
all the vertices from tail to head in order. For thesimplex model, it means the vertices in the stored order. Example:
list edge[2].vertices
list edge ee where ee.vertex[1].on_constraint 1
Edge midv
Edge read-only attribute. In thequadratic model, gives the id of the midpoint vertex of an edge. Example:
print edge[23].midv
Edge facets
Edge read-only attribute. Generates facets attached to an edge, in order around the edge when meaningful, with facet
orientation agreeing with edge orientation. Examples:
list edge[2].facets
foreach edge ee do print max(ee.facets,area)
Edge valence
Edge read-only integer attribute. The valence of an edge is the number of facets adjacent to it. Examples:
list edges where valence == 1
refine edge where valence != 2
Bare edge
Edge read-write boolean attribute. Declaring an edge "bare" indicates that an edge does not have an adjacent facet
(soapfilm model). Best declared in the datafile, by adding the keyword bare to the line defining an edge. Useful
in avoiding warning messages. Bare edges are useful to show wires, frameworks, outlines, axes, etc. in graphics.
Example:
list edge where bare
Edge constraints
Edge read-write attribute. An edge may be put on a level set constraint. For such an edge, any vertices and edges
generated by refining the edge will inherit the constraint. An edge may be put on constraints in the edges section of the
datafile by listing the constraint numbers after the keyword constraint on the line defining the edge. Putting an edge
on a constraint does not put its existing vertices on the constraint. In commands, the status of an edge can be read with
the on_constraint attribute. The status can be changed with the set orunset commands. Examples:
61
62
list edge where on_constraint 2

set edge constraint 1 where id == 4 or id == 6
unset edge constraint 3
Edge on_constraint
See the on_constraint attribute for general elements.

Edge e_constraint_list
This read-only attribute gives access to the list of constraints an edge is on. e_constraint_list[1] is the number of
constraints in the list, followed by the numbers of the constraints. Note that for named constraints, the internally
assigned numbers are used.
Edge boundary
Edge read-write attribute. If an edge is on a parametric boundary, then any edges and vertices generated from the
edge will inherit the boundary. By default, new vertex parameter values are calculated by extrapolating from one
end of the edge. This avoids wrap-around problems that would arise from interpolating parameter values. But if the
interp_bdry_param toggle is on, then interpolation is used. The status of whether an edge is on a boundary can be
queried with the Boolean attribute on_boundary. Edges can be unset from boundaries, and set on them. Examples:
list edges where on_boundary 1
unset edges boundary 2
Edge on_boundary
See on_boundary for general elements.

Edge e_boundary
Edge read-only integer attribute. Internal edge attribute holding the id numbers of the boundary an edge is on. Recall
that named boundaries have internal id numbers, which are used here.
Edge color
Edge read-write attribute. Color for graphics. The default color is black. Color may be set in the datafile, or with theset
command. Ingeomview, the edge color will show up only for edges satisfying the show edge condition, and then they
will have to compete with the edges geomview draws, unless you turn off geomviews drawing of edges with "ae" in
the geomview window. Examples:
set edge color red where length > 1
show edge where color != black
Edge edge_vector
Edge read-only attribute. The components of the edge vector in thelinear model can be accessed as edge attributes x,y,z
or x1,x2,x3,..., or x[1],x[2],x[3]. edge_vector is another way to refer to the vector as a vector; it is useful for expressions
like dot products where x wont work. In a command, the vector between edge endpoints is used inquadratic model
orlagrange model. But when used in an integral, the tangent is evaluated at the Gaussian integration points. Not
defined in the simplex model. Example to list nearly vertical edges:
62
63
list edges where z^2 > 10*(x^2 + y^2)

print edge[1].edge_vector * edge[2].edge_vector
Edge no_refine
Edge and facet read-write Boolean attribute. An edge with the no_refine attribute will not be refined by ther command.
This is useful for avoiding needless refining of lines or planes that are used only for display. The no_refine attribute
may be specified on the datafile line for an edge, or theset command may be used. Examples:
set edge no_refine where fixed
unset edge[2] no_refine
list edge where no_refine
print edge[3].no_refine
Edge no_transform
Edge and facet read-write Boolean attribute. An edge with the no_refine attribute will not be duplicated by the
view_transform mechanism; only the original element will occur. For example, you might have edges that form a
display of coordinate axes, which you would not want duplicated. Example:
set edge no_transform where valence == 0
Edge wrap
Edge read-write attribute. When a symmetry group is in effect (such as thetorus model) and an edge crosses the
boundary of a fundamental domain, the edge is labelled with the group element that moves the edge head vertex to
its proper position relative to the tail vertex. The label is internally encoded as an integer, the encoding peculiar to
each symmetry group. Edge wrappings are set in thedatafile. Thetorus model has its own peculiar wrap representation
in the datafile: * for no wrap, + for positive wrap, and - for negative wrap. Wraps are maintained automatically by
Evolver during surface manipulations. The numeric edge wrap values can be queried with attribute syntax. Example:
list edge where wrap != 0
Unfortunately, the torus model wraps come out rather opaquely, since one cannot print hex. The torus wrap number
is the sum of numbers for the individual directions: +x = 1; -x = 31; +y = 64; -y = 1984; +z = 4096; -z = 127040.
Caution: even though this attribute can be written by the user at runtime, only gurus should try it.
Edge wrap_list
Edge read-write integer array attribute. In a torus or symmetry model, this holds the integer used to encode the wrap
of an edge. Note that this is implemented as an array of length 1 rather than just a value; I forget why, maybe so it
could be length 0 if not needed. Example:
print edge[3].wrap_list[1]
Edge show
Edge andfacet read-only Boolean attribute giving the current status of an edge or facet according to the show edge or
show facet criterion in effect.
Edge orientation
Edge integer read-write attribute. Controls the sign of oriented integrals on an edge. Value +1 or -1. Useful when
triangulation manipulations create an edge going the wrong way. Example:
63
64
set edge[2] orientation -1
Edge frontbody
Edge read-only integer attribute. In the string model, this is the id number of the body attached to the front of the edge,
that is, the body on the facet that has positive orientation with respect to the edge. Invalid in thesoapfilm model.
Edge backbody
Edge read-only integer attribute. In the string model, this is the id number of the body attached to the back of the edge,
that is, the body on the facet that has negative orientation with respect to the edge. Invalid in thesoapfilm model.
Edge dihedral
Edge read-only real attribute. The angle in radians between the normals of two facets on an edge. Zero if there are not
exactly two facets. This attribute is not stored, but recalculated each time it is used. If there are not exactly two facets
on the edge, the value is 0.
Edge noncontent
Edge read-write boolean attribute. When set, indicates this facet should not be used in volume calculations in the
soapfilm model or facet area calculations in the string model. Useful, for example, if you want to have edges be
part of a body boundary for display purposes, but want to use constraint integrands for greater accuracy in volume
calculations. Example:
set edge noncontent where on_constraint 1
Edge named quantity
See cf named quantities for general elements.

Edge on_quantity

Edge named method instances

Edge on_method_instance

Edge e_method_list
Edge read-only integer array attribute. Internal edge attribute holding the id numbers of the method instances that this
edge contributes to. Size expands as needed. Read-only. Use ">set edge method .../a>" or"unset edge method .." to
change the status of a edge.
Edge extra attributes
See extra attributes for general elements.
64
65
4.9 Facet attributes

Facet id

Facet oid

Facet original

Facet tension or density
Facet read-write attribute. Energy per unit area of facet; surface tension. Default 0 instring model, 1 insoapfilm model.
May be set in the datafile by adding "tension value" to the line defining the facet. The density is inherited by any facets
generated by refining. "Tension" and "density" are synonyms. Examples:
set facet tension 3 where original == 1
list facet where density < .4
Facet area
Facet read-only attribute. The area of the facet. Example:

list facet where area < .1
Fixed facet
Facet read-write attribute. For a facet to be "fixed" means that any vertex, edge, or facet created by refining a facet will
inherit the fixed attribute. Fixing a facet in the datafile or at the command prompt does not fix any edges or vertices.
A face may be declared fixed in the datafile by putting fixed on the line defining the face, after the coordinates. From
the command prompt, one can fix or unfix facets with thefix and unfix commands.
Facet vertices
Facet read-only attribute. Generates vertices around a facet, oriented as the facet boundary. "vertex" and "vertices" are
synonymous. In the string model, if the facet is not a closed loop of edges, the vertices will be generated in order from
one end. If the given facet has negative orientation, then the vertices will be generated accordingly. Example:
list facet[3].vertex
Facet edges
Facet read-only attribute. Generates edges around a facet, oriented as the facet boundary. "edge" and "edges" are
synonymous. In the string model, if the edges of the facet do not make a closed loop, then the edges will be listed in
order starting from one end. If the given facet has negative orientation, the edges will be listed accordingly. Example:
65
66
list facet[3].edges
list facet[-3].edges
Facet bodies
Facet body generator attribute. Generates the bodies adjacent to a facet, in frontbody-backbody order. Example:
list facet[2] bodies
Frontbody
Facet read-write attribute. The id of the body of which the facet is on the positively oriented boundary. Useful after
creating a new body with the new_body command. As a read attribute, the value is 0 if there is no such body. Examples:
newb := new_body; set facet frontbody newb where color == red
print facet[2].frontbody
Frontbody also works for adding edges to a facet in the string model, but the added edge must be attach to one end of
the edge arc, or close the arc.
Backbody
Facet read-write attribute. The id of the body of which the facet is on the negatively oriented boundary. Useful
after creating a new body with the new_body command. As a read attribute, the value is 0 if there is no such body.
Examples:
newb := new_body; set facet[1] frontbody newb;
set facet backbody newb where id == 2 or id == 4;
print facet[4].backbody
Backbody also works for adding edges to a facet in the string model, but the added edge must be attach to one end of
the edge arc, or close the arc.
f_body_list
Facet internal array attribute. Contains the frontbody and backbody ids of the facet. Not too useful directly; listed here
because it shows up in list_attributes.
Facet valence
Facet read-only attribute. The valence of a facet is the number of edges (or vertices) that it contains. Most useful in
the string model. Example:
list facets where valence != 3
Facet constraints
Facet read-write attribute. Putting a facet on a constraint means that every vertex, edge, or facet generated by refining
the facet will inherit that constraint. Setting a facet on a constraint does not set any of its existing edges or vertices
on the constraint. Facets may be put on constraints in the datafile by listing the constraint numbers after the keyword
constraint on the line defining the facet, or with the set command. They may be removed with theunset command.
Examples:
66
67
list facets where on_constraint 1

set facet[2] constraint 2
unset facet constraint 1
Facet on_constraint
See on_constraint for general elements.

f_constraint_list
This read-only attribute gives access to the list of constraints a facet is on. f_constraint_list[1] is the number of constraints in the list, followed by the numbers of the constraints. Note that for named constraints, the internally assigned
numbers are used.
Facet boundary
Facet read-write attribute. If a facet is on a parametric boundary, then any facets, edges, and vertices generated from
the facet will inherit the boundary. By default, new vertex parameter values are calculated by extrapolating from one
vertex of the facet. This avoids wrap-around problems that would arise from interpolating parameter values. But if
the interp_bdry_param toggle is on, then interpolation is used. The status of whether a facet is on a boundary can be
queried with the Boolean attribute on_boundary. The actual boundary number is stored in the attribute f_boundary,
which can be read but should not be written directly. Facets can beunset from boundaries, and set on them (but care is
needed to do this properly). Examples:
list facets where on_boundary 1
unset facet[2] boundary 2
set facet[3] boundary 2
Facet f_boundary
Facet read-only integer attribute. The number of any parameterized boundary the facet is on. Note that named boundaries have internal numbers, and those are used here.
Facet color
Facet read-write attribute. Color of both sides of facet for graphics. Default is white. Datafile example:
Faces
1 1 2 3 color red
Command examples:
list facets where color == red
set facet[3] color green
set facet color red where area > 2
Facet frontcolor
Facet read-write attribute. Color of positive side of facet for graphics. Default is white. Datafile example:
Faces
1 1 2 3 frontcolor green backcolor red
67
68
Command examples:
list facets where frontcolor == red
set facet[3] frontcolor green
set facet frontcolor red where area > 2
Facet backcolor
Facet read-write attribute. Color of negative side of facet for graphics. Default is white. Set also when the "color"
attribute is set. Datafile example:
Faces
1 1 2 3 frontcolor green backcolor red
Command examples:
list facets where backcolor == red
set facet[3] backcolor green
set facet backcolor red where area > 2
<! Facet opacity >

Facet opacity
Facet read-write attribute for transparency. Syntax:

set facet opacity value where condition
where value is between 0 (clear) and 1 (opaque). Screen graphics will show transparency, but PostScript output will
not. Hitting the O key in the graphics window will toggle transparency, if the opacity attribute has been assigned
values. Datafile example:
Faces
1 1 2 3 opacity 0.5
Command examples:
set facet opacity 0.6
set facet opacity 0.6 where original == 2
Facet facet_normal
Facet read-only attribute. The components of the facet normal vector may be referred to as x,y,z or x1,x2,x3 or
x[1],x[2],x[3] in the linear model. Length is equal to facet area. facet_normal is the internal name of the attribute, and
is useful to refer to the entire vector in array expressions such as dot product. Inquadratic model orlagrange model,
only the three facet corner vertices are used to calculate the normal. When used in integrals, the normal is calculated
at each integration points. Not defined in simplex model.
Facet no_display
Facet read-write attribute. When set, suppresses the display of the facet in graphics. Can be set in thedatafile by
adding nodisplay to the line defining the facet. Can also be manipulated by theset andunset commands. No_display is
a synonym provided since thats what I kept typing in. Example:
set facet nodisplay where color != red
68
69
Facet no_refine
Facet read-write Boolean attribute. Giving a facet the no_refine attribute has no effect except that edges created within
the facet by refining will inherit the no_refine attribute. So to avoid refinement of a plane, all edges and facets in the
plane must be given the no_refine attribute. The no_refine attribute may be specified on the datafile line for a facet, or
theset command may be used. Examples:
set facet no_refine where fixed
unset facet[2] no_refine
list facet where no_refine
print facet[3].no_refine
Facet no_transform
Facet read-write Boolean attribute. read-write Boolean attribute. An edge or facet with the no_transform attribute will
not be duplicated by the view_transform mechanism; only the original element will occur. For example, you might
have facets that form a display of outer walls which you would not want duplicated. Example:
set facet no_transform where color == brown
Facet orientation
Facet read-write attribute. Controls the sign of oriented integrals on a facet. Value +1 or -1. Useful when triangulation
manipulations create a facet with an undesired orientation. Example:
set facet[123] orientation -1
Also see the reverse_orientation command to physically reverse a facets orientation.

Facet noncontent
Facet read-write attribute. When set, indicates this facet should not be used in volume calculations. Useful, for
example, if you want to have facets be part of a body boundary for display purposes, but want to use constraint
integrands for greater accuracy in volume calculations. Example:
set facet noncontent where on_constraint 1
Facet phase
Facet read-write attribute. If there is aphasefile, this attribute determines the edge tension of an edge between two
facets in the string model. Example:
list facet where phase == 1
Facet named quantity

Facet on_quantity
69
70

Facet named method instances

Facet on_method_instance

Facet extra attributes
See extra attributes for general elements.

Facet f_method_list
Facet array attribute. Internal name for the array holding the id numbers of the method instances to which this facet
contributes. Read-only.
Facet f_next_bfacet
Facet read-only attribute. Used internally when iterating over the facets comprising a body.
Facet f_next_vfacet
Vertex integer attribute. Internal attribute used for the linked list that is used when iterating over the facets adjacent to
a vertex. Not accessible to user since it uses internal element id type.
4.10 Body attributes

Body id

Body original

Body facets
Body read-only attribute. Generates facets bounding a body, with proper facet orientation with respect to the body.
Example:
list body[1].facets
Body density
Body read-write attribute. Density used for gravitational potential energy. It can be set in thebodies section of the
datafile, or with theset command, or by assignment. Command examples:
70
71
print body[2].density
set body density 3
body[2].density := 5
Body volume
Body read-only attribute. Actual volume of a body. This is the sum of three parts, in the soapfilm model:
An integral over the facets bounding the body. This is z dx dy normally, but (x dy dz + y dz dx + z dx dy)/3 if
symmetric_content is in effect.
Any constraint content edge integrals applying to the body.
The bodys volconst attribute.
In the string model, the parts are
An integral over the edges bounding the bodys facet. This is ydx.
Any constraint content vertex integrals applying to the body.
The bodys volconst attribute.
Body volumes can be displayed with thev command, or with standard attribute syntax. Example:
print body[1].volume
foreach body where volume > 2 do print id
Body target
Body read-write attribute. The target volume of a volume constraint. May be set in the datafile, by theb command, or
the set command. A volume constraint may be removed by the unset, or with the b command. Command examples:
set body[1] target 23
unset body target where id == 2
print body[2].target
Body volfixed
Body Boolean read-only attribute. Value is 1 if the volume of the body is fixed, 0 if not.
Body volconst
Body read-write attribute. A constant added to the calculated volume. Useful for correcting for omitted parts of body
boundaries. Also used internally as a correction in thetorus model , which will use the target volume to calculate
volconst internally. In the torus model, the target volume should be set within 1/12 of a torus volume of the actual
volume for each body, so the correct volconst can be computed. Each volconst will be adjusted proportionately when
the volume of a fundamental torus domain is change by changing the period formulas. Volconst can be set in the
datafilebodies section, or interactively by theset command or by assignment. Examples:
print body[1].volconst
set body[2] volconst 1.2
body[2].volconst := 1.2
71
72
It is best to avoid using volconst except in the torus model. Rather, useedge content integrals so that the proper
adjustments will be made if the boundary of the surface is moved, orrebody is done.
Body actual_volume
Body datafile attribute. Actual_volume is a number that can be specified in the datafile definition of a body in the rare
circumstances where the torus model volume volconst calculation gives the wrong answer; volconst will be adjusted
to give this volume of the body.
Body pressure
Body read-write real attribute. If a body has a prescribed volume, this is a read-only attribute, which is the Lagrange
multiplier for the volume constraint. If a body is given a prescribed pressure, then there is an energy term equal
to pressure times volume. A body cannot have a prescribed volume and a prescribed pressure at the same time.
Prescribed volume or pressure can be set in thebodies section of the datafile. If pressure is prescribed, then the value
can be changed interactively with theb command, the set command, or by assignment. Examples:
print body[2].pressure
body[2].pressure := 1.3
set body[2] pressure 1.3
Body phase
Body read-write attribute. If there is aphasefile, this attribute determines the facet tension of an edge between two
bodies in the soapfilm model. Example:
list body where phase == 1
Body centerofmass
Body read-write boolean attribute. Boolean body attribute. Applies to the "connected" bodies mode of graphical
display in the torus model. When this is set for a body, the center of mass of the body as displayed is remembered,
and the next time a body is graphed, its wrap is such that its new center of mass is near its previous center of mass.
This prevents bodies near the boundaries of the fundamental region from jumping back and forth as they shift slightly
during evolution. Default off. Example:
set bodies centerofmass
Body quantities
There are no named methods currently implemented for bodies, so named quantities and methods do not apply.
4.11 Facetedge attributes

Facetedge id

72
73
Facetedge oid

Facetedge edge
Facetedge read-only attribute. Generates the single edge of the facetedge. Example:
print facetedge[1].edge[1].id
Facetedge facet
Facetedge read-only attribute. Generates the single facet of the facetedge. Example:
print facetedge[1].facet[1].id
Facetedge nextedge
Facetedge internal attribute. Oriented id number of the next facetedge in the facet edge loop. May be 0 in the string
model for the last edge of a facet. Not available to users except in the output of a "list facetedges ... " command.
Facetedge prevedge
Facetedge internal attribute. Oriented id number of the previous facetedge in the facet edge loop. May be 0 in the
string model for the first edge of a facet. Not available to users except in the output of a "list facetedges ... " command.
Facetedge nextfacet
Facetedge internal attribute. Oriented id number of the next facetedge in the loop of facets around the edge. Not
available to users except in the output of a "list facetedges ... " command.
Facetedge prevfacet
Facetedge internal attribute. Oriented id number of the previous facetedge in the loop of facets around an edge. Not
available to users except in the output of a "list facetedges ... " command.
73
Chapter 5
The Model
This chapter describes the mathematical model of a surface that the Surface Evolver incorporates. See the Datafile
chapter for how to set up the initial surface and the Operations chapter for how to manipulate the surface. Keywords
are often given in upper case in this manual, even though case is not significant. In the Index, keywords are in
typewriter font.
5.1 Dimension of surface

In the standard Evolver surface, the surface tension resides in the two-dimensional facets (this is referred to as the
soapfilm model.). However, it is possible to have the tension reside in the edges. This is referred to as the string
model and is declared by the keyword STRING in the datafile. See slidestr.fe for an example. It is possible in the
soapfilm model to also have edges with tension in order to represent a situation where singular curves have energy.
This chapter is written mainly for the soapfilm model. For differences in the string model, see the String Model
section below. For a surface of more than two dimensions, see Simplex model.
5.2 Quadratic model

By default, edges and facets are linear. It is possible to represent edges as quadratic curves and facets as quadratic
patches by putting the QUADRATIC keyword in the datafile or by using the M command. If this is done, then each
edge is endowed with a midpoint vertex. Some basic features are implemented for the quadratic representation, such
as area, volume, constraints, and basic named quantity methods.
5.3 Lagrange model

For higher accuracy than the linear or quadratic models provide, there is a very limited implementation of higher order
elements. In the Lagrange model of order n, each edge is defined by Lagrange interpolation on n + 1 vertices evenly
spaced in the parameter domain, and each facet is defined by interpolation on (n + 1)(n + 2)/2 vertices evenly spaced
in a triangular pattern in the parameter domain. That is, the elements are Lagrange elements in the terminology of
finite element analysis. The Lagrange model may be invoked with the command "lagrange n".
The Lagrange model is limited to certain named quantities and methods, so the Evolver must be started with the
-q option, or you must give the convert_to_quantities command. No triangulation manipulations are available: no
refining, equiangulation, or anything. Use the linear or quadratic model to establish your final triangulation, and just
use the Lagrange model to get extra precision.
Lagrange elements are normally plotted subdivided on their vertices, but if the smooth_graph flag is on, they are
plotted with 8-fold subdivision.
74
75
5.4 Simplex model

This model represents a surface solely with vertices and facets. It was put in to enable representation of arbitrary
dimension surfaces, but many Evolver features are not available with it. Here each facet is represented as an oriented
list of k + 1 vertices, where k is the dimension of the surface. Edges may be specified as k 1 dimensional simplices,
but they are used only to compute constraint integrals; a complete list of edges is not needed. Bodies still exist for
hypersurfaces.
The datafile must have the keyword SIMPLEX_REPRESENTATION in the first section, and the phrase SURFACE_DIMENSION k if k = 2. If the domain is not 3-dimensional, then SPACE_DIMENSION n must also be included.
The EDGES section is optional. Each facet should list k + 1 vertex numbers. Non-simplicial facets are not allowed.
See the sample datafile simplex3.fe.
Most features are not implemented. Vertices may be FIXED. Constraints are allowed, but no integrands of any
kind. No TORUS domain. No QUADRATIC representation. No surface changing except iteration and refining. Refining
subdivides each simplex edge, with the edge midpoint inheriting the common attributes of the edge endpoints. Refining
will increase the number of facets by a factor of 2k .
The simplex_to_fe command will convert a 1 or 2 dimensional simplex model to a string or soapfilm model, in any
ambient dimension.
5.5 Dimension of ambient space

By default, surfaces live in 3 dimensional space. However, the phrase SPACE_DIMENSION n in the datafile sets the
dimension to n. This means that all coordinates and vectors have n components. The only restriction is that Evolver has
to be compiled with the MAXCOORD macro defined to be at least n in Makefile or in model.h. Change MAXCOORD
and recompile if you want more than four dimensions.
Graphics will display only the first three dimensions of spaces with more than three dimensions, except for geomview, which has a four-dimensional viewer built in (although its use is awkward now).
5.6 Riemannian metric

The ambient space can be endowed with a Riemannian metric by putting the keyword METRIC or CONFORMAL_METRIC in the datafile followed by the elements of the metric tensor. A conformal metric tensor is a multiple
of the identity matrix, so only one element is needed. Only one coordinate patch is allowed, but the quotient space
feature makes this quite flexible. Edges and facets are linear in coordinates, they are not geodesic. The metric is used
solely to calculate lengths and areas. It is not used for volume. To get a volume constraint on a body, you will have to
define your own quantity constraint. See quadm.fe for an example of a metric.
One special metric is available built-in. It is the Klein model of hyperbolic space in n dimensions. The domain
is the unit disk or sphere in Euclidean coordinates. Including the keyword KLEIN_METRIC in the top section of the
datafile will invoke this metric. Lengths and areas are calculated exactly, but as with other metrics you are on your
own for volumes and quantities.
5.7 Torus domain.

By default, the domain of a surface is Euclidean space R3 . However, there are many interesting problems dealing with
periodic surfaces. The Evolver can take as its domain a flat torus with an arbitrary parallelpiped as its unit cell, i.e. the
domain is a parallelpiped with its opposite faces identified. This is indicated by the TORUS keyword in the datafile.
The defining basis vectors of the parallelpiped are given in the TORUS_PERIODS entry of the datafile. See twointor.fe
for an example.
75
76
Certain features are not available with the torus domain, namely quadratic representation, constraints, boundaries,
quantities, and gravity. (Actually, you could put them in, but they will not take into account the torus wrapping.)
However, there are new named quantity methods that will do quadratic model torus domain.
Volumes and volume constraints are available. However, if the torus is completely partitioned into bodies of
prescribed volume, then the volumes must add up to the volume of the unit cell and the TORUS_FILLED keyword
must be given in the datafile. Or just dont prescribe the volume of one body.
Volumes are somewhat ambiguous. The volume calculation method is accurate only to one torus volume, so it is
possible that a body whose volume is positive gets its volume calculated as negative. Evolver adjusts volumes after
changes to be as continuous as possible with the previous volumes as possible, or with target volumes when available.
You can also set a bodys volconst attribute if you dont like the Evolvers actions.
Level set constraints can be used in the torus model, but be cautious when using them as mirror symmetry planes
with volumes. The torus volume algorithm does not cope well with such partial surfaces. If you must, then use
y=const symmetry planes rather than x=const, and use the -q option or do convert_to_quantities. Double-check that
your volumes are turning out correctly; use volconst if necessary.
Vertex coordinates are given as Euclidean coordinates within the unit cell, not as linear combinations of the basis
vectors. The coordinates need not lie within the parallelpiped, as the exact shape of the unit cell is somewhat arbitrary.
The way the surface wraps around in the torus is given by saying how the edges cross the faces of the unit cell. In the
datafile, each edge has one symbol per dimension indicating how the edge vector crosses each identified pair of faces,
and how the vector between the endpoints needs to be adjusted to get the true edge vector:
does not cross face
crosses in same direction as basis vector, so basis vector added to edge vector
crosses in opposite direction, so basis vector subtracted from edge vector.
There are several commands for ways of displaying a torus surface:
*
+
-
raw_cells - Graph the facets as they are. Good for checking your datafile if its all triangles. Faces that get subdivided
may be unrecognizable.
connected - Each bodys facets are unwrapped in the torus, so the body appears in one connected piece. Nicest option.
clipped - Shows the unit cell specified in the datafile. Facets are clipped on the parallelpiped faces.
For a torus mode surface, if clipped mode is in effect, the center of the clip box is set with the display_origin[] array
whose dimension is the dimension of the ambient space. This array does not exist by default, it has to be created by
the user in the top of the datafile with the syntax
display_origin x y z
where x y z are the coordinates for the desired center of the clip box. At runtime, the array elements may be changed
as normal for an array:
display_origin[2] := 0.5
Changing display_origin will automatically cause the graphics to re-display.

The displayed parallelogram unit cell can be different from the actual unit cell if you put an array called display_periods in the top of the datafile, in addition to the regular periods. For a string model example,
parameter shear = 1
torus_filled
periods
4 0
shear 4
display_periods
40
04
76
77
This will always display a square, no matter how much the actual unit cell is sheared. This feature works well for
shears; it may not work nicely for other kinds of deformation. Display_periods works better for the string model than
the soapfilm model. For the soapfilm model, it seems to do horizontal shears best, but it cant cope with large shears,
so if your shear gets too large, I advise resetting your fundamental region to less shear, say with the unshear command
in unshear.cmd.
5.8 Quotient spaces and general symmetry

As a generalization of the torus domain, you may declare the domain to be the quotient space of Rn with respect to
some symmetry group. This cannot be completely specified in the datafile, but requires you to write some C routines
to define group operations. Group elements are represented by integers attached to edges (like the wrap specifications
in the torus model). You define the integer representation of the group elements. See the files quotient.c and symtest.fe
for an example. See khyp.c and khyp.fe for a more intricate example modelling an octagon in Klein hyperbolic space
identified into a genus 2 surface. The datafile requires the keyword SYMMETRY_GROUP followed by your name for
the group in quotes. Edges that wrap have their group element specified in the datafile by the phrase wrap n, where
n is the number of the group element. The wrap values are accessible at run time with the wrap attribute of edges. The
group operations are accessible by the functions wrap_inverse(w) and wrap_compose(w1,w2).
Using any Hessian commands with any symmetry group other than the built-in torus model requires converting to
all named quantities with either the -q startup option or the convert_to_quantities command.
Volumes of bodies may not be calculated correctly with a symmetry group. The volume calculation only knows
about the built-in torus model. For other symmetry groups, if you declare a body, it will use the Euclidean volume
calculation. It is up to you to design an alternate volume calculation using named quantities and methods.
Currently implemented symmetries:
5.8.1
TORUS symmetry group
This is the underlying symmetry for the torus model. Although the torus model has a number of special features built
in to the Evolver, it can also be accessed through the general symmetry group interface. The torus group is the group
on n-dimensional Euclidean space generated by n independent vectors, called the period vectors. The torus group uses
the torus periods listed in the datafile top section.
Datafile declaration:
symmetry_group "torus"
periods
200
030
003
Group element encoding: The 32-bit code word is divided into 6-bit fields, one field for the wrap in each dimension,
with low bits for the first dimension. Hence the maximum space dimension is 5. Within each bitfield, 1 codes for
positive wrap and 011111 codes for negative wrap. The coding is actually a 2s complement 5-bit integer,so higher
wraps could be represented.
5.8.2
ROTATE symmetry group
This is the cyclic symmetry group of rotations in the x-y plane, where the order of the group is given by the internal
variable rotation_order. There is also an internal variable generator_power (default 1) such that the angle of rotation is
2*pi*generator_power/rotation_order. Generator_power does not exist until the user assigns a value to it. Note: Since
this group has fixed points, some special precautions are necessary. Vertices on the rotation axis must be labelled with
the attribute axial_point in the datafile. Edges out of an axial point must have the axial point at their tail, and must have
zero wrap. Facets including an axial point must have the axial point at the tail of the first edge in the facet.
Sample datafile declaration:
77
78
symmetry_group "rotate"
parameter rotation_order = 6
Group element encoding: An element is encoded as the power of the group generator.
5.8.3
FLIP ROTATE symmetry group
This is the cyclic symmetry group of rotations in the x-y plane with a flip z -> -z on every odd rotation, where the
order of the group is given by the internal variable rotation_order, which had better be even. Note: Since this group
has points that are fixed under an even number of rotations, some special precautions are necessary. Vertices on the
rotation axis must be labelled with the attribute double_axial_point in the datafile. Edges out of an axial point must
have the axial point at their tail, and must have zero wrap. Facets including an axial point must have the axial point at
the tail of the first edge in the facet.
symmetry_group "flip_rotate"
parameter rotation_order = 6
Group element encoding: An element is encoded as the power of the group generator.
5.8.4
CUBOCTA symmetry group
This is the full symmetry group of the cube. It can be viewed as all permutations and sign changes of (x,y,z).
symmetry_group "cubocta"
Group element encoding: The low bit of wrap denotes reflection in x; second lowest bit reflection in y; third lowest
bit reflection in z; next two bits form the power of the (xyz) permutation cycle; next bit tells whether to then swap x,y.
(By John Sullivan; source in quotient.c under name pgcube)
5.8.5
CUBELAT symmetry group
Cubelat is the full symmetry group of the unit cubic lattice. That is, all permutations and sign changes of the coordinates, together with adding or subtracting integers.
symmetry_group "cubelat"
Group element encoding: wrap&1,2,4 give sign changes for x,y,z; (wrap&24)/8 is power of (xyz) to cycle
(wrap&32)/32 tells whether to then swap x,y; that is, wrap&63 is the same as for cubocta. Then wrap/64 is a torus
wrap as in the torus symmetry group: three six-bit signed fields for translation in each coordinate. Translation is to be
applied after other operations.
5.8.6
XYZ symmetry group
The orientation-preserving subgroup of cubocta. Included temporarily since it is referred to in a paper. See cubocta
for details.
78
5.8.7
79
GENUS2 symmetry group
This is a symmetry group on the Klein model of hyperbolic space whose quotient group is a genus 2 hyperbolic
manifold. The fundamental region is an octagon.
symmetry_group "genus2"
Group element encoding: There are 8 translation elements that translate the fundamental region to one of its neighbors. Translating around a vertex gives a circular string of the 8 elements. The group elements encoded are substrings
of the 8, with null string being the identity. Encoding is 4 bits for start element, 4 bits for stop (actually the one after
stop so 0 0 is identity). See khyp.c for more.
5.8.8
DODECAHEDRON symmetry group
This is the symmetry group of translations of hyperbolic 3 space tiled with right-angled dodecahedra. The elements
of the group are represented as integers. There are 32 generators of the group so each generator is represented by five
bits. Under this scheme any element that is the composition of up to five generators can be represented. If you want to
use this group, youll have to check out the source code in dodecgroup.c, since somebody else wrote this group and I
dont feel like figuring it all out right now.
Klein_metric
symmetry_group "dodecahedron"
5.8.9
CENTRAL SYMMETRY symmetry group
This is the order 2 symmetry group of inversion through the origin, X -> -X.
symmetry_group "central_symmetry"
Group element encoding: 0 for identity, 1 for inversion.
5.8.10
SCREW SYMMETRY symmetry group
This is the symmetry group of screw motions along the z axis.The global parameter screw_height is the translation
distance (default 1), and the global parameter screw_angle is the rotation angle in degrees (default 0).
parameter screw_height = 4.0
parameter screw_angle = 180.0
symmetry_group "screw_symmetry"
Group element encoding: The integer value is the power of the group generator.
5.8.11
QUARTER_TURN symmetry group
3D torus with quarter turn in identification of top and bottom. x and y periods taken to be 1. z period is the user-defined
variable quarter_turn_period. Generators x,y,z. x and y as in regular torus mode. z is vertical translation with quarter
turn: (x,y,z)->(-y,x,z). Relations: xz = zy1 , yz = zx. Numerical representation: as in torus, for powers of x,y,z with
generators applied in that order. Sample datafile declaration:
parameter quarter_turn_period = 2.0
symmetry_group "quarter_turn"
Group element encoding: Same as torus.

79
80
5.9 Symmetric surfaces

Symmetric surfaces can often be done more easily by evolving just one fundamental domain. This is easiest if the
fundamental region is bounded by planes of mirror symmetry. These planes become constraint surfaces for the fundamental domain (see Constraints below). Multiple images of the fundamental domain may be displayed by using
transformation matrices (see view_transforms in the Datafile chapter).
5.10 Level set constraints

A constraint is a restriction on the motion of vertices. It may be represented as a level set of a function or as a
parametric manifold. The term constraint as used in the Evolver refers to the level set formulation, and boundary
refers to the parametric formulation. The term volume constraint is something entirely else, refering to bodies.
A level set constraint may have several roles:
1. Vertices may be required to lie on a constraint (equality constraint) or on one side (inequality constraint). A
constraint may be declared GLOBAL, in which case it applies to all vertices. See mound.fe for an example. The
formula defining the function may contain attributes of vertices, so in particular vertex extra attributes may be used to
customize one formula to individual vertices.
2. A constraint may have a vectorfield associated with it that is integrated over edges lying in the constraint to give
an energy. The vectorfield is called the energy integrand; see the Datafile section 6.3.29 for syntax. This is useful for
specifying wall contact angles and for calculating gravitational energy. See mound.fe for an example.
3. A constraint may be declared CONVEX, in which case edges in the constraint have an energy associated with
them that is proportional to the area between the straight edge and the curved wall. This energy (referred to as gap
energy) is meant to compensate for the tendency for flat facets meeting a curved wall to minimize their area by
lengthening some edges on the wall and shortening others, with the net effect of increasing the net gap between the
edges and the wall. See tankex.fe for an example.
4. A constraint may have a vectorfield associated with it that is integrated over edges lying in the constraint to give
a volume contribution to a body whose boundary facets contain the edges. This vectorfield is known as the content
integrand. It is useful for getting correct volumes for bodies without completely surrounding them with otherwise
useless facets. It is important to understand how the content is added to the body in order to get the signs right. The
integral is evaluated along the positive direction of the edge. If the edge is positively oriented on a facet, and the facet
is positively oriented on a body, then the integral is added to the body. This may wind up giving the opposite sign to
the integrand from what you think may be natural. Integrals are not evaluated over edges that are FIXED. See tankex.fe
for an example. In the string model, the content integrand is evaluated on vertices on the constraint.
5. Edge integrands (content and energy) are evaluated by Gaussian quadrature. The number of points used is
controlled by the INTEGRAL_ORDER option in the datafile.
6. If a level set constraint is declared NONNEGATIVE or NONPOSITIVE in the datafile, the vertices subject to
the constraint must stay in that part of the domain of the level set function. It is usually unwise to give edge integrals
to edges on one-sided constraints, or to declare them CONVEX. Whether a vertex exactly satisfies the constraint may
be queried with the vertex hit_constraint attribute. The g iteration step will check for a vertex wanting to leave a
one-sided constraint it has hit, but hessian commands do not; therefore it is wise to intersperse g with hessian or
hessian_seek when there are one-sided constraints involved. Example: Suppose one wanted to keep a bubble inside a
spherical tank of radius 5. Then one would define the constraint in the datafile
constraint 1 nonpositive
formula: x^2 + y^2 + z^2 = 25
For purposes of evaluating nonnegativity or nonpositivity, all terms are shifted to the left side of the formula. One
would then apply this constraint to all vertices, edges, and facets of the bubble surface. If you define the real-valued
vertex extra attribute one_sided_lagrange, the Lagrange multipliers for vertices hitting one-sided constraints will
be recorded. One_sided_lagrange may be defined as an array. If a vertex hits more constraints than the size of
one_sided_lagrange, then the first ones that fit will be recorded.
Several element attributes are useful in connection with level-set constraints:
80
81
on_constraint - Boolean attribute that tells whether the element has been put on a constraint.
hit_constraint - Boolean vertex attribute that tells whether the constraint exactly satisfies the constraint. Meant
for one-sided constraints; always 1 for exact constraints.
value_of_constraint - Vertex attribute that is the value of the constraint formula at the vertex.
v_constraint_list, e_constraint_list, f_constraint_list, b_constraint_list - Vector attribute whose first element is
the number of constraints the element is on, followed by a list of the constraint numbers (using the internal
numbers for named constraints).
constraint normal - the unit normal vector of the constraint at a vertex.
The type of a constraint can be queried at runtime as expressions
constraint[number].nonnegative
constraint[name].nonnegative
constraint[number].nonpositive
constraint[name].nonpositive
constraint[number].fixed
constraint[name].fixed
which have value 1 if the constraint is of the give type, else the value is 0. "number" may be an expression; "name" is
the unquoted name of the constraint, if it has one. Example:
print constraint[floorcon].nonnegative
Example of listing all the equality constraints that vertices are on:
foreach vertex vv do
for ( inx := 1 ; inx < vv.v_constraint_list[1] ; inx++ )
if constraint[vv.v_constraint_list[i+1]].fixed then
printf "Vertex %d on equality constraint %d
vv.id,vv.v_constraint_list[i+1];
n",
5.11 Boundaries
Boundaries are parameterized submanifolds. Vertices, edges, and facets may be deemed to lie in a boundary. For a
vertex, this means that the fundamental parameters of the vertex are the parameters of the boundary, and its coordinates
are calculated from these. Vertices on boundaries may move during iteration, unless declared fixed. See cat.fe for an
example. Edges on boundaries have energy and content integrals in the same manner as constraint edges, but they are
internally implemented as named quantities.
A delicate point is how to handle wrap-arounds on a boundary such as a circle or cylinder. Subdividing a boundary
edge requires a midpoint, but taking the average parameters of the endpoints can give nonsense. Therefore the average
coordinates are calculated, and that point projected on the boundary as continued from one endpoint, i.e. extrapolation
is used.
Extrapolating instead of interpolating midpoint parameters solves the problem of wrap-arounds on a boundary
such as a circle or cylinder. However if you do want interpolation, you can use the keyword INTERP_BDRY_PARAM
in the top of the datafile, or use the toggle command interp_bdry_param.
Interpolation requires that both endpoints of an edge be on the same boundary, which cannot happen where edges
on different boundaries meet. To handle that case, it is possible to add extra boundary information to a vertex by
declaring two particular vertex extra attributes, extra_boundary and extra_boundary_param:
interp_bdry_param
define vertex attribute extra_boundary integer
define vertex attribute extra_boundary_param real[1]
81
82
Then declare attribute values on key vertices, for example

vertices
1 0.00 boundary 1 fixed extra_boundary 2 extra_boundary_param 2*pi
If the extra_boundary attribute is not set on a vertex when wanted, Evolver will silently fall back on interpolation.
A general guideline is to use constraints for two-dimensional walls and boundaries for one-dimensional wires. If
you are using a boundary wire, you can probably declare the vertices and edges on the boundary to be FIXED. Then
the boundary becomes just a guide for refining the boundary edges.
NOTE: A vertex on a boundary cannot also have constraints.
5.12 Energy.
The Evolver works by minimizing the total energy of the surface. This energy can have several components:
1. Surface tension. Soap films and interfaces between different fluids have an energy content proportional to their
area. Hence they shrink to minimize energy. The energy per unit area can also be regarded as a surface tension, or
force per unit length. Each facet has a surface tension, which is 1 unless the datafile specifies otherwise (see TENSION
option for faces). Different facets may have different surface tensions. Facet tensions may be changed interactively
with the set facet tension ... command. The contribution to the total energy is the sum of all the facet areas times
their respective surface tensions. The surface tension of a facet may also be specified as depending on the phases of
the bodies it separates. See PHASE.
2. Gravitational potential energy. If a body has a density (see DENSITY option for bodies), then that body
contributes its gravitational energy to the total. The acceleration of gravity G is under user control. Letting be the
body density, the energy is defined as

E=
GzdV
(5.1)
body
but is calculated using the Divergence Theorem as

E=
z2
G k dS.
2
body sur f ace
(5.2)
This integral is done over each facet that bounds a body. If a facet bounds two bodies of different density, then the
appropriate difference in density is used. Facets lying in the z = 0 plane make no contribution, and may be omitted
if they are otherwise unneeded. Facets lying in constraints may be omitted if their contributions to the gravitational
energy are contained in constraint energy integrals.
3. Named quantities. See the section on named quantities below.
4. Constraint edge integrals. An edge on a constraint may have an energy given by integrating a vectorfield F
over the oriented edge:
F dl.
(5.3)
E=
edge
The integral uses the innate orientation of the edge, but if the orientation attribute of the edge is negative, the value is
negated. This is useful for prescribed contact angles on walls (in place of wall facets with equivalent tension) and for
gravitational potential energy that would otherwise require facets in the constraint.
6. Convex constraints. Consider a soap film spanning a circular cylinder. The Evolver must approximate this
surface with a collection of facets. The straight edges of these facets cannot conform to the curved wall, and hence the
computed area of the surface leaves out the gaps between the outer edges and the wall. The Evolver will naturally try
to minimize area by moving the outer vertices around so the gaps increase, ultimately resulting in a surface collapsed
to a line. This is not good. Therefore there is provision for a gap energy to discourage this. A constraint may be
declared CONVEX in the datafile. For an edge on such a constraint, an energy is calculated as

E = k S Q
(5.4)
/6
82
83
is the projection of the edge on the tangent plane of the constraint at the tail vertex of
where S is the edge vector and Q
the edge. The constant k is a global constant called the gap constant. A gap constant of 1 gives the best approximation
to the actual area of the gap. A larger value minimizes gaps and gets vertices nicely spread out along a constraint.
The gap energy falls off quadratically as the surface is refined. That is, refining once reduces the gap energy by a
factor of four. You can see if this energy has a significant effect on the surface by changing the value of k with the k
command.
7. Compressible bodies. If the ideal gas mode is in effect (see PRESSURE keyword for datafile), then each body
contributes an energy
E = PambV0 ln(V /V0 )
(5.5)
where Pamb is the ambient pressure, V0 is the volume prescribed in the datafile, and V is the actual volume of the body.
To account for work done against the ambient pressure, each body also makes a negative contribution of
E = PambV.
(5.6)
8. Prescribed pressure. Each body with a prescribed pressure P contributes energy

E = PV.
(5.7)
where V is the actual volume of the body. This can be used to generate surfaces of prescribed mean curvature, since
mean curvature is proportional to pressure.
9. Squared mean curvature. An average mean curvature around each vertex can be calculated, and the integral
of the square of this average counted as energy. This component is included by the SQUARE_CURVATURE phrase in
the datafile. A curvature offset h0 (so the energy is (h h0 )2 ) can be specified with H_ZERO value in the datafile. The
weighting factor can be changed with the A command by changing the value of the adjustable constant square curvature
modulus. For details of exactly how the integral of squared curvature is calculated, see the Technical Reference section.
(This feature is not implemented for the quadratic model.) Several variants of squared mean curvature are available,
toggled by effective_area and normal_curvature commands. See the technical reference chapter for details of the
formulas. [HKS] reports some experimental results.
10. Mean curvature integral. The energy can include the integral of the signed mean curvature by including the
MEAN_CURVATURE_INTEGRAL keyword with weighting factor in the datafile.
11. Squared Gaussian curvature. The Gauss map of a surface maps the surface to a unit sphere by mapping a
point to its normal. The Gaussian curvature at a point is the Jacobian of this map. This definition can be extended
to piecewise linear surfaces. The Evolver can include the integral of the squared Gaussian curvature in the energy by
including the SQUARE_GAUSSIAN_CURVATURE keyword with weighting factor in the datafile.
12. Crystalline integrands. The Evolver can model energies of crystalline surfaces. These energies are proportional to the area of a facet, but they also depend on the direction of the normal. The energy is given by the largest
dot product of the surface normal with a set of vectors known as the Wulff vectors. Surface area can be regarded as
a crystalline integrand whose Wulff vectors are the unit sphere. See the datafile section on Wulff vectors for more.
A surface has either crystalline energy or surface tension, not both. Use is not recommended since nonsmoothness
makes Evolver work poorly.
5.13 Named quantities and methods

These are an effort to provide a more systematic way of adding new energies and constraints.
A method is a way of calculating a scalar value from some particular type of element (vertex, edge, facet, body).
Each method is implemented internally as a set of functions for calculating the value and its gradient as a function
of vertex positions. Methods are referred to by their names. See the Named Methods and Quantities chapter for the
full list available, and the specifications each needs. Many hessians are available for named quantities that are not
available for their old-fashioned counterparts. The command line option -q will automatically convert everything to
named quantities, although some things cannot be converted yet. The Evolver command convert_to_quantities will
accomplish the same effect interactively.
83
84
Adding a new method involves writing C routines to calculate the value and the gradient as a function of vertex
coordinates, and adding a structure to the method name array in quantity.c. All the other syntax for invoking it from
the datafile is already in place.
A method instance is a particular use of a method, with whatever particular parameters may be needed. A method
instance may be created explicitly and given a name, or they may be created implicitly by defining quantities to use
methods. See the Datafile chapter for details. Every method instance has a modulus, which is multiplied times the
basic
causes the entire instance calculation to be omitted.
A quantity is the sum total of various method instances although usually just one instance is involved. Any
quantity may be declared to be one of three types: 1) energy quantities which are added to the overall energy of the
surface; 2) fixed quantities that are constrained to a fixed target value (by Newton steps at each iteration); and 3)
info_only quantities whose values are merely reported to the user. Each quantity has a modulus, which is just a
scalar multiplier of the sum of all instance values. A modulus of 0 will turn off calculation of all the instances.
The sample datafile knotty.fe contains some examples. Quantity values can be seen with the v command.
It is planned that eventually all energies and global constraints will be converted to this system. However, existing
syntax will remain valid wherever possible. Starting Evolver with the -q option will do this conversion now.
5.14 Pressure
Pressure is force per unit area perpendicular to a surface. In the Evolver, pressure can arise in three ways:
1. If a body has a volume constraint, then the boundary surface is unlikely to be a minimal surface. Hence
pressure is needed to counteract the desire of the surface to shrink. When there are volume constraints, the Evolver
automatically calculates the pressure needed. The value of the pressure can be seen with the v command.
2. A body may have a prescribed pressure. Then the appropriate force is added to the forces on the vertices when
calculating the motion of the surface. This is the way to get a surface of prescribed mean curvature, since pressure =
surface tension mean curvature. NOTE: Prescribed volume and prescribed pressure on the same body are illegal.
CAUTION: Prescribed pressure can make a surface expand to infinity if the pressure is too large.
3. The Evolver can treat bodies as being made up of an isothermal ideal gas, that is, bodies can be compressible.
This happens if the PRESSURE keyword comes in the first part of the datafile. The pressure given there is the ambient
pressure outside all bodies. Each body must have a volume specified, which is the volume of the body at the ambient
pressure.
5.15 Volume or content

The terms volume and content are used for the highest dimensional measure of a region in n-space: area in R2 ,
volume in R3 , etc. Bodies may have a volume specified in the datafile, which then becomes a volume constraint. The
volume of a body can be written as

V=
1dV,
(5.8)
body
which by the Divergence Theorem can be written a surface integral:

V=
body sur f ace
zk dS.
(5.9)
This integral is evaluated over all the boundary facets of a body.

The part of the boundary of a body lying on a constraint need not be given as facets. In that case, Stokes Theorem
can be used to convert the part of the surface integral on the constraint to a line integral over the edges where the body
surface meets the constraint. The line integral integrands are given as constraint content integrands in the datafile.
An alternate surface integral is
V=
1
3
body sur f ace
84
(xi + yj + zk) dS.
(5.10)
85
This is used if SYMMETRIC_CONTENT is specified in the datafile. It is useful if the constraint content integral (which
is evaluated by an approximation) leads to asymmetric results on what should be a symmetric surface.
As with surface area, the gap between flat facets and curved constraints causes problems. You can use constraint
content integrals to overcome this. See the tank and sphere examples in the tutorial section.
5.16 Diffusion
The Evolver can simulate the real-life phenomenon of gas diffusion between neighboring bubbles. This diffusion is
driven by the pressure difference across a surface. This is invoked by the keyword DIFFUSION in the first part of
the datafile, followed by the value of the diffusion constant. The amount diffused across a facet during an iteration is
calculated as
dm = (scale factor)(diffusion constant)(facet area)(pressure difference).
(5.11)
The scale factor is included as the time step of an iteration. The amount is added to or subtracted from the prescribed
volumes of the bodies on either side of the facet.
The internal variable diffusion_coefficient references the diffusion constant.
If you want finer control over the rate of diffusion across various surfaces, you can define the edge_diffusion edge
attribute in the string model or the facet_diffusion facet attribute in the soapfilm model and give individual values for
edges or facets as you desire. If the attribute is defined, then its value is used instead of the global diffusion constant.
Diffusion can be toggled at runtime with the diffusion toggle command.
5.17 Motion
The heart of the Evolver is the iteration step that reduces energy while obeying any constraints. The surface is changed
by moving the vertices. No changes in topology or triangulation are made (unless certain options are toggled on). The
idea is to calculate the velocity at each vertex and move the vertex in that direction. If the conjugate gradient option is
in effect, the direction of motion is adjusted accordingly.
The force is the negative gradient of the energy. At each vertex, the gradient of the total energy (see the Energy
section above) as a function of the position of that vertex is calculated. By default, the velocity is taken to be equal
to the force. If the area normalization option is in effect, then the force is divided by 1/3 the area of the neighboring
vertices to get an approximation to the mean curvature vector for velocity. If relevant, the gradients of quantity constraints (such as body volumes), and level set constraints are calculated at each vertex. Forces and quantity constraint
gradients are projected onto the tangent space of the relevant level set constraints at each vertex. To enforce quantity
constraints, a restoring motion along the volume gradients is calculated (there is found a constant for each quantity
such that when each vertex is moved by the sum of (constant)(quantity gradient) for all quantities including that
vertex, then the quantity come out right, at least in a linear approximation). Then multiples of the quantity gradients
are added to the forces to keep quantities constant (for volumes, the multiplying factors are the pressures of the bodies,
which are actually found as the Lagrange multipliers for the volume constraints).
The actual motion of a vertex is the quantity restoring motion plus a universal scale factor times the velocity. The
scale factor can be fixed or it can be optimized (see command m ). If optimized, then the Evolver calculates the energy
for several values of the scale (doubling or halving each time) until the minimum energy is passed. The optimum scale
is then calculated by quadratic interpolation. One can also move by a multiple of the optimum scale by assigning a
value to the variable scale_scale. A Runge-Kutta step is also available.
The unfixed vertices are then moved. If jiggling is in effect, each non-fixed vertex is moved in a random direction
from a Gaussian distribution with standard deviation = (temperature)(characteristic length), where the characteristic
length starts out as 1 and is halved at each refinement.
Finally, all level set constraints are enforced, including those on fixed vertices. Vertices violating an inequality or
equality constraint are projected to the constraint (Newtons method). Several projection steps may be needed, until
the violation is less that a certain tolerance or a certain number of steps are done (which generates a warning message).
85
86
The default constraint tolerance is 1e-12, but it can be set with the CONSTRAINT_TOLERANCE option in the datafile,
or setting the constraint_tolerance variable.
5.18 Hessian
There is another motion available in restricted circumstances that tries to get to the minimum in one step by calculating
the second derivative matrix (Hessian) for energy, and then solving for the minimum energy. The command to do one
iteration is hessian. There are some restrictions on its use. The energies it applies to are area, constraint edge
integrals, simplex model facet areas, and the named quantity methods so labelled earlier. Note particularly it does
not apply to the string model or to ordinary gravity. One can fake strings in the soapfilm model with edge_length
quantity, and use gravity_method instead of ordinary gravity. The constraints that are handled are body volumes, level
set constraints, and fixed named quantities involving the afore mentioned methods. Unfixed vertices on parameterized
boundaries are not handled. Some of these restrictions will be removed in future versions. If Evolver complains that
it cannot do Hessian on your surface, try restarting Evolver with the -q option, which converts everything to named
quantities if it can.
The Hessian iteration should be tried only when the surface is extremely close to a minimum (or some critical
point). I advise evolving with other methods (like conjugate gradient) until the energy has converged to 8 places or so.
If you do try the Hessian from too far away, it is likely to explode your surface. But if you do check_increase ON,
the surface will be restored to its previous state if the Hessian iteration would increase its energy.
The needed closeness to a minimum is largely due to motions tangential to the surface as the triangulation rearranges itself. There is a toggle hessian_normal that constrains motion to be along the surface normal (volume gradient,
to be precise). Points without a well-defined normal (such as along triple junctions or at tetrahedral points) are not
constrained.
Running the Hessian method will produce warning messages when the matrix is not positive definite. If the
constraints dont make the net Hessian positive semidefinite, then you will get a message like
WARNING: Constrained Hessian not positive definite. Index 3
where the index is the number of negative eigenvalues. This may show up in the first few iterations. Dont worry about
it unless it shows up when the Hessian method has converged all the way, or causes the surface to blow up. If you are
worried about blowing up, give the command check_increase. You can check positive definiteness without moving
with commands mentioned in the next subsection.
The criterion for treating a value as zero in solving the Hessian is set by the variable hessian_epsilon. Its default
value is 1e-8.
To get a feel for the Hessian method, try it out with cat.fe.
Some experimental stuff using the Hessian matrix can be accessed with the command hessian_menu. This brings
up a menu of various things. Some of the more interesting: choice E will find the lowest eigenvalue and corresponding
eigenvector. Useful at a saddle point. After E, use choice S to move along eigenvector direction to minimum energy.
Choice P finds the number of eigenvalues above and below a value.
For more details, see the Hessian section of the Technical Reference chapter.
5.19 Eigenvalues and eigenvectors

If the surface has reached a saddle point, it is nice to be able to analyze the Hessian. A positive definite Hessian proves
the current surface to be a stable local minimum. Negative eigenvalues indicate directions of energy decrease. Zero
eigenvalues may indicate symmetries such as translation or rotation. For background, see any decent linear algebra
text, such as Strang [SG], especially chapter 6. For more on stability and bifurcations, see Arnold [AV] or Hale and
Kocak [HK].
Several commands are available to analyze eigenvalues. The inertia of the Hessian with respect to some probe
value may be found with eigenprobe value, which reports the number of eigenvalues less than, equal to, or greater
than the given probe value. Eigenvalues near the probe value may be found with the lanczos value command, which
will print out 15 approximate eigenvalues near the probe value. The nearest is very accurate, but others may not be.
86
87
Also, the multiplicities are often spurious. Since lanczos starts with a random vector, it can be run multiple times to get
an idea of the error. For more accurate eigenvalues and multiplicities, there is the ritz(value, n) command, which takes
a random n-dimensional subspace and applies shifted inverse iteration to it. It reports eigenvalues as they converge to
machine precision. You can interrupt it by hitting the interrupt key (usually CTRL-C), and it will report the current
values of the rest. The saddle command will find the lowest eigenvalue, and, if negative, will seek in the corresponding
direction for the lowest energy.
The default Hessian usually has many small eigenvalues due to the approximate translational freedom of vertices
tangentially to the surface. This is an especially troublesome problem in quadratic mode, where there tend to be lots
of tiny negative eigenvalues on the order of 0.00001 that are extremely hard to get rid of by iteration. The cure for
this is to restrict the motion of the vertices to be perpendicular to the surface. The hessian_normal command toggles
this mode. The normal direction is defined to be the volume gradient. Points where this is not well-defined, such
as points on triple lines, are not restricted. This mode is highly recommended to speed the convergence of regular
Hessian iteration also.
You may want to approximate the eigenvalues and eigenvectors of a smooth surface. This requires some careful
thought. First of all, hessian_normal mode should be used, since smooth surfaces are considered to move only in
normal directions. Second, you may notice that eigenvalues change when you refine. This is due to the fact that an
eigenvector is a critical value for the Rayleigh quotient
X T HX
.
XT X
(5.12)
Now the presence of X T X is a clue that a metric is implicitly involved, and it is necessary to include the metric
explicitly. Eigenvectors should really be regarded as critical points of the Rayleigh quotient
X T HX
.
X T MX
(5.13)
The default metric is M = I, which gives unit weight to each vertex. However, for smooth surfaces, eigenfunctions are
calculated by using the L2 metric, which is integration on the surface:
< f , g >=
f (x)g(x) dx.
(5.14)
There are a couple of ways to approximate the L2 metric on an Evolver surface. One is to take a diagonal M where
the weight of each vertex is just the area associated to the vertex, 1/3 of the area of the adjacent facets, to be precise.
I refer to this as the star metric. Another way is to take the true L2 metric of the linear interpolations of functions
defined by their values at the vertices. I call this the linear metric. The linear_metric command toggles the use of
metrics in the calculation of eigenvalues and eigenvectors. By default, the metric is a 50-50 mix of the star metric and
the linear metric, since that gave the most accurate approximations to the smooth surface eigenvalues in a couple of
test cases. But you may control the mix by setting the variable linear_metric_mix to the proportion of linear metric that
you want.
In quadratic mode, linear_metric actually uses quadratic interpolation. No star metric is used, because eigenvalues
with star metric converge like h2 while eigenvalues with pure quadratic interpolation converge like h4 .
If you want to actually see the eigenvectors, you need to go to hessian_menu. Do choice 1 to initialize, then choice
V with the approximate value of your choice. This will do shifted inverse iteration, and you will probably need less
than 50 iterations to get a good eigenvector. Then use choice 4 to move by some multiple of the eigenvector. Choice 7
undoes moves, so dont be afraid.
5.20 Mobility
There is a choice to be made in the conversion of the forces on vertices into velocities of vertices. Technically, force
is the gradient of energy, hence a covector on the manifold of all possible configurations. In the Evolver, that global
covector can be represented as a covector at each vertex. The velocity is a global vector, which is represented as a
87
88
vector at each vertex. Conversion from the global covector to the global vector requires multiplication by a metric
tensor, i.e. singling out a particular inner product on global vectors and covectors. The tensor converting from force to
velocity is the mobility tensor, represented as the mobility matrix M in some coordinate system. Its inverse, converting
from velocity to force, is the resistance tensor S = M 1 . The same inner product has to be used in projecting the
velocity tangent to the constraints, whether they be level set constraints on vertices or constraints on body volumes or
quantity integrals. There are several choices implemented in the Evolver, corresponding to several different physical
pictures of how the medium resists the motion of the surface through it.
5.20.1
Vertex mobility
This is the default mobility, in which the velocity is equal to the force. Hence M and S are the identity matrices in
standard coordinates. The physical interpretation of this is that there is a resistance to motion of each vertex through
the medium proportional to its velocity, but not for the edges. This does not approximate motion by mean curvature,
but it is very easy to calculate.
5.20.2
Area normalization
In motion by mean curvature, the resistance to motion is really due to the surfaces, not the vertices. One way to
approximate this is to say the resistance to motion of a vertex is proportional to the area associated with the vertex. So
this scheme makes the resistance of a vertex equal to 1/3 of the area of the star of facets around it (or 1/2 the area of
the star of edges in the string model). This is easy to calculate, since it is a local calculation for each vertex. S and M
are diagonal matrices.
5.20.3
Area normalization with effective area
Simple area normalization as described in the previous paragraph isnt whats really wanted in certain circumstances.
It has equal resistance for motion in all directions, both parallel and normal to the surface. If a vertex is a triple
junction and migrating along the direction of one of the edges, it shouldnt matter how long that edge is. Therefore, if
the effective area mode is in effect, the area associated with a vertex is the area of its star projected normal to the force
at the vertex. This is a little more complicated calculation, but it is still local. S and M are block diagonal matrices,
with one block for each vertex. At a free edge not on any constraint, the force is tangent to the surface, the resistance
is zero, and the mobility is infinite. But this accurately describes a popping soapfilm.
5.20.4
Approximate polyhedral curvature
Following a suggestion of Gerhard Dzuik and Alfred Schmidt, the inner product of global vectors is taken to be the
integral of the scalar product of their linear interpolations over the facets (or edges in the string model). This has the
advantage that the rate of area decrease of the surface is equal to the rate volume is swept out by the surface, which is
a characteristic of motion by mean curvature. A big disadvantage is that the matrices M and S are no longer local. See
chapter 7 for the details. S is a sparse matrix with entries corresponding to each pair of vertices joined by an edge, and
M is its dense inverse.
5.20.5
Approximate polyhedral curvature with effective area
The previous section did not make any distinction between motion parallel and perpendicular to the surface. A better
approximation is to count only motion perpendicular to the surface. This can be done by projecting the interpolated
vectorfields normal to the facets before integrating their scalar product. Now the rate of area decrease is equal to the
rate geometric volume is swept out, as opposed to the slightly flaky way one had to calculate volume sweeping in the
previous paragraph. Again S is a sparse matrix with entries corresponding to each pair of vertices joined by an edge,
and M is its dense inverse.
88
5.20.6
89
User-defined mobility
The user may define a mobility tensor in the datafile. There is a scalar form with the keyword MOBILITY and a tensor
form with MOBILITY_TENSOR. When in effect, this mobility is multiplied times the velocity to give a new velocity.
This happens after any of the previous mobilities of this section have been applied and before projection to constraints.
The formulas defining the mobility may include adjustable parameters, permitting the mobility to be adjusted during
runtime.
5.21 Stability
The timestep of an iteration should not be so large as to amplify perturbations of the surface. Short wavelength
perturbations are most prone to amplification. This section contains a sketch of the stability characteristics of the
various mobility modes, enough to let the user relate the maximum timestep to the minimum facet or edge size. Two
examples are discussed: a zigzag string and a nearly flat surface with equilateral triangulation. Effective area is not
included, as it is an insignificant correction for nearly flat surfaces. The general moral of this section is that the
maximum time step in iteration is limited by the length of the shortest edge or the area of the smallest facet, except in
one case.
5.21.1
Zigzag string
Let the amplitude of the perturbation about the midline be Y and the edge length L. Then the force on a vertex is
F = 4Y /L for small Y . Let the timestep (the Evolver scale factor) be t. Let V be the vertex velocity. Then the
critical timestep for amplification of the perturbation is given by V t = 2Y , or t = 2Y /V .
Vertex mobility. Here V = F, so t = L/2.
Area normalization. Here the vertex star has length 2L, so V = F/L and t = L2 /2.
Approximate curvature. It turns out that the zigzag is an eigenvector of the mobility matrix M for the largest
eigenvalue 3/L, so V = 3F/L and t = L2 /6. This is a major disadvantage of approximate curvature. If perturbation
instability is the limitation on the timestep, it will take three times as many iterations as with area normalization to do
the same evolution.
5.21.2
Perturbed sheet with equilateral triangulation
Consider a plane surface triangulated with equilateral triangles of area A. The perturbation consists of a tiling of
hexagonal
dimples with their centers of height
Y above their peripheries. The force at a central vertex turns out to be

3Y and the force at a peripheral vertex 3Y /2.
Vertex mobility. The critical time step is given by
( 3Y + 3Y /2)t = 2Y,
(5.15)
so t = 4/3 3. Note that this is independent of the triangle size. This is consistent with experience in evolving with
optimizing scale factor, where the optimum time step is in the range 0.2 - 0.3 indenpendent of triangle size. This is a
definite advantage of this version of mobility, since different parts of the surface can have different size triangulations
and one size time step can work for all.
Area normalization. Thestar area of each vertex is 6A, so the velocities become 3Y /2A and 3Y /4A, and
the critical time step t = 4/6 3A. Hence on a surface with varying size triangles, the timestep is limited by the area
of the smallest triangles.
Approximate area. This force turns out to be an eigenvector of the mobility matrix with eigenvalue 2/A. Hence
the velocity is four times that of the area normalization, and the critical time step four times shorter.
89
90
5.22 Topology changes

The term topology of the surface refers to the topology of the union of the facets. The term triangulation refers
to the specfic way facets subdivide the surface. Some operations, such as iteration, change neither. Some, such as
refinement, change the triangulation but not the topology. Some, such as short edge elimination, can change both.
Some operations, such as vertex popping, are meant to change the topology.
5.23 Refinement
Refinement of a triangulation refers to creating a new triangulation by subdividing each triangle of the original
triangulation. The scheme used in the Evolver is to create new vertices at the midpoints of all edges and use these to
subdivide each facet into four new facets each similar to the original. (See the Technical Reference chapter for simplex
refinement.)
Certain attributes of new elements are inherited from the old elements in which they were created. Fixedness,
constraints, and boundaries are always inherited. Torus wrapping on edges is inherited by some but not all new edges.
Surface tension and displayability are inherited by the new facets. Extra attributes are inherited by the same type of
element.
Refinement can change surface area and energy if there are curved constraints or boundaries. Likewise for body
volumes.
5.24 Adjustable parameters and variables

The user may define named variables or parameters and assign them values. They may be used in expressions
whenever the expression does not have to be a constant, such as constraint formulas or boundary parameterizations.
Their values may be changed with the A command, or by assignment, e.g. foo := 3.4 from the command prompt.
Variables may be created by defining them in the top section of the datafile as parameter foo = 3.4, or by assigning
a value from the command prompt. All parameters are real-valued. Changing the value of a parameter declared in the
top of the datafile will cause automatic recalculation of the surface, on the assumption that such a parameter is used in
a constraint or boundary, unless AUTORECALC has been toggled off.
It is possible to make a parameter one of the optimization variables (along with the vertex coordinates) by declaring
it with optimizing_parameter instead of parameter in the datafile. Energy gradients and hessiansi with respect to
optimizing parameters are calculated numerically rather than symbolically, so there is a loss of speed and accuracy.
At runtime, a parameter may be toggled to be optimizing or not with the FIX and UNFIX commands. That is, fix
radius would make the radius variable non-optimizing (fixed value).
5.25 The String Model

This section lists the differences you need to be aware of when using the string model.
In general, a face takes on the role of a body, and an edge takes on the role of a facet. Coordinates are still in 3D.
If you want cells of prescribed volume, always put z = 0 for the vertices and for each face make a body with just that
one face as boundary.
The face section of the datafile is optional.
Faces are not subdivided on input.
Face area is calculated solely using x,y coordinates, even though vertices are still 3D.
Content refers to area.
Constraint energy and content integrands are scalars (have only one component, E1 or C1 as the case may be, and
are evaluated at vertices on constraints. Tails of edges count negative, and heads positive.
90
Chapter 6
The Datafile
6.1 Datafile organization
The initial configuration of the surface is read from an ASCII datafile. The datafile is organized into six parts:
1.
2.
3.
4.
5.
6.
Definitions and options

Vertices
Edges
Faces
Bodies
Commands
In the syntax descriptions below, keywords will be in upper case. const_ expr means a constant expression, and
expr means any expression. n or k means an integer, which may be signed if it is being used as an oriented element
label. Square brackets denote optional items. | means or.
6.2 Lexical format

For those who know about such things, the datafile is read with a lexical analyzer generated by the lex program. The
specification is in datafile.lex.
6.2.1
Comments
Comments may be enclosed in /* */ pairs (as in C) and may span lines. // indicates the rest of the line is a comment,
as in C++.
6.2.2
Lines and line splicing
The file is made up of lines. Line breaks are significant. The next physical line can be spliced onto the current line by
having
be the last character of the current line. Line splicing is not effective in // comments. Blank lines and comment lines
may be placed freely anywhere in the datafile. The various combinations of CR and NL that various computer systems
use are all recognized.
91
6.2.3
92
Including files
The standard C language method of including other files is available. The file name must be in quotes. If the file is
not in the current directory, EVOLVERPATH will be searched. Includes may be nested to something like 10 deep.
Example:
#include common.stuff
6.2.4
Macros
Simple macros (no parameters) may be defined as in C:

#DEFINE identifier string
identifier must be an identifier without other special meaning to the parser (see the keyword list below). string is the
rest of the logical line, not including comments. It will be substituted for identifier whenever identifier occurs as a
token subsequently. Substitutions are re-scanned. No checks for recursiveness are made. There is a maximum length
(currently 500 characters) on a macro definition. Note: macro identifiers are separate tokens, so if -M translates into
-2, this will be read as two tokens, not a signed number.
The keyword keep_macros in the datafile will keep macro definitions active during runtime, until the next datafile
is loaded.
6.2.5
Case
Case is not significant in the datafile, and at runtime is significant only for single-letter commands.
6.2.6
Whitespace
Whitespace consists of spaces, tabs, commas, colons, and semicolons. So its fine if you want to use commas to
separate coordinate values. CTRL-Z is also whitespace, for benefit of files imported from DOS.
6.2.7
Identifiers
Identifiers follow standard C rules (composed of alphanumeric characters and _ with the leading character not a digit)
and must not be keywords. Identifiers are used for macros and adjustable constants. Use at least two characters, since
single characters can be confused with commands. To find out if a name is already in use as a keyword or user-defined
name, use the is_defined function, which has the syntax
is_defined( stringexpr)
The stringexpr must be a quoted string or other string expression. The return value is 0 if the name is undefined, 1 if
defined.
6.2.8
Strings
Literal strings of characters are enclosed in double quotes, and use the standard C backslash escape conventions for
special characters. Successive quoted strings are concatenated into one string when read.
6.2.9
Numbers
Recognized number representations include integers, fixed point, scientific notation, hexadecimal, and binary numbers,
such as
2
-3
.5
23.
5e-10
+0.7D2 0x4FA5 11101b
92
6.2.10
93
Keywords
All words mentioned in this manual as having special meaning to the Evolver should be regarded as reserved words
and are not available for use by the user as identifiers for variables, commands, quantity names, etc.
6.2.11
Colors
The colors of edges and facets are recorded as integers. How these integers translate to colors on the screen is determined by how the graphics drivers are written. The following synonyms are supplied for the integers 0 through 15, and
it is hoped that the graphics drivers will be written to display these correctly: BLACK, BLUE, GREEN, CYAN, RED,
MAGENTA, BROWN, LIGHTGRAY, DARKGRAY, LIGHTBLUE, LIGHTGREEN, LIGHTCYAN, LIGHTRED, LIGHTMAGENTA, YELLOW, and WHITE. The special color value CLEAR (-1) makes a facet transparent. These tokens are simply
translated to integer values wherever they occur, so these are reserved words.
6.2.12
Expressions
Variable expressions are used in constraint and boundary formulas and integrands. Constant expressions can be used
wherever a real value is needed. Expressions are given in algebraic notation with the following tokens:
x1,x2,...
x,y,z,w
p1,p2
constant
G
identifier
identifier[ expr]...
E, PI
+,-,*,/,%,mod
imod,idiv
=
(,)
^
**
?:
coordinates
coordinates
parameters for boundaries
any integer or real number
current gravitational constant
user-defined variable
indexed array
special constants
real arithmetic
integer arithmetic
treated as low-precedence grouping and functional notation
raise to real power
raise to real power
conditional expression, as in C language
functions:
abs
sqr
sin,cos,tan
acos,asin,atan
atan2
sqrt
log,exp
sinh,cosh,tanh
asinh,acosh,atanh
pow
ceil,floor
ellipticK,ellipticE
incompleteEllipticE
incompleteEllipticF
minimum,maximum
usrn
absolute value
square
trig functions
inverse trig functions (acos, asin arguments clamped to [-1,1])
inverse tangent, atan2(y,x)
square root; argument must be nonnegative
natural log, exponentiation base e
hyperbolic functions
inverse hyperbolic functions
pow(x,y): raise x to real power y
round up or down to integer
Complete elliptic functions
Incomplete elliptic function of (, m)
Incomplete elliptic function of (, m)
of two arguments, i.e. minimum(a,b)
user-defined functions
93
94
User-defined functions can be defined in C in userfunc.c. They are meant for situations where expression interpretation is too slow, or functions such as elliptic integrals are wanted. Currently, they are automatically functions of the
coordinates. Do not give any arguments in the expression; for example (usr1 + usr3)/usr10.
Expressions are parsed into evaluation trees, which are interpreted when needed. Constant subtrees are folded into
constant nodes at parse time. For using compiled functions in certain places, see the section on dynamic load libraries
at the end of the Install chapter.
Constant expressions must evaluate to values when they are read, i.e. they have no parameters or coordinates.
In parsing an expression, the longest legal expression is used. This permits coordinates to be specified by several
consecutive expressions with no special separators.
NOTES: A + or - preceded by whitespace and followed by a number is taken to be a signed number. Thus 3 5 and 3-5 are single expressions, but 3 -5 is not. This is for convenience in separating multiple expressions listed
on the same line for vertex coordinates, metric components, etc. If in doubt about how a - will be interpreted, or if
you get error messages about not enough values, check out the minus signs.
The mod operator % or mod does a real modulus operation:
x%y = x f loor(x/y) y.
(6.1)
The integer operator idiv rounds its operands toward zero before doing integer division (as implemented in C). imod
rounds its operands down:
x imod y = f loor(x) f loor( f loor(x)/ f loor(y)) f loor(y).
(6.2)
6.3 Datafile top section: definitions and options

Each line starts with a keyword. The order is immaterial, except that identifiers must be defined before use and
quantities must be defined before referring to them in constraints. None of these are required, but those marked as
default will be assumed unless an overriding option is present.
6.3.1
Macros
#DEFINE identifier string
See Macros section above.
6.3.2
Version check
evolver_version "2.10"
If a datafile contains features present only after a certain version of the Evolver, the datafile can contain a line of
the above form. This will generate a version error message if the current version is earlier, or just a syntax error if run
on an Evolver version earlier than 2.10.
6.3.3
Element id numbers
The presence of the keyword

keep_originals
in the top of the datafile has the same effect as the -i command line option, which is to keep the id numbers internally
the same as in the datafile, instead of renumbering them in the order they are read in.
94
6.3.4
95
Variables
PARAMETER identifier = const_ expr ON_ASSIGN_CALL procedure_name
This declares identifier to be a variable with the given initial value. The value may be changed with the A command,
or by assignment. Variables may be used in any subsequent expression or constant expression. Changing variables
defined heres results in automatic recalculation of the surface, unless AUTORECALC has been toggled off.
The optional attribute ON_ASSIGN_CALL names a procedure to be called whenever the value of the variable
changes; the procedure has no arguments and must have previously been declared as a procedure. For example,
procedure tester();
parameter aaa = 4
parameter bbb = 5 on_assign_call tester
procedure tester() aaa := bbb;
OPTIMIZING_PARAMETER identifier = const_ expr PDELTA = const_ expr PSCALE = const_ expr
ON_ASSIGN_CALL procedure_name
This declares a variable as above with the additional property that it is subject to optimization. That is, it joins
the vertex coordinates in the set of independent variables. It is different from coordinates in the sense that its gradient
is calculated by finite differences rather than analytically. Hence it may be used in any kind of expression where a
variable is permitted. Hessians of optimizing parameters are implemented. The optional pdelta value is the parameter
difference to use in finite differences; the default value is 0.0001. The optional pscale value is a multiplier for the
parameters motion, to do "impedance matching" of the parameter to the surface energy. These attributes may be set on
any parameter, for potential use as an optimizing parameter. At runtime, a parameter may be toggled to be optimizing
or not with the FIX and UNFIX commands. That is, fix radius would make the radius variable non-optimizing (fixed
value). "Optimising_parameter" is a synonym.
At runtime, one may use the p_force attribute of the variable to find the rate of change of energy with respect to
the variable before constraint corrections, and the p_velocity attribute to find the rate of change of the variable with
respect to the scale factor of the g command. Example:
g
print height.p_velocity
6.3.5
Arrays
It is possible to define multidimensional arrays of integers or reals with the syntax

DEFINE variablename REAL|INTEGER|STRING [ expr]...
This syntax works both in the datafile header and at the command prompt. If the array already exists, it will be resized,
with old elements kept as far as possible. Do not resize with a different number of dimensions. Example:
define fvalues integer[10][4]
define basecoord real[10][space_dimension]
In the top of the datafile, arrays may be initialized with nested bracket initializers following the definition. For
example:
define qwerty integer[4] = 23, 45, 12, 2
define vmat real[3][2] = 1,2,3,4,5,6
define primcol string[3] = "red","blue","green"
95
96
Initializers need not be complete; missing values will be zero.

Array sizes may be changed at run time by executing another definition of the attribute, but the number of dimensions must be the same. Array entry values are preserved as far as possible when sizes are changed.
The PRINT command may be used to print whole arrays or array slices in bracketed form. Example:
print fvalues
print fvalues[4]
In the run-time command language, there are some basic whole-array operations that permit arrays on the left side
of an assignment statement:
array := scalar
array := array
array := scalar * array
array := array + array
array := array - array
array := array * array
Here "array" on both sides of the assignment means a single whole array; not an array-producing expression or array
slice. But "scalar" can be any expression that evaluates to a single value. For multiplication, the arrays must be twodimensional with properly matching sizes. These operations also apply to element attributes that are arrays. For the
inner product of vectors, there is an infix operator dot_product.
6.3.6
Dimensionality
The dimensionality of the surface itself can be declared:

STRING
The Evolver is to use the string model (see the String Model section).
SOAPFILM
The Evolver is to use the standard two-dimensional surface model. (default)

Alternately,
SURFACE_DIMENSION const_ expr
Surface is of given dimension. Dimension over 2 is valid only in the simplex model.
6.3.7
Domain
EUCLIDEAN (default)
The surface lives in Euclidean space.

SPACE_DIMENSION const_ expr
The surface lives in Euclidean space of the given dimension. Default is 3. The dimension must be at most the
value of MAXCOORD in model.h, which is 4 in the distributed version.
TORUS
The surface lives in a 3 dimensional flat torus. Surface is equivalent to a periodic surface.
TORUS_FILLED
96
97
Indicates the entire torus is filled by bodies to enable the program to avoid degenerate matrices when adjusting
constrained volumes.
PERIODS
expr
expr
expr
expr
expr
expr
expr
expr
expr
Used with TORUS domain. Specifies that the side vectors (basis vectors) of the unit cell parallelpiped follow on
the next myverbatim. Each vector is given by its components. The size of this matrix depends on the space dimension.
Default is unit cube. Adjustable parameters may be used in the expressions, so the fundamental domain may be
changed interactively with either the A command or by assigning new values to the parameters. Be sure to do a
recalc to get the period matrix re-evaluated.
SYMMETRY_GROUP " name"
The domain is the quotient of Rn by a user-defined symmetry group. The user must link in C functions that define
the group operations. See quotient.c for an example. name is a double-quoted name that is checked against a name
in quotient.c to be sure the right symmetry group is linked in.
SYMMETRIC_CONTENT
For body volume integrals, use an alternate surface integral

V=
1
3

body sur f ace
(xi + yj + zk) dS.
(6.3)
It is useful if unmodelled sides of a body are radial from the origin, or if constraint content integrals (which is evaluated
by an approximation) lead to asymmetric results on what should be a symmetric surface.
6.3.8
Length method
This item, length_method_name, specifies the name of the pre-defined method to use as the method to compute edge
lengths in place of the default edge_area method. It is optional. Usage automatically invokes the all quantities mode.
The principle usage so far is to use exact circular arcs in two-dimensional foams. Syntax:
volume_method_name quoted_method_name
For example,
string
space_dimension 2
length_method_name "circular_arc_length"
The special method null_length may be used to get a zero value.
6.3.9
Area method
This item, area_method_name, specifies the name of the pre-defined method to use as the method to compute facet
areas in place of the default edge_area method in the string model or the facet_area method in the soapfilm model.
In the string model, it is synonymous with volume_method_name. It is optional. Usage automatically invokes the all
quantities mode. Developed for using exact circular arcs in two-dimensional foams. Syntax:
97
98
area_method_name quoted_method_name
For example,
string
space_dimension 2
area_method_name "circular_arc_area"
The special method null_area may be used to get a zero value.
6.3.10
Volume method
This item, volume_method_name, specifies the name of the pre-defined method to use as the method to compute body
volumes (or facet areas in the string model) in place of the default edge_area or facet_volume methods. It is optional.
Usage automatically invokes the all quantities mode. Syntax:
volume_method_name quoted_method_name
For example,
string
space_dimension 2
volume_method_name "circular_arc_area"
6.3.11
Representation
LINEAR
(default)
Facets are flat triangles.

QUADRATIC
Facets are quadratic patches. See the Quadratic Representation section.

LAGRANGE n
The surface is in the Lagrange order n representation. Best not to try to create a Lagrange representation input file
by hand. This phrase is in here so dump files of the Lagrange representation may be reloaded.
SIMPLEX_REPRESENTATION
Facets are defined by oriented vertex list rather than edge list. See the section above on Simplex Model.
6.3.12
Hessian special normal vector
In using Hessian commands, it may be useful to have the perturbations follow a specified direction rather than the
usual surface normal. The direction vectorfield is specified in the datafile header section with the syntax
HESSIAN_SPECIAL_NORMAL_VECTOR
c1: expr
c2: expr
c3: expr
98
99
Vertex attributes may be used in the component expressions, which permits elaborate calculations to be done beforehand to specify the vectorfield. For example, one could do
define vertex attribute pervec real[3]
HESSIAN_SPECIAL_NORMAL_VECTOR
c1: pervec[1]
c2: pervec[2]
c3: pervec[3]
6.3.13
Dynamic load libraries
To load a dynamic library of compiled functions, the syntax is

LOAD_LIBRARY " filename"
where the double-quoted filename is the library. The current directory and the EVOLVERPATH will be searched for the
library. For details on how to set up and use a dynamic load library, see the Installation chapter.
6.3.14
Extra attributes
It is possible to dynamically define extra attributes for elements, which may be single values or up to eight-dimensional
arrays. The definition syntax is
DEFINE elementtype ATTRIBUTE name type [ [ dim] ... ]
where elementtype is vertex, edge, facet, or body, name is an identifier of your choice, and dim is an optional expression for the dimension. Type is REAL or INTEGER (internally there is also a ULONG unsigned long type also). The
type may be followed by FUNCTION followed by a procedure in brackets to be evaluated whenever the value of the
attribute is read; in the formula, self may be used to refer to the element in question to use its attributes, in particular
to at some point assign a value to the attribute. There is no practical distinction between real and integer types at the
moment, since everything is stored internally as reals. But there may be more datatypes added in the future. Extra
attributes are inherited by elements of the same type generated by subdivision. Examples:
define edge attribute charlie real
define vertex attribute newx real
define vertex attribute vmat real[3][2]
vertices
1 2 0 0 newx 3 vmat 1,2,3,4,5,6
The command language can use the name with the same syntax as built-in attributes, and can define extra attributes
at run time:
set vertex newx x
define edge attribute vibel real[2]
set edge[2] vibel[1] 3; set edge[2] vibel[2] 4
print vertex[3].newx
Attribute array sizes may be changed at run time by executing another definition of the attribute, but the number of
dimensions must be the same.
The total number of elements of an extra attribute may be retrieved at runtime with the sizeof function, with the
syntax
SIZEOF( name)
print vertex[34].newx;
print facet[1].knots[3][2];
99
6.3.15
100
Surface tension energy
The surface tension energy is always included in the total energy by default. It can be turned off only by giving the
facets (or edges in the string model) the density 0 attribute.
AREA
(default)
Surface energy of a facet is its area times its surface tension.

WULFF
" filename"
Specifies that a crystalline integrand is to be used. The next token should be a double-quoted filename (with path)
of a file giving the Wulff vectors of the integrand. The format of the file is one Wulff vector per line with its three
components in ASCII decimal format separated by spaces. The first blank line ends the specification. Some special
integrands can be used by giving a special name in place of the file name. Currently, these are "hemisphere" for a
Wulff shape that is an upper unit hemisphere, and "lens" for two unit spherical caps of thickness 1/2 glued together on
a horizontal plane. These two dont need separate files.
PHASEFILE
" filename"
This permits the surface tension of a grain boundary to depend on the phases or types of the adjacent grains. The
information is read from an ASCII file. The first line of the file has the number of different phases. Each line after
consists of two phase numbers and the surface tension between them. Lines not starting with a pair of numbers are
taken to be comments.If a pair of phases is not mentioned, the surface tension between them is taken to be 1.0. Facets
in the string model or bodies in the soapfilm model can be labelled with phases with the PHASE phrase.
6.3.16
Squared mean curvature
SQUARE_CURVATURE const_ expr
This phrase indicates that the integral of squared mean curvature will be included in the energy with a weight given by
const_ expr. The weight can be changed with the A command by changing the value of the adjustable constant square
curvature modulus.
6.3.17
Integrated mean curvature
MEAN_CURVATURE_INTEGRAL const_ expr
This phrase indicates that the integral of mean curvature will be included in the energy with a weight given by
const_ expr. The weight can be changed with the A command by changing the value of the adjustable constant mean
curvature modulus.
6.3.18
Gaussian curvature
GAUSS_CURVATURE const_ expr
This phrase indicates that the integral of Gaussian curvature will be included in the energy with a weight given by
const_ expr.
100
6.3.19
101
Squared Gaussian curvature
SQUARE_GAUSSIAN_CURVATURE const_ expr
This phrase indicates that the integral of squared Gaussian curvature will be included in the energy with a weight given
by const_ expr. The weight can be changed with the A command by changing the value of the adjustable constant
square_Gaussian_modulus. Synonyms: squared_gaussian_curvature, sqgauss.
6.3.20
Ideal gas model

PRESSURE const_ expr
This specifies that bodies are compressible and the ambient pressure is the given value. The default is that bodies
with given volume are not compressible.
6.3.21
Gravity
GRAVITY_CONSTANT const_ expr
Specifies gravitational constant G. Default 1.0.
6.3.22
Gap energy
GAP_CONSTANT const_ expr
Multiplier for convex constraint gap energy. Default 1.0. Synonym: spring_constant
6.3.23
Knot energy
There are a bunch of named quantity methods for knot energies, and that syntax should be used. But there are a couple
of synonyms floating around.
INSULATING_KNOT_ENERGY const_ expr
The total energy will include the knot energy method knot_energy with multiplier const_ expr. Abbreviation for a
named quantity.
CONDUCTING_KNOT_ENERGY const_ expr
The total energy will include the knot energy method edge_knot_energy with multiplier const_ expr. Abbreviation
for a named quantity.
6.3.24
Mobility and motion by mean curvature

GRADIENT_MODE (default)
Velocity is equal to the force.

AREA_NORMALIZATION
The velocity of a vertex is the force divided by 1/3 area of neighboring facets (or 1/2 length of neighboring edges
in string model) to approximate motion by mean curvature.
101
102
APPROXIMATE_CURVATURE
Calculates vertex velocity from force by means of polyhedral linear interpolation inner product. Do not use together
with AREA_NORMALIZATION.
EFFECTIVE_AREA
For both area normalization and approximate curvature modes, the resistance to motion is only on the component of
velocity normal to the surface.
6.3.25
Annealing
JIGGLE
Continuous jiggling. Default none.

TEMPERATURE const_ expr
Gives temperature for jiggling. Default 0.05.
6.3.26
Diffusion
DIFFUSION const_ expr
Specifies that diffusion between bodies is in effect with diffusion constant as given. Default 0.
6.3.27
Named method instances
These are methods of calculating scalar values for geometric elements that are referred to by name. They are used by
named quantities (see next subsection). For which models each is valid in (linear, quadratic, Lagrange, simplex, etc.),
see Chapter 4.
METHOD_INSTANCE name METHOD methodname [ MODULUS constexpr]
[ ELEMENT_MODULUS attrname] [ GLOBAL] parameters
This is an explicit definition of a named method instance for a named quantity. Such an explicit definition is useful
if your quantity has several method instances, and you want to see their individual values or apply the instances to
different elements. The modulus multiplies the method value to give the instance value. The default modulus is 1.
GLOBAL makes the method apply to all elements of the appropriate type. Non-global instances may be applied to
elements individually.
Each method may have various parameters to specialize it to an instance. Currently the only parameters specified
here are scalar integrands, with syntax
SCALAR_INTEGRAND: expr
vector integrands, with syntax
VECTOR_INTEGRAND:
Q1:
expr
Q2:
expr
Q3:
expr
2-form integrands, with syntax
102
103
FORM_INTEGRAND:
Q1:
expr
Q2:
expr
Q3:
expr
$. . . $
where the form components are listed in lexicographic order, i.e. in 4D the six components 12,13,14,23,24,34 would
be listed as Q1 through Q6.
The expressions may use attributes for individual elements (density, length, extra attributes, etc.)
Some methods use global parameters (a holdover that will be done away with eventually.) The instance definition does
not have to be on one line.
See the Named Methods and Quantities chapter for a list of the methods available and the specifications each
needs.
6.3.28
Named quantities
QUANTITY name ENERGY| FIXED =value | INFO_ONLY| CONSERVED

[ MODULUS constexpr ] [ LAGRANGE_MULTIPLIER constexpr ] [ TOLERANCE constexpr ]
methodlist | FUNCTION methodexpr
These are an effort to provide a more systematic way of adding new energies and constraints. A method is a
way of calculating a scalar value from some particular type of element (vertex, edge, facet, body). A quantity is the
sum total of various methods applied to various elements, although usually just one method is involved. The name is
an identifier. Any quantities may be declared to be one of three types: 1) energy quantities are added to the overall
energy of the surface; 2) fixed quantities that are constrained to a fixed target value (by a single Newton step at each
iteration), 3) information quantities whose values are merely reported to the user. 4) conserved quantities which are
not evaluated as such, but gradients and hessians treat them as fixed quantities in calculating directions of motion.
Each quantity has a modulus, which is just a scalar multiplier of the whole quantity. A modulus of 0 will turn off
an energy quantity. The default modulus is 1. Adding a new method involves writing C routines to calculate the value
and the gradient as a function of vertex coordinates, and adding a structure to the method name array in quantity.c.
For fixed quantities, the optional Lagrange multiplier value supplies the initial value of the Lagrange multiplier
(the "pressure" attribute of the quantity). It is meant for dump files, so on reloading no iteration need be done to have
a valid Lagrange multiplier.
For fixed quantities, the tolerance attribute is used to judge convergence. A surface is deemed converged when the
sum of all ratios of quantity discrepancies to tolerances is less than 1. This sum also includes bodies of fixed volume.
If the tolerance is not set or is negative, the value of the variable target_tolerance is used, which has a default value of
0.0001.
Conserved quantities are useful for eliminating unwanted degrees of freedom in hessian calculations, particularly
rotation. It works best to apply the quantity to vertices rather than edges or facets. Conserved quantities are incompatible with optimizing parameters, since gradients for optimizing parameters are found by finite differences, which
dont work here.
The methodlist version of the quantity definition may contain one or more method instances. To incorporate a
previously explicitly defined instance, include
METHOD instancename
To instantiate a method in the quantity definition, you essentially incorporate the instance definition, but without an
instance name:
METHOD methodname [MODULUS constexpr] [GLOBAL] parameters
See the previous subsection for method details. Usually the second, implicit definition will be more convenient, as
there is usually only one method per quantity. If GLOBAL_METHOD is used instead of GLOBAL, then the method is
applied to all elements of the proper type. It is equivalent to using the GLOBAL keyword in the method specification.
Nonglobal instances must be applied individually to elements. That is done by simply adding the quantity or instance
name to the line defining the element. If a quantity name is used, then all method instances of that quantity of the
103
104
appropriate type are applied to the element. Original attachments of quantities are remembered, soIf an edge method
is applied to a facet, then edges created from refining that facet will inherit the edge method. Orientable methods can
be applied with negative orientation to elements in the datafile by following the name with a dash. The orientation in
a set command follows the orientation the element is generated with.
The quantity may also be defined by an arbitrary function of method instances. The keyword FUNCTION indicates
this, and is followed by the defining function expression. The method instances involved must all have been previously
defined as named method instances.
Quantity values can be seen with the Q or A command, or may be referred to as "quantityname.value " in commands. Do not use just quantityname since quantityname alone is interpreted as an element attribute. One can assign
values to quantityname.target or quantityname.modulus .
Examples:
Hooke energy:
quantity hooke ENERGY modulus 10 global_method hooke_energy
A little sample datafile:

// test of quantity integrands on constraints
method_instance len method edge_length
quantity lenny info_only method len
quantity sam info_only method facet_scalar_integral
scalar_integrand z
vertices
1 0 0 1 fixed
2 1 1 2 fixed
3 0 1 3 fixed
edges
1 1 2 lenny
2 23
3 3 1 len
faces
1 1 2 3 sam
The sample datafile knotty.fe contains more examples.

The keyword EVERYTHING_QUANTITIES in the top section of the datafile causes all areas, volumes, etc. to be
converted to named quantities and methods. It is equivalent to the command line option -q, or the convert_to_quantities
command.
6.3.29
Level set constraints
CONSTRAINT n [GLOBAL] [CONVEX] [NONNEGATIVE| NONPOSITIVE] [NONWALL] [CONTENT_RANK n]

EQUATION | FORMULA | FUNCTION expr
[ ENERGY
E1:
expr
E2:
expr
E3:
expr]
[ CONTENT
104

C1:
C2:
C3:
105
expr
expr
expr]
This defines constraint number n, where n is a positive integer. GLOBAL means the constraint automatically
applies to all vertices (but not automatically to edges or faces). GLOBAL constraints count in the number limit. If
CONVEX is given, then an additional gap energy is attributed to edges on the constraint to prevent them from trying to
short-circuit a convex boundary; see the Constraints and Energy sections above and the k command. If NONNEGATIVE
or NONPOSITIVE is given, then all vertices will be forced to conform appropriately to the constraint at each iteration.
The EQUATION expression defines the zero level set which is the actual constraint. It may be written as an equation,
since = is parsed as a low-precedence minus sign. Do not use > or < to indicate inequalities; use NONNEGATIVE
or NONPOSITIVE and an expresson. Conditional expressions, as in C language, are useful for defining constraints
composed of several surfaces joined smoothly, such as a cylinder with hemispherical caps.
NONWALL indicates this constraint is to be ignored in vertex and edge popping, and the various star squared
mean curvature methods will not regard this constraint as a mirror plane.
The formula may include any expressions whose values are known to the Evolver, given the particular vertex. Most
commonly one just uses the coordinates (x,y,z) of the vertex, but one can use variables, quantity values, or vertex extra
attributes. Using a vertex extra attribute is a good way to customize one formula to individual vertices. For example,
if there were a vertex extra attribute called zfix, one could force vertices to individual z values with one constraint
with the formula z = zfix, after of course assigning proper values to zfix for each vertex. Be sure to fix up the extra
attribute after refining or otherwise creating new vertices, since new vertices will have a default value of 0 for the extra
attribute.
NOTE: One-sided constraints can cause the optimal scale factor algorithm to misbehave. It may be necessary to
use a fixed scale factor. See the m command below.
IMPORTANT NOTE: Do not let two constraints with the same formula apply to a vertex; that leads to a singular
matrix inversion when trying to project the vertex onto the constraints. For example, do not have a vertex subject to
X1 = 0 and also have a global X1 NONNEGATIVE.
ENERGY signifies that vertices or edges on the constraint are deemed to have an energy. In the SOAPFILM model,
the next three lines give components of a vectorfield that will be integrated along each edge on the constraint. In the
STRING model, just one component is needed, which is evaluated at each vertex on the constraint. The main purpose
of this is to permit facets entirely on the constraint to be omitted. Any energy they would have had should be included
here. One use is to get prescribed contact angles at a constraint. This energy should also include gravitational potential
energy due to omitted facets. Integrals are not evaluated on FIXED edges.
CONTENT signifies that vertices (STRING model ) or edges (SOAPFILM model ) on the constraint contribute to
the area or volume of bodies. If the boundary of a body that is on a constraint is not given as facets, then the body
volume must get a contribution from a content integral. It is important to understand how the content is added to the
body in order to get the signs right. The integral is evaluated along the positive direction of the edge. If the edge is
positively oriented on a facet, and the facet is positively oriented on a body, then the integral is added to the body.
This may wind up giving the opposite sign to the integrand from what you think may be natural. Always check a new
datafile when you load it to be sure the integrals come out right.
If a vertex (string model) or an edge (soapfilm model) is on multiple constraints with content integrals (say where
walls meet) then if content ranks are present, the content integral with the least rank will contribute to the content on
the negative side body and the highest rank content will contribute to the content of the positive side body.
Warning: These integrals are evaluated only for edges which are on the constraints and both of whose endpoints
are on the constraints. It is a bad idea to put any of these integrals on one-sided constraints, as both endpoints must
actually hit the constraint to count.
6.3.30
Constraint tolerance
CONSTRAINT_TOLERANCE const_ expr
105
106
This is the tolerance within which a vertex is deemed to satisfy a constraint. Default 1e-12.
6.3.31
Boundaries
BOUNDARY n PARAMETERS k [CONVEX] [CONTENT_RANK n]
X1
expr
X2
expr
X3
expr
[ ENERGY
E1
expr
E2
expr
E3
expr]
[ CONTENT
C1
expr
C2
expr
C3
expr]
This defines boundary number n, where n is a positive integer and k is the number of parameters (1 or 2). If
CONVEX is given, then an additional energy is attributed to edges on the boundary to prevent them from trying to
short-circuit a convex boundary; see the k menu option below. The following three lines have the functions for the
three coordinates in terms of the parameters P1 and maybe P2. Energy and content integrals for boundaries are
implemented with the same syntax as for constraints.
6.3.32
Numerical integration precision

INTEGRAL_ORDER_1D n
INTEGRAL_ORDER_2D n
Sets the degree of polynomial done exactly by numerical integration. Edge integrals are done by k-point Gaussian
quadrature. This give exact values for polynomials of degree n = 2k 1. Default is k = 2, which will do cubic
polynomials exactly. No limit on order, as weights and abscissas are calculated by Evolver.
Facet integrals are done with 1, 3, 7, 12, or 28 point integration, corresponding to n =1, 2, 5, 6, or 11.
6.3.33
Scale factor
SCALE const_ expr
[ FIXED]
Sets the initial scale factor. If FIXED is present, sets fixed scale factor mode. Default is scale = 0.1 and optimizing
mode.
SCALE_LIMIT const_ expr
Sets upper bound on scale factor to prevent runaway motions. Default value is 1. If you use surface tensions and
densities not near unity, you may have to set this value.
106
6.3.34
107
Mobility
MOBILITY_TENSOR
expr
expr
expr
expr
expr
expr
expr
expr
expr
or
MOBILITY expr
The force vector is multiplied by the mobility scalar or tensor to get the velocity. Good for, say, having grain
boundary mobility depend on temperature.
6.3.35
Metric
METRIC
expr
expr
expr
expr
expr
expr
expr
expr
expr
or
CONFORMAL_METRIC
expr
or
KLEIN_METRIC
The user may define a background metric for the string model only. The keyword METRIC is follwed by the N 2
components of the metric tensor, where N is the dimension of space. The components do not have to obey any
particular line layout; they may be all on one line, or each on its own line, or any combination. It is up to the user
to maintain symmetry. A conformal metric is a scalar multiple of the identity matrix, and only the multiple need be
given. A conformal metric will run about twice as fast. The Klein metric is a built-in metric for hyperbolic n-space
modelled on the unit disk or ball.
6.3.36
Autochopping
AUTOCHOP const_ expr
This turns on autochopping of long edges. The constant is the maximum edge length.
6.3.37
Autopopping
AUTOPOP
This turns on autopopping of short edges and improper vertices.
6.3.38
Total time
TOTAL_TIME const_ expr
This permits setting the initial time of a surface evolving with a fixed scale. Used primarily when resuming from a
dump file.
107
6.3.39
108
Runge-Kutta
RUNGE_KUTTA const_ expr
This turns on doing iteration with the Runge-Kutta method.
6.3.40
Homothety scaling
HOMOTHETY const_ expr
This turns on doing homothety scaling each iteration. The scaling is a uniform scaling from the origin to keep the total
volume of all bodies at the given value.
6.3.41
Viewing matrix
VIEW_MATRIX
const_ expr const_ expr


For the specification of the initial viewing transformation matrix of the surface. The matrix is in homogeneous coordinates with translations in the last column. The size of the matrix is one more than the space dimension. This
matrix will be part of all dump files, so the view can be saved between sessions. This matrix only applies to internal
graphics (Postscript, Xwindows, etc.) and not external graphics (geomview). The elements may be read at runtime by
view_matrix[i][j], where the indices start at 1.
6.3.42
View transforms
VIEW_TRANSFORMS integer
[ color color]
[ swap_colors swap_colors]


...
For the display of several transformations of the surface simultaneously, a number of viewing transformation matrices
may be given. The transforms apply to all graphics, internal and external, and are prior to the viewing transformation
matrix for internal graphics. The identity transform is always done, so it does not need to be specified. The number of
matrices follows the keyword VIEW_TRANSFORMS. Each matrix is in homogeneous coordinates. The size of each
matrix is one more than the space dimension. Individual matrices need no special separation; Evolver just goes on an
expression reading frenzy until it has all the numbers it wants. Each matrix may be preceded by a color specification
that applies to facets transformed by that matrix. The color applies to one transform only; it does not continue until the
next color specification. If SWAP_COLORS is present instead, facet frontcolor and backcolor will be swapped when
this matrix is applied. Transforms may be activated or deactivated interactively with the transforms on or transforms
off. The internal variable transform_count records the number of transforms, and the transform matrices are accessible
at runtime as a three-dimensional matrix view_transforms[][][]. See the next paragraph for a more sophisticated way to
control view transforms.
108
6.3.43
109
View transform generators
VIEW_TRANSFORM_GENERATORS integer
SWAP_COLORS


...
Listing all the view transforms as in the previous paragraph is tedious and inflexible. An alternative is to list just a few
matrices that can generate transforms. See the transform_expr command for instructions on entering the expression
that generates the actual transforms. Special Note: in torus mode, the period translations are automatically added to
the end of the list. So in torus mode, these are always available, even if you dont have view_transform_generators in
the datafile. If SWAP_COLORS is present, facet frontcolor and backcolor will be swapped when this matrix is applied.
The internal variable transform_count records the number of transforms, and the transform matrices are accessible at
runtime as a three-dimensional matrix view_transforms[][][].
6.3.44
Zoom parameter
ZOOM_RADIUS constexpr
ZOOM_VERTEX constexpr
Sets the current parameters for the zoom command. Used in dump files, rather than users original datafiles.
6.3.45
Alternate volume method
VOLUME_METHOD_NAME "methodname"
Sets the method used to calculate the volume under a facet (or area under an edge in 2D) to the named method
(given in quotes). Automatically converts everything to quantities.
6.3.46
Fixed area constraint
FIXED_AREA expr or AREA_FIXED expr
Obsolete method of constraining the total area to a given value. Do not use anymore. Use the facet_area method
in a fixed named quantity.
6.3.47
Merit factor
MERIT_FACTOR expr
If the keyword MERIT_FACTOR is present, then the i command will print the ratio total_area3 /total_volume2 ,
which measures the efficiency of area enclosing volume. This is a holdover from the Evolvers early days of trying to
beat Kelvins partition of space.
6.3.48
Parameter files
PARAMETER name PARAMETER_FILE string
A parameter can be initialized with a set of values from a file, but I forget at the moment how it is all supposed to
work.
109
6.3.49
110
Suppressing warnings
Undesired warnings may be suppresseed by including lines with the syntax:

SUPPRESS_WARNING number
where number is the number of the warning. Meant to suppress irritating warning messages that you know
are irrelevant. Warnings can be restored with the syntax
UNSUPPRESS_WARNING number
6.3.50
Giving hints on memory allocation
It is possible to give Evolver instructions on how many of various data structures to allocate at the beginning of
the datafile. Doing this is entirely optional, but for large surfaces, it may speed loading and cut down on memory
fragmentation due to expanding element lists. Syntax:
vertices_predicted number
edges_predicted number
facets_predicted number
facetedges_predicted number
bodies_predicted number
method_instances_predicted number
quantities_predicted number
These can be used independently of each other; if you dont expect a lot of named quantities, you can predict just the
vertices through bodies.
6.4 Element lists

The lists of geometric elements follow a general format. Each element is defined on one line. The first entry on a
line is the element number. Numbering need not be consecutive, and may omit numbers, but be aware that internally
elements will be renumbered in order. The original number in the datafile is accessible as the original attribute of an
element. After the element number comes the basic defining data, followed by optional attributes in arbitrary order.
Besides the particular attributes for each element type listed below, one may specify values for any extra attributes
defined earlier. The syntax is attribute name followed by the appropriate number of values. Also an arbitrary number
of named quantities or method instances may be listed. These add method values for this element to the named
quantity. The named quantity or instance must have been declared in the top section of the datafile. See the Named
quantity section above.
6.5 Vertex list

The vertex list is started by the keyword VERTICES at the start of a line. It is followed by lines with one vertex
specification per line in one of these formats:
k x y z [ FIXED] [ CONSTRAINT c1 [c2]] [ BARE] [quantityname ...] [methodname ...]
k p1 [p2] BOUNDARY b [ FIXED] [ BARE] [quantityname ...] [methodname ...]
Here k is the vertex number, a positive integer. Vertices do not need to be listed in order, and there may be gaps
in the numbering. However, if they are not in consecutive order, then the numbering in dump files will be different.
x, y, and z are constant expressions for coordinates in the domain; p1 and p2 are constant expressions for parameter
110
111
values. If FIXED is given, then the vertex never moves, except possibly for an initial projection to constraints. If
CONSTRAINT is given, then one or two constraint numbers must follow. (Actually, you can list as many constraints
as you want, as long as those that apply exactly at any time are consistent and independent.) The given coordinates
need not lie exactly on the constraints; they will be projected onto them.
If BOUNDARY is given, then the boundary parameter values are given instead of the coordinates. The vertex
coordinates will be defined by the coordinate formulas of boundary number b. A vertex may be on only one boundary.
The BARE attribute is just an instruction to the checking routines that this vertex is not supposed to have an adjacent
facet in the soapfilm model, so spurious warnings will not be generated. This is useful when you want to show bare
wires or outline fundamental domains.
6.6 Edge list

The edge list is started by the keyword EDGES at the start of a line. It is followed by lines with one edge specification
per line in this format (linespliced here):
k v1 v2 [midv] [s1 s2 s3] [ FIXED] [ BOUNDARY b] [ CONSTRAINTS c1 c2 ...]
[ TENSION | DENSITY const_ expr] [ COLOR n] [ BARE] [ NO_REFINE] [ NONCONTENT]
[quantityname ...] [methodname ...]
Here k is the edge number, with numbering following the same rules as for vertices. v1 and v2 are the numbers
of the tail and head vertices of the edge. In the quadratic model, the edge midpoint may be listed as a third vertex
midv (by default, a midpoint will be created). In a TORUS model, there follow three signs s1 s2 s3 indicating
how the edge wraps around each unit cell direction: + for once positive, * for none, and - for once negative. FIXED
means that all vertices and edges resulting from subdividing this edge will have the FIXED attribute. It does not mean
that the endpoints of the edge will be fixed. (Note: EFIXED is an obsolete version of FIXED left over from when
FIXED did fix the endpoints.) Likewise the BOUNDARY and CONSTRAINT attributes will be inherited by all edges
and vertices derived from this edge. If a constraint has energy or content integrands, these will be done for this edge.
IMPORTANT: If a constraint number is given as negative, the edge energy and content integrals will be done in the
opposite orientation. In the string model, the default tension is 1, and in the soapfilm model, the default tension
is 0. However, edges may be given nonzero tension in the soapfilm model, and they will contribute to the energy.
NO_REFINE means this edge will not be subdivided by the r command.
If the simplex model is in effect, edges are one less dimension than facets and given by an ordered list of vertices.
Only edges on constraints with integrals need be listed.
The BARE attribute is just an instruction to the checking routines that this ede is not supposed to have an adjacent
facet in the soapfilm model, so spurious warnings will not be generated. This is useful when you want to show bare
wires or outline fundamental domains.
The NONCONTENT attribute indicates the edge is not to be used in any volume calculations in the soapfilm model
or area calculations in the string model.
6.7 Face list

The face list is started by the keyword FACES at the start of a line. It is followed by lines with one facets specification
per line in this format:
k e1 e2 ... [ FIXED] [ TENSION | DENSITY const_ expr] [ BOUNDARY b]
[CONSTRAINTS c1 [c2]] [ NODISPLAY] [ NO_REFINE]
[ COLOR n] [ FRONTCOLOR n] [ BACKCOLOR n] [ PHASE n] [ NONCONTENT]
[quantityname ...] [methodname ...]
Here k is the face number, with numbering following the same rules as for vertices. There follows a list of oriented
edge numbers in counterclockwise order around the face. A negative edge number means the opposite orientation
111
112
of the edge from that defined in the edge list. The head of the last edge must be the tail of the first edge (except if
youre being tricky in the STRING model). There is no limit on the number of edges. The face will be automatically
subdivided into triangles if it has more than three edges in the soapfilm model. The TENSION or DENSITY value is
the energy per unit area (the surface tension) of the facet; the default is 1. Density 0 facets exert no force, and can
be useful to define volumes or in displays. Fractional density is useful for prescribed contact angles. NODISPLAY
(synonym NO_DISPLAY) prevents the facet from being displayed. The COLOR attribute applies to both sides of a
facet; FRONTCOLOR applies to the positive side (edges going counterclockwise) and BACKCOLOR to the negative
side. The PHASE number is used in the string model to determine the surface tension of edges between facets of
different phases, if phases are used. The NONCONTENT attribute means the face will not be used in the volume
calculation for any body it is on. The FIXED, BOUNDARY, CONSTRAINT, DENSITY and NODISPLAY attributes will
be inherited by all facets, edges, and vertices resulting from the subdivision of the interior of the face. NO_REFINE
has no effect on refining the facet itself, but does get inherited by edges created in the interior of the facet.
If the simplex model is in effect, the edge list should be replaced by an oriented list of vertex numbers.
The faces section is optional in the STRING model.
6.8 Bodies
The body list is started by the keyword BODIES at the start of a line. It is followed by lines with one body specification
per line in this format:
k f1 f2 f3 .... [VOLUME v] [VOLCONST v] [PRESSURE p] [DENSITY d]
[PHASE n] [ACTUAL_VOLUME v] [quantityname ...] [methodname ...]
Here k is the body number, and f1 f2 f3 ... is an unordered list of signed facet numbers. Positive sign indicates
that the facet normal (as given by the right-hand rule from the edge order in the facet list) is outward from the body
and negative means the normal is inward. Giving a VOLUME value v means the body has a volume constraint, unless
the ideal gas model is in effect, in which case v is the volume at the ambient pressure. VOLCONST is a value added
to the volume; it is useful when the volume calculation from facet and edge integrals differs from the true volume by a
constant amount, as may happen in the torus model. Beware, though, when changing things that affect body volume;
you may have to reset volconst. ACTUAL_VOLUME is a number that can be specified in the rare circumstances where
the torus volume volconst calculation gives the wrong answer; volconst will be adjusted to give this volume of the body.
Giving a PRESSURE value p means that the body is deemed to have a constant internal pressure p; this is useful for
prescribed mean curvature problems. It is incompatible with prescribed volume. Giving a DENSITY value d means
that gravitational potential energy (gravitational constant G) will be included. v, p, and d are constant expressions.
To endow a facet with VOLUME, PRESSURE, or DENSITY attributes in the STRING model, define a body with
just the one facet.
The PHASE number is used in the soapfilm model to determine the surface tension of facets between bodies of
different phases, if phases are used.
The BODIES section is optional.
6.9 Commands
Encountering the keyword READ in the datafile causes the Evolver to switch from datafile mode to command mode
and read the rest of the datafile as command input. This feature is useful for automatic initialization of the surface with
refining, iteration, defining your own commands, etc.
112
Chapter 7
Language syntax
7.1
General language syntax
Commands are entered one line at a time, parsed, and executed. Multi-line commands may be entered by enclosing
them in braces. If a line ends while nested in braces or parenthesis, Evolver will ask for more input. It will also ask
for more if the line ends with certain tokens (such as +) that cannot legally end a command. Unclosed quotes will
also ask for more, and embedded newlines will be omitted. Explicit continuation to the next line may be indicated by
ending a line with a backslash (linesplicing). You may want to use the read command to read long commands from a
file.
Successfully parsed commands are saved in a history list, up to 100 commands. They may be accessed with !! for
the last command or !string for the latest command with matching initial string. !number will repeat a command by
number. The command will be echoed. The saved history list may be printed with the history command.
Some single-letter commands require interactive input. For those, there are equivalent commands listed below that
take input information as part of the command. This is so commands may be read from a script without having to put
input data on additional lines after the command, although that can still be done for the single-letter versions.
General note: Some commands will prompt you for a value. A null response (just RETURN) will leave the old
value unchanged and return you to the command prompt. On options where a zero value is significant, the zero must
be explicitly entered. Commands that need a real value will accept an arbitrary expression.
Many commands that change the surface or change the model will cause energies and volumes to be recalculated.
If you suspect a command has not done this, the recalc command will recalculate everything. It will also update any
automatic display.
In the following command syntax description, keywords are shown in upper case, although case is irrelevant in
actual commands, except for single-letter commands. Square brackets enclose optional parts of commands.
section General control structures
7.1.1
Command separator
command ; command ...

Several commands on the same line may be separated by a semicolon. Semicolons also are needed to separate
commands inside compound commands. A semicolon is not needed after the last command. Example:
g 10; r; g 10; u
7.1.2
Compound commands
{ command ; ... }
113
114
Curly braces group a list of commands into one command. This is useful in the various control structures. A
semicolon is necessary after a } if there is a following command (note this is different from the C language). Do
not use a semicolon after the first command in an IF THEN ELSE command. An empty compound command { } is
legal. The scope of a variable name may be restricted to a compound command by declaring the name to be local, for
example,
{ local inx;
for ( inx := 1 ; inx <= 5 ; inx += 1 )
{ print inx;
};
};
The use of local variables prevents pollution of global namespace and permits recursive functions. Local variables
can shadow global variables of the same name. Note that a local declaration is not a type declaration, just a scope
declaration.
7.1.3
Command repetition
command expr
For execution of a command a number of times. Be sure to leave a space between a single-letter command and the
expression lest your command be interpreted as one identifier. To avoid several types of confusion, only certain types
of commands are repeatable:
1. single letter commands that dont have optional arguments (l,t,j,m,n,w,P,M,G have optional arguments)
2. command list in braces
3. user-defined procedure names
4. redefined single letter commands
Example:
{ g 10; u; V; } 100
7.1.4
Piping command output
command | stringexpr
The output of the command is piped to a system command. The stringexpr needs to be a quoted string or a string
variable. It is interpreted as a system command.
Examples:
list facets | "more"
list vertices | "tee vlist" ; g 10
list edges | "cat >edgefile"
7.1.5
Redirecting command output
command >> stringexpr

command >>> stringexpr
The output of the command is redirected to a file, appending with >> and overwriting with >>>. The stringexpr
needs to be a quoted string or a string variable.
Redirection with > to replace current contents is not available due to the use of > as an comparison operator.
Examples:
{ { g 10; u } 15 } >> "logfile"
list vertices >>> "vlist.txt"
114
7.1.6
115
Flow of control
This section explains the basic programming constructs that control the sequence of execution of commands.
IF expr THEN command [ ELSE command ]
For conditional execution of commands. expr is true if nonzero. Do not use a semicolon to end the first command.
Examples:
if denom == 0 then return;
if vertex[1] fixed then print "fixed\n" else print "unfixed\n";
if max(edges,length) > 0.02 then { r; g 100 } else g 4;
WHILE expr DO command

DO command WHILE expr
For iterated execution of command controlled by a logical expression. Expression is true if nonzero. Example:
while max(edges,length) > 0.02 do { r; { g 40; u} 5 }
FOR ( command1 ; expr ; command2 ) command3
This is the Evolvers version of the C language "for" construct. The first command is the initialization command;
note that it is a single command, rather than an expression as in C. If you want multiple commands in the initialization,
use a compound command enclosed in curly braces. The middle expression is evaluated at the start of each loop
iteration; if its value is true (i.e. nonzero) then the loop is executed; otherwise the flow of control passes to the
command after command3. The command2 is executed at the end of each loop iteration; again, it is a single command.
The body of the loop is the single command command3, often a compound] command. Note: Command3 should end
with a semicolon, unless it is the if clause of an if-then statement. Examples:
for ( inx := 1 ; inx < 3 ; inx++ )
print facet[inx].area;
for ( {inx := 1; factorial := 1;} ; inx < 7 ; inx++ )
{ factorial *= inx;
printf "factorial %d is %d\n",inx,factorial;
};
ABORT
Causes immediate termination of the executing command and returns to the command prompt. Meant for stopping
execution of a command when an error condition is found. There will be an error message output, giving the file and
line number where the abort occurred, but it is still wise to have a script or procedure or function print an error message
using errprintf before doing the abort command, so the user knows why.
BREAK
115
116
Exits the innermost current loop. Note: Commands with repetition counts do not qualify as loops.
BREAK n
Exits the innermost n loops. Note: Commands with repetition counts do not qualify as loops.
CONTINUE
Skips the rest of the body of the current loop, and goes to the next iteration. Note: Commands with repetition
counts do not qualify as loops.
CONTINUE n
Exits the innermost n-1 loops, and skips to the generator of the nth innermost loop. Note: Commands with
repetition counts do not qualify as loops.
RETURN
Exits the current command. If the current command is a user-defined command called by another command, the
parent command continues. This is essentially a return from a subroutine.
7.1.7
User-defined procedures
Users may define their own command words with the syntax
identifier := { commandlist }
Examples:
gg := { g 10; u; V; }
groom := {
refine edges where length > 0.4 and not fixed;
delete edges where length < 0.1;
u; V; u; V;
};
Users may define their own procedures with arguments with the syntax
PROCEDURE identifier ( type arg1, type arg2, ... )
{ commands}
Right now the implemented types for arguments are real and integer. The argument list can be empty. Example:
procedure proc1 ( real ht, real wd )
{
prod := ht*wd; // this would make prod a global variable
return;
}
Note that the procedure arguments act as local variables, i.e. their scope is the procedure body, and they have stack
storage so procedures may be recursive. Procedure prototypes may be used to declare procedures before their bodies
are defined with the same syntax, just replacing the body of the procedure with a semicolon. Prototype syntax:
116
117
procedure identifier ( type arg1, type arg2, ... ) ;
Note that a procedure is used as a command, and a function is used in a numerical expression.
7.1.8
User-defined functions
Users may define their own functions that have arguments and return values with the syntax
FUNCTION type identifier ( type arg1, type arg2, ... )
{ commands }
Right now the implemented types for the return value and arguments are real and integer. The argument list can be
empty. The return value is given in a return expr statement. Example:
function real proc1 ( real ht, real wd )
{ local prod;
prod := ht*wd;
return prod;
}
Note that the function arguments act as local variables, i.e. their scope is the function body, and they have stack storage
so functions may be recursive. Function prototypes may be used to declare functions before their bodies are defined
with the same syntax, just replacing the body of the function with a semicolon. Prototype syntax:
function type identifier ( type arg1, type arg2, ... ) ;
Note that a procedure is used as a command, and a function is used in a numerical expression.
7.2 Expressions
Expressions are of three types: numeric, numeric array, and string. String expressions are either quoted strings or
are created by the SPRINTF command; successive quoted strings will be concatenated into one string. Numeric
expressions are always floating-point, not integer. Boolean values for conditions are also floating point, 0 for false,
nonzero for true (1 for true result of booleanoperation). Integer numeric values are all converted to real values internally
for expression evaluation.
Numeric values:
constant
An explicit number: integer, fixed point, scientific, hexadecimal

(0x12AC), or binary notation (0110110b).
Current gravitational constant
e, pi
Special constants.
identifier
Name of a numeric variable.
identifier [expr]...
A fully indexed array name, one array entry.
element.attribute
Numeric element attribute. The element may be specified by an element

generator (see section 11.2), an element name from a generator (again
see section 11.2), or it may be absent if it is an attribute of the current
element in an element generator.
117
118
functionname(arglist)
Function call.
max, min, avg, sum, count
Aggregate expressions.
quantity.attribute
An attribute of a named quantity: value, target, fixed, pressure, modulus,

tolerance.
instance.attribute
An attribute of a named method instance.
constraintname
A constraint name used where a value is expected is converted to its

internal identifying number.
boundaryname
A parameterized boundary name used where a value is expected is converted to its internal identifying number.
parameter.attribute
Optimizing parameter attributes: pdelta, pscale, or fixed.
colorname
The standard color names black, blue, ..., white are treated as the integers
0 to 15.
togglename
A toggle name used in an expression has value 1 for ON and 0 for OFF.
This does not apply to single-letter toggles like U, but those all have word
equivalents, such as conj_grad.
Numeric array expressions: These may be used in array arithmetic or with the print command.
arrayname
A non-indexed array name.
arrayname[expr]...
A partially indexed array name, a slice of the array.
element.arrayname
A non-indexed array attribute of an element.
element.arrayname[expr]...
A partially indexed element array attribute, a slice of the attribute array.
Operators, high to low precedence: Scalar means an operator acts on single numeric values.
1. ()
Grouping and functional notation.
2. , **
Raise to real power, e.g. 2[]4 = 16. Scalar.
3. Unary minus, as in -x. Scalar and array.
4. *
Multiplication. Scalar and array.
/
Real division. Scalar.
%, mod
Real modulus, i.e. remainder after division. 11 % 3 = 2. Scalar.
imod
Integer modulus. Scalar. Real operands are rounded down to integer
before modulus.
idiv
Integer division, e.g. 34 imod 5 is 6. Scalar. Real operands are rounded
towards zero.
5. +,Addition and subtraction. Scalar and array.
6. ==,>,<,<=,>=,!= Comparison. Scalar.
7. not, !
Logical NOT. Scalar.
8. and, &&
Logical AND, short-circuit evaluation. Scalar.
9. or, ||
Logical OR, short-circuit evaluation. Scalar.
10. ? :
Conditional expression, as in C language. That is, expr1 ? expr2 : expr3
evaluates to expr2 if expr1 is true (nonzero) and to expr2 if expr1 is false
(zero). Scalar.
118
119
NOTE: In the datafile, a + or - preceded by whitespace and followed by a number is taken to be a signed
number. Thus 3 - 5" and 3-5" are single expressions, but 3 -5" is not. This is for convenience in separating multiple
expressions listed on the same line in the datafile.
NOTE: The boolean AND and OR operators use short-circuit evaluation; i.e. the second operand is evaluated only
if necessary.
Increment and decrement: Pre- and post-increment and decrement operators ++ and --, as in the C language, work
on single variables only as stand-alone commands. ++var or var++ will increase the value of var by 1, and var or var-will decrease the value of var by 1. Examples:
change_count++;
--left_to_do;
for ( inx := 1 ; inx <= 10 ; inx++ )
are legal, but "zsum := zarray[inx++]" is not, since "inx++" is a command, not a value in Evolver.
String expressions: A string expression evaluates to a string of characters. At present, the only ways to produce
strings are:
1. double-quoted string literals, e.g. "this is a string". The following standard C escape sequences are recognized:
\n newline
\r carriage return
\t tab
\b backspace
\q double-quote mark
\c the character sl c elsewise, e.g. \\ gives \ and \" gives ".
2. string variables, either internal like datafilename, or user-defined.
3. output from sprintf.
In Microsoft Windows file paths, use / as the directory separator, since backslash is an escape character. Windows has
always accepted / as a directory separator.
7.3 Aggregate expressions

The maximum, minimum, sum, count, and average of an expression over a set of elements may be done with aggregate
expressions. The histogram and loghistogram commands have the same form. The general form is
aggregate( generator, expr)
where aggregate may be max, min, sum, count, or avg. The generator generates elements as discussed above. Max
and min over logical expressions may be used for logical any and all. The max of an empty set is large negative, and
the min of an empty set is large positive, so beware when using WHERE clauses..
Examples:
print max(edge,length);
ave_area := avg(facet where not fixed,area);
histogram(facet where color==green, area);
show facets ff where max(ff.edge ee,max(ee.facet fff, fff.id == 4))}
119
Chapter 8
Internal variables
Here is an alphabetical list of variables defined by Evolver.
ambient_pressure_value
Value of the external pressure in the ideal gas model. Read-write.
area_fixed
The target value for the old fixed area constraint. Does not activate constraint. Obsolete. Better now to use a named quantity with the facet_area
method. Read-write.
autochop_length
The cutoff length for the autochop command. Writing it does not toggle
autochop on. Read-write.
background
Background color used on certain screen graphics. Read-write.
body_count
Number of bodies. Read-only.
body_dissolve_count
Number of facets dissolved by dissolve bodies command. Read-only.

Reset by the reset_counts command.
breakflag
When set to a non-zero value, causes the command interpreter to abort

and return to the command prompt. Software equivalent of hitting the
keyboard interrupt (typically CTRL-C). The break doesnt happen immediately, but at a certain point in the interpreter loop when it periodically
checks for user interrupt. Meant for bailing out of nested commands,
since return only breaks out of the current procedure. Obsolete, maybe.
Use the abort command instead. Read-write.
brightness
Median gray level used in PostScript output and screen graphics. Readwrite.
check_count
Number of errors found by the most recent C command. Read-only.
clock
Process elapsed time in seconds since starting Evolver. Read-only.
120
121
console_title
This string variable is the title that is displayed on the Evolver command
console window. The default value is "Surface Evolver - datafilename"
(with the name of the current datafile, of course). Useful when you are
simultaneously running various instances, and you want to tell which is
running in which console window. Just assigning a string to console_title
automatically changes the title on the window. Read-write.
colorfile
String variable holding the file name for the colormap option of the P
command. Read-write.
constraint_tolerance
Constraint value regarded as equivalent to zero.Read-write.
cpu_counter
CPU cycles since booting (available so far on x86 only). Read-only.
curvature_power
The power of curvature used in the string model when squared_curvature

is declared in the top of the datafile or the sqcurve_string named method
is used. Read-write.
datafilename
String containing the name of the current datafile. Read-only.
date_and_time
A string containing the current date and time. Read-only.
delete_count
Sum of edge_delete_count and facet_delete_count. Read-only. Reset by

the reset_counts command.
detorus_epsilon
The tolerance for detorus to identify vertices when detorus_sticky is toggled on. Read-write.
diffusion_coeff
Variable controlling the rate of diffusion of volume across surfaces when

diffusion is in effect. If you want finer control over the rate of diffusion
across various surfaces, you can define the edge_diffusion edge attribute
in the string model or the facet_diffusion facet attribute in the soapfilm
model and give individual values for edges or facets as you desire. If
the attribute is defined, then its value is used instead of diffusion_coeff.
Read-write.
dissolve_count
Sum of vertex_dissolve_count, edge_dissolve_count,
edge_count
Number of edges. Read-only.
edge_delete_count
Number of edges deleted by delete edges command. Read-only. Reset

by the reset_counts command.
edge_dissolve_count
Number of edges dissolved by dissolve edges command. Read-only.

edge_pop_count
Number of edges popped by pop edges or O command. Read-only.

121
122
edge_refine_count
Number of edges refined by refine edges command. Read-only. Reset

edgeswap_count
Number of edges flipped by edgeswap command. Read-only. Reset by

the reset_counts command.
eigenneg
Number of negative eigenvalues in last Hessian factoring. Read-only.
eigenpos
Number of positive eigenvalues in last Hessian factoring. Read-only.
eigenvalues
Array containing the list of eigenvalues produced by the ritz command.

Read-only.
eigenzero
Number of zero eigenvalues in last Hessian factoring. Read-only.
equi_count
Number of edges flipped by equiangulate or u commands. Read-only.

estimated_change
estimated change during g step when estimate toggle is on. Read-only.
facet_count
Number of facets. Read-only.
facet_delete_count
Number of facets deleted by delete facets command. Read-only. Reset

facet_dissolve_count
Number of facets dissolved by dissolve facets command. Read-only. Reset by the reset_counts command.
facet_refine_count
Number of facets refined by refine facets command. Read-only. Reset

facetedge_count
Number of facetedges. Read-only.
fix_count
Number of elements fixed by fix command. . Read-only. Reset by the

reset_counts command.
graphics_title
This string variable is the title that is displayed on the Evolver graphics window. The default value is the datafile name. Useful when you
are simultaneously running various instances, and you want to tell which
is running in which graphics window. Just assigning a string to graphics_title automatically changes the title on the window. There are also
graphics_title2 and graphics_title3 variables if you have a second or third
graphics window. Read-write.
gravity_constant
Value of the gravitational constant; also set by the G command. Readwrite.
122
123
hessian_epsilon
Magnitude regarded as zero by hessian command when factoring the

Hessian matrix. If a zero appears on the diagonal at the pivot during
factoring, a check is made to see if the rest of the row is zero. If it is,
then a 1 is placed on the diagonal; else an error is reported. Read-write.
hessian_slant_cutoff
Makes hessian commands treat vertices whose normal vector is nearly

perpendicular to constraints as fixed. hessian_slant_cutoff is the cosine of angle. Works on vertices with one degree of freedom in hessian_normal mode. Read-write.
high_boundary
Highest parameterized boundary number (including named boundaries).

Read-only.
high_constraint
Highest level-set constraint number (including named constraints).

Read-only.
inverse_periods[expr][expr]
Inverse of the torus_periods matrix; 1-based indices. Read-only.
integral_order
Order of polynomial done by 1D and 2D Gaussian integration Much better to set 1D and 2D separately. Read-write.
integral_order_1d
Order of polynomial done by 1D Gaussian integration. Read-write.
integral_order_2d
Order of polynomial done by 2D Gaussian integration. Read-write.
iteration_counter
Value of index of current iteration loop. Read-only.
jiggle_temperature
Current temperature for jiggling. Read-write.
lagrange_order
Order of Lagrange model. Set with the lagrange command. Read-write.
last_eigenvalue
Lowest eigenvalue from last saddle or ritz command. Read-only.
last_error
Number of last error or warning message. Read-write. Writable so you

can set it equal to 0 to clear the error indication.
last_hessian_scale
Stepsize from last saddle or hessian_seek command. Read-only.
linear_metric_mix
Fraction of linear interpolation in Hessian metric, as opposed to vertex

weighting. Read-write.
memory_arena
Total memory allocated to the programs heap. Windows versions only.

Read-only.
memory_used
Total memory used in the programs heap. Windows versions only.

Read-only.
notch_count
Number of edges notched by notch command. Read-only. Reset by the

123
124
pickenum
Number of last edge picked in graphics display. Read-write.
pickfnum
Number of last facet picked in graphics display. Read-write.
pickvnum
Number of last vertex picked in graphics display. This vertex number is

used by the graphics window F key command to set the vertex at the
center of rotation, which is why it is useful to be writable. Read-write.
pop_count
Sum of vertex_pop_count and edge_pop_count. Read-only. Reset by the

pop_edge_to_tri_count
Number of edges flipped to triangles by pop_edge_to_tri command.

Read-only. Reset by the reset_counts command.
pop_quad_to_quad_count
Number of quadrilaterals flipped by pop_quad_to_quad command.

pop_tri_to_edge_count
Number of triangles flipped to edges by pop_tri_to_edge command.

ps_bareedgewidth
Width of bare edges in PostScript output, in absolute terms relative to an

image width of 3. Read-write.
ps_conedgewidth
Width of constraint or boundary edges in PostScript output, in absolute

terms relative to an image width of 3. Read-write.
ps_fixededgewidth
Width of fixed edges in PostScript output, in absolute terms relative to

an image width of 3. Read-write.
ps_gridedgewidth
Width of edges in PostScript output for which none of the special edge
categories apply, in absolute terms relative to an image width of 3. Readwrite.
ps_labelsize
Relative width of element labels in PostScript output. Default is 3; a

value of 1 gives small but still readable labels relative to an image width
of 3. Read-write.
ps_tripleedgewidth
Width of edges with valence three or more in PostScript output, in absolute terms relative to an image width of 3. Read-write.
ps_stringwidth
Normal width of string model edges in PostScript output, in absolute

terms relative to an image width of 3. Read-write.
quadratic_metric_mix
Fraction of quadratice interpolation in Hessian metric in the quadratic

model. (Rather useless to change from the default value of one.) Readwrite.
random
Random real number between 0 and 1, different each time. Read-only.
124
125
random_seed
Seed for random number generator. Defaults to 1 at start of datafile.

Re-initializes generator whenever written. Read-write.
refine_count
Sum of number of edge_refine_count and facet_refine_count Read-only.

scale
Current scale factor. Read-write.
scrollbuffersize
Set command window scroll buffer size on Windows machines. Meant

for non-NT Windows that dont have menu properties option for this.
Read-write.
self_sim_coeff
Variable used as magnitude coefficent when self_similar is toggled on.

Read-write.
simplex_representation
Whether facets are represented as simplices (Boolean). Read-only.
space_dimension
Dimension of ambient space. Read-only.
string_curve_tolerance
In the quadratic model, the smoothness of graphing of curved quadratic

edges can be controlled with the internal variable string_curve_tolerance,
which is the desired angle in degrees between successive graphed segments making up the edge. Read-write.
surface_dimension
Dimension of surface, 1 for string model, 2 for soapfilm model. Readonly.
symmetry_group
Whether any symmetry active (Boolean). Read-only.
t1_edgeswap_count
Number of edges flipped by t1_edgeswap command. Read-only. Reset

target_tolerance
Default value of fixed quantity error tolerance, defaults to 0.0001. Readwrite.
thickness
Thickness to separate facet sides of different colors when doing 3D

graphics, to prevent weird stippling effects. Read-write.
torus
Whether torus domain is in effect. (Boolean) Read-only.
torus_filled
Whether torus_filled mode is in effect (Boolean). Read-only.
torus_periods[expr][expr]
Torus period vectors, as in datafile top section; 1-based indices. Note that
the rows are the period vectors. Read-only.
total_area
Total area of the surface (soapfilm model). Read-only.
total_energy
Total energy of the surface. Read-only.
total_length
Total length of the surface (string model). Read-only.
125
126
total_time
Elapsed time in the form of total scale factors. Read-only.
transform_count
Number of view transformations active. Read-only.
unfix_count
Number of elements unfixed by unfix command Read-only. Reset by the

vertex_count
Number of vertices. Read-only.
vertex_dissolve_count
Number of vertices dissolved by dissolve vertices command. Read-only.

vertex_pop_count
Number of vertices popped by pop vertices or o or O command.

view_matrix[][]
Matrix that transforms from world coordinates (those of the surface) to

screen coordinates. The file saveview.cmd in the Evolver distribution has
a command saveview which can be used to save the current view_matrix
in a file which can be read back in to restore a given view. Read-write.
view_transforms_unique_point[]
Vector used as point coordinate by transform_expr command in

view_transforms_use_unique_point mode for discarding duplicate view
transforms. By default, "duplicate" means same matrix, but in this
mode, "duplicate" is judged by the image of the coordinates in the
view_transforms_unique_point vector. Read-write.
Number of items satisfying last where clause. Read-only. Reset by the
where_count
8.1 Utility functions

sizeof(name)
Number of entries in an array or an array extra attribute name (which is

not in quotes). Can also be applied to a string or string variable to get the
number of characters in the string.
is_defined(stringexpr)
Boolean function. Returns 1 if the identifier in string is known to the

Evolver, as keyword or variable name or quantity name or whatever, 0
if not. This function is evaluated at run-time, but variables in the whole
command are parsed before the command is executed, so a command
like if is_defined("newvar") then newvar := 1 else newvar := 2 will give
newvar the value 1 even if this is its first appearance. A better way in
scripts to test is to use the define command to define the variable without
initialization, and then test to see if it has the default value, i.e. 0 for a
numeric variable and a sizeof 0 for a string variable.
126
valid_element(indexed_element)
127
Boolean function. Returns 1 or 0 depending on whether an element of a

given index exists. Examples:
if valid_element(edge[12]) then refine edge[12]
if valid_element(body[2]) then set body[2].facet color red
valid_constraint(expr)
Boolean function. Returns 1 if the integer argument is the number of

a valid level-set constraint (recall that named constraints have internal
numbers, which are used here).
valid_boundary(expr)
Boolean function. Returns 1 if the integer argument is the number of a

valid parametrized boundary (recall that named boundaries have internal
numbers, which are used here).
127
Chapter 9
Single-letter commands
The oldest and most commonly used commands are just single letters. Case is significant for these. Single letters are
always interpreted as commands, so you may not use single letters for variable names.
It is possible to reassign a single letter to your own command by the syntax
letter :::= command
but this should only be used in special circumstances, such as redefining r to do additional actions along with refinement. The old meaning can be restored with a null assignment, letter :::=. Use single quotes around the letter to
get the old meaning, i.e. r will do a standard refine when r has been redefined. Redefinitions are cleared when a new
surface is loaded. Be careful when using redefined commands in defining other commands. Redefinition is effective
on execution of the redefinition command, not on parsing. Redefinition is not retroactive to uses in previously defined
commands.
9.1
Single-letter command summary
This is a brief summary; see the next section for full descriptions. There are conceptually five groups of commands:
1. Reporting:
C
c
e
i
v
X
z
Run consistency checks.

Report count of elements.
Extrapolate.
Information on status.
Report volumes.
List element extra attributes.
Do curvature test.
2. Model characteristics:
128

A
a
b
f
G
J
k
M
m
p
Q
U
W
129
Set adjustable constants.

Toggle area normalization.
Set body pressures.
Set diffusion constant. Needs value.
Set gravity. Needs value.
Toggle jiggling on every move.
Set boundary gap constant.
Toggle linear/quadratic model.
Toggle fixed motion scale. Needs value to set, not to unset.
Set ambient pressure.
Report or set quantities.
Toggle conjugate gradient method.
Homothety toggle.
3. Surface modification:
g Go one iteration step. Often followed by a repetition count.
j
Jiggle once. Requires value for jiggle size.
K Skinny triangle long edge divide. Requires value.
l
Subdivide long edges. Requires value.
N Set target volumes to actual.
n Notch ridges and valleys. Requires value.
O Pop non-minimal edges.
o Pop non-minimal vertices.
r
Refine triangulation.
t
Remove tiny edges. Requires value.
u Equiangulate.
V Vertex averaging.
w Weed out small triangles. Requires value.
y Torus duplication.
Z Zoom in on vertex.
4. Output:
D
d
P
s
Toggle display every iteration.

Dump surface to datafile.
Create graphics display file.
Show triangulation graphically.
5. Miscellaneous:
F Toggle command logging.
H,h,? Help screen.
q,x Exit.
The G, j, l, m, n, t, and w commands require real values, which may be entered on the same line, or given in response
to a prompt if not.
9.2
A
Alphabetical single-letter command reference

Adjustable values. Displays current values and allows you to enter new values. New value is entered as the
number of the constant (from the list) and the new value. Values of named quantities, their moduli, and the
target values of fixed named quantities also appear here. The modulus and target value may be changed. Only
explicitly user-defined named quantities appear here, unless show_all_quantities is toggled on.
129
130
Toggles area normalization of vertex forces and other gradient. Be sure you have a small enough scale factor,
or else things tend to blow up. Reduce scale factor temporarily after refinement, since triangle areas are cut by
a factor of 4 but the old creases remain. When this option is ON, there is an optional check that can be made
for facets that move too much. This is done by computing the ratio of the length of the normal change to the
length of the old normal. If this exceeds the user-specified value, then all vertices are restored to their previous
position. The user should reduce the motion scale factor and iterate again.
Permits user to change body prescribed volumes or pressures. Prints old value for each body and prompts for
new.
Runs various internal consistency checks. If no problems, just prints Checks completed. The number of errors
found is stored in the variable check_count. The checks are:
Element list integrity - checks that data structures are intact. This kind of error is probably an Evolver bug
and should be reported.
Facet-edge check - that if a facet adjoins an edge, then the edge adjoins the facet, and that the three edges
around a facet link up. This kind of error is probably an Evolver bug and should be reported.
Facet-body check - whether adjacent facets have the same body on the same side. Probably a user problem
due to mis-oriented faces in body definitions in the datafile, or due to the surface getting kinked up at triple
lines.
Collapsed elements - check if endpoints of an edge are the same, and whether neighboring facets share
more than one edge and two vertices. Not illegal, but you probably want to avoid.
After "C" or "check" command finishes, there are some variables that hold the number of errors of various types
that were found:
bad_next_prev_count - bad links in element linked lists.
inconsistent_bodies_count - violations of adjacent facets having same bodies.
edge_loop_count - edges that are loops on single vertices.
edges_same_vertices_count - edge pairs with the same endpoints.
facets_same_vertices_count - facet pairs with the same endpoints.
bad_error_count - sum of the various types of errors that I consider serious enough that you should revise
your evolution to avoid them. Bad links within element lists, and bad links between elements.
Prints count of elements and memory used (just for element structures, not everything) and prints various model
parameters. Synonym: counts.
Toggles displaying every iteration. Default is ON.
Dumps data to ASCII file in same format as initial data file. You will be prompted for a filename. An empty
reponse will use the default dump name, which is the datafile name with a .dmp extension. Useful for checking
your input is being read correctly, for saving current configuration, and for debugging.
Extrapolates total energy to infinite refinement if at least two refinements have been done. Uses last energy values
at three successive levels of refinement. Synonym: extrapolate.
Toggle logging of commands in file. If starting logging, you will be prompted for the name of a log file. Any
existing file of that name will be overwritten. Logging stops automatically when the surface is exited. Only
commands that change the surface are logged.
Set diffusion constant. Prints old and prompts for new.
130
131
Toggles gravity on or off. Gravity starts ON if any body has DENSITY; otherwise OFF. If followed by a value,
sets gravity to that value. Otherwise prints old value of gravitational constant and prompts for new. Optionally
takes new gravity value on command line.
Same as go command. Do iteration step. Each iteration calculates the force on each vertex due to its contribution
to the total energy and moves the vertex by a multiple of the force. There is a global scale factor that multiplies
the force to give the displacement. If area normaliztion is turned on, the force at each vertex is also divided by
the total area of facets adjacent to the vertex to better approximate motion by mean curvature (but this seems
to be often numerically badly behaved due to long skinny facets). If any bodies have prescribed volumes, the
vertices are also displaced to bring the volumes back to near the prescribed values.
If scale optimizing (see command m) is ON, then different global scale values will be tried until a quadratic
interpolation can be done to find the optimal value. (This can blow up under certain circumstances.) The scale
factor used for motion is then multipied by the scale_scale variable (default value 1).
The output consists of the number of iterations left (for people who wonder how close their 1000 iterations are to
ending), the area and energy, and the scale factor. The user can abort repeated iterations by sending an interrupt
to the process (SIGINT, to be precise; CTRL-C or whatever on your keyboard).
h,H,?
i
Prints help screen listing these commands.
Prints miscellaneous information:

Total energy
Total area of facets
Count of elements and memory usage
Area normalization, if on
LINEAR or QUADRATIC model
Whether conjugate gradient on
Order of numerical integration
Scale factor value and option (fixed or optimizing)
Diffusion option and diffusion constant value
Gravity option and gravitational constant value
Jiggling status and temperature
Gap constant (for gap energy, if active)
Ambient pressure (if ideal gas model in effect)
Toggles jiggling on every iteration. If jiggling gets turned on, prompts for temperature value.
Jiggles all vertices once. Useful for shaking up surfaces that get in a rut, especially crystalline integrands. You
will be prompted for a temperature which is used as a scaling factor, if you dont give a temperature with
the command. Default value is the value of the jiggle_temperature internal variable, which starts as 0.05. The
actual jiggle is a random displacement of each vertex independently with a Gaussian distribution. See the longj
command below for a user-definable perturbation.
Finds skinny triangles whose smallest angle is less than a specified cutoff. You will be prompted for a value.
Such triangles will have their longest edge subdivided. Should be followed with tiny edge removal (t) and
equiangulation (u). By default, the long edge is subdivided at its midpoint, but if you do k_altitude_mode on
then it will be subdivided at the foot of the altitude from the opposite vertex.
131
132
Sets gap constant for gap energy for CONVEX boundaries. Adds energy roughly proportional to area between
edge and boundary. You will be prompted for a value. Normal values are on the order of magnitude of unity.
Value k = 1 is closest to true area. Use 0 to eliminate the energy.
Subdivides long edges, creating new facets as necessary. You will be prompted for a cutoff edge length, if you
dont give a value with the command. Existing edges longer than the cutoff will be divided once only. Newly
created edges will not be divided. Hence there may be some long edges left afterward. If you enter h, you will
get a histogram of edge lengths. If you hit ENTER with no value, nothing will be done. It is much better to
refine than to subdivide all edges. A synonym for "l value" is "edge_divide value".
Sets model type to LINEAR, QUADRATIC, or LAGRANGE. Changing from LINEAR to QUADRATIC adds vertices
at the midpoints of each edge. Changing from QUADRATIC to LINEAR deletes the midpoints. Likewise for
Lagrange. Optionally takes new model type ( 1 (linear), 2 (quadratic), or > 2 (Lagrange of given order) ) on
command line.
Toggles quadratic search for optimal global motion scale factor. If search is toggled OFF, you will be prompted
for a scale factor. If you give a value with the command, then you are setting a fixed scale factor. Values around
0.2 work well when the triangulation is well-behaved and area normalization (command a) is off. In optimizing
mode, a scale factor getting small, say below 0.01, indicates triangulation problems. Too large a scale factor
will show up as total energy increasing. If you have motion by area normalization ON (command a), use a small
scale factor, like 0.001, until you get a feel for what works.
Normalize body prescribed volumes to current actual volumes.
Notching ridges and valleys. Finds edges that have two adjacent facets, and those facets normals make an angle
greater than some cutoff angle. You will be prompted for the cutoff angle (radians) if you dont give a value with
the command. Qualifying edges will have the adjacent facets subdivided by putting a new vertex in the center.
Should follow with equiangulation. In the string model, it will refine edges next to vertices with angle between
edges (parallel orientation) exceeding the given value. Optionally takes cutoff angle on command line.
Pop non-minimal edges. Scans for edges with more than three facets attached. Splits such edges into triple-facet
edges. Splits propagate along a multiple edge until they run into some obstacle. It also tries to pop edges on
walls properly. Try octa.fe for an example. For finer control, use the pop command.
Pop non-minimal vertices. This command scans the surface for vertices that dont have the topologies of one of
the three minimal tangent cones. These are popped to proper local topologies. The algorithm is to replace the
vertex with a sphere. The facets into the original vertex are truncated at the sphere surface. The sphere is divided
into cells by those facets, and the largest cell is deleted, which preserves the topology of the complement of the
surface. A special case is two cones meeting at a vertex; if the cones are broad enough, they will be merged,
otherwise they will be split. In case of merging cones, if both cone interiors are defined to be part of the same
body, then no facet is placed across the neck created by the merger; if they are different bodies or no bodies,
a facet will be placed across the neck. Only vertices in the interior of a surface, not fixed or on constraints or
boundaries, are tested. This command tends to create a lot of small edges and skinny triangles. Try popstr.fe
and octa.fe for examples.
Produce graphics output files. The view is the same as seen with the s command. Several formats are currently
available, and a menu will appear. These are now Pixar, geomview, PostScript, SoftImage, and a personal
format called Triangle. You may optionally give the menu item number on the command line. If you are doing
a torus surface, you will be asked for a display option, for which see the Torus section. You will be prompted
for a filename. For Pixar format, you will be asked whether vertices should have normal vectors for normal
interpolation (calculated as the average normal of all facets around a vertex); whether inner facets (adjacent to
two bodies) outer facets (adjacent to zero or one body), or all facets are to be plotted; and whether different
colors should be used for different bodies. If so, you are asked for the name of a file with a colormap in the
format of RGB values, one set per line, values between 0 and 1. (This map may not be effective on all devices.)
132
133
You may also be asked if you want thickening. If you do, each facet will be recorded twice, with opposite
orientations, with vertices moved from their original positions by the thickening distance (which the option lets
you enter) in the normal direction. The normal used at each vertex is the same as used for normal interpolation,
so all the facets around a planar vertex will have that vertex moved the same amount. Triple junctions will
be separated. Thickening is good for rendering programs that insist on consistently oriented surfaces, or that
cant see the backside of a surface. The default thickening distance is one one-thousandth of the diameter of the
object.
For file formats, see the section on Graphics file formats.
For those of you that have geomview, the relevant commands are here. Geomview uses a pipe interface at the
moment. Besides options for starting and stopping simultaneous geomview, there are options for starting a
named pipe without invoking geomview. You will be told the name of the pipe, and it is up to you to start a
pipe reader. Evolver blocks until a pipe reader is started. This is useful for having a second instance of Evolver
feed a second surface to geomview by having geomview load the pipe. Commands in the geomview command
language may be sent with the command ggeomview string.
This command can be followed by a number to pick a menu option without displaying the menu. Thus P 8 starts
geomview.
See also the GEOMVIEW and POSTSCRIPT commands.
p
Sets ambient pressure in ideal gas model. A large value gives more incompressible bodies.
Single letter main command. Report current values of user-defined method instances and named quantities.
If theshow_all_quantities toggle is on, then internal quantities and method instances are also shown. This is
particularly informative if convert_to_quantities has been done (same as -q command line option), since then
internal values such as constraint integrals are in the form of method instances.
Exit program. You will be given a chance to have second thoughts. You may also load a new datafile. Automatically closes graphics if you really quit. Does not save anything.
Refines the triangulation. Edges are divided in two, and SOAPFILM facets are divided into four facets with
inherited attributes. Edges and facets with the no_refine attribute set are not refined. Reports the number of
structures and amount of memory used.
Shows the surface on the screen. The proper display routines must have been linked in for the machine the
display is on. Goes into the graphics command mode (see below). Torus surfaces have display options you
will be asked for the first time, for which see the Torus section. The graphics window may be closed with the
close_show command.
Eliminates tiny edges and their adjacent facets. You will be prompted for a cutoff edge length if you dont give
a value with the command. If you enter h, you will get an edge length histogram. If you hit ENTER without a
value, nothing will happen. Some edges may not be eliminable due to being fixed or endpoints having different
attrtibutes from the edge.
This toggles conjugate gradient mode, which will usually give faster convergence to the minimum energy than
the default gradient descent mode. The only difference is that motion is along the conjugate gradient direction.
The scale factor should be in optimizing mode. The history vector is reset after every surface modification, such
as refinement or equiangulation. After large changes (say, to volume), run without conjugate gradient a few
steps to restore sanity.
This command, called equiangulation, tries to polish up the triangulation. In the soapfilm model, each edge
that has two neighboring facets (and hence is the diagonal of a quadrilateral) is tested to see if switching the
quadrilateral diagonal would make the triangles more equiangular. For a plane triangulation, a fully equiangulated triangulation is a Delaunay triangulation, but the test makes sense for skew quadrilaterals in 3-space
also. It may be necessary to repeat the command several times to get complete equiangulation. The tt edgeswap
133
134
command can force flipping of prescribed edges. In the simplex model, equiangulation works only for surface
dimension 3. There, two types of move are available when a face of a tetrahedron violates the Delaunay void
condition: replacing two tetrahedra with a common face by three, or the reverse operation of replacing three
tetrahedra around a common edge by two, depending on how the condition is violated. This command does not
work in the string model.
V
Vertex averaging. For each vertex, computes new position as area-weighted average of centroids of adjacent
facets. Only adjacent facets with the same constraints and boundaries are used. Preserves volumes, at least
to first order. See the rawv command below for vertex averaging without volume preservation, and rawestv
for ignoring likeness of constraints. Does not move vertices on triple edges or fixed vertices. Vertices on
constraints are projected back onto their constraints. All new positions are calculated before moving. For
sequential calculation and motion, see the vertex_average command.
Shows prescribed volume, actual volume, and pressure of each body. Also shows named quantities. Only
explicitly user-defined named quantities are shown, unless show_all_quantities has been toggled on. Pressures
are really the Lagrange multipliers. Pressures are computed before an iteration, so the reported values are
essentially are one iteration behind. Synonym: show_vol.
Toggles homothety. If homothety ON, then after every iteration, the surface is scaled up so that the total volume
of all bodies is 1. Meant to be used on surfaces without any constraints of any kind, to see the limiting shape as
surface collapses to a point.
Tries to weed out small triangles. You will be prompted for the cutoff area value if you dont give a value with
the command. If you enter h, you will get a histogram of areas to guide you. If you hit RETURN with no value,
nothing will be done. Some small triangles may not be eliminable due to constraints or other such obstacles.
The action is to eliminate an edge on the triangle, eliminating several facets in the process. Edges will be tried
for elimination in shortest to longest order. WARNING: Although checks are made to see if it is reasonable to
eliminate an edge, it is predicated on facets being relatively small. If you tell it to eliminate all below area 5,
Evolver may eliminate your entire surface without compunction.
List the current element extra attributes, including name, size, and type. Some internal attributes are also listed,
beginning with double underscore.
Exit program. You will be given a chance to have second thoughts. You may also load a new datafile. Automatically closes graphics if you really quit. Does not save anything.
Torus duplication. In torus model, prompts for a period number (1,2, or 3) and then doubles the torus unit cell in
that direction. Useful for extending simple configurations into more extensive ones.
Zooms in on a vertex. Asks for vertex number and radius. Number is as given in vertex list in datafile. Beware
that vertex numbers change in a dump (but correct current zoom vertex number will be recorded in dump).
Eliminates all elements outside radius distance from vertex 1. New edges at the radius are made FIXED. Meant
to investigate tangent cones and intricate behavior, for example, where wire goes through surface in the overhand
knot surface. Zooming is only implemented for surfaces without bodies.
Do curvature test on quadratic model. Supposed to be useful if youre seeking a surface with monotone mean
curvature. Currently checks angle of creases along edges and samples curvature on facet interiors. Orientation
is with respect the way facets were originally defined.
134
Chapter 10
Toggles
Various features can be toggled on or off by giving the toggle name with ON or OFF. Just using the toggle name alone
is equivalent to using it with ON. This is different from the single-letter command toggles, which always change the
state. The toggle names below have brief descriptions of their actions in the ON state. Toggles will usually print their
previous state. The current value of a toggle may be found by the command "print togglename".
A few toggles are initialized to an "unset" state, which prints as -1. These will generally prompt for a value when
needed.
AMBIENT_PRESSURE
Toggles ideal gas mode, where there is a fixed external pressure. The external pressure can be set with the pressure
phrase in the top of the datafile, or at runtime with the p command, e.g. "p 10". Default is OFF.
APPROXIMATE_CURVATURE
Use polyhedral curvature (linear interpolation over facets for metric) for mean curvature vector. Actually establishes the inner product for forms or vectors to be integration over facets of euclidean inner products of linear
interpolation of values at vertices. synonyms: approx_curv, approx_curvature. Default is OFF.
AREA_NORMALIZATION
Convert force on vertex to velocity for motion by the g command by dividing the force by one-third of the area of
the facets adjecent to the vertex in the soapfilm model, or by half the length of the adjacent edges in the string model.
This is meant to make the velocity a discrete approximation of the mean curvature when the energy is surface tension.
Similar to the a command. Default is OFF.
ASSUME_ORIENTED
Tells squared mean curvature routines that they can assume the surface is locally consistently oriented. Significant
only for extreme shapes. Default is OFF.
AUGMENTED_HESSIAN
Solves constrained Hessians by putting the body and quantity constraint gradients in an augmented matrix with
the Hessian, and using sparse matrix techniques to factor. Vastly speeds things up when there are thousands of sparse
constraints, as in a foam. The default state is unset (prints as a value of -1), in which case augmentation is used for 50
or more constraints, but not for less.
AUTOCHOP
Do autochopping of long edges each iteration. By an abuse of syntax, it also doubles as a variable for the cutoff
length. Use "autochop := expr" to set the autochop length and toggle autochopping on. Or the read-write variable
autochop_length can be used without affecting the toggle state. Each iteration, any edge that is projected to become
longer than the cutoff is bisected. If any bisections are done, the motion calculation is redone. Default is OFF.
AUTODISPLAY
Toggle automatic display every time the surface changes. Same effect as the D command. Default is ON.
135
136
AUTOPOP
Toggles automatic deletion of short edges and popping of improper vertices each iteration. Before each iteration,
any edge projected
to shorten to under the critical length is deleted by identifying its endpoints. The critical length is
calculated as Lc = 2t, where t is the time step or scale factor. Hence this should be used only with a fixed scale,
not optimizing scale factor. The critical length is chosen so that instabilities do not arise in motion by mean curvature.
If any edges are deleted, then vertices are examined for improper vertices as in the o command. Useful in string
model. Default is OFF.
Autopop is also implemented for small facets as of Evolver version 2.30. The critical area is calculated as Ac =
2t P/2, where the perimeter P is the sum of the lengths of the three sides of the facet.
See also the immediate_pop and autopop_quartic toggles.
AUTOPOP_QUARTIC
Modifiesthe autopop mode. The critical length for edges is set to Lc = 2 4 t and the critical area for facets is set
to Ac = 2 4 t P/2 where P is the facet perimeter; meant for quantities such as laplacian_mean_curvature where
velocity is proportional to fourth derivative of surface. Default is OFF.
AUTORECALC
Toggles automatic recalculation of the surface whenever adjustable parameters or energy quantity moduli are
changed. Default is ON.
BACKCULL
Prevents display of facets with normal away from viewer. May have different effects in different graphics displays.
For example, to see the inside back of a body only, "set frontcolor clear" alone works in 2D displays, but needs backcull
also for direct 3D. Default is OFF.
BEZIER_BASIS
When Evolver is using the Lagrange model for geometric elements, this toggle replaces the Lagrange interpolation
polynomials (which pass through the control points) with Bezier basis polynomials (which do not pass through interior
control points, but have positive values, which guarantees the edge or facet is within the convex hull of the control
points). Default is OFF>
BIG_ENDIAN
Controls the order of bytes in binary_printf numerical output. Big-endian is most significant byte first. To change to
little-endian, use little_endian, not little_endian off. Default is set to the native mode of the machine Evolver is running
on.
BLAS_FLAG
Toggles using BLAS versions of some matrix routines, if the Evolver program has been compiled with the -DBLAS
option and linked with some BLAS library. For developer use only at the moment. Default is OFF.
BOUNDARY_CURVATURE
When doing integrals of mean curvature or squared curvature, the curvature of a boundary vertex cannot be defined
by its neighbor vertices, so the area of the boundary vertex star instead is counted with an adjacent interior vertex.
Default is OFF.
BREAK_AFTER_WARNING
Causes Evolver to cease execution of commands and return to command prompt after any warning message.
The break does not happen until the executing command or subcommand completes; use break_on_warning for an
instantanous break. Same effect as command line option -y. Default is OFF.
BREAK_ON_WARNING
Causes Evolver to cease execution of commands and return to command prompt immediately after any warning
message. Does not delay until completion of current command as break_after_warning does. Default is OFF.
BUNCH_KAUFMAN
136
137
Toggles Bunch-Kaufman factoring in the alternative minimal degree factoring method (ysmp off). This factors the
Hessian as LBLT where L is lower triangular with ones on the diagonal, and B is block diagonal, with 1x1 or 2x2
blocks. Supposed to be more stable when factoring indefinite hessians. Default is OFF.
CALCULATE_IN_3D
The squared mean curvature named methods star_sq_mean_curvature, star_eff_area_sq_mean_curvature,

star_normal_mean_curvature, and star_perp_sq_mean_curvature work in any dimension space, but if for some reason
the space has an ambient dimension greater than 3, and you want to restrict the calculation of curvature to the first
three coordinates, the toggle calculate_in_3d will do that.
CHECK_INCREASE
Toggles checking for increase of energy in an iteration loop. If energy increases, then the iteration loop is halted.
Meant for early detection of instabilities and other problems causing the surface to misbehave. Useful in doing a
multiple iteration with a fixed scale. Caution: there are circumstances where an energy increase is appropriate, for
example when there are volume or quantity constraints and conforming to the constraints means an energy increase
initially. Default is OFF.
CIRCULAR_ARC_DRAW
If on, then in quadratic string mode, an edge is drawn as a circular arc (actually 16 subsegments) through the
endpoints and midpoint, instead of a quadratic spline. Default OFF.
CLIP_VIEW
Toggles showing a clipping plane in graphics. In the graphics window, the l key (lower case L) will make
mouse dragging translate the clipping plane, and the k key will make mouse dragging rotate the clipping plane. Also,
clipping plane coefficients may be set manually in the array clip_coeff[10][4], enabling up to 10 simultaneous clipping
planes. Each set of 4 clip coefficients c1,c2,c3,c4 determines a clip volume c1*x + c2*y + c3*z <= c4. Default OFF.
CLIPPED, CLIPPED_CELLS
Sets torus quotient space display to clip to fundamental region. Not an on-off toggle. 3-way toggle with
RAW_CELLS and CONNECTED. The origin (lower left back corner) of the clip parallelogram can be set by setting
entries in the display_origin vector. Default is unset, so Evolver prompts the user when graphics are first displayed. The
setting persists when loading a new surface. But loading a torus model when a non-torus model is currently displayed
will not prompt.
COLORMAP
Use colormap from file. Use colorfile := "filename" to set file. Default is OFF.
CONF_EDGE
Calculation of squared curvature by fitting sphere to edge and adjacent vertices (conformal curvature). Default is
OFF.
CONJ_GRAD
Use conjugate gradient method. See also RIBIERE. Default is OFF.

CONNECTED
Sets quotient space display to do each body as a connected, wrapped surface. Not an on-off toggle. 3-way toggle
with CLIPPED and RAW_CELLS. Synonym: connected_cells. Default is unset, so Evolver prompts the user when
graphics are first displayed. The setting persists when loading a new surface. But loading a torus model when a
non-torus model is currently displayed will not prompt.
Since slight motions during evolution can cause the wrap to change suddenly, there is a body boolean attribute
centerofmass that causes the center of mass of a body to be remembered, and the next time a body is plotted, the wrap
is adjusted so the center of mass is close to the previous center of mass. Example: set body centerofmass
DETORUS_STICKY
Controls whether the detorus command will identify coincident vertices, edges, and facets. The tolerance for
identification is given by the variable detorus_epsilon, which defaults to 1e-6. Default ON.
137
138
DEBUG
Print YACC debug trace of parsing of commands. Dont do this! Default OFF.
DETURCK
Motion by unit velocity along normal, instead of by curvature vector. Default OFF.
DIFFUSION
Activates diffusion step each iteration. Default OFF.

DIRICHLET_MODE
When the facet_area method is being used to calculate areas in hessian commands, this toggles using an approximate facet_area hessian that is positive definite. This permits hessian iteration to make big steps in a far-from-minimal
surface without fear of blowing up. However, since it is only an approximate hessian, final convergence to the minimum can be slow. Linear model only. Does convert_to_quantities implicitly. Another variant of this is triggered by
sobolev_mode. Default OFF.
DIV_NORMAL_CURVATURE
Toggle to make sq_mean_curvature energy calculate the mean curvature by the divergence of the normal vectors
at the vertices of a facet. Default OFF.

EFFECTIVE_AREA
In area normalization, the resistance factor to motion is taken to be only the projection of the vertex star area
perpendicular to the motion. If squared mean curvature is being calculated, this projected area is used in calculating
the curvature. Default OFF.
ESTIMATE
Activates estimation of energy decrease in each gradient descent step g command). For each g iteration, it prints
the estimated and actual change in energy. The estimate is computed by the inner product of energy gradient with
actual motion. Useful only for a fixed scale factor much less than optimizing, so linear approximation is good. The
internal variable estimated_change records the estimated value. Default OFF.
FACET_COLORS
Enables coloring of facets in certain graphics interfaces (e.g. xgraph). If off, facet color is white. Default is ON.
FORCE_DELETION
In the soapfilm model, overrides the refusal of the delete command to delete edges or facets when that would create
two edges with the same endpoints. Sometimes it is necessary to have such edges, for example in pinching off necks.
But usually it is a bad idea. Also see star_finagle. Default is OFF.
FORCE_EDGESWAP Toggle. Makes the u or equiangulate or edgeswap commands skip certain tests and do the swap
anyway. The skipped tests are: (1) the two vertices of the flipped edge are distinct, (2) creating two facets with the
same vertices. Meant only for rare cases when you really know what you are doing. You should not leave this toggle
in the ON state; turn it OFF after you have done the recalcitrant edge swap.
FORCE_POS_DEF
If this is on during YSMP factoring of Hessian and the Hessian turns up indefinite, something will be added to the
diagonal element to make it positive definite. Left over from some experiment probably. Default is OFF.
FULL_BOUNDING_BOX
Causes bounding box in PostScript output to be the full window, rather than the actual extent of the surface within
the window. Default is OFF.
FUNCTION_QUANTITY_SPARSE For named quantities defined as functions of named methods, this toggles the use
of sparse matrices in calculating hessians. Default is OFF.

GRAVITY
Includes gravitational energy in total energy. Default ON if there are bodies with density.
138
139
GV_BINARY
Toggles sending data to geomview in binary format, which is faster than ascii. Default is binary on SGI, ascii on
other systems, since there have been problems with binary format on some systems. Ascii is also useful for debugging.
Default OFF.
H_INVERSE_METRIC
Replaces force by Laplacian of force. For doing motion by Laplacian of mean curvature. Default OFF.
HESSIAN_DOUBLE_NORMAL
When hessian_normal is also on and the space dimension is even, then the normal vector components in the last
half of the dimensions are copies of those in the first half. Sounds weird, huh? But it is useful when using a string
model in 4D to calculate the stability of cylindrically symmetric surfaces. Default OFF.
HESSIAN_DIFF
Calculates Hessian by finite differences. Very slow. For debugging Hessian routines. Default OFF.
HESSIAN_NORMAL
Constrains hessian iteration to move each vertex perpendicular to the surface. This eliminates all the fiddly sideways movement of vertices that makes convergence difficult. Perpendicular is defined as the volume gradient, except
at triple junctions and such, which are left with full degrees of freedom. Default is ON.
If you wish to run in hessian_normal mode but exempt particular vertices from the restriction, you can set the
vertices no_hessian_normal attribute, for example
set vertex no_hessian_normal where z > 1.2
HESSIAN_NORMAL_ONE
If this and hessian_normal are on, then the normal at any point will be one-dimensional. This is meant for soap
films with Plateau borders, where there are triple junctions with tangent films. Ordinary hessian_normal permits lateral
movement of such triple junctions, but hessian_normal_one does not. Valid only for the string model in 2D and the
soapfilm model in 3D. The normal vector is computed as the eigenvector of the largest eigenvalue of the sum of the
normal projection matrices of all the edges or facets adjoining the vertex. Default OFF.
HESSIAN_NORMAL_PERP
If this is on, then the Hessian linear metric uses only the component of the normal perpendicular to the facet or
edge. This raises eigenvalues slightly. Default OFF.
HESSIAN_QUIET
Suppresses status messages during Hessian operations. Off is for debugging. Default ON.
HESSIAN_SPECIAL_NORMAL
When hessian_normal is on, this toggles using a special vectorfield for the direction of the perturbation, rather
than the usual surface normal. The vectorfield is specified in the hessian_special_normal_vector section of the datafile
header. Beware that hessian_special_normal also applies to the normal calculated by the vertexnormal attribute and
the normal used by regular vertex averaging. Default OFF.
HOMOTHETY
Adjust total volume of all bodies to fixed value after each iteration by uniformly scaling entire surface. Default
OFF.
IMMEDIATE_AUTOPOP
Modifies the autopop mode. Causes deletion of a short edge or small facet immediately upon detection, before
proceeding with further detection of small edges or facets. Original behavior was to do all detection before any
elimination, which could cause bad results if a lot of edges got short simultaneously. Default is OFF for backward
compatibility, but you should probably turn it on.
INTERP_BDRY_PARAM
139
140
For edges on parameterized boundaries, calculate the parameter values of new vertices (introduced by refining) by
interpolating parameter values, rather than extrapolating from one end. Useful only when parameters are not periodic.
Default OFF.
INTERP_NORMALS
Display using interpolated vertex normals for shading for those graphics interfaces that support it. Default OFF.
ITDEBUG
Prints some debugging information during a g step. For gurus only. Default OFF.
JIGGLE
Toggles jiggling on every iteration. Default OFF.

K_ALTITUDE_MODE Toggles the K command using the altitude from a vertex rather than the median to subdivide a
skinny triangle. Default OFF.

KRAYNIKPOPEDGE
Toggles edge-popping mode (O or pop commands) in which poppable edges look for adjacent facets of different
edge_pop_attribute values to split off from the original edge; failing that it reverts to the regular mode of popping the
edge. This is meant to give the user greater control on how edge popping is done. It is up to the user to declare the
edge_pop_attribute integer facet attribute and assign values. Default OFF.
KRAYNIKPOPVERTEX
Toggles 3D vertex popping mode in which a poppable vertex is examined to see if it is a special configuration of
six edges and 9 facets. If it is, a special pop is done that is much nicer than the default pop. Default OFF.
KUSNER
Calculation of squared curvature by edge formula rather than vertex formula. Default OFF.
LINEAR_METRIC
Eigenvalues and eigenvectors of the Hessian are defined with respect to a metric. This command toggles a metric
that imitates the smooth surface natural metric of L2 integration on the surface. Use with HESSIAN_NORMAL to get
eigenvalues and eigenvectors similar to those on smooth surfaces. Default OFF.
LITTLE_ENDIAN
Controls the order of bytes in binary_printf numerical output. Big-endian is most significant byte first. To change
to little-endian, use big_endian, not big_endian off. Default is the native mode of the machine Evolver is running on.
MEMDEBUG
Prints memory allocation/deallocation messages, and has c command print memory usage. For debugging. Default OFF.
METIS_FACTOR
Toggles experimental Hessian matrix factoring using the METIS library of Karypis and Kumar. Not in the public
distribution. Default OFF.
METRIC_CONVERT
If a Riemannian metric is defined, whether to use the metric to do gradient form to vector conversions. Synonym:
metric_conversion. Default OFF.
NO_DUMP
This is a per-variable toggle, or boolean attribute. When set, it prevents the value of a variable from being written
out by the "dump" or "d" commands. Useful with the replace_load and add_load commands when reloading dumps of
the current file and you want to preserve variable values that would otherwise be overwritten by loading the dump file,
which has variables declared in the top of the datafile by default. no_dump variables are instead written in the "read"
section of the dump file, so the dump file will load as a stand-alone file Works on global variables and arrays. Syntax
(run-time commands; not in top of datafile):
140
141
variable.no_dump on
variable.no_dump off
The no_dump declaration must come after the variable exists. Example:
frame := 1;
frame.no_dump;
dump "temp.dmp";
frame := 2;
NORMAL_CURVATURE
Calculation of squared curvature by taking area of vertex to be the component of the volume gradient parallel to
the mean curvature vector. Default OFF.
NORMAL_MOTION
Projects motion to surface normal, defined as the volume gradient. May be useful with squared curvature if vertices
tend to slither sideways into ugly patterns. Default OFF.
OLD_AREA
In the string model with area normalization, at a triple vertex Evolver normally tries to calculate the motion so that
Von Neumanns law will be obeyed, that is, the rate of area change is proportional to the number of sides of a cell. If
old_area is ON, then motion is calculated simply by dividing force by star area. Default OFF.
PINNING
Check for vertices that cant move because adjacent vertices are not on same constraint when they could be.
Obscure. Default OFF.
POP_DISJOIN
Changes the behavior of popping edges and vertices to act like merging Plateau borders, i.e. produce disjoined
films instead of films joined with cross-facets. In the edge case, if four facets meet along an edge and two opposite
bodies are the same body, then popping the edge will join the bodies if pop_disjoin is in effect. In the vertex case,
if the vertex has one body as an annulus around it, then the vertex will be separated into two vertices so the annulus
becomes a continuous disk. This is all done regardless of the angles at which facets meet. Applies to pop, o, and O
commands. Default OFF.
POP_ENJOIN
Changes the behavior of popping vertices in the soapfilm model so that when two distinct cones are detected
meeting at a common vertex, the popping result is a widening of the cone vertex into a neck rather than a disjoining of
the cones. meet. Applies to pop and o commands. Default OFF.
POP_TO_EDGE
The non-minimal cone over a triangular prism frame can pop in two ways. If this toggle is on, then popping to an
edge rather that a facet will be done. Default OFF.
POP_TO_FACE
The non-minimal cone over a triangular prism frame can pop in two ways. If this toggle is on, then popping to a
facet rather that an edge will be done. Default OFF.
PS_COLORFLAG
Makes the POSTSCRIPT command do CMYK color instead of RGB color in the file it creates.. Default OFF.
PS_COLORFLAG
Makes the POSTSCRIPT command do color. Default OFF.

PS_CROSSINGFLAG
Makes the POSTSCRIPT command put breaks in background edges in the string model. Default state is unset
(prints as -1).
141
142
PS_GRIDFLAG
Makes the POSTSCRIPT command draw all edges of displayed facets, not just those satisfying the current edgeshow condition. Default is OFF.
PS_LABELFLAG
Makes the POSTSCRIPT command print labels on elements. Useful for debugging a new datafile, but not when
there are too many elements. Default is OFF.
QUANTITIES_ONLY
Inactivates all energies except named quantities. Meant for programmers debugging use. Default OFF.
QUIET
Suppresses all normal output messages automatically generated by commands. Good while running scripts, or for
loading datafiles with long read sections. Explicit output from print, printf, and list commands will still appear, as
will prompts for user input. Applies to redirected output as well as console output. An error or user interrupting a
command will turn QUIET off, for sanity. Default OFF.
QUIETGO
Suppresses only iteration step output. Default OFF.

QUIETLOAD
Suppresses echoing of files being read in. This applies to the read section at the end of the datafile and any files
read in with the read command. This toggle does not get reset at the start of a new datafile. This toggle can be set with
the -Q command line option, to suppress echoing in the first datafile loaded. Default is OFF.
POST_PROJECT
Introduces extra projections to volume and fixed quantity constraints each iteration. If convergence fails after 10
iterations, you will get a warning message, repeated iterations will stop, and the variable iteration_counter will be
negative. Default OFF.
RAW_CELLS
Sets quotient space display for plain, unwrapped facets. Not an on-off toggle. 3-way toggle with CLIPPED and
CONNECTED. Default is unset, so Evolver prompts the user when graphics are first displayed. The setting persists
when loading a new surface. But loading a torus model when a non-torus model is currently displayed will not prompt.
RGB_COLORS
Toggles graphics to use user-specified red-green-blue components of color for elements rather than the color attribute indexing the pre-defined 16 color palette. The individual element rgb values are in element extra attributes:
ergb for edges, frgb for facets, and fbrgb for facet backcolor. It is up to the user to define these attributes; if they are
not defined, then they are not used and do not take up space. If frgb is defined but not fbrgb, then frgb is used for both
front and back color. The attributes are real of dimension 3 or 4; if 4 dimensional, the fourth component is passed
to the graphics system as the alpha value, but probably wont have any effect. The value range is 0 to 1. Be sure to
initialize the rgb attributes, or else you will get an all-black surface. The attribute definitions to use are:
define edge attribute ergb real[3]
define facet attribute frgb real[3]
define facet attribute fbrgb real[3]
Default is OFF.
RIBIERE
Makes the conjugate gradient method use the Polak-Ribiere version instead of Fletcher-Reeves. (The toggle
doesnt turn on conjugate gradient.) Polak-Ribiere seems to recover much better from stalling. Ribiere is the default mode.
ROTATE_LIGHTS
When ON, this makes the lights rotate with the object in the graphics display. Default OFF.
142
143
RUNGE_KUTTA
Use Runge-Kutta method in iteration step (fixed scale factor only). Default is OFF.
SELF_SIMILAR
If squared mean curvature energy is being used, this scales the velocity toward a self-similar motion. Applies
only when the old top-of-datafile squared_curvature declaration is used, or the sqcurve named method. The global
read-write variable self_sim_coeff is used as a multiplier. Default is OFF.
SEPTUM_FLAG Under some circumstances, popping a vertex can leave two regions connected by a tunnel; if the
septum_flag toggle is on, it will force a surface to be placed across the tunnel. Default is OFF.
SLICE_VIEW
Toggles showing a plane cross-section of the surface in graphics. In the graphics window, the l key (lower case
L) will make mouse dragging translate the slicing plane, and the k key will make mouse dragging rotate the slicing
plane. Also, slicing plane coefficients may be set manually in the array slice_coeff[4]. The set of 4 clip coefficients
c1,c2,c3,c4 determines a plane c1*x + c2*y + c3*z = c4. Default is OFF.
SMOOTH_GRAPH
In Lagrange model, causes edges and facets to be plotted with 8-fold subdivision rather than Lagrange order
subdivision. Default is OFF.
SHADING
Toggles facet shading in certain graphics interfaces (xgraph, psgraph). Darkness of facet depends on angle of
normal from vertical, simulating a light source above surface. Default is ON.
SHOW_ALL_EDGES
Evolver toggle command. Controls the showing of all edges in the graphics window, regardless of the current
"show edges ..." condition. Same as the e key in the graphics window.
SHOW_ALL_QUANTITIES
By default, only explicitly user-defined named quantities are displayed by the Q or v commands. If
show_all_quantities is on, then all internal named quantities (created by the -q option or by convert_to_quantities)
are also shown. Default is OFF.
SHOW_BOUNDING_BOX
Toggles display of the bounding box in the graphics display. Equivalent to hitting the o key in the graphics
window.
SHOW_INNER
Display interior facets, those on 2 bodies. Default is OFF.

SHOW_OUTER
Display outer facets, those on 0 or 1 body. Default is OFF.

SOBOLEV_MODE
When the facet_area method is being used to calculate areas in hessian commands, this toggles using an approximate facet_area hessian that is positive definite. This permits hessian iteration to make big steps in a far-from-minimal
surface without fear of blowing up. However, since it is only an approximate hessian, final convergence to the minimum can be slow. Linear model only. Does convert_to_quantities implicitly. Another variant of this is triggered by
dirichlet_mode. A detailed explanation is in the Technical Reference chapter. Default is OFF.
SPARSE_CONSTRAINTS
Toggles using sparse matrix techniques to accumulate and handle body and quantity gradients in iteration and
hessian commands. Now the default.
SQUARED_GRADIENT
Causes the hessian_seek command to minimize the square of the gradient of the energy rather than minimize the
energy itself. Useful for converging to unstable critical points. Default is OFF.
143
144
STAR_FINAGLE
In the soapfilm model, the delete command for edges or facets normally will not do the deletion if it would result
in the creation of two edges with the same endpoints. Some simple configurations that cause this are detected and
handled automatically, namely a "star" configuration in which there are three facets forming a triangle adjacent to the
edge being deleted. Such a star is automatically removed by deleting one of its internal edges before deleting the
original edge. But sometimes there are more complicated configurations that such unstarring wont handle, and then
Evolver will not delete the edge unless the force_deletion toggle is on. An alternative is to first refine the edges that
would have the common endpoints, and this is what the star_finagle toggle enables. Default is OFF.
THICKEN
Display thickened surface, for use when different sides of a facet are colored differently. Use thickness := expr
to set thickness. The default thickness is 0.001 times the maximum linear dimension of the surface. Default is OFF.
But looks kind of obsolete, since none of the graphics displays pay any attention to it, but take care of different colors
automatically. But the thickness variable is still used.
VERBOSE
Prints action messages from pop edge, pop vertex, delete, notch, refine, dissolve, edgeswap, and some other
commands. Meant for debugging, if some action does not have the desired effect. Default is OFF.
VIEW_TRANSFORMS_USE_UNIQUE_POINT
When view transforms are generated with transform_expr, Evolver weeds out duplicate transforms. By default,
"duplicate" means the same transform matrix, but there are circumstances where different transform matrices carry the
surface to the same spot. The view_transforms_use_unique_point toggle enables a mode whereby two transform matrices are deemed identical if they transform the point given by the vector view_transforms_unique_point[] to the same
image point. The standard use is to make view_transforms_unique_point[] a vertex on the surface being transformed,
for example
view_transforms_unique_point := vertex[5].__x;
view_transforms_use_unique_point on;
transform_expr "abababa";
The vector view_transforms_unique_point[] is pre-defined, so the use does not need to define it. Default is OFF.
VISIBILITY_TEST
Toggles an occluded-triangle test for graphics output that uses the Painters Algorithm to produce 2D output
(PostScript, Xwindows). This can greatly reduce the size of a PostScript file, but inspect the output since the implementation of the algorithm may have flaws. Default is OFF.
VOLGRADS_EVERY
Toggles recalculating volume constraint gradients every projection step during constraint enforcement. Good for
stiff problems. Default is OFF.
YSMP
Toggles between Yale Sparse Matrix Package routines for factoring hessians, and my own minimal degree factoring. Default is OFF.
ZENER_DRAG
Toggles Zener drag feature, in which the velocity of the surface is reduced by a magnitude given by the variable
zener_coeff, and the velocity is set to zero if it is smaller than zener_coeff. Default is OFF.
144
Chapter 11
General commands
11.1 Variable assignment
Values can be assigned to variables. The variable names must be two or more letters, in order they not be confused
with single-letter commands. Note that := is used for assignment, not =. Numeric, array, string, and command
values may be assigned.
Numeric assigment. This involves a single numeric value being assigned to a variable or array element or attribute.
Syntax:
variable := expr
arrayname[ expr]...[ expr] := expr
element.attribute := expr
quantityname.attribute := expr
variable.attribute := expr
See the sections on individual element types for which attributes are writable. The quantity attributes are target, modulus, tolerance, and volconst. The variable attributes are pdelta and pscale; these are meant for optimizing parameters.
The C-style arithmetic assignments +=, -=, *=, and /= work for numeric assignment. For example, val += 2 is
equivalent to val := val + 2.
Array assignment. This involves whole arrays or array slices. An array slice is a partially indexed array; for
example if mata has dimensions 4 by 5 by 6, then mata[2] is a 5 by 6 array consisting of all the elements of the
form mata[2][i][j]. There are two forms of array assignment: assigning a single value (a scalar) to every element, or
assigning corresponding elements. Syntax for scalar assignment:
arrayname := expr
arrayname[ expr]... := expr
element.arrayattribute := expr
element.arrayattribute[ expr] := expr
Examples of scalar assignment:
define mata integer[5][5];
mata := 0;
mata[3] := 1;
In array-to-array assignment, the sizes of the arrays on the two sides do not have to agree; Evolver will assign only
the elements that exist on both sides. But the number of dimensions has to agree on both sides, except when assigning
a single value to all the elements. Syntax:
145
146
arrayname := arrayexpr
arrayname[ expr]... := arrayexpr
element.arrayattribute := arrayexpr
element.arrayattribute[ expr] := arrayexpr
Examples:
define mata integer[5][5];
define matb real[5][5];
define matc real[3][4][6];
mata := matb;
mata[2] := matb[4];
mata := matc[2];
mata := 4*matb;
Array initialization syntax works for runtime assignments to arrays and array slices, including element attributes
that are arrays. The entries in the initializer must be single numbers, not arrays. The number of dimensions on the left
and right side of the assignment must agree, but the sizes in each dimension need not agree. Missing elements on the
right side are regarded as zero. Examples:
vertex[1].__x := { 1.2, 3.1, 4.7 }
define mat real[2][2];
mat := {{1,2},{3,4}}
The right side is evaluated each time the assignment is executed, so the entries on the right can be any expressions that
evaluate to numbers.
The C-style arithmetic assignments +=, -=, *=, and /= work for array assignment, but note that all work elementwise (i.e. *= and /= are not matrix multiplication and division).
String assignment. Currently, strings may be assigned only to string variables. String arrays and string attributes are
not supported. Syntax:
variablename := stringexpr
The stringexpr on the right side can be a quoted string, string variable, or sprintf. One cannot assign a string to a
variable that is already numeric, or vice versa, since string and numeric values are used in different places in the
language syntax.
Command assignment. New command words can be created by assigning a command to an identifier. Command
words can be re-defined new commands. Syntax:
identifier := command
This makes identifier an abbreviation for command. Subsequently, identifier may be used as a command itself. You
are strongly urged to enclose the command on the right side in braces so the parser can tell its a command and not an
expression. Also multiline commands then dont need linesplicing. Examples:
gg := { g 10; u };
gg 12;
Single letter assignment. Single-letter commands (including those not already in use) can be re-defined with the
triple-colon assignment sign :::=. The original meaning of the letter command can be accessed by putting the letter in
single quotes. I find this most useful in making the r command automatically carry out extra steps. Syntax:
letter :::= { commandlist }

Example:
r :::= { r; u; V; cutlength /= 2; }
146
147
Permanent assignment, If ::= is used instead of :=, then the command becomes permanent and will not be
forgotten when a new surface is begun. Such a command can only refer to permanent variables, permanent commands,
and internal variables. See the permload command for an example of use. This is not a recommended technique, since
keeping straight which commands use which variables gets tricky.
11.2 Element generators and SQL-type queries on sets of elements

Many commands operate on a set of elements defined by an element generator. Generators are also used in the
aggregate expressions sum, avg, max, min, and count. An element generator has the syntax
elementgen name where expr
elementgen may be
1. A multiple element generator, which can be
A. An element type, vertex, edge, facet, facetedge, or body, which generates all elements of that type in the
surface. But new elements created during a loop will not be generated, so "refine edges" will refine all
existing edges just once.
B. A single element subelement. The implemented subelements are: of a vertex: edge, facet (in no particular
order)
i. of an edge: vertex (in tail, head order), facet (in geometric order)
ii. of a facet: vertex, edge, body (all in order around the facet)
iii. of a body: facet (in no particular order)
2. A single element, which can be
A. an element name of an active generator
B. an indexed element type, vertex, edge, facet, or body. Indexing starts at 1. The index may be negative, in
which case the generated element has negative orientation.
C. an indexed subelement of an element (error if no element of that index). Indexing starts at 1. The indexing
is the same as the order produced by the foreach generator. Indexed subelements of an edge or facet follow
the orientation of the edge or facet.
"name" is an optional identifier which can be used in the body of a loop to refer to the generated element. expr is
interpreted as a boolean expression, 0 for false, nonzero for true. Only elements for which expr is true are generated.
The where expr clause is optional. The innermost generator generates a default element, which can have its attributes
referred to just by attribute name. But be sure to remember that in a nested iteration, an unqualified element type
generates all elements of that type, not just those associated with the parent element. Examples:
list facet where color == red
foreach edge ee where ee.length < .3 do list ee.vertex
print facet[2].edge[1].vertex[2].id
foreach facet ff do { printf "facet %g:\n"; list ff.edge }
print max(edge where on_constraint 1, length)
The commands that use generators are:

DELETE
DISSOLVE
EDGESWAP
EQUIANGLULATE
FIX
FOREACH
HISTOGRAM
LIST
LOGHISTOGRAM
POP
POP_TRI_TO_EDGE
POP_EDGE_TO_TRI
POP_QUAD_TO_QUAD
RAW_VERTEX_AVERAGE
RAWEST_VERTEX_AVERAGE
REFINE
147
SET
SHOW
SHOW_EXPR
T1_EDGESWAP
UNFIX
UNSET
VERTEX_AVERAGE
148
11.3 Array operations.

There are some basic whole-array operations that permit arrays on the left side of an assignment statement:
array := scalar
array := array
array := scalar * array
array := array + array
array := array - array
array := array * array
Here "array" on both sides of the assignment means a single whole array or array slice. An "array" on the right
side may itself be a compound array expression. "Scalar" can be any expression that evaluates to a single value.
For multiplication, the arrays must be two-dimensional with properly matching sizes; multiplication of two onedimensional arrays is the scalar or dot product. These operations also apply to element attributes that are arrays.
Array initialization syntax works for runtime assignments to arrays and array slices, including element attributes
that are arrays. The entries in the initializer must be single numbers, not arrays. The number of dimensions on the left
and right side of the assignment must agree, but the sizes in each dimension need not agree. Missing elements on the
right side are regarded as zero. Examples:
vertex[1].__x := { 1.2, 3.1, 4.7 }
define mat real[2][2];
mat := {{1,2},{3,4}}
The right side is evaluated each time the assignment is executed, so the entries on the right can be any expressions
that evaluate to numbers. Assignment with += and -= also work, as does *= and /=, but note that all of these work
element-wise (i.e. *= and /= are not matrix multiplication and division).
11.4 Keyword commands

This section contains an alphabetical list of commands base on keywords. Single letter commands, toggle commands,
and parameter-setting commands have their own sections elsewhere. Most of these commands are verbs; that is, a
command consists of the keyword possibly followed by arguments or other stuff. Some have the form of procedure
calls, since that seemed like a good idea whenever I wrote them. Some are function calls, which mean they return
a value; they are listed here because their main purpose is some action and the return value is a success code or
something else that it is useful for the calling routine to know.
ABORT
Causes immediate termination of the executing command and returns to the command prompt. Meant for stopping
execution of a command when an error condition is found. There will be an error message output, giving the file and
line number where the abort occurred, but it is still wise to have a script or procedure or function print an error message
using errprintf before doing the abort command, so the user knows why.
ADDLOAD string
Loads a new datafile without deleting the current surface, permitting the simultaneous loading of multiple copies
of the same datafile or different datafiles. String is either a sting literal in double quotes, or a string variable name such
as datafilename. Elements in the new datafile are re-numbered to not conflict with existing elements. This is actually
the same as the default behavior of Evolver when loading a single datafile. Thus the -i command line option or the
keep_originals keyword is not obeyed for the new datafile. The read section of the new datafile is not executed; this
permits a datafile to use the addload command in its read section to load more copies of itself. The loading script is
148
149
responsible for all initialization that would ordinarily be done in the read section of the new datafile. Declarations
in the top section of the new datafile will overwrite any existing declarations. This is usually not a problem when
loading multiple copies of the same datafile, but requires attention when loading different datafiles. For example,
numbered constraints are a bad idea; use named constraints instead. For variables you dont want written, you can set
the no_dump property of a variable to prevent it from being dumped in the top of the datafile; it will be dumped in the
bottom section instead. Example (as commands, not in the top of the datafile):
htvar := 2.3
htvar.no_dump on
See the sample datafile addload_example.fe for an example of how to load and distinguish between multiple copies
of the same surface.
ALICE expr
A special private command.

AREAWEED expr
Weeds out facets smaller than given area. Same as w command, except does not need interactive response.
BINARY_OFF_FILE
Produces one frame file for my 3D movie program evmovie (see http://www.susqu.edu/brakke/evmovie/evmoviedoc.html for more details). Syntax:
binary_off_file string
where string is the name of the output file, either a double-quoted string, a string variable, or a string-generating
expression (typically using sprintf).
BINARY_PRINTF
For printing formatted binary output to files. Syntax:

BINARY_PRINTF string, expr, expr,...
Prints to standard output using a binary interpretation of the standard C formats:

non-format character are copied verbatim as single bytes.
The byte order for numbers can be set with the big_endian or little_endian toggles. NOTE: Either big_endian or
little_endian must be set for binary_printf to work! The format string can be a string variable or a quoted string. There
is a limit of 1000 characters on the format string, otherwise there is no limit on the number of arguments. Meant to
be use with redirection to a file. In Microsoft Windows, the output file type is temporarily changed from TEXT to
BINARY so newline bytes dont get converted. Example:
binary_printf "%ld%ld%ld",vertex_count,edge_count,facet_count >> "out.bin"
BODY_METIS expr
Partitions the set of bodies into expr parts using the METIS library of Karypis and Kumar, (http://wwwusers.cs.umn.edu/karypis/metis/) if this library has been compiled into the Evolver. Meant for experiments in partitioning the surface for multiprocessors. The partition number of a body is left in the body extra attribute bpart, which
is created if it does not already exist. BODY_METIS uses the PMETIS algorithm.
BREAKPOINT string expr
The user may set a breakpoint in an already loaded script with the "set breakpoint" command. The syntax is
BREAKPOINT scriptname linenumber
149
150
where scriptname is the name of the function or procedure and linenumber is the line number in the file
where the breakpoint is to be set. There must be executable code on the line, or you will get an error. linenumber may
be an expression.
Breakpoints may be unset individually with
UNSET BREAKPOINT scriptname linenumber
or as a group with
UNSET BREAKPOINTS
When a breakpoint is reached, Evolver will enter into a subcommand prompt, at which the user may enter any Evolver
commands (although some commands, such as load would be very unwise). To exit from the subcommand prompt,
use q or exit or quit.
BURCHARD expr
A special private command.

CHDIR stringexpr
Changes the current directory, used for searching for files before EVOLVERPATH is used. In MS-Windows,
use a front slash / or a double backslash \\ instead of a single backslash as the path character. Example: chdir
"/usr/smith/project"
CHECK
Runs internal surface consistency checks; the same as the C command, but with no message if there are no errors.
The number of errors is recorded in the variable check_count.
CLOSE_SHOW
Closes the native graphics window started by the s or SHOW commands. Does not affect geomview or OpenGL
version. Synonym: show_off.
CONVERT_TO_QUANTITIES
This will do a conversion of old-style energies to new-style named quantities. This has the same effect as the
-q command line option, but can be done from the Evolver command prompt. If everything has been converted to
quantities one way or another, the internal variable everything_quantities is set to 1. Convert_to_quantities cannot be
undone. Useful when hessian complains about not being able to do a type of energy. It is also useful when setting
up a datafile, since the Q command will show all the internal quantities individually (when the show_all_quantities
toggle is on), so you can tell what each constraint integral is and so forth. A few energies dont convert yet. It is my
intention that this will be the default sometime in the near future, if it can be made sufficiently fast and reliable.
DEFINE
The DEFINE command may be used to define element attributes, arrays, array attributes, or single variables. The
same definition syntax works in the top of the datafile or as a runtime command. Note that array indexing starts at 1,
not 0. Zero is a legal dimension for an array, if you want an empty array, say as a syntax placeholder. Array sizes may
be redefined dynamically with preservation of data as possible, but the number of dimensions must remain the same.
There is runtime checking on array bounds.
It is possible to define multidimensional arrays of integers or reals with the syntax
DEFINE variablename REAL|INTEGER [ expr]...
150
151
This syntax works both in the datafile header and at the command prompt. If the array already exists, it will be resized,
with old elements kept as far as possible. Do not resize with a different number of dimensions. Example:
define fvalues integer[10][4]
define basecoord real[10][space_dimension]
print fvalues
print fvalues[4]
It is possible to dynamically define extra attributes for elements, which may be single values or up to eightdimensional arrays. Array element attributes may be defined with the syntax
DEFINE elementtype ATTRIBUTE name REAL|INTEGER [ [ dim] ... ]
where elementtype is vertex, edge, facet, or body, name is an identifier of your choice, and [dim] is an optional
expression for the dimension (with the brackets). There is no practical distinction between real and integer types at the
moment, since everything is stored internally as reals. But there may be more datatypes added in the future. It is not
an error to redefine an attribute that already exists, as long as the definition is the same. Extra attributes are inherited
by elements of the same type generated by subdivision. Examples:
define edge attribute charlie real
define vertex attribute newx real[3]
define facet attribute knots real[5][5][5]
The value of an extra attribute can also be calculated by user-supplied code. The attribute definition is followed by
the keyword "function" and then the code in brackets. In the code, the keyword "self" is used to refer to the element
the attribute is being calculated for. Example: To implement the lowest z value of a facet as an attribute:
define facet attribute minz real function
{ self.minz := min(self.vertex,z);}
These attributes can also be indexed. Due to current parser limitations on parsing executable code, this type of extra
attribute definition cannot occur in the top section of the datafile, although the non-function version can to declare the
attribute name, and the function part added in a re-definition in the READ section of the datafile.
It is also possible to use DEFINE to declare a variable without giving it an initial value. This is primarily a
mechanism to be sure a variable is defined before use, without overwriting an existing value. If the variable is new, the
initial value is zero or the null string. Syntax:
DEFINE name REAL|INTEGER|STRING
The DEFINE command may also be used to make runtime definitions of level-set constraints, boundaries, named
quantities, and method instances. The syntax is the same as in the top of the datafile, except that the word "define"
comes first. Multi-line definitions should be enclosed in brackets and terminated with a semicolon. Or they can be
enclosed in quotes and fed to the exec command. Of course, using exec means the parser doesnt know about the
define until the exec is executed, so you cannot use the defined item in commands until then.
define constraint floorcon formula z = 0
define constraint frontcon formula x = 1 energy: e1: -y/2 e2: x/2
exec "define boundary newboundary parameters 1
x: sin(p1)
y: cos(p1)
z: 3"
exec "define quantity qarea info_only method facet_area global"
151
e3: 0;
152
DELETE generator
Deletes edges by shrinking the edge to zero length (as in the tiny edge weed command t) and facets by eliminating
one edge of the facet. Facet edges will be tried for elimination in shortest to longest order. Edges will not be deleted if
both endpoints are fixed, or both endpoints have different constraints or boundaries from the edge. Delete will also fail
if it would create two edges with the same endpoints, unless the force_deletetion toggle is on; also see star_finagle.
Example:
delete facets where area < 0.0001
DELETE_TEXT (text_id)
Command to delete a text string from the graphics display. Syntax:

delete_text(text_id)
where text_id is the value returned by the call to display_text that created the string.
DETORUS
Main prompt command that converts the displayed surface to a real surface. It is meant for situations like the torus
model where one wants to unwrap the torus in reality in order to write an export file for some other program. After
detorus, the torus model is not in effect. Detorus also works with view transforms. Beware that after detorus vertices
and edges may be removed from boundaries and constraints, so considerable patching up may be necessary to get an
evolvable surface. There is a toggle detorus_sticky that makes detorus merge vertices, edges, and facets that coincide;
it defaults to on. The tolerance for identifying vertices is given by the variable detorus_epsilon, which defaults to 1e-6.
DIRICHLET
Does one iteration of minimizing the Dirichlet integral of the surface. The current surface is the domain, and the
Dirichlet integral of the map from the current surface to the next. This is according to a scheme of Konrad Polthier
and Ulrich Pinkall [PP]. At minimum Dirichlet integral, the area is minimized also. Works only on area with fixed
boundary; no volume constraints or anything else. Seems to converge very slowly near minimum, so not a substitute
for other iteration methods. But if you have just a simple soap film far, far from the minimum, then this method
can make a big first step. DIRICHLET_SEEK will do an energy-minimizing search in the direction. See section in
Technical Reference chapter.
DISPLAY_TEXT (x,y,string)
Causes the display of simple text on the graphics display. Currently implemented for OpenGL and PostScript
graphics. Syntax:
text_id := display_text(x,y,height,string)
The x,y coordinates of the start of the string are in window units, i.e. the window coordinates run from (0,0) in the
lower left to (1,1) in the upper right. height is the font height in window units, so 0.04 is a reasonable value. The
return value should be saved in a variable in case you want to delete the text later; even if you dont want to delete it
with delete_text, you must have something on the left of the assignment for syntax purposes. No font size control or
font type or color implemented. Meant for captioning images, for example a timer in frames of a movie.
DISSOLVE generator
Removes elements from the surface without closing the gap left. The effect is the same as if the line for the element
were erased from a datafile. Hence no element can be dissolved that is used by a higher dimensional element. (There
are two exceptions: dissolving an edge on a facet in the string model, and dissolving a facet on a body in the soapfilm
model.) Thus "dissolve edges; dissolve vertices" is safe because only unused edges and vertices will be dissolved. No
152
153
error messages are generated by doing this. Good for poking holes in a surface. The number of elements dissolved is
available afterwards in the read-only variable dissolve_count. Example:
dissolve facets where original == 2
DUMP filename
Dumps current surface to named file in datafile format. The filename can be a quoted string or a string variable or
expression. With no filename, dumps to the default dump file, which is the datafile name with .dmp extension. See
no_dump for suppressing dumping of particular variables.
DUMP_MEMLIST
Lists the currently allocated memory blocks. For my own use in debugging memory problems.
EDGE_MERGE (id1,id2)
Merges two edges into one in a side-by-side fashion. Meant for joining together surfaces that bump into each other.
Should not be used on edges already connected by a facet, but merging edges that already have a common endpoint(s)
is fine. Syntax:
edge_merge(integer,integer)
Note the arguments are signed integer ids for the elements, not element generators. The tails of the edges are merged,
and so are the heads. Orientation is important. Example:
edge_merge(3,-12)
EDGESWAP edge_generator
If any of the qualifying edges are diagonals of quadrilaterals, they are flipped in the same way as in equiangulation, regardless of whether equiangularity is improved. edgeswap edge will try to swap all edges, and is not
recommended, unless you like weird things.
edgeswap edge[22]
edgeswap edge where color == red
EDGEWEED expr
Weeds out edges shorter than given value. Same as t command, except does not need interactive response.
EIGENPROBE expr
EIGENPROBE(expr,expr)
Prints the number of eigenvalues of the energy Hessian that are less than, equal to, and greater than the given value.
It is OK to use an exact eigenvalue (like 0, often) for the value. Useful for probing stability. Second form will further
do inverse power iteration to find an eigenvector. The second argument is the limit on the number of iterations. The
eigenvalue will be stored in the last_eigenvalue variable, and the eigenvector can be used by the move command. The
direction of the eigenvector is chosen to be downhill in energy, if the energy gradient is nonzero.
ERRPRINTF string, expr, expr,...
Same as printf, except it sends its output to stderr instead of stdout. Useful in reporting error messages in scripts
that have their output redirected to a file.

EPRINT
153
154
Function that prints an expression and returns the value. Syntax:

eprint expr
Meant for debugging; probably an archaic leftover from when the command language was not as developed.
Example:
print sum(facet, eprint area)
will print out all the facet areas and then the sum.
EQUIANGULATE edge_generator
This command tests the given edges to see if flipping them would improve equiangularity. It is the u command
applied to a specified set of edges. It differs from the edgeswap command in that only edges that pass the test are
flipped.
EXPRINT commandname
Prints the original text of a user-defined command.

EXEC
Executes a command in string form. Good for runtime generation of commands. Syntax:
exec stringexpr
Example:
exec sprintf "define vertex attribute prop%d real",propnumber
FACET_CROSSCUT (facet_id,tail_id,head_id)
Function. In the string model, it subdivides a facet by constructing a diagonal edge between given vertices. The
return value is the id number of the new edge. The new facet will be on the same body as the old facet. Syntax:
newedge := facet_crosscut(facet_id,tail_id,head_id)
FACET_MERGE (id1,id2)
Merges two soapfilm-model facets into one in a side-by-side fashion. Meant for joining together surfaces that
bump into each other. The pairs of vertices to be merged are selected in a way to minimize the distance between
merged pairs subject to the orientations given, so there are three choices the algorithm has to choose from. It is legal
to merge facets that already have some endpoints or edges merged. Syntax:
facet_merge(integer,integer)
Note the syntax is a function taking signed integer facet id arguments, not element generators. IMPORTANT: The
frontbody of the first facet should be equal to the backbody of the second (this includes having no body); this is the
body that will be squeezed out when the facets are merged. If this is not true, then facet_merge will try flipping the
facets orientations until it finds a legal match. Example:
facet_merge(3,-12)
FIX generator [ name ] [ WHERE expr ]
Sets the FIXED attribute for qualifying elements. Example:

fix vertices where on_constraint 2
FIX variable
Converts the variable to a non-optimizing parameter.

154
155
FIX named quantity
Converts an info_only named quantity to a fixed quantity constraint.

FLUSH_COUNTS
Prints of various internal counters that have become nonzero. The counters are:
equi_count
edge_refine_count
body_dissolve_count
facet_pop_count
fix_count
notch_count
edge_delete_count
facet_refine_count
edge_dissolve_count
vertex_pop_count
where_count
unfix_count
facet_delete_count
notch_count
edge_pop_count
edgeswap_count
t1_edgeswap_count
Normally, these counts are accumulated during the execution of a command and printed at the end of the command.
Flush_counts can be used to display them at some point within a command. Flush_counts is usually followed by
reset_counts, which resets all these counters to 0.
FOREACH generator DO command
Repeat a command for each element generated by the generator. Examples:

foreach vertex do print x^2 + y^2 + z^2;
foreach edge ee where ee.dihedral > .4 do
{ printf "id %g\\n",id;
foreach ee.vertex do printf " %g %g %g\\n",x,y,z;
};
FREE_DISCARDS
Frees deleted elements in internal storage. Ordinarily, deleting elements does not free their memory for re-use
until the command completes, so that element iteration loops do not get disrupted. If for some reason this behavior
leads to excess memory usage or some other problem, the user may use the free_discards command to free element
storage of deleted elements. Just be sure not to do this inside any element iteration loop that might be affected.
HELP
Main prompt command. Prints what Evolver knows about an identifier or keyword. User-defined variables, named
quantities, named methods, named constraints, and element attributes are identified as such. Information for syntax
keywords comes from a file evhelp.txtt in the doc subdirectory of your Evolver installation, so that subdirectory should
be on your EVOLVERPATH environment variable. Syntax:
help string
The keyword need not be in quotes, unless there are embedded blanks. After printing the help section exactly
matching the keyword, a list of related terms is printed. These are just the keywords containing your keyword as a
substring.
HESSIAN
Attempt to minimize energy in one step using Newtons method with the Hessian matrix of the energy. See the
Hessian section of the Model chapter, and the Hessian section of the Technical Reference chapter.
HESSIAN_MENU
155
156
Brings up a menu of experimental stuff involving the Hessian matrix. Not all of it works well, and may disappear
in future versions. A one-line prompt with options appears. Use option ? to get a fuller description of the choices.
A quick summary of the current options:
1. Fill in hessian matrix.
Allocation and calculation of Hessian matrix.

2. Fill in right side. (Do 1 first)
Calculates gradient and constraint values.

3. Solve. (Do 2 first)
Solves system for a motion direction.

4. Move. (Do 3, A, B, C, E, K, or L first)
Having a motion direction, this will move some stepsize in that direction. Will prompt for stepsize.
7. Restore original coordinates.
Will undo any moves. So you can move without fear.

9. Toggle debugging. (Dont do this!)
Prints Hessian matrix and right side as they are calculated in other options. Produces copious output, and is meant
for development only. Do NOT try this option.
B. Chebyshev (For Hessian solution ).
Chebyshev iteration to solve system. This option takes care of its own initialization, so you dont have to do steps
1 and 2 first. Not too useful.
C. Chebyshev (For most negative eigenvalue eigenvector).
Chebyshev iteration to find most negative eigenvalue and eigenvector. Will ask for number of iterations, and will
prompt for further iterations. End by just saying 0 iterations. Prints Rayleigh quotient every 50 iterations. After
finding an eigenpair, gives you the chance to find next lowest. Last eigenvector found becomes motion for step 4. Last
eigenvalue is stored in the last_eigenvalue variable. Self initializing. Not too useful.
E. Lowest eigenvalue. (By factoring. Do 1 first)
Uses factoring to probe the inertia of H I until it has the lowest eigenvalue located within .01. Then uses inverse
iteration to find eigenpair. The eigenvalue is stored in the last_eigenvalue variable.
F. Lowest eigenvalue. (By conjugate gradient. Do 1 first)
Uses conjugate gradient to minimize the Rayleigh quotient. The eigenvalue is stored in the last_eigenvalue variable.
L. Lanczos. (Finds eigenvalues near probe value. )
Uses Lanczos method to solve for 15 eigenvalues near the probe value left over from menu choices P or V.
These are approximate eigenvalues, but the first one is usually very accurate. Do not trust apparent multiplicities.
From the main command prompt, you can use the lanczos expr command. The first eigenvalue printed is stored in the
last_eigenvalue variable.
R. Lanczos with selective reorthogonalization.
Same as L, but a little more elaborate to cut down on spurious multiplicities by saving some vectors to reorthogonalize the Lanczos vectors. Not quite the same as the official selective reorthogonalization found in textbooks. Last
eigenvalue is stored in the last_eigenvalue variable.
Z. Ritz subspace iteration for eigenvalues. (Do 1 first)
Same as ritz main command. Will prompt for parameters. The first eigenvalue printed is stored in the
last_eigenvalue variable.
X. Pick Ritz vector for motion. (Do Z first)
For moving by the various eigenvectors found in option Z. Particularly useful for multiple eigenvalues.
P. Eigenvalue probe. (By factoring. Do 1 first)
Reports inertia of H I for user-supplied values of . The Hessian H includes the effects of constraints. Will
prompt repeatedly for . Null response exits. From the main command prompt, you can use the eigenprobe expr
command.
S. Seek along direction. (Do 3, A, B, E, C, K, or L first)
Can do this instead of option 4 if you want Evolver to seek to lowest energy in an already found direction of
motion. Uses the same line search algorithm as the optimizing g command.
156
157
Y. Toggle YSMP/alternate minimal degree factoring.
Default Hessian factoring is by Yale Sparse Matrix Package. The alternate is a minimal degree factoring routine
of my own devising that is a little more aware of the surface structure, and maybe more efficient. If YSMP gives
problems, like running out of storage, try the alternate. This option is available at the main prompt as the ysmp toggle.
U. Toggle Bunch-Kaufman version of min deg.
YSMP is designed for positive definite matrices, since it doesnt do any pivoting or anything. The alternate minimal
degree factoring method, though, has the option of handling negative diagonal elements in a special way. This option
is available at the main prompt as the bunch_kaufman toggle.
M. Toggle projecting to global constraints in move.
Toggles projecting to global constraints, such as volume constraints. Default is ON. Dont mess with this. Actually,
I dont remember why I put it in.
G. Toggle minimizing square gradient in seek.
For converging to unstable critical points. When this is on, option S will minimize the square of the energy
gradient rather than minimizing the energy. Also the regular saddle and hessian_seek commands will minimize
square gradient instead of energy.
=. Start an Evolver command subshell. Useful for making PostScript files of eigenmodes, for example, without
permanenently changing the surface. To exit the subshell, just do q or "quit".k 0. Exit hessian.
Exits the menu. q also works.
HESSIAN_SEEK maxscale
Seeks to minimize energy along the direction found by Newtons method using the Hessian. maxscale is an
optional upper bound for the distance to seek. The default maxscale is 1, which corresponds to a plain hessian step.
The seek will look both ways along the direction, and will test down to 1e-6 of the maxscale before giving up and
returning a scale of 0. This command is meant to be used when the surface is far enough away from equilibrium that
the plain hessian command is unreliable, as hessian_seek guarantees an energy decrease, if it moves at all.
HISTOGRAM(element [ WHERE expr1 ], expr2 )
LOGHISTOGRAM(element [ WHERE expr1 ], expr2 )
Prints a histogram of the values of expr2 for the qualifying elements. The syntax is the same as the aggregate
expressions. It uses 20 bins, and finds its own maximum and minimum values, so the user does not have to specify
binsize. The log version will lump all zero and negative values into one bin. Example:
histogram(edge,dihedral*180/pi)
HISTORY
Print the saved history list of commands.

IS_DEFINED
To find out if a name is already in use as a keyword or user-defined name, use the is_defined function, which has
the syntax
is_defined(stringexpr)
The stringexpr must be a quoted string or other string expression. The return value is 0 if the name is undefined, 1
if defined. This function is evaluated at parse-time, not run-time, so a command like if is_defined(newvar) then newvar
:= 1 else newvar := 0 will give newvar the value 0 if this is its first appearance.
KEYLOGFILE stringexpr
Starts recording all incoming keystrokes to the file specified by stringexpr, which must be a quoted string or a
string variable or expression. Appends to an existing file. To end logging, use keylogfile off. For logging input and
output, use logfile instead.
157
158
LAGRANGE expr
Converts to the Lagrange model of the given order. Note that lagrange 1 gives the Lagrange model of order 1,
which has a different internal representation than the linear model. Likewise, lagrange 2 does not give the quadratic
model.
LANCZOS expr
LANCZOS (expr,expr)
Does a little Lanczos algorithm and reports the nearest approximate eigenvalues to the given probe value. In the
first form, expr is the probe value, and 15 eigenvalues are found. In the second form, the first argument is the probe
value, the second is the number of eigenvalues desired. The output begins with the number of eigenvalues less than,
equal to, and greater than the probe value. Then come the eigenvalues in distance order from the probe. Not real
polished yet. Beware that multiplicities reported can be inaccurate. The eigenvalue nearest the probe value is usually
very accurate, but others can be misleading due to incomplete convergence. Since the algorithm starts with a random
vector, running it twice can give an idea of its accuracy. The first eigenvalue printed is stored in the last_eigenvalue
variable.
LINEAR
Converts to the LINEAR model. Removes midpoints from quadratic model, and interior vertices from the Lagrange
model.
LIST
List elements on the screen in the same format as in the datafile, Piping to more can be used for long displays.
Syntax:
LIST generator
Examples:
list vertices | "more";
list vertices where x < 1 and y > 2 and z >= 3;
list edges where id == 12;
The LIST command is also used to print out individual constraint, boundary, quantity, or method instance definitions,
and it can be used to list all the element attributes that internally use the extra attribute mechanism for storage. Syntax:
LIST constraintname
LIST CONSTRAINT constraintnumber
LIST boundaryname
LIST BOUNDARY boundarynumber
LIST quantityname
LIST instancename
LIST ATTRIBUTES
Examples:
list constraint 1;
list floorcon;
list cmquant;
list attributes;
LIST ATTRIBUTES
158
159
Prints a list of the "extra attributes" of each type of element. Besides user-defined extra attributes, this list also
contains the predefined attributes that make use of the extra attribute mechanism (being of variable size), such as coordinates, parameters, forces, and velocities. It does not list permanent, fixed-size attributes such as color or fixedness,
or possible attributes that are not used at all.
LIST BOTTOMINFO
Prints what would be dumped in the read section at the end of a dumpfile: command definitions and various
toggle states.
LIST PROCEDURES
Prints the names of user-defined commands.

LIST TOPINFO
Prints the first section of the datafile on the screen. This is everything before the vertices section.
LOAD filename
Terminates the current surface and loads a new datafile. The filename is the datafile name, and can be either a
quoted string or a string variable. Since the automatic re-initialization makes Evolver forget all commands (including
whatever command is doing the load), this command is not useful except as the last action of a command. For loading
a new surface and continuing with the current command, see replace_load. Wildcard matching is in effect on some
systems (Windows, linux, maybe others), but be very careful when using wildcards since there can be unexpected
matches.
LOGFILE stringexpr
Starts recording all input and output to the file specified by stringexpr, which must be a quoted string or a string
variable or expression. Appends to an existing file. To end logging, use logfile off. For logging just incoming
keystrokes, use keylogfile instead.
LOGHISTOGRAM(element [ WHERE expr1 ], expr2 )
Prints a histogram of the values of expr2 for the qualifying elements, with bins in equal ratios. It will lump all zero
and negative values into one bin. The syntax is the same as the aggregate expressions. It uses 20 bins, and finds its
own maximum and minimum values, so the user does not have to specify binsize.
Example:
loghistogram(facet,area)
LONGJ
Long jiggle once. This provides long wavelength perturbations that can test a surface for stability. The parameters
are a wavevector, a phase, and a vector amplitude. The user will be prompted for values. Numbers for vectors should
be entered separated by blanks, not commas. An empty reply will accept the defaults. A reply of r will generate
random values. Any other will exit the command without doing a jiggle. In the random cases, a random amplitude A
and a random wavelength L are chosen from a sphere whose radius is the size of the object. The wavelength is inverted
to a wavevector w. A random phase is picked. Then each vertex v is moved by A sin(v w + ). More control over
perturbations may be had with the "set vertex coord ..." type of command.
MATRIX_DETERMINANT(matrix)
Main prompt function. For computing the determinant of a square matrix. Syntax:
retval := MATRIX_DETERMINANT( matrix)
159
160
Here matrix is the name of the square matrix, which could be a stand-alone matrix or a element array extra attribute, or
even a square slice of higher-dimensional matrix. The return value is 1 if the inversion is successful and 0 if matrix1
is singular. Examples:
define mata real[5][5]
// fill in mata
...
det := matrix_determinant(mata)
MATRIX_INVERSE(expr,expr)
Main prompt function. For computing the inverse of a square matrix. Currently applies only to global matrices,
not element attribute matrices. Syntax:
retval := MATRIX_INVERSE(matrix1, matrix2)
Here matrix1 is the name of the original matrix, and matrix2 is the name of the inverse matrix. They may be the
same matrix to get an in-place inverse. The return value is 1 if the inversion is successful and 0 if matrix1 is singular.
Examples:
define mata real[5][5]
define matb real[5][5]
... // fill in values of mata
retval := matrix_inverse(mata,matb)
retval := matrix_inverse(mata,mata)
METIS expr
KMETIS expr
Partitions the set of facets (edges in string model) into expr parts using the METIS library of Karypis and Kumar,
(http://www-users.cs.umn.edu/karypis/metis/) if this library has been compiled into the Evolver. Meant for experiments in partitioning the surface for multiprocessors. The partition number of a facet is left in extra attribute fpart (edge
epart for string model), which is created if it does not already exist. METIS uses the PMETIS algorithm, KMETIS
uses the KMETIS algorithm.
MOVE expr
Moves the surface along the previous direction of motion by the stepsize given by expr. The previous direction
can be either from a gradient step or a hessian step (hessian, saddle, hessian_seek, etc.). The stepsize does not affect
the current scale factor. A negative step is not a perfect undo, since it cannot undo projections to constraints. Move
sometimes does not work well with optimizing parameters and hessian together.
NEW_VERTEX ( expr, expr, ... )
For creating a new vertex. The syntax is that of a function instead of a verb, since it returns the id number of the
new vertex. The arguments are the coordinates of the vertex. The new vertex is not connected to anything else; use the
new_edge command to connect it. Usage:
newid := NEW_VERTEX( expr, expr, ... )
Examples:
newid := new_vertex(0,0,1)
newid := new_vertex(pi/2,0,max(vertex,x))
NEW_EDGE ( expr, expr )
160
161
For creating a new edge. The syntax is that of a function instead of a verb, since it returns the id number of the
new edge. The arguments are the ids of the tail and head vertices. Usage:
newid := NEW_EDGE( expr, expr )
The new edge has the same default properties as if it had been created in the datafile with no attributes, so you will
need to explicitly add any attributes you want. Example to create a set of coordinate axes in 3D:
newv1 := new_vertex(0,0,0); fix vertex[newv1];
newe1 := new_edge(newv1,newv2); fix edge[newe1];
set edge[newe1] no_refine; set edge[newe1] bare;
NEW_FACET ( expr, expr, ... )
For creating a new facet. The syntax is that of a function instead of a verb, since it returns the id number of the
new edge. The arguments are the oriented ids of the edges around the boundary of the facet, in the same manner that
a face is defined in the datafile. The number of edges is arbitrary, and they need not form a closed loop in the string
model. In the soapfilm model, if more than three edges are given, the new face will be triangulated by insertion of a
central vertex. In that case, the returned value will be the original attribute of the new facets. In the simplex model,
the arguments are the ids of the facet vertices. Usage:
newid := NEW_FACET(expr, expr, ... )
The new facet has the same default properties as if it had been created in the datafile with no attributes, so you will
need to explicitly add any attributes you want. Example:
newf := new_facet(1,2,-3,-4); fix facet where original == newf;
NEW_BODY
For creating a new body. The syntax is that of a function instead of a verb, since it returns the id number of the
new body. There are no arguments. Usage:
newid := NEW_BODY
The body is created with no facets. Use the set facet frontbody and set facet backbody commands to install the bodys
facets. The new body has the same default properties as if it had been created in the datafile with no attributes, so you
will need to explicitly add any attributes you want, such as density or target volume. Example:
newb := new_body
set facet frontbody newb where color == red
NOTCH expr
Notches ridges and valleys with dihedral angle greater than given value. Same as n command.
OMETIS expr
Computes an ordering for Hessian factoring using the METIS library of Karypis and Kumar,
(http://www-users.cs.umn.edu/karypis/metis/) if this library has been compiled into the Evolver (which it isnt in
the public distribution, yet). The optional expr is the smallest partition size, which defaults to 100. To actually use
METIS ordering during factoring, use the toggle metis_factor.
161
162
OPTIMIZE expr
Set g iteration mode to optimizing with upper bound of expr on the scale factor. Synonym: optimise.
PAUSE
Pauses execution until the user hits ENTER. Useful in scripts to give the user a chance to look at some output
before proceeding.
PERMLOAD filename
Loads a new datafile and continues with the current command after the read section of the datafile finishes. The
filename is the datafile name, and can be either a quoted string or a string variable. Since the automatic re-initialization
makes Evolver forget all non-permanent variables, care should be taken that the current command only uses permanently assigned variables (assigned with ::= ). Useful for writing scripts that run a sequence of evolutions based on
varying parameter values. Using permload is a little tricky, since you dont want to be redefining your permanent
commands and variables every time you reload the datafile, and your permanent command cannot refer directly to
variables parameterizing the surface. One way to do it is to read in commands from separate files. For example, the
catenoid of cat.fe has height controlled by the variable zmax. You could have a file permcat.cmd containing the overall
series script command
run_series ::= {
for ( height ::= 0.5 ; height < 0.9 ; height ::= height + .05 )
{ permload "cat"; read "permcat.gogo";}
}
and a file permcat.gogo containing the evolution commands
u; zmax := height; recalc; r; g 10; r; g 10; hessian;
printf "height: %f area: %18.15f\n",height,total\_area >> "permcat.out";
Then at the Evolver command prompt,
Enter command: read "permcat.cmd"
Enter command: run_series
For loading a new surface and not continuing with the current command, see load. Wildcard matching is in effect on
some systems (Windows, linux, maybe others), but be very careful when using wildcards since there can be unexpected
matches.
NOTE: permload does not work well; I suggest using replace_load instead.
POP
Pops an individual edge or vertex or set of edges or vertices, giving finer control than the universal popping of
the O and o commands. The specified vertices or edges are tested for not being minimal in the soap film sense. For
vertices, this means having more than four triple edges adjacent; higher valence edges are automatically popped. For
edges, this means having more than three adjacent facets when not on constraints or otherwise restricted. It tries to act
properly on constrained edges also, but beware that my idea of proper behavior may be different from yours. Normally,
popping puts in new edges and facets to keep originally separated regions separate, but that behavior can be changed
with the pop_disjoin toggle. Also, the pop_to_edge and pop_to_face toggles cause favoring of popping outcomes.
Examples: Examples:
pop edge[2]
pop edge where valence==5
162
163
Under some circumstances, popping a vertex can leave two regions connected by a tunnel; if the septum_flag toggle
is on, it will force a surface to be placed across the tunnel.
POP_EDGE_TO_TRI
This command does a particular topological transformation common in three-dimensional foam evolution. An
edge with tetrahedral point endpoints is transformed to a single facet. A preliminary geometry check is made to be
sure the edge satisfies the necessary conditions, one of which is that the triple edges radiating from the endpoints have
no common farther endpoints. If run with the verbose toggle on, messages are printed when a specified edge fails to
be transformed. This command is the inverse of the pop_tri_to_edge command. Works in linear and quadratic mode.
Examples:
pop_edge_to_tri edge[2]
pop_edge_to_tri edge where valence==3 and length < 0.001
POP_QUAD_TO_QUAD
This command does a particular topological transformation common in three-dimensional foam evolution. A
quadrilateral bounded by four triple edges is transformed to a quadrilateral oriented in the opposite direction. The
shortest pair of opposite quadrilateral edges are shrunk to zero length, converting the quadrilateral to an edge, then
the edge is expanded in the opposite direction to form the new quadrilateral. The new quadrilateral inherits attributes
such as color from the first quadrilateral, although all the facet numbers are different. A preliminary geometry check
is made to be sure the edge satisfies the necessary conditions, one of which is that the triple edges radiating from the
quadrilateral corners have no common farther endpoints. If run with the verbose toggle on, messages are printed when
a specified quadriteral fails to be transformed. The specified facet can be any one of the facets of the quadrilateral
with a triple line on its border. It doesnt hurt to apply the command to all the facets of the quadrilateral, or to facets
of multilple quadrilaterals. Quadrilaterals may be arbitrarily subdivided into facets; in particular, they may have some
purely interior facets. Works in linear and quadratic mode. Examples:
pop_quad_to_quad facet[2]
pop_quad_to_quad facet where color==red
POP_TRI_TO_EDGE
This command does a particular topological transformation common in three-dimensional foam evolution. A facet
with three tetrahedral point vertices is transformed to a single facet. A preliminary geometry check is made to be sure
the edge satisfies the necessary conditions, one of which is that the triple edges radiating from the vertices have no
common farther endpoints. If run with the verbose toggle on, messages are printed when a specified edge fails to be
transformed. This command is the inverse of the pop_edge_to_tri command. Works in linear and quadratic mode.
Examples:
pop_tri_to_edge facet[2]
pop_tri_to_edge facet where color == red
PRINT expr
Prints the value of the expression. PRINT expr can also be used inside an expression, where it prints the expression
and evaluates to the value of its expression. Will do string expressions. Examples:
print datafilename
print max(edge,length)
print max(vertex, print (x^2+y^2+z^2) )
163
164
PRINT procedure_name
Prints the command assigned to the named procedure. These are reconstructed from the parse tree, so may not be
identical in appearance to the original commands.
PRINT arrayslice rint array rray
The arrayslice option takes an array name or a partially indexed array name. If more than one element results, the
slice is printed in nested curly braces. The arrayslice can also be that of an array attribute of an element. Remember
array indexing starts at 1. Examples:
define parts real[3][2][3];
print parts
print parts[3][2]
define vertex attribute oldcoords real[3]
print edge[2].vertex[1].oldcoords
PRINT WARNING_MESSAGES
This prints accumulated warning messages. Useful to review warning messages generated early in loading a
datafile which scroll off the screen too quickly to be read.
PRINTF string, expr, expr,...
Prints to standard output using the standard C sprintf function. Only string and numeric expressions are valid.
Evolver invokes the system printf function, so any formats recognized by printf are valid. Before calling the system
printf function, Evolver will scan the format string so that it knows to pass integer format items as integers and floating
point format items as floating point. Evolver will also check the number of expressions to make sure it agrees with the
number of formats. Full printf format specification is rather extensive, and can vary among systems, so one should
consult printf documentation for your particular system. But here is a short summary of commonly used formats. 
The following standard C escape sequences are recognized:
\n newline
\r carriage return
\t tab
\b backspace
\q double-quote mark
An individual format has the syntax %[flags][width][.precision][length]specifier where the specifier letter can be
d Signed decimal integer
e Scientific notation (mantissa/exponent) using e character
E Scientific notation (mantissa/exponent) using E character
f Decimal floating point
g Use the shorter of cf %e or cf %f
G Use the shorter of cf %E or cf %f
o Signed octal
s String of characters
u Unsigned decimal integer
x Unsigned hexadecimal integer
X Unsigned hexadecimal integer (capital letters)
% A % followed by another % character will write % to stdout.
The optional flags can be
164
165
Left-justify within the given field width; Right justification is the default
(see width sub-specifier).
+ Forces preceding the result with a plus or minus sign (+ or -) even for
positive numbers. By default, only negative numbers are preceded with
a - sign.
space If no sign is going to be written, a blank space is inserted before the
value.
# Used with o, x or X specifiers the value is preceeded with 0, 0x or 0X
respectively for values different than zero. Used with e, E and f, it forces
the written output to contain a decimal point even if no digits would
follow. By default, if no digits follow, no decimal point is written. Used
with g or G the result is the same as with e or E but trailing zeros are not
removed.
0 Left-pads the number with zeroes (0) instead of spaces, where padding
is specified (see width sub-specifier).
The optional width can be
number Minimum number of characters to be printed. If the value to be printed
is shorter than this number, the result is padded with blank spaces. The
value is not truncated even if the result is larger.
* The width is not specified in the format string, but as an additional integer
value argument preceding the argument that has to be formatted.
The optional .precision field can be
.number For integer specifiers (d, i, o, u, x, X): precision specifies the minimum
number of digits to be written. If the value to be written is shorter than
this number, the result is padded with leading zeros. The value is not
truncated even if the result is longer. A precision of 0 means that no
character is written for the value 0.
For e, E and f specifiers: this is the number of digits to be printed after
the decimal point.
For g and G specifiers: This is the maximum number of significant digits
to be printed.
For s: this is the maximum number of characters to be printed. By default
all characters are printed until the ending null character is encountered.
For c type: it has no effect.
When no precision is specified, the default is 1. If the period is specified
without an explicit value for precision, 0 is assumed.
.* The precision is not specified in the format string, but as an additional
integer value argument preceding the argument that has to be formatted.
The format string can be a string variable or a quoted string. The format string is limited to 1000 characters, but
otherwise there is no limit on the number of arguments. Example:
-
printf "Datafile: %s area: %21.15g\n",datafilename,total_area;
QUADRATIC
Changes to the quadratic model by inserting midpoints in edges.

QUIT, BYE
Exits Evolver or starts new datafile. Same as q single-letter command.

RAWESTV
165
166
Does vertex averaging on all vertices without regard for conserving volume or whether averaged vertices have like
constraints. But doesnt move vertices on boundaries. Use rawest_vertex_average for doing one vertex at a time.
RAWEST_VERTEX_AVERAGE generator
Vertex average a set of vertices. Each vertex is moved to the average position of its edge-connected neighbors
without regard for volume preservation or any constraints. Vertices are projected back to level-set constraints after
motion. Example:
rawest_vertex_average vertex[23]
READ filename
Reads commands from the named file. The filename can be either a quoted string or a string variable. Echoing of
input can be suppressed with the quietload toggle.
RAWV
Does vertex averaging on all vertices without conserving volume on each side of surface. Will only average vertices
with those of like type of constraints. Doesnt move vertices on boundaries. Use raw_vertex_average for doing one
vertex at a time.
RAW_VERTEX_AVERAGE generator
Vertex average a set of vertices. Each vertex is moved to the average position of its edge-connected neighbors
without regard for volume preservation. Only averages with vertices on the same level-set constraints. Vertices are
projected back to level-set constraints after motion. Example:
raw_vertex_average vertex[23]
REBODY
Recalculates connected bodies. Useful after a body has been disconnected by a neck pinching off. Facets of an old
body are divided into edge-connected sets, and each set defines a new body (one of which gets the old body number).
The new bodies inherit the attributes of the old body. If the original body volume was fixed, then the new bdoies
target volumes become the new actual volumes. If the original body had a volconst, then the new bodies will inherit
the same value; this will probably lead to incorrect volumes, so you will need to adjust it by hand. In commands, you
may specify the new bodies descended from an original body by using the original atttribute.
RECALC
Recalculates everything. Useful after changing some variable or something and recalculation is not automatically
done.
REFINE generator
Subdivides edges by putting a vertex in the middle of each edge and splitting neighboring facets in two. It is the
same action as the long edge subdivide command (command l). Facets will be subdivided by putting a vertex in the
center and creating edges out to the old vertices. It is strongly suggested that you follow this with equiangulation to
nicen up the triangulation. Edge refinement is better than facet refinement as facet refinement can leave long edges
even after equiangulation. Example:
refine edges where not fixed and length > .1
RENUMBER_ALL
Reassigns element id numbers in accordance with order in storage, i.e. as printed with the LIST commands.
Besides renumbering after massive topology changes, this can be used with the reorder_storage command to number
elements as you desire. Do NOT use this command inside an element generator loop!
166
167
REORDER_STORAGE
Reorders the storage of element data structures, sorted by the extra attributes vertex_order_key, edge_order_key,
facet_order_key, body_order_key, and facetedge_order_key. Originally written for testing dependence of execution
speed on storage ordering, but could be useful for other purposes, particularly when renumber_all is used afterwards.
Example:
define
define
define
define
define
vertex attribute vertex\_order\_key real

edge attribute edge\_order\_key real
facet attribute facet\_order\_key real
body attribute body\_order\_key real
facetedge attribute facetedge\_order\_key real
reorder := {
set vertex vertex\_order\_key x+y+z;
set edge ee edge\_order\_key min(ee.vertex,vertex\_order\_key);
set facetedge fe facetedge\_order\_key fe.edge[1].edge\_order\_key;
set facet ff facet\_order\_key min(ff.vertex,vertex\_order\_key);
set body bb body\_order\_key min(bb.facet,facet\_order\_key);
reorder\_storage;
}
REPLACE_LOAD
Main prompt command. Replaces the current surface with a new surface from a datafile without a total reinitialization. Syntax:
replace_load
filename
where filename is a double-quoted string or a string variable. The replace_load command actually dissolves all the
elements of the current surface and then does the addload command to read in the desired datafile. Thus only the top
section and the elements of the new file are read in; the "read" section of the new datafile is not read. All variables,
constraints, quantities, and commands from the original file are remembered, although they may be re-initialized in the
top of the new datafile. This command can be used in loops to repeatedly evolve a surface under different conditions,
for example
run := { printf "Height
Energy\n" >>> "results.txt";
for ( htvar := 1.0 ; htvar <= 2.0 ; htvar += 0.1 )
{ replace_load datafilename;
height := htvar;
gogo;
printf "%f
%f\n",height,total_energy >> "results.txt";
}
}
It is necessary that htvar NOT appear in the top of the datafile, so that it does not get re-initialized when replace_load
is done. You can set the no_dump property of a variable to prevent it from being dumped in the top of the datafile; it
will be dumped in the bottom section instead. Example (as commands, not in the top of the datafile):
htvar := 2.3
htvar.no_dump on
167
168
Replace_load is meant as a replacement for permload, which I never have been able to get to work right.
RESET_COUNTS
Resets to 0 various internal counters that have become nonzero. The counters are:
equi_count
facet_delete_count
facet_refine_count
vertex_pop_count
facet_pop_count
where_count
fix_count
t1_edgeswap_count
edge_delete_count
edge_refine_count
notch_count
edge_dissolve_count
body_dissolve_count
edge_pop_count
edgeswap_count
unfix_count
notch_count
Normally, a count is set to 0 at the start of a command that potentially affects it, accumulated during the execution
of the command, and printed at the end of the command. To be precise, each counter has a "reported" bit associated
with it, and if the "reported" bit is set when the appropriate command (such as u) is encountered, the counter will be
reset to 0 and the "reported" bit cleared. The "reported" bit is set by either flush_counts or the end of a command. The
idea is to have the counts from previous commands available to subsequent commands as long as possible, but still
have the counter reflect recent activity.
RESET_PROFILING
For those systems that have CPU clock cycle profiling active (see profiling), reset_profiling will set all the counters
back to 0.
REVERSE_ORIENTATION generator
Reverses the internal orientation of selected edges or facets, as if they had been entered in the datafile with the
opposite orientation. Useful, for example, when edges come in contact with a constraint and you want to get them
all oriented in the same direction. Relative orientations of constraint and quantity integrals change to compensate,
so energy, volumes, etc. should be the same after the command, but it would be wise to check in your application.
Examples:
reverse_orientation edge[7]
reverse_orientation facets where backbody != 0
RITZ(expr,expr)
Applies powers of inverse shifted Hessian to a random subspace to calculate eigenvalues near the shift value. First
argument is the shift. Second argument is the dimension of the subspace. Prints out eigenvalues as they converge to
machine accuracy. This may happen slowly, so you can interrupt it by hitting whatever your interrupt key is, such as
CTRL-C, and the current values of the remaining eigenvalues will be printed out. Good for examining multiplicities
of eigenvalues. The lowest eigenvalue is subsequently available as last_eigenvalue, and all the eigenvalues produced
are available in the eigenvalues[] array. Example:
ritz(0,5)
SADDLE
Seek to minimum energy along the eigenvector of the lowest eigenvalue of the Hessian. The eigenvalue is stored
in the last_eigenvalue variable.
168
169
SET generator [ name ] attrib expr [ WHERE expr ]
SET generator.attrib expr [ WHERE expr ]
Sets an attribute to the indicated value for qualifying elements. The new value expression may refer to the element
in question. The second form permits the use of . for people who use . instinctively in field names. SET can
change the following attributes: constraint, coordinates, density, orientation, non-global named quantity or named
method, user-defined extra attributes, body target volume, body volconst, fixed, pressure, color, frontcolor, backcolor,
boundary, and opacity (for the appropriate type elements). Fixed, named quantity, and named method attributes are
just toggled on; they do not need the first expr. Setting the pressure on a body automatically unfixes its volume. For
constraint, the expr is the constraint number. If using set to put a vertex on a parametric boundary, set the vertexs
boundary parameters p1, p2, etc. first. For color, the color value is an integer. The integers from 0 to 15 can be referred
to by the predefined names BLACK, BLUE, GREEN, CYAN, RED, MAGENTA, BROWN, LIGHTGRAY, DARKGRAY,
LIGHTBLUE, LIGHTGREEN, LIGHTCYAN, LIGHTRED, LIGHTMAGENTA, YELLOW, and WHITE. The special color
value CLEAR (-1) makes a facet transparent. Edge colors may not show up in geomview unless you do show edges
where 1 to have Evolver explicitly feed edges to geomview instead of letting geomview just show facet edges on its
own. On certain systems (Silicon Graphics and Pixar at present) the opacity value sets the transparency of ALL facets.
A value of 0.0 corresponds to transparent, and a value of 1.0 to opaque. Examples:
set facets density 0.3 where original == 2
set vertices x 3*x where id < 5 // multiplies x coordinate by 3
set body target 5 where id == 1 // sets body 1 target volume to 5
set vertices constraint 1 where id == 4
set facet color clear where original < 5
SET name attrib expr
Inside a FOREACH loop with a named element, one can set an attribute for the named element. A . may be used
as in sl name.attrib by people who like that sort of thing.
SET quantityname attrib expr
SET instancename attrib expr
Set the value of a named quantity or named method instance attribute. For a named quantity, the settable attributes
are target, modulus, volconst, and tolerance. For a named method instance, only modulus. There is no implicit
reference to the quantity in the expression, so say
set myquant target myquant.value
rather than set myquant target value.

SET BACKGROUND color
Sets the backgound color for native graphics (not geomview). The color is an integer 0-15 or a color name listed
in the general SET command below. Example:
set background black
SHELL
Invokes a system subshell for the user.
169
170
SIMPLEX_TO_FE
Converts a simplex model surface to a string or soapfilm model surface. Only works for dimension 1 or 2 surfaces,
but works in any ambient dimension.
SOBOLEV
Uses a positive definite approximation to the area Hessian to do one Newton iteration, following a scheme due
to Renka and Neuberger [RN]. See section in Technical Reference chapter for more. Works only on area with fixed
boundary; no volume constraints or anything else. Seems to converge very slowly near minimum, so not a substitute
for other iteration methods. But if you have just a simple soap film far, far from the minimum, then this method can
make a big first step. SOBOLEV_SEEK will do an energy-minimizing search in the direction.
SPRINTF string, expr,...
Prints to a string using the standard C printf function. Otherwise same as PRINTF. Can be used wherever a string
expression is called for. Examples:
dump sprintf "file%g.dmp",count
message := sprintf "Total area is %g
n",total_area
STABILITY_TEST
Calculates the largest eigenvalue of the mobility matrix. Generates a random vector and applies mobility matrix
20 times. Prints out ratios of successive vector lengths. Useful in investigating stability when using approximate
curvature.
SUBCOMMAND
Invokes a subsidiary command interpreter. Useful if you want to pause in the middle of a script to give the user the
chance to enter commands. A subcommand interpreter gives the prompt Subcommand: instead of Enter command:
. Subcommands may be nested several deep, in which case the prompt will display the subcommand level. To exit
a subcommand prompt, use q, quit, or exit. The abort command will return to the prompt on the same subcommand
level.
SYSTEM stringexpr
Invokes a subshell to execute the given command. Command must be a quoted string or a string variable. Will
wait for command to finish before resuming.
T1_EDGESWAP edgegenerator
Does a T1 topological transition in the string model. When applied to an edge joining two triple points, it reconnects edges so that opposite faces originally adjacent are no longer adjacent, but two originally non-adjacent faces
become adjacent.
\_/
/ \
=>
\/
|
/\
It will silently skip edges it is applied to that dont fulfill the two triple endpoint criteria, or whose flipping is
barred due to fixedness or constraint incompatibilities. The number of edges flipped can be accessed through the
t1_edgeswap_count internal variable. Running with the verbose toggle on will print details of what it is doing. Syntax:
170
171
T1_EDGESWAP edge_generator
Examples:
t1_edgeswap edge[23]
t1_edgeswap edge where length < 0.1
UNFIX elements [ name ] [ WHERE expr ]
Removes the FIXED attribute for qualifying elements. Example:

unfix vertices where on_constraint 2
UNFIX variable
Converts the variable to an optimizing parameter.

UNFIX named quantity
Converts a fixed named quantity to an info_only quantity.

UNSET elements [ name ] attrib [ WHERE expr ]
Removes the attribute for qualifying elements. Attributes here are FIXED, VOLUME, PRESSURE, DENSITY, nonglobal named quantities or named methods, FRONTBODY, BACKBODY, CONSTRAINT n, or BOUNDARY n. A use
for the last is to use a boundary or constraint to define an initial curve or surface, refine to get a decent triangulation,
then use unset vertices boundary 1" and "unset edges boundary 1" to free the curve or surface to evolve. The form
unset facet bodies ... is also available to disassociate given facets from their bodies.
UNSET elements [ name ] CONSTRAINT expr [ WHERE expr ]
Removes constraint of given number from qualifying elements.

UTEST
Runs a test to see if the triangulation is Delaunay. Meant for higher dimensions and simplex model.
WHEREAMI
If Evolver is at a debugging breakpoint, then whereami will print a stack trace of the sequence of commands
invoked to get to the current breakpoint.
VERTEX_AVERAGE vertexgenerator
The action is the same as the V command, except that each new vertex position is calculated sequentially, instead
of simultaneously, and an arbitrary subset of vertices may be specified. Fixed vertices do not move. Examples:
vertex_average vertex[2]
vertex_average vertex where id < 10
vertex_average vertex vv where max(vv.facet,color==red) == 1
VERTEX_MERGE
Merges two soapfilm-model vertices into one. Meant for joining together surfaces that bump into each other.
Should not be used for vertices already joined by an edge. Syntax:
vertex_merge(integer,integer)
171
172
Note the syntax is a function taking integer vertex id arguments, not element generators. Example:
vertex_merge(3,12)
WRAP_VERTEX(v,w)
In a symmetry group model, transforms the coordinates of vertex number n by symmetry group element w and
adjusts wraps of adjacent edges accordingly.
ZOOM [ integer expr ]
Zooms in on vertex whose id is the given integer, with radius the given expression. Same as Z command, but not
interactive. If id and radius omitted, then uses current values.
172
Chapter 12
Script examples
Here are some longer examples using the command language. Since the language expects one command per line,
long commands have line breaks only within braces so Evolver can tell the command continues. I tend to break
long commands up into several little named commands. The best way to use these longer commands is to put their
definitions at the end of your datafile in a "read" section, or have them in a separate file you can read with the "read"
command.
1. A script to evolve and regularly save dumps in different files:
nn := 1
while nn < 100 do { g 100; ff := sprintf "stage%d.dmp",nn; dump ff; nn:=nn+1}
2. Your own iterate command to print out what you want, like energy change each step:
olde := total_energy;
gg := { g; printf "Change: %10.5g\n",total_energy-olde; olde:=total_energy}
3. A command to print a file listing vertices and edges (happens to be the OFF format for geomview):
aa:=foreach vertex vv do { printf "%f %f %f \n",x,y,z }
bb:=foreach facet ff do
{ printf "3 ";
foreach ff.vertex do printf "%d ",id-1;
printf "\n"
};
do_off:= { printf "OFF\n%d %d %d\n",vertex_count,facet_count,edge_count;aa;bb}
This prints to stdout, so you would use it with redirection:

Enter command: do_off | "cat >test.off"
4. A command to duplicate the fundamental region of a symmetric surface across a plane of mirror symmetry.
// Producing datafile surface duplicated across a mirror.
// This is set up to do constraint 1 only.
// Mirrored constraints given x numbers.
// Resulting datafile needs mirrored constraints added by hand.
// Usage: mm | "cat >filename.fe"
173
174
// Mirror plane aa*x + bb*y + cc*z = dd

// You have to set these to the proper coefficients for constraint 1
// before doing mm.
aa := 1; bb := 0; cc := 0; dd := 0;
ma := foreach vertex do
{ printf "%g %g %g %g ",id,x,y,z;
if fixed then printf "fixed ";
if on_constraint 1 then printf "constraint 1 " ;
printf "\n";
};
mb := { voff := vertex_count };
mc := foreach vertex do if not on_constraint 1 then
{ lam := 2*(dd-aa*x-bb*y-cc*z)/(aa*aa+bb*bb+cc*cc);
printf "%g %g %g %g ",id+voff,x+lam*aa,y+lam*bb,z+lam*cc;
if fixed then printf "fixed ";
if on_constraint 1 then printf "constraint 1x " ;
printf "\n";
};
md := foreach edge ee do
{ printf "%g ",id;
foreach ee.vertex do printf "%g ",oid;
if ee.fixed then printf "fixed ";
if ee.on_constraint 1 then printf "constraint 1 " ;
printf "\n";
};
me := { eoff := edge_count };
mf := foreach edge ee do if not on_constraint 1 then
{ printf "%g ",id+eoff;
foreach ee.vertex do
{ if on_constraint 1 then { printf "%g ",oid;}
else { printf "%g ", id+voff;}
};
if ee.fixed then printf "fixed ";
if ee.on_constraint 1 then printf "constraint 1x " ;
printf "\n";
};
mg := foreach facet ff do
{ printf "%g ",id;
foreach ff.edge do printf "%g ",oid;
if ff.fixed then printf "fixed ";
if ff.on_constraint 1 then printf "constraint 1 " ;
printf "\n";
};
mh := { foff := facet_count};
mi := foreach facet ff do if not on_constraint 1 then
{ printf "%g ",id+foff;
foreach ff.edge do
{ if on_constraint 1 then { printf "%g ",id;}
else printf "%g ",(oid<0?-(id+eoff):id+eoff);
};
174
175
if ff.fixed then printf "fixed ";

if ff.on_constraint 1 then printf "constraint 1x " ;
printf "\n";
};
mm := { list topinfo;
printf "\nVertices\n";
ma;mb;mc;
printf "\nEdges\n";
md;me;mf;
printf "\nFaces\n";
mg;mh;mi;
};
175
Chapter 13
Operation
13.1
System command line
Syntax:
evolver [-ffilename] [-a-] [-d] [-e] [-i] [-m] [-pn] [-q] [-Q] [-x] [-w] [-y] [datafile]
The current directory and EVOLVERPATH will be searched for the datafile. If the datafile is not found, then a new
search with extension .fe is done. Wildcard matching is in effect on some systems (Windows, linux, maybe others),
but be very careful when using wildcards since there can be unexpected matches. If the datafile is still not found, or
no datafile is given on the command line, the user will be prompted to supply a datafile name.
Options:
-a- Do not enable automatic conversion to named methods and quantities mode when a situation requiring it arises.
-d Prints YACC debugging trace as datafile is parsed. May be helpful if you cant figure out why your datafile doesnt
get read in properly AND you know YACC.

-e Echo input. Meant for echoing commands of piped input to screen so the user can follow what is going on in case
Evolver is being controlled by another process.

-f Specifies the name of a file to be used as command input. At the end of this file, input reverts to stdin. The effect is
the same as redirecting input from the file, except that -f will echo commands to the screen and revert to stdin at
the end. Also note that errors will cause input to revert to stdin.
-i Keeps elements numbers as listed in the datafile, instead of renumbering them consecutively. A datafile can enable
this by including the keyword keep_originals in the top section.
-m Turn memory debugging on at start of program. Same effect as memdebug command.
-pn Forces use of n processes for an Evolver compiled in multi-processor mode. n may be larger or smaller than the
physically available number of processors. The default is 1 processor. This should be regarded as experimental;
there is still too much overhead to be useful.
-q Convert everything to named quantities internally. There are some things for which no quantities exist yet, produc-
ing error messages.

-Q Suppresses echoing of the read section of the datafile, and of files input with the read command. Same as the
quietload toggle.
-w Causes Evolver to exit whenever a warning occurs. Meant to be used when Evolver is run in a shell script.
176
177
-x Causes Evolver to exit whenever an error occurs. Meant to be used when Evolver is run in a shell script.
-y Causes Evolver to cease execution of commands and return to command prompt after any warning message. Same
effect as break_after_warning runtime toggle.
13.2
Initialization
The following steps occur when a new datafile is read in:

1. Any previous surface has all memory deallocated.
2. Defaults are initialized.
3. The datafile is read in.
4. Any non-triangular faces are divided into triangles in the soapfilm model.
5. The order of facets around each edge is determined geometrically.
6. Vertices are projected to their constraints.
7. Checks as described under the C command are run.
8. Initial areas, energies, and volumes are calculated.
9. The viewing transformation matrix is reset to center the surface on the screen.
10. The main command interface is started.
13.3
Error handling
There are several categories of errors:

WARNING Something has happened that you should know about, but it is possible to proceed.
EXPRESSION ERROR There is an error in parsing an expression.
PARSING ERROR Error in datafile syntax, but parsing will continue as best it can.
DATAFILE ERROR Error in datafile semantics, but parsing will continue as best it can.
ERROR The current operation cannot continue. It is abandoned and you return to the command prompt.
FATAL ERROR The program cannot continue. Exits immediately.
All error messages go to stderr. Errors in the datafile report line numbers and print the current line so far. Note that
the real error may have been at the end of the previous line. If command input is being taken from a file at the time an
error occurs, the line number of the offending command will be printed and command input will revert to stdin. If the
-x option was given when Evolver was started, then Evolver will exit immediately with a nonzero error code.
177
178
13.4 Interrupts
Evolver operation may be interrupted with the standard keyboard interrupt, CTRL-C usually (SIGINT for you unix
gurus). During repeated iteration steps, this will set a flag which causes the repetition to cease after the current step.
Otherwise, the current command is aborted and control returns to the main prompt. If Evolver receives SIGTERM (say
from the unix kill command), it will dump to the default dump file and exit. This is useful for stopping background
processes running on a script and saving the current state. The same thing will happen with SIGHUP, so losing a
modem connection will save the current surface.
Note: In Windows NT/95/98, the second interrupt doesnt do anything much since Windows creates a separate
thread to handle the interrupt, and I cant find any way to force the offending thread to stop and longjmp back to where
it should. So if the Evolver is really, really stuck, you may just have to kill the whole program.
178
Chapter 14
Graphics
14.1
Graphics commands
Graphics commands control the display of the surface, both the viewing transformation and which elements are seen.
Keep in mind that there are two ways to display graphics on many systems: a native or internal mode entirely under
the control of the Evolver (such as the X-windows display), and an external mode where information is provided to an
external program such as geomview. Remember that hardcopy output (such as Postscript) will show what the native
mode shows, not what geomview shows. Some general commands that affect both modes can be made from the main
prompt Enter command:. Others, that control the native graphics, are made in a special graphics command mode
entered by the s or show commands.
Note: native mode graphics are assumed to be slow, and dont automatically redraw by default. In particular, there
usually is no polling for window redraw events from the operating system, so after moving or resizing a graphics
window you may have to give an explicit s command to redraw.
The display consists entirely of facets and edges. Special edges (fixed edges, boundary edges, constraint edges,
triple edges) are always shown, unless you make their color CLEAR. The individual facet edges can be toggled with
the graphics command e or the geomview command ae.
Picking: If you use geomview or a Windows OpenGL version of the Evolver, you can right mouse click on a
vertex, edge, or facet in the geomview window, and Evolver will print the number of the items picked. These numbers
are saved in the internal variables pickvnum, pickenum, pickfnum. The F key command on the graphics window sets
the rotation and scaling center to the pickvnum vertex. Pickvnum is settable with ordinary assignment commands, so
the user can zoom in on any vertex desired.
Note: Since vertices are not drawn individually, Evolver reports a vertex as picked only when two edges with a
common vertex are simultaneously picked. Therefore a vertex at the end of a single edge cannot be picked.
This section lists the graphics mode commands first and then the general commands.
The graphics mode prompt is Graphics command:. A graphics command is a string of letters followed by
ENTER. Each letter causes an action. Some commands may be preceded by an integer count of how many repetitions
to do. Example command: rr15u2z. Rotation commands may be preceded by a real number giving the degrees of
rotation; an integer will give a repetition count with the default angle of 6 degrees. If you mean to specify the angle,
be sure to include a decimal point.
Commands in graphics command mode:
Repeatable commands:
u
Tip up. Rotates image six degrees about horizontal axis.
Tip down. Rotates image six degrees other way.
179
180
Rotate right. Rotates by six degrees about vertical axis.
Rotate left. Rotates by six degrees the other way.
Rotate clockwise by six degrees about center of screen.
Rotate counterclockwise by six degrees about center of screen.
Zoom. Expands image by factor of 1.2.
Shrink. Contracts image by factor of 1.2.

Move image half screen width in appropriate direction. May not work on all terminals. For MSWindows versions, arrow keys work when the mouse is in the display window.
arrow keys
Non-repeatable commands:
R
Reset viewing angles to original defaults and rescale the image to fit the viewing window.
Toggle showing all the facet edges.
Toggles display of bounding box. Useful for visualizing orientation.

changed by setting the bounding_box_color variable.
Reset mode of displaying torus surfaces. Choice of raw cell, connected bodies, or clipped unit cell. See the Torus
section for more explanation.
Toggles display of facets entirely on constraints. Meant to show traces where surfaces meet constraints. w
stands for wall. Applies to facets run up against inequality constraints.
Toggles display of facets on boundaries or equality constraints attribute.
Toggles showing of convex and concave edges in different colors. v stands for valleys.
Increments color number used for facet edges.
Decrements color number used for facet edges.
h,?
Prints help screen for graphics commands.
q,x
Exit from graphics mode, and return to main command mode.
The color of the bounding box can be
OpenGL graphics
Ideally, you have a version of the Evolver that uses OpenGL for its screen graphics. OpenGL is standard on Windows
NT, Unix, Linux, and Mac OS X. Tbe graphics display is invoked with the s command, which leaves you at the
graphics prompt, which you should quit q right away since graphics commands are better given in the graphics
window. Besides all the standard screen graphics commands, there are many geomview-like features. The left mouse
button moves the surface continuously, and the right mouse button picks vertices, edges, and facets. With the graphics
window in the foreground, these keyboard commands are active:
h Print a help screen on the console window
r Rotate mode for left mouse button
t Translate mode for left mouse button
z Zoom mode for left mouse button
180
181
c Clockwise/counterclockwise spin mode for left mouse button

l Clipping plane translate mode (lower case l)
k Clipping plane rotate mode
L Turn off clipping plane
+ Widen edges
- Narrow edges
R Reset the view
p Toggle orthogonal/perspective projection
s Toggle cross-eyed stereo
e Toggle showing all edges, regardless of "show edge" condition
f Toggle showing all facets, regardless of "show facet" condition
F Move the rotate/zoom origin to the last picked vertex
G Start another graphics window with independent camera.
o Toggle drawing a bounding box.
g Toggle Gourard (smooth) shading.
x Close the graphics window.
arrow keys Translate a bit.
And some more advanced commands most users will never use, but are listed here for completeness:
H Print advanced help.
a Toggle using OpenGL element arrays.
i Toggle interleaved elements in OpenGL arrays.
I Toggle indexed OpenGL arrays.
S Toggle OpenGL triangle strips.
Y Toggle strip coloring (I was curious as to what OpenGL triangle strips would look like).
D Toggle using a display list.
Q Toggle printing drawing statistics.
181
182
Graphics-related variables
background
Background color of screen graphics.

bounding_box_color
The color of the bounding box, if the bounding box is shown.

pickvnum, pickenum, pickfnum
The id numbers of the most recently picked vertex, edge, and facet respectively. Pickvnum is used by the F key
command to set the center of rotation.
thickness
Some graphics displays do not permit the front and back colors of a facet to be set separately. In this case, two
facets are displayed, separated by the distance thickness.
window_aspect_ratio
The ratio of the the vertical to horizontal dimensions of the display window. If set, this locks the aspect ratio to the
given value. The window may be resized with the mouse, but the aspect ratio will stay the same. The unset value of
window_aspect_ratio is 0; setting window_aspect_ratio to 0 will unlock the aspect ratio. Applies also to the PostScript
bounding box, if full_bounding_box is on. Currently implemented only in Evolver with GLUT graphics. Caveat: the
window doesnt always fully follow the mouse; just keep trying.
Graphics commands in main command mode
The following commands are entered at the main command prompt:
GEOMVIEW [ON|OFF]
Toggles geomview display. Same as P menu options 8 and 9.

GEOMVIEW stringexpr
Sends the string to geomview. Useful for Evolver scripts using geomview to make movies.
GEOMPIPE [ON|OFF]
Toggles sending geomview output through named pipe. Same as P menu options A and B.
GEOMPIPE stringexpr
Redirects Evolvers geomview output through the command stringexpr, which is a quoted string or a string variable.
Good for debugging, but be sure to have gv_binary off so the output is human readable.
OOGLFILE stringexpr
Writes a file containing OOGL-formatted graphics data for the surface as a POLY or CPOLY quad file. This is a
non-interactive version of the P 2 command. The string gets ".quad" appended to form the filename. This command
does not ask any of the other questions the P 2 command asks; it uses the default values, or whatever the last responses
were to the previous use of the interactive P 2 command. Good for use in scripts. Example:
ooglfilename := sprintf "frame%d",framecounter;
ooglfile ooglfilename;
framecounter += 1;
POSTSCRIPT stringexpr
Creates a PostScript file of name stringexpr for the current surface. If the filename is missing a .ps or .eps
extension, a .ps extension will be added. This is the same as the P 3 command, except there are no interactive responses
needed, so it is useful in scripts. The PostScript output options are controlled by the pscolorflag, gridflag, crossingflag,
and labelflag toggles. The median gray level can be set with the variable brightness. The relative label size may be
controlled by the variable ps_labelsize, whose default value is 3.0. The linewidth of edges may be controlled by the
user. Widths are relative to the image size, which is 3 units square. If the real-valued edge extra attribute ps_linewidth
is defined, that value is used as the edge width. Otherwise some internal read-write variables are consulted for various
types of edges, in order:
182
183
ps_stringwidth - edges in the string model, default 0.002

ps_bareedgewidth - "bare" edges, no adjacent facets, default 0.0025
ps_fixededgewidth - "fixed" edges, default 0.002
ps_conedgewidth - edges on constraints or boundaries, default 0.002
ps_tripleedgewidth - edges with three or more adjacent facets, default 0.0015
ps_gridedgewidth - other edges, default 0.001
The bounding box listed in the PostScript file is normally the actual extent of the surface in the window (i.e. the
bounding box is never bigger than the window, but may be smaller). The full_bounding_box toggle will force the
bounding box to be the full window. This is useful in controlling the image size while making a series of images of
different views or evolution stages of a surface.
SHOW elements [ name ] [ WHERE expr ]
Shows surface with screen graphics and goes into graphics command mode. The condition will prescibe which
facets or edges will be displayed with graphics commands. The prescription will remain in effect until the next show
query. It will also be saved in any dump file. Example:
show facets where area < .01
The default for facets is to show all facets. This can be done with show facets. But since many graphics displays
(such as geomview) like to handle showing edges themselves, the default for edges is just to show triple junctions,
borders, and other special edges. To force showing of all edges, do show edges where 1. Such edges will be
displayed by geomview despite the status of its ae command.
SHOWQ
Will show surface and return immediately to main command mode, without going into graphics command mode.
SHOW_EXPR elements [ name ] [ WHERE expr ]
This sets the display prescription without actually doing any graphics. Useful when graphics are not available, as
when working remotely, and you want to generate graphical file output with a prescription, or in start-up commands
at the end of a datafile.
SHOW_TRANS string
Applies string of graphics commands to the image transformation matrix without doing any graphic display. The
string must be in double quotes or be a string variable, and is the same format as is accepted by the regular graphics
command prompt. Example:
show_trans "rrdd5z"
TRANSFORMS [ON|OFF]
Toggles displaying of multiple viewing transforms. These are either those listed explicitly in the datafile or those
generated with a previous transform_expr command.
TRANSFORM_DEPTH n
Quick way of generating all possible view transforms from view transform generators, to a given depth n. I.e.
generates all possible products of n or less generators.
TRANSFORM_EXPR string
If view transform generators were included in the datafile, then a set of view transforms may be generated by
a expression with a syntax much like a regular expression. An expression generates a set of transforms, and are
compounded by the following rules. Here a lower-case letter stands for one of the generators, and an upper-case letter
a
Generates set I, a.
!a
Generates set a.
AB Generates all ordered products of pairs.
for an expression.
The precedence order is that nA is higher than AB
nA Generates all n-fold ordered products.
A|B Generates union of sets A and B.
(A) Grouping; generates same set as A.
which is higher that A|B. Note that the expression string must be enclosed in double quotes or be a string variable.
The "!" character suppresses the identity matrix in the set of matrices so far. Examples:
183
184
transform_expr "3(a|b|c)"
(All products of 3 or fewer generators.)
transform_expr "abcd"
(Generates sixteen transforms) transform_expr "!a!a!a!a!"
(Generates one transform)

All duplicate transforms are removed, so the growth of the sets do not get out of hand. Note that the identity
transform is always included. The letter denoting a single generator may be upper or lower case. The order is the same
order as in the datafile. If 26 generators are not enough for somebody, let me know.
The current transform_expr may be used as a string, and the number of transformations generated can be accessed
as transform_count. For example,
print transform_expr
VIEW_4D [ON|OFF]
For geomview graphics, sends a full 4D file in 4OFF and 4VECT format. Default is OFF, so geomview gets just
the 3D projection.
14.2
Graphics output file formats
This section explains the file formats produced by the P command.
14.2.1
Pixar
A file extension .quad is automatically added to the filename.

This format lists the facets as quadrilaterals, with the third vertex being repeated. Two options are available:
1. Interpolated normals. This requires recording the normal vector at each vertex.
2. Colors. This requires recording color values (red,green,blue,alpha) at each vertex.
The file starts with a keyword:
CNPOLY
NPOLY
CPOLY
POLY
colors and normals

normals
There follows one facet per line. Each vertex is
colors
vertices only.
printed as three coordinates, three normal vector components (if used), and four color values (if used).
The coordinates are the true coordinates, not transformed for viewing. Facets are not ordered back to front, since
Pixar handles viewing and hidden surfaces itself.
14.2.2
Geomview
The P command menu option 8 starts the geomview program and feeds it data through a pipe. The file fed into the
pipe follows the geomview command file format. Parts of it may be in binary format for speed. A copy of the file may
be created by using the P menu option A to create a named pipe, and then using cat on the pipe to read the pipe into
your own file. End the file with P option B.
Some handy geomview keyboard commands (to be entered with mouse in the geomview image window):
r Set mouse to rotate mode.
t Set mouse to translate mode.
z Set mouse to zoom mode.
ab Toggle drawing bounding box.
ae Toggle facet edge drawing.
af Toggle drawing facets.
Remember that what you see in geomview is not what you will get with a Postscript output, since Evolver knows
nothing about the commands you give geomview.
184
14.2.3
185
PostScript
A PostScript file can be created with the POSTSCRIPT command or the P command option 3. The output file is
suitable for direct input to a PostScript device. If the output filename you give is missing a .ps or .eps extension, a
.ps extension will be added. The view is from the current viewing angle as established with the s command or the
show_trans string command. Edges and facets are drawn back to front.
The P 3 command is interactive. You are given a choice whether to draw all the grid lines (all the facet edges).
Even if you dont choose to, all special edges (fixed, boundary, triple edges, bare edges) will still be drawn. You
can prevent an edge from being drawn by giving it the color CLEAR. Color data is optionally included; you will be
prompted for your choice. For a string model in at least 3 dimensions, you will also be asked whether you want to
show edge crossings with breaks in the background edge, which helps in depth perception. There is also a label option,
which prints element numbers, for easy cross-referencing with datafiles. Edge orientation is indicated by the labels
being slightly displaced toward the edge head, and face labels are signed according to which side you are seeing. By
responding with i or o at the labels prompt, you can get either current id numbers or original id numbers.
The POSTSCRIPT command is not interactive. You give the filename with the command, and the output options
are controlled by the pscolorflag, gridflag, crossingflag, and labelflag options.
The linewidth of edges may be controlled by the user. Widths are relative to the image size, which is 3 units square.
If the real-valued edge extra attribute ps_linewidth is defined, that value is used as the edge width. Otherwise some
internal read-write variables are consulted for various types of edges, in order:
ps_stringwidth - edges in the string model, default 0.002
ps_bareedgewidth - "bare" edges, no adjacent facets, default 0.0025
ps_fixededgewidth - "fixed" edges, default 0.002
ps_conedgewidth - edges on constraints or boundaries, default 0.002
ps_tripleedgewidth - edges with three or more adjacent facets, default 0.0015
ps_gridedgewidth - other edges, default 0.002
The relative label size may be controlled by the variable ps_labelsize, whose default value is 3.0.
14.2.4
Triangle file
This is a format I put in to get the data I wanted for a certain external application. This lists one facet per line.
Coordinates are viewing transformed to the screen plane. Facets are listed back to front. The format for a line is x1 y1
x2 y2 x3 y3 w1 w2 w3 d Here (x1,y1), (x2,y2), and (x3,y3) are the vertex coordinates, and w1,w2, and w3 are types
of edges:
type 0: ordinary edge, no constraints or boundaries, two adjacent facets.
type 1: edge on constraint or boundary, but not fixed.
type 3: fixed edge.
type 4: edge adjacent to at least 3 facets.
These values can be used to weight different types of edges to illustrate the structure of the surface. d is the cosine of
the angle of the facet normal to the screen normal. Useful for shading facets.
14.2.5
SoftImage file
This consists of two files, one with extension .mdl and one with .def. The extensions are added automatically.
185
Chapter 15
Named Methods and Quantities

15.1 Introduction
Named methods and quantities are a systematic scheme of calculating global quantities such as area, volume, and
surface integrals that supplements the original ad hoc scheme in the Evolver. Briefly, named methods are built-in functions, and named quantities are combinations of instances of methods. See the ringblob.fe datafile for an example. The
original ad hoc calculations are still the default where they exist, but all new quantities are being added in the named
quantity scheme. Some new features will work only with named quantities, in particular many Hessian calculations.
To convert everything to named quantities, start Evolver with the -q option or use the convert_to_quantities command.
This has not been made the default since named quantities can be slower than the originals. It is planned that eventually
all energies and global constraints will be converted to named quantity system. However, existing syntax will remain
valid wherever possible. Starting Evolver with the -q option will do this conversion now.
The sample datafiles qcube.fe, qmound.fe, and ringblob.fe contains some examples of named quantities and instances. The first two are quantity versions of cube.fe and mound.fe. These illustrate the most general and useful
methods, namely facet_vector_integral, facet_scalar_integral, and edge_vector_integral, rather than the faster but more
specialized methods such as facet_area. My advice is that the user stick to the old implicit methods for area, volume,
and gravitational energy, and use named quantities only for specialized circumstances.
15.2 Named methods

A method is a way of calculating a scalar value from some particular type of element (vertex, edge, facet, body). Each
method is implemented internally as a set of functions for calculating the value and its gradient as a function of vertex
positions. The most common methods also have Hessian functions. Methods are referred to by their names. See the
Implemented methods list below for the available methods. Adding a new method involves writing C routines to
calculate the value and the gradient (and maybe the Hessian) as functions of vertex coordinates, adding the function
declarations to quantity.h, and adding a structure to the method declaration array in quantity.c. All the other syntax for
invoking it from the datafile is already in place.
15.3 Method instances

A method instance is the sum of a particular method applied to a particular set of geometric elements. Some methods
(like facet_area) are completely self-contained. Others (like facet_vector_integral) require the user to specify some
further information. For these, each instance has a specification of this further information. Method instances are
defined in the datafile, and may either be unnamed parts of named quantity definitions or separate named method
instances for inclusion in named quantities. The separate named version is useful if you want to inspect instance
values for the whole surface or individual elements.
186
187
An instance total value can be printed with the A command, or may be referred to as "instancename.value" in
commands. The instance name itself may be used as an element attribute. For example, supposing there is an instance
named moment, which applies to facets. Then typical commands would be
print moment.value
print facet[3].moment
list facet where moment > 0.1
Every method instance has a modulus, which is multiplied times the basic method value to give the instance value.
A modulus of 0 causes the entire instance calculation to be omitted whenever quantities are calculated. The modulus
may be set in the datafile or by assignment. Example commands:
print moment.modulus
moment.modulus := 1.3
The declaration of a method instance may cause it to use a different modulus for each element by specifying an
element extra attribute to use for that purpose. The extra attribute has to have already been declared. Example:
define facet attribute mymod real
quantity myquant energy method facet_area global element_modulus mymod
Of course, it is up to the user to properly initialize the values of the extra attribute.
Some methods, those that logically depend on the orientation of the element, can be applied with a relative orientation. When applied to individual elements in the datafile, a negative orientation is indicated by a - after the instance
name. When applied at runtime with the set command, the orientation will be negative if the element is generated with
negative orientation, i.e. set body[1].facet method_instance qqq. The methods currently implementing this feature are:
edge_vector_integral, string_gravity,facet_vector_integral, facet_2form_integral, facet_volume,facet_torus_volume,
simplex_vector_integral, simplex_k_vector_integral,
edge_k_vector_integral, gravity_method, and full_gravity_method.
15.4 Named quantities

A named quantity is the sum total of various method instances, although usually just one instance is involved. The
instances need not apply to the same type of element; for example, both facet and edge integrals may be needed to
define a volume quantity. Each named quantity is one of three types:
energy quantities which are added to the total energy of the surface;
fixed quantities that are constrained to a fixed target value (by Newton steps at each iteration); and
conserved quantities are like fixed, but the value is irrelevant. The quantity gradient is used to eliminate a degree of
freedom in motion. Rarely used, but useful to eliminate rotational degree of freedom, for example. Will not
work with optimizing parameters, since they do gradients by differences.
info_only quantities whose values are merely reported to the user. This type is initially set in a quantitys datafile
declaration. A quantity can be toggled between fixed and info_only with the "fix quantityname" and "unfix
quantityname" commands.
The value of a quantity may be displayed with the A or v commands, or as an expression "quantityname.value".
Furthermore, using the quantity name as an element attribute evaluates to the sum of all the applicable component
instance values on that element. For example, supposing there is a quantity named vol, one could do
print vol.value
print facet[2].vol
histogram(facet,vol)
187
188
Each quantity has a modulus, which is just a scalar multiplier for the sum of all instance values. A modulus of 0
will turn off calculation of all the instances. The modulus can be set in the datafile declaration, or by assignment:
quantityname.modulus := 1.2
Each fixed quantity has a target value, to which the Evolver attempts to constraint the quantity value. Each time
an iteration is done ( g command or the various Hessian commands), Newtons Method is used to project the surface
to the constrained values. The target value can be displayed with the A or v commands, or as "quantityname.target". It
can be changed by assignment. Example:
print qname.target
qname.target := 3.12
A quantity can have a constant value added to it, similar to the body attribute volconst. This quantity attribute is
also called volconst. It is useful for adding in known values of say integrals that are omitted from the actual calculation.
It can be set in the quantitys datafile definition, or by an assignment command.
Each fixed quantity has a Lagrange multiplier associated to it. The Lagrange multiplier of a constraint is the rate
of energy change with respect to the constraint target value. For a volume constraint, the Lagrange multiplier is just
the pressure. Lagrange multipliers are calculated whenever an iteration step is done. They may be displayed by the v
command in the "pressure" column, or as an expression "quantityname.pressure".
A fixed quantity can have a tolerance attribute, which is used to judge convergence. A surface is deemed converged
when the sum of all ratios of quantity discrepancies to tolerances is less than 1. This sum also includes bodies of fixed
volume. If the tolerance is not set or is negative, the value of the variable target_tolerance is used, which has a default
value of 0.0001.
Function quantities. Instead of being a simple sum of methods, a named quantity can be an arbitrary function of
method values. The datafile syntax has "function expression" instead of a method list. For example:
named
method_instance qwerty method facet_scalar_integral
scalar_integrand: x^2
quantity foobar energy function qwerty^3
Note the method name is used plain, without a "value" suffix. Also note that the method values used are global
values, not element-wise. Quantity functions can do Hessian operations, if the component methods have Hessians.
Such hessians can become quite memory consuming in default dense matrix form; there is a toggle command function_quantity_sparse that will cause sparse matrices to be used.
15.5 Implemented methods

The currently implemented methods are listed here, grouped somewhat by nature. Within each group, they are more
or less in order of importance.
0-dimensional
vertex_scalar_integral
1-dimensional
edge_tension, edge_length
density_edge_length
edge_scalar_integral
edge_vector_integral
edge_general_integral
edge_area
188
189
edge_torus_area
string_gravity
hooke_energy
hooke2_energy
hooke3_energy
local_hooke_energy
dihedral_hooke
sqcurve_string
sqcurve_string_marked
sqcurve2_string
sqcurve3_string
sqcurve_gauss_curv_cyl
sqcurve_mean_curv_cyl
sq_torsion
metric_edge_length
klein_length
circular_arc_length
circular_arc_area
spherical_arc_length
spherical_arc_area
2-dimensional
facet_tension, facet_area
density_facet_area
facet_volume
facet_scalar_integral
facet_vector_integral
facet_2form_integral
facet_2form_sq_integral
facet_general_integral
facet_torus_volume
gravity_method
full_gravity_method
facet_area_u
density_facet_area_u
gap_energy
metric_facet_area
klein_area
dirichlet_area
sobolev_area
189
190
pos_area_hess
spherical_area
stokes2d
stokes2d_laplacian
2-D Curvatures
mean_curvature_integral
sq_mean_curvature
eff_area_sq_mean_curvature
normal_sq_mean_curvature
star_sq_mean_curvature
star_eff_area_sq_mean_curvature
star_normal_sq_mean_curvature
star_perp_sq_mean_curvature
gauss_curvature_integral
sq_gauss_curvature
circle_willmore
laplacian_mean_curvature
General dimensions
simplex_vector_integral
simplex_k_vector_integral
edge_k_vector_integral
Knot energies
knot_energy
uniform_knot_energy
uniform_knot_energy_normalizer
uniform_knot_normalizer1
uniform_knot_normalizer2
edge_edge_knot_energy, edge_knot_energy
edge_knot_energy_normalizer
simon_knot_energy_normalizer
facet_knot_energy
facet_knot_energy_fix
bi_surface
buck_knot_energy
proj_knot_energy
circle_knot_energy
sphere_knot_energy
sin_knot_energy
190
191
curvature_binormal
ddd_gamma_sq
edge_min_knot_energy
true_average_crossings
true_writhe
twist
writhe
curvature_function
knot_thickness
knot_thickness_0
knot_thickness_p
knot_thickness_p2
knot_thickness2
knot_local_thickness
Elastic stretching energies

dirichlet_elastic
linear_elastic
general_linear_elastic
linear_elastic_B
relaxed_elastic
relaxed_elastic1
relaxed_elastic2
relaxed_elastic_A
relaxed_elastic1_A
relaxed_elastic2_A
SVK_elastic_A
Weird and miscellaneous

wulff_energy
area_square
stress_integral
carter_energy
charge_gradient
johndust
ackerman
191
192
15.6 Method descriptions

The descriptions below of the individual methods give a mathematical definition of the method, what type of element
it applies to, definition parameters, which types of models it applies to, any restrictions on the dimension of ambient
space, and whether the method has a Hessian implemented. Fuller mathematical formulas for some of the methods
may be found in the Technical Reference chapter. Unless specifically noted, a method has the gradient implemented,
and hence may be used for an energy or a constraint. The definition parameters are usually scalar or vector integrands
(see the Datafile chapter for full syntax). Some methods also depend on global variables as noted. The sample datafile
declarations given are for simple cases; full syntax is given in the Datafile chapter. Remember in the samples that for
quantities not declared global, the quantity has to be individually applied to the desired elements.
15.6.1
0-dimensional methods
These are methods that apply to single vertices, not assuming any adjacent edges or facets.
vertex_scalar_integral. Description: Function value at a vertex. This actually produces a sum over vertices,
but as a mathematician, I think of a sum over vertices as a point-weighted integral. Element: vertex. Parameters:
scalar_integrand. Models: linear, quadratic, Lagrange, simplex. Ambient dimension: any. Hessian: yes. Example
datafile declaration:
quantity point_value energy method vertex_scalar_integral
scalar_integrand: x^2 + y^2 - 2x + 3
15.6.2
These methods are meant for curves, whether in the string model or surface boundaries in the string model. Those that
are applied to elements assume there are adjacent edges.
edge_tension or edge_length. Description: Length of edge. Quadratic model uses Gaussian quadrature of order
integral_order_1D. Element: edge. Parameters: none. Models: linear, quadratic, Lagrange. Ambient dimension: any.
Hessian: yes. Example datafile declaration:

quantity len energy method edge_length global
density_edge_length. Description: Length of edge. Quadratic model; uses Gaussian quadrature of order integral_order_1D. Element: edge. Parameters: none. Models: linear, quadratic, Lagrange. Ambient dimension: any.

quantity len energy method density_edge_length global
edge_scalar_integral. Description: Integral of a scalar function over arclength. Uses Gaussian quadrature of order integral_order_1D. Element: Parameters: Models: linear, quadratic, Lagrange. Ambient dimension: any. Hessian:
yes. Type: edge. Parameters: scalar_integrand. Example datafile declaration:
quantity edge_sint energy method edge_scalar_integral
scalar_integrand: x^2 - 3*y + 4
192
193
edge_vector_integral. Description: Integral of a vectorfield over an oriented edge. Uses Gaussian quadrature of
order integral_order_1D. Element: edge. Parameters: vector_integrand. Models: linear, quadratic, Lagrange. Ambient
dimension: any. Hessian: yes. Orientable: yes. Example datafile declaration:
quantity edge_vint energy method edge_vector_integral
vector_integrand:
q1: 0
q2: 0
q3: z^2/2
edge_general_integral. Description: Integral of a scalar function of position and tangent over an edge. The
components of the tangent vector are represented by continuing the coordinate indices. That is, in 3D the position
coordinates are x1,x2,x3 and the tangent components are x4,x5,x6. For proper behavior, the integrand should be
homogeneous of degree 1 in the tangent components. Uses Gaussian quadrature of order integral_order_1D. Element:
edge. Parameters: scalar_integrand. Models: linear, quadratic, Lagrange. Ambient dimension: any. Hessian: yes.
Example datafile declaration: the edge length in 3D could be calculated with this quantity:
quantity arclength energy method edge_general_integral
scalar_integrand: sqrt(x4^2 + x5^2 + x6^2)
edge_area. Description: For calculating the area of a body in the string model. Implemented as the exact integral
ydx over the edge. Valid for torus model, but not general symmetry groups. Element: edge. Parameters: none.
Models: linear, quadratic, Lagrange. Ambient dimension: 2. Hessian: yes. Example datafile declaration:
quantity cell1_area fixed = 1.3 method edge_area
edge_torus_area. Description: For 2D torus string model body area calculations. Contains adjustments for torus
wraps. Element: edge. Parameters: none. Models: torus; string; linear,quadratic,Lagrange. Ambient dimension: 2.
Hessian: no. Example datafile declaration:
quantity cell_area fixed = 1.3 method edge_torus_area
string_gravity. Description: To calculate the gravitational potential energy of a body in the string model. Uses
differences in body densities. Does not use gravitational constant G as modulus (unless invoked as internal quantity
by convert_to_quantities). Element: edge. Parameters: none. Models: string linear, quadratic, lagrange. Ambient
dimension: 2. Hessian: yes. Orientable: yes. Example datafile declaration:
quantity cell_grav energy modulus 980*8.5 method string_gravity
hooke_energy. Description: One would often like to require edges to have fixed length. The total length of some
set of edges may be constrained by defining a fixed quantity. This is used to fix the total length of an evolving knot,
for example. But to have one constraint for each edge would be impractical, since projecting to n constraints requires
inverting an n x n matrix. Instead there is a Hookes Law energy available to encourage edges to have equal length. Its
form per edge is
E = |L L0 | p ,
(15.1)
193
194
where L is the edge length, L0 is the equilibrium length, embodied as an adjustable parameter hooke_length, and the
power p is an adjustable parameter hooke_power. The default power is p = 2, and the default equilibrium length
is the average edge length in the initial datafile. You will want to adjust this, especially if you have a total length
constaint. A high modulus will decrease the hooke component of the total energy, since the restoring force is linear
in displacement and the energy is quadratic (when p = 2). As an extra added bonus, a hooke_power of 0 will give
E = log |L L0 |. See hooke2_energy for individual edge equilibrium lengths. Element: edge. Parameters: none.
Models: linear. Ambient dimension: any. Hessian: yes. Example datafile declaration:
parameter hooke_length 0.3 // will apply to all edges
parameter hooke_power 2 // the default
quantity slinky energy method hooke_energy global
hooke2_energy. Description: Same as hooke_energy, but each edge has an equilibrium length extra attribute
hooke_size (which the user need not declare). If the user does not set hooke_size by the time the method is first called,
the value will default to the current length. Hooke_size is not automatically adjusted by refining. Element: edge.
Parameters: none. Models: linear. Ambient dimension: any. Hessian: yes. Example datafile declaration:
parameter hooke2_power 2 // the default
define edge attribute hooke_size real
quantity slinky energy method hooke2_energy global
...
read
r;r;set edge hooke_size length
hooke3_energy.
Description: Same as hooke2_energy, but uses an elastic model instead of a spring. The
energy is energy = 0.5*(length-hooke_size)2/hooke_size.

The exponent can be altered from 2 by setting the parameter
hooke3_power. Element: edge. Parameters: none. Models: linear. Ambient dimension: any. Hessian: yes. Example
parameter hooke3_power 2 // the default
quantity slinky energy method hooke3_energy global
...
read
r;r;set edge hooke_size length
local_hooke_energy. Description: Energy of edges as springs with equilibrium length being average of lengths
of neighbor edges. Actually, the energy is calculated per vertex,
)
(
L1 L2 2
E=
(15.2)
L1 + L2
where L1 and L2 are the lengths of the edges adjacent to the vertex. Meant for loops of string. (by John Sullivan) If
you set the variable local_hooke_flag nonzero then local_hooke_energy will not be evaluated at vertices with valence
not equal to 2. Element: vertex. Parameters: none. Models: linear. Ambient dimension: any. Hessian: no. Example
quantity slinky energy method local_hooke_energy global
dihedral_hooke. Description: Energy of an edge is edge length times square of angle between normals of adjacent
facets. Actually, e = (1 - cos(angle))*length. Element: edge. Parameters: none. Models: linear. Ambient dimension:
any. Hessian: yes . Example datafile declaration:
194
195
quantity bender energy method dihedral_hooke global
sqcurve_string. Description: Integral of squared curvature in string model. Assumes two edges per vertex, so
dont use with triple points; see sqcurve_string_marked for more complicated topologies. The value zero at endpoint
of curve. The value is calculated as if the exterior angle at the vertex is evenly spread over the adjacent half-edges.
More precisely, if s1 and s2 are the adjacent edge lengths and t is the exterior angle, value = 4*(1 - cos(t))/(s1+s2).
Other powers of the curvature can be specified by using the parameter parameter_1 in the instance definition. Also see
sqcurve2_string for a version with intrinsic curvature, and sqcurve3_string for a version that uses a slightly different
formula to encourage equal length edges. Element: vertex. Parameters: parameter_1. Models: linear. Ambient
dimension: any. Hessian: yes. Example datafile declaration:
quantity sq energy method sqcurve_string global
parameter_1 3
sqcurve2_string. Description: Integral of squared curvature in string model, but with an intrinsic curvature. The
value is zero at endpoint of curve. The value is calculated as if the exterior angle at the vertex is evenly spread over
the adjacent half-edges. More precisely, if s1 and s2 are the adjacent edge lengths, h0 is the intrinsic curvature, and t is
the exterior angle, then value = (sin(t)/((s1 + s2)/2) h0 )2 . The intrinsic curvature is specified by a global variable
h_zero or a real-valued vertex attribute named h_zero.
Element: vertex. Models: linear. Ambient dimension: 2 Hessian: no. Example datafile declaration:
define vertex attribute intrinsic_curvature real
quantity sq2 energy method sqcurve2_string global
sqcurve3_string. Description: Same as sqcurve_string, but uses a slightly different formula to encourage equal
length edges The value zero at endpoint of curve. The value is calculated as if the exterior angle at the vertex is evenly
spread over the adjacent half-edges. More precisely, if s1 and s2 are the adjacent edge lengths, h0 is the intrinsic
curvature, and t is the exterior angle, value = 2*(1 - cos(t))*(1/s1+1/s2). Element: vertex. Models: linear. Ambient
dimension: any Hessian: yes. Example datafile declaration:
quantity sq3 energy method sqcurve3_string global
sq_gaussian_curv_cyl. Description: Integral of the squared gaussian curvature of a surface of revolution. The
generating curve is set up in the string model, and this method applied to its vertices. The axis of rotation is the x-axis.
Element: vertex. Models: linear string. Ambient dimension: 2 Hessian: yes. Example datafile declaration:
define vertex attribute h_zero real
quantity sqgausscyl energy method sq_gauss_curv_cyl global
sq_torsion. Integral of squared torsion for curves. The torsion is approximated by looking at triples of adjacent
are the edge vectors, then the sin of the angle the osculating plane twists by (from A B to B C)
is
edges; if A, B, C
S=
A B C||
B||
||A B|| ||B C||
(This is analogous to T = T T T /2 for tangent vector T and curvature .) Then the torsion is
t = arcsin(S)/||B||
195
196
and the integral of the square of the torsion is

t 2 ||B|| = arcsin(S)2 /||B||
This function assumes the edges in each component are consistently oriented. Since this method is meant to be used
on boundary wires of surfaces, it uses a sq_torsion_mark attribute on edges to tell which edges are to be included.
Example top of datafile declaration:
define edge attribute sq_torsion_mark integer
quantity sqtor energy method sq_torsion
Then mark the edges you want included, for example using quad.fe
edges
1 1 2 constraints 1,2 sq_torsion_mark 1 quantity sqtor
 Element: vertex. Parameters: none. Models: linear. Ambient dimension: 3. Hessian: no.
sq_mean_curv_cyl. Description: Integral of the squared mean curvature of a surface of revolution. The generating curve is set up in the string model, and this method applied to its vertices. The axis of rotation is the x-axis. This
method will do intrinsic curvature by means either of a global variable h_zero or a real-valued vertex attribute h_zero.
Element: vertex. Models: linear string. Ambient dimension: 2 Hessian: yes. Example datafile declaration:
define vertex attribute h_zero real
quantity sqcyl energy method sq_mean_curv_cyl global
sqcurve_string_marked. Description: Integral of squared curvature in string model. Same as sqcurve_string, but
only "marked" edges are used, so the topology of edges can be more complicated than a loop or curve. The marking
is done by declaring an integer-valued edge attribute named sqcurve_string_mark and setting it to some nonzero value
for those edges you want to be involved, usually two at each vertex to which this method is applied. Value zero at
vertex with only one marked edge. Value is calculated as if the exterior angle at the vertex is evenly spread over
the adjacent half-edges. More precisely, if s1 and s2 are the adjacent edge lengths and t is the exterior angle, value
= 4*(1 - cos(t))/(s1+s2). Other powers of the curvature can be specified by using the parameter parameter_1 in the
instance definition. Element: vertex. Parameters: parameter_1. Models: linear. Ambient dimension: any. Hessian:
yes. Example datafile declaration:
define edge attribute sqcurve_string_mark integer
quantity sqmark energy method sqcurve_string_marked
metric_edge_length. Description: In the string model with a Riemannian metric, this is the length of an edge.
Element: edge. Parameters: none. Models: linear,quadratic,simplex. Ambient dimension: any. Hessian: yes. Example
string
space_dimension 2
metric
1+x^2 y
y 1+y^2
quantity mel energy method metric_edge_length global
196
197
klein_length. Description: Edge length in Klein hyperbolic plane model. Does not depend on klein_metric being declared. Vertices should be inside unit sphere. Element: edge. Parameters: none. Models: linear. Ambient
dimension: 2. Hessian: no. Example datafile declaration:
quantity kleinlen energy method klein_length global
circular_arc_length. Description: Edge length, modelling the edge as a circular arc through three points, hence
useful only in the quadratic model. If not in the quadratic model, it evaluates as the edge_length method. The presence
of this quantity has the side effect of automatically toggling circular_arc_draw, causing edges to display as circular
arcs in the quadratic model. Element: edge. Parameters: none. Models: quadratic. Ambient dimension: 2. Hessian:
quantity arclen energy method circular_arc_lenght global
circular_arc_area. Description: Area between an edge and the y axis, with the edge modelled as a circular arc
through three points. Useful in the quadratic model; in other models it is the same as facet_area. Element: facet.
Parameters: none. Models: linear. Ambient dimension: any. Orientable: yes. Hessian: yes. Example datafile
declaration:
constraint 1 formula: x^2 + y^2 + z^2 = 1
quantity spharea energy method circular_arc_area global
spherical_arc_length. Description: Edge length, modelling the edge as a spherical great circle arc between its
two endpoints, which are assumed to lie on an arbitrary radius sphere centered at the origin. This method is meant for
modelling string networks on spheres, and is suitable for use with the length_method_name feature for substituting
the default edge length calculation method. Note that this method is an exact spherical calculation in the linear model,
so there is no need to refine edges or use higher order models for accuracy. But no special graphing yet, so you might
want to refine when making pictures. Element: edge. Parameters: none. Models: linear. Ambient dimension: 3.
parameter rad = 2
constraint 1
formula: x\^2 + y\^2 + z\^2 = rad\^2
length\_method\_name "spherical\_arc\_length"
spherical_arc_area_n, spherical_arc_area_s. Description: Area on a sphere between an edge (considered as a

great circle arc) and the north (or south) pole. This is an exact calculation in the linear model. Meant for calculating the
areas of facets in the string model with the string network confined to a sphere of arbitrary radius centered at the origin.
There are two versions of this method, since calculation of facet areas by means of edges necessarily has a singularity
somewhere on the sphere. Spherical_arc_area_n has its singularity at the south pole, and spherical_arc_area_s has its
singularity at the north pole. Thus spherical_arc_area_s will work accurately for facets not including the north pole
in there interiors; a facet including the north pole will have its area calculated as the negative complement of its true
area, so a body defined using it could get the correct area by using a volconst of a whole sphere area. If the singular
pole falls on an edge or vertex, then results are unpredictable. With these caveats, these methods are suitable for use
with the area_method_name feature for substituting the default edge area method. Element: edge. Parameters: none.
Models: linear. Ambient dimension: 3. Orientable: yes. Hessian: yes. Example datafile declaration:
197
198
parameter rad = 2
constraint 1
formula: x\^2 + y\^2 + z\^2 = rad\^2
area\_method\_name "spherical\_arc\_area\_s"
15.6.3
These methods are meant for two-dimensional surfaces. If they are applied to vertices or facets, they assume there are
adjacent facets.
facet_tension, facet_area. Description: Area of facet. Does not multiply by facet density; density_facet_area
does that. Quadratic model uses Gaussian cubature of order integral_order_2D. Beware that this is an approximation
to the area, and if the facets in the quadratic or Lagrange model get too distorted, it can be a bad approximation.
Furthermore, facets can distort themselves in seeking the lowest numerical area. By default, changing the model to
quadratic or Lagrange will set an appropriate integral_order_2D. Element: facet. Parameters: none. Models: linear,
quadratic, Lagrange, simplex. Ambient dimension: any. Hessian: yes. Example datafile declaration:
quantity farea energy method facet_area global
density_facet_area. Description: Area of facet, multiplied by its density. Otherwise same as facet_area. Element:
facet. Parameters: none. Models: linear, quadratic, Lagrange, simplex. Ambient dimension: any. Hessian: yes.
Example datafile declaration:
quantity farea energy method density_facet_area global
facet_volume. Description: Integral
z dx dy over an oriented facet. Valid in the torus domain. Not valid for other
symmetry groups. Element: facet. Parameters: none. Models: linear, quadratic, Lagrange. Ambient dimension: 3.
Hessian: yes. Orientable: yes. Example datafile declaration:
quantity vol fixed = 1.3 method facet_volume
facet_scalar_integral. Description: Integral of a scalar function over facet area. Uses Gaussian cubature of order
integral_order_2D. Element: facet. Parameters: scalar_integrand. Models: linear, quadratic, Lagrange. Ambient
dimension: any. Hessian: yes. Example datafile declaration:

quantity fint energy method facet_scalar_integral global
scalar_integrand: x^2+y^2
facet_vector_integral. Description: Integral of a vectorfield inner product with the surface normal over a facet.
The normal is the right-hand rule normal of the facet as defined in the datafile. Uses Gaussian cubature of order
integral_order_2D. Element: facet. Parameters: vector_integrand. Models: linear, quadratic, Lagrange, simplex.
Ambient dimension: any. Hessian: yes. Orientable: yes. Example datafile declaration, for volume equivalent:
quantity fvint energy method facet_vector_integrand
vector_integrand:
q1: 0
q2: 0
q3: z
198
199
facet_2form_integral. Description: Integral of a 2-form over a facet. Meant for ambient dimensions higher than
3. Uses Gaussian cubature of order integral_order_2D. Element: facet. Parameters: form_integrand (components in
lexicographic order). Models: linear, Lagrange, simplex. Ambient dimension: any. Hessian: yes. Orientable: yes.
Example datafile declaration in 4D:
quantity formex energy method facet_2form_integral
form_integrand:
q1: x2 // 12 component
q2: 0
// 13 component
q3: x4 // 14 component
q4: 0
// 23 component
q5: 0
// 24 component
q6: x3*x2 // 34 component
facet_2form_sq_integral. Description: Integral of the square of a 2-form over a facet. Meant for ambient dimensions higher than 3. Uses Gaussian cubature of order integral_order_2D. Element: facet. Parameters: form_integrand
(components in lexicographic order). Models: linear. Ambient dimension: any. Hessian: no. Orientable: no. Example
datafile declaration in 4D:
space_dimension 4
// symplectic area
// Correspondence: z1 = (x1,x2) z2 = (x3,x4)
#define DENOM ((x1^2+x2^2+x3^2+x4^2)^2)
quantity symplectic_sq energy method facet_2form_sq_integral global
form_integrand:
q1: -2*(x3^2 + x4^2)/DENOM // dx1 wedge dx2 term
q2: 2*(x2*x3-x1*x4)/DENOM // dx1 wedge dx3 term
q3: 2*(x1*x3+x2*x4)/DENOM // dx1 wedge dx4 term
q4: -2*(x1*x3+x2*x4)/DENOM // dx2 wedge dx3 term
q5: 2*(x2*x3-x1*x4)/DENOM // dx2 wedge dx4 term
q6: -2*(x1^2 + x2^2)/DENOM // dx3 wedge dx4 term
facet_general_integral. Named method. Description: Integral of a scalar function of position and normal vector
over a 2D facet in any ambient dimension space. Uses Gaussian cubature of order integral_order_2D. The 2D facet
is represented by a 2-vector. The components of the 2-vector are named as xn, where the number n ranges from N+1
to N+N*(N+1)/2, continuing the coordinate indices. That is, in 4D the position coordinates are x1,x2,x3,x4 and the
2-vector components are x5 = x1 x2, x6 = x1 x3, x7 = x1 x4, x8 = x2 x3, x9 = x2 x4, and x10 = x3
x4. For proper behavior, the integrand should be homogeneous of degree 1 in the 2-form components. Element: facet.
Parameters: scalar_integrand. Models: linear, quadratic, Lagrange. Ambient dimension: 3. Hessian: yes. Example:
The facet area could be calculated with this quantity:
space_dimension 4
quantity surfacearea energy method facet_general_hi_d_integral
scalar_integrand: sqrt(x5^2 + x6^2 + x7^2 + x8^2 + x9^2 + x10^2)
facet_torus_volume. Description: For 3D soapfilm model, calculates body volume integral for a facet, with
corrections for edge wraps. Element: facet. Parameters: none. Models: linear,quadratic,lagrange. Ambient dimension:
3. Hessian: yes. Orientable: yes. Example datafile declaration:
199
200
quantity body_vol energy method facet_torus_volume
Description: Gravitational energy, integral z2 /2dx dy over a facet,

where is difference in adjacent body densities. Note: this method uses the gravitational constant as the modulus if
invoked as full_gravity_method. Just gravity_method does not automatically use the gravitational constant. Element:
facet. Parameters: none. Models: linear, quadratic, Lagrange. Ambient dimension: 3. Hessian: yes. Orientable: yes.
gravity_method, full_gravity_method.
quantity grav energy modulus 980*8.5 method gravity_method
facet_area_u, density_facet_area_u. Description: Area of facet. In quadratic model, it is an upper bound of

area, by the Schwarz Inequality. For the paranoid. Same as facet_area in linear model. Sets integral_order_2D to
6, since it doesnt work well with less. Using the density_facet_area_u name automatically incorporates the facet
tension, but facet_area_u doesnt. Element: facet. Parameters: none. Models: linear, quadratic. Ambient dimension:
any. Hessian: yes. Example datafile declaration:
quantity area_u energy method facet_area_u global
gap_energy. Description: Implementation of gap energy, which is designed to keep edges from short-cutting
curved constraint surfaces. This method serves the same purpose as declaring a constraint convex. Automatically
incorporates the gap constant set in the datafile or by the k command. Element: edge. Parameters: none. Models:
linear. Ambient dimension: any. Hessian: no. Example datafile declaration:
quantity gappy energy method gap_energy global
metric_facet_area. Description: For a Riemannian metric, this is the area of a facet. Element: edge. Parameters:
none. Models: linear,quadratic,simplex. Ambient dimension: any. Hessian: yes. Example datafile declaration:
metric
1+x^2 0 z
0 1+y^2 0
z 0 1+z^2
quantity mfa energy method metric_facet_area global
klein_area. Description: Facet area in Klein hyperbolic 3D space model. Does not depend on klein_metric being
declared in the datafile. Vertices should be inside the unit sphere. Element: facet. Parameters: none. Models: linear.
Ambient dimension: 3. Hessian: no. Example datafile declaration:
quantity kleinarea energy method klein_area global
dirichlet_area. Description: Same as the facet_tension method, but the Hessian is modified to be guaranteed
positive definite, after the scheme of Polthier and Pinkall [PP]. The energy is taken to be the Dirichlet integral of the
perturbation from the current surface, which is exactly quadratic and positive definite. Hence the hessian command
always works, but final convergence may be slow (no faster than regular iteration) since it is only an approximate
Hessian. Also see the dirichlet command. Element: facet. Parameters: none. Models: linear. Ambient dimension: any.
200
201
quantity dirarea energy method dirichlet_area global
sobolev_area. Description: Same as the facet_tension method, but the Hessian is modified to be guaranteed
positive definite, after the scheme of Renka and Neuberger. [RN]. Hence the hessian command always works, but
final convergence may be slow (no faster than regular iteration) since it is only an approximate Hessian. Also see
the sobolev command. Element: facet. Parameters: none. Models: linear. Ambient dimension: any. Hessian: yes.
quantity sobarea energy method sobolev_area global
pos_area_hess. Description: Same as the facet_area method, but the Hessian can be adjusted various ways by
setting the variables fgagfa_coeff, gfa_2_coeff, gfagfa_coeff, and gfaafg_coeff. This will make sense if you look at the
Dirichlet section of the Technical Reference chapter of the printed manual. The default values of the coefficients are
-1, 1, -1, and 0 respectively. Element: facet. Parameters: none. Models: linear. Ambient dimension: any. Hessian:
quantity parea energy method pos_area_hess global
spherical_area. Description: The spherical area of a triangle projected out to a unit sphere. The user must have
the vertices on the unit sphere. First meant for minimal surfaces in 4D that are to be mapped to surfaces of constant
mean curvature in 3D. Element: facet. Parameters: none. Models: linear. Ambient dimension: any. Hessian: no.
constraint 1 formula: x^2 + y^2 + z^2 = 1
quantity spharea energy method spherical_area global
stokes2d. Description: Square of the Laplacian of z viewed as a function of (x, y). Meant for the calculation
of two-dimensional Stokes flow of a fluid (i.e. slow steady-state flow where inertia is not significant) by having the
Evolver surface be the graph of the velocity potential and minimizing the viscous dissipation, which is the square of
the Laplacian of z. Boundary conditions are handled by declaring a vertex attribute "stokes_type" of type integer, and
assigning each boundary vertex one of these values:
0 - vertex is not on a wall; treat as if on a mirror symmetry plane.
1 - vertex is on a slip wall.
2 = vertex is on a nonslip wall; normal derivative of potential is zero.
Boundary values of z should be set to constants between 0 and 1 on various sections of boundary that represent
walls.
Element: vertex. Parameters: none. Models: linear. Ambient dimension: 3. Hessian: yes. Example datafile
declaration:
quantity dissip energy method stokes2d global
stokes2d_laplacian. Description: The Laplacian of z viewed as a function of (x, y). This is auxiliary to the
stokes2d method. It is the same Laplacian, unsquared, with the same boundary rules. Meant for calculating pressures
and such after stokes2d energy has been minimized. Element: vertex. Parameters: none. Models: linear. Ambient
dimension: 3. Hessian: yes. Example datafile declaration:

quantity laplac info\_only method stokes2d\_laplacian global
201
15.6.4
202
Surface curvature methods
These methods implement various types of surface curvatures.

mean_curvature_integral. Description: Integral of signed scalar mean curvature of a 2D surface. The computation is exact, in the sense that for a polyhedral surface the mean curvature is concentrated on edges and singular there,
but the total mean curvature for an edge is the edge length times its dihedral angle. Element: edge. Parameters: none.
Models: linear. Ambient dimension: any. Hessian: yes. Example datafile declaration:
quantity mci energy method mean_curvature_integral
The method mean_curvature_integral_a does the same thing, but uses a numerical formulation which may be better
behaved.
sq_mean_curvature.
Description: Integral of squared mean curvature of a surface. There are several methods implemented for calculating the integral of the squared mean curvature of a surface. The older
methods sq_mean_curvature, eff_area_sq_mean_curvature, and normal_sq_mean_curvature, are now deprecated,
since they dont have Hessians and the newer methods star_sq_mean_curvature, star_eff_area_sq_mean_curvature,
star_normal_sq_mean_curvature, and my current favorite star_perp_sq_mean_curvature, do have Hessians and can
now handle incomplete facet stars around vertices. But read the following for general remarks on squared curvature
also. The integral of squared mean curvature in the soapfilm model is calculated as follows: Each vertex v has a
star of facets around it of area Av . The force due to surface tension on the vertex is
Fv =
Av
.
v
(15.3)
Since each facet has 3 vertices, the area associated with v is Av /3. Hence the average mean curvature at v is
hv =
1 Fv
,
2 Av /3
(15.4)
and this vertexs contribution to the total integral is

Ev = h2v Av /3 =
1 Fv2
.
4 Av /3
(15.5)
Philosophical note: The squared mean curvature on a triangulated surface is technically infinite, so some kind of
approximation scheme is needed. The alternative to locating curvature at vertices is to locate it on the edges, where
it really is, and average it over the neighboring facets. But this has the problem that a least area triangulated surface
would have nonzero squared curvature, whereas in the vertex formulation it would have zero squared curvature.
Practical note: The above definition of squared mean curvature seems in practice to be subject to instablities. One
is that sharp corners grow sharper rather than smoothing out. Another is that some facets want to get very large at the
expense of their neighbors. Hence a couple of alternate definitions have been added.
Curvature at boundary: If the edge of the surface is a free boundary on a constraint, then the above calculation
gives the proper curvature under the assumption the surface is continued by reflection across the constraint. This
permits symmetric surfaces to be represented by one fundamental region. If the edge of the surface is a fixed edge or
on a 1-dimensional boundary, then there is no way to calculate the curvature on a boundary vertex from knowledge
of neighboring facets. For example, the rings of facets around the bases of a catenoid and a spherical cap may be
identical. Therefore curvature is calculated only at interior vertices, and when the surface integral is done, area along
the boundary is assigned to the nearest interior vertex. However, including ignore_fixed or ignore_constraints in the
method declaration will force the calculation of energy even at fixed points or ignoring constraints respectively.
If the parameter h_zero is defined, then the value per vertex is the same as for the following method,
eff_area_sq_mean_curvature.
Element: vertex. Parameters: ignore_constraints, ignore_fixed. Models: linear. Ambient dimension: any. Hessian:
no. Example datafile declaration:
202
203
quantity sqc energy method sq_mean_curvature global
eff_area_sq_mean_curvature. Description: Integral of squared mean curvature of a surface, with a slightly

different definition from sq_mean_curvature or normal_sq_mean_curvature. The area around a vertex is taken to be
the magnitude of the gradient of the volume. This is less than the true area, so makes a larger curvature. This also
eliminates the spike instability, since a spike has more area gradient but the same volume gradient. Letting Nv be the
volume gradient at vertex v,
1 Fv
hv =
,
(15.6)
2 ||Nv ||/3
and
3 Fv2
Av .
4 ||Nv ||2
Ev = h2v Av /3 =
(15.7)
The facets of the surface must be consistently oriented for this to work, since the evolver needs an inside and
outside of the surface to calculate the volume gradient. There are still possible instabilities where some facets grow
at the expense of others.
If the parameter h_zero is defined, then the value per vertex is
(
)2
Fv Nv
2
Ev = (hv h0 ) Av /3 = 3
h0 Av .
(15.8)
2||Nv ||2
This does not reduce to the non-h_zero formula when h_zero has the value zero, but users should feel lucky to have
any h_zero version at all.
If the vertex is on one or several constraints, the Fv and Nv are projected to the constraints, essentially making the
constraints act as mirror symmetry planes. A constraint that should not be considered as a mirror plane should be
given the attribute nonwall in its declaration.
WARNING: For some extreme shapes, Evolver may have problems detecting consistent local surface orientation.
The assume_oriented toggle lets Evolver assume that the facets have been defined with consistent local orientation.
Element: vertex. Parameters: none. Models: linear. Ambient dimension: any. Hessian: no. Example datafile
declaration:
quantity effsq energy method eff_area_sq_mean_curvature global
normal_sq_mean_curvature.
Description: Integral of squared mean curvature of a surface, with a
slightly different definition from sq_mean_curvature or eff_area_sq_mean_curvature. To alleviate the instability of
eff_area_sq_mean_curvature, normal_sq_mean_curvature considers the area around the vertex to be the component of
the volume gradient parallel to the mean curvature vector, rather than the magnitude of the volume gradient. Thus
1 Fv2
2 Nv Fv /3
[
]
3 Fv Fv 2
Ev =
Av .
4 Nv Fv
hv =
(15.9)
(15.10)
This is still not perfect, but is a lot better.

[
3 Fv2
Ev = (hv h0 ) Av /3 =
h0
2 Nv Fv
2
]2
Av
3
(15.11)
constraints act as mirror symmetry planes.
203
204
declaration:
quantity nsq energy method normal_sq_mean_curvature global
mix_sq_mean_curvature. Description: Integral of squared mean curvature of a surface, with a combination of

formulas for approximating mean curvature:
hv =
1 Fv Nv
1 Fv2
q+
(1 q)
2 Nv Fv /3
2 Nv Nv /3
(15.12)
where Fv is the force or area gradient at vertex v, N is the unit normal as determined by the volume gradient, and q is
the value of the user-defined variable cf sq_mean_mix. Then the energy is
Ev = h2v A/3
(15.13)
where A is the total area adjacent to a facet.

Ev = (hv h0 )2 Av /3.
(15.14)
constraints act as mirror symmetry planes.
declaration:
parameter mean_sq_mix = 0.5
quantity mixsq energy method mix_sq_mean_curvature global
star_sq_mean_curvature. Description: Integral of squared mean curvature over a surface. This is a different
implementation of sq_mean_curvature which is more suitable for parallel calculation and has a Hessian. But it assumes
a closed surface, i.e. each vertex it is applied to should have a complete star of facets around it. This method does not
use the h_zero parameter.
The curvature calculation works in any dimension space. If for some reason the space has an ambient dimension
greater than 3, and you want to restrict the calculation of curvature to the first three coordinates, the toggle calculate_in_3d will do that.
Element: vertex. Parameters: none. Models: linear. Ambient dimension: any. Hessian: yes. Example datafile
declaration:
quantity starsq energy method star_sq_mean_curvature global
star_eff_area_sq_mean_curvature. Description: Integral of squared mean curvature over a surface. This is a

different implementation of eff_area_sq_mean_curvature which is more suitable for parallel calculation and has a
Hessian. But assumes a closed surface, i.e. each vertex it is applied to should have a complete star of facets around it.
This method does not use the h_zero parameter.
204
205
declaration:
quantity seffsq energy method star_eff_area_sq_mean_curvature global
star_normal_sq_mean_curvature. Description: Integral of squared mean curvature over a surface. This is

a different implementation of normal_sq_mean_curvature which is more suitable for parallel calculation and has a
Hessian. But it assumes a closed surface, i.e. each vertex it is applied to should have a complete star of facets around
it. This method can use h_zero.
declaration:
quantity stnsq energy method star_normal_sq_mean_curvature global
star_perp_sq_mean_curvature. Description: Integral of squared mean curvature over a surface. This is my

current favorite implementation of squared mean curvature. It is an implementation specifically designed to agree
with the mean curvature computed as the gradient of area when normal motion is on (either the normal_motion toggle
for g iteration, or Hessian with hessian_normal). Thus if you get zero squared mean curvature with this method,
then switch to area energy, the hessian will report exact convergence. Likewise if you do prescribed curvature and
then convert to area minimization with a volume constraint. This method has a Hessian. This method does not require
a complete circle of vertices around a vertex; boundary edges are treated as if they are on mirror symmetry planes,
which is usually true. This method can use the h_zero parameter or vertex attribute for prescribed mean curvature.
The actual formula for the energy at a vertex is
hv =
1 Fv Nv
2 Nv Nv /3
[
Ev = (hv h0 )2 Av /3 =
3 Fv Fv
h0
2 Nv Fv
(15.15)
]2
Av
3
(15.16)
where Fv is the area gradient at the vertex, Nv is the volume gradient, and Av is the area of the adjacent facets. If the
vertex is on one or several constraints, then Fv and Nv are projected to the constraints, essentially making the constraints
act as mirror symmetry planes. The positive orientation of the surface is determined by the positive orientation of the
first facet of the vertexs internal facet list.
declaration:
quantity stnsq energy method star_normal_sq_mean_curvature global
gauss_curvature_integral. Description: This computes the total Gaussian curvature of a surface with boundary.
The Gaussian curvature of a polyhedral surface may be defined at an interior vertex as the angle deficit of the adjacent
205
206
angles. But as is well-known, total Gaussian curvature can be computed simply in terms of the boundary vertices,
which is what is done here. The total Gaussian curvature is implemented as the total geodesic curvature around the
boundary of the surface. The contribution of a boundary vertex is
E = ( i ) .
(15.17)
The total over all boundary vertices is exactly equal to the total angle deficit of all interior vertices plus 2, where
is the Euler characteristic of the surface. For reasons due to the Evolvers internal architecture, the sum is actually
broken up as a sum over facets, adding the vertex angle for each facet vertex on the boundary and subtracting for
each boundary edge. Boundary vertices are deemed to be those that are fixed or on a parametric boundary. Alternately,
one may define a vertex extra attribute gauss_bdry_v and an edge extra attribute gauss_bdry_e and set them nonzero
on the relevant vertices and edges; this overrides the fixed/boundary criterion. Element: facet. Parameters: none.
Models: linear. Ambient dimension: any. Hessian: no. Example datafile declaration:
quantity gint energy method gauss_curvature_integral global
star_gauss_curvature Description: Computes the angle deficit around vertices to which this method is applied.
The angle deficit is 2*pi minus the sum of all the adjacent angles of facets. No compensation is made for vertices on
the boundary of a surface; you just get big deficits there. Deficits are counted as positive, following the convention
for gaussian curvature. Element: vertex. Parameters: none. Models: linear. Ambient dimension: any. Hessian: no.
quantity total_deficit energy method star_gauss_curvature global
sq_gauss_curvature. Description: Computes the integral of the squared Gaussian curvature. At each vertex, the
Gaussian curvature is calculated as the angle defect divided by one third of the total area of the adjacent facets. This is
then squared and weighted with one third of the area of the adjacent facets. This method works only on closed surfaces
with no singularities due to the way it calculates the angle defect. Element: vertex. Parameters: none. Models: linear.
Ambient dimension: any. Hessian: no. Example datafile declaration:
quantity sqg energy method sq_gauss_curvature global
circle_willmore. Description: Alexander Bobenkos circle-based discrete Willmore energy, which is conformally
invariant. At each vertex, energy is (sum of the angles between facet circumcircles) - 2*pi. More simply done as edge
quantity, since angles at each end are the same. For edge e, if adjacent facet edge loops are a,e,d and b,c,-e, then circle
angle for edge has
cos() = (< a, c >< b, c > < a, b >< c, d > < b, c >< d, a >)/|a|/|b|/|c|/|d|
(15.18)
For now, assumes all vertices are faceted, and fully starred. Severe numerical difficulties: Not smooth when angle
beta is zero, which is all too common. Set of zero angles should be codimension 2, which means generally avoided,
but still crops up. Element: edge. Parameters: none. Models: linear. Ambient dimension: 3. Hessian: no. Example
quantity bobenko energy method circle_willmore global
laplacian_mean_curvature. Description: Calculates the velocity of a vertex as the Laplacian of the mean curvature of the surface, meant to model the surface diffusion of atoms in sintering. The mean curvature at each vertex
206
207
is calculated as a scalar, in the same way as for area_normalized area gradient, i.e. area gradient dotted with volume
gradient, divided by the area of the surrounding facets. Then finite differences are used to calculate the Laplacian of
the mean curvature. This calculates velocity only; the energy is always 0. This method should only be used with fixed
scale in the g command.
The relative speed of vertices can be controlled by the vertex attribute lmc_mobility, which the user should declare
if wanted. If the user wants to access the values of mean curvature the method finds, the user should define the vertex
scalar attribute lmc_mean_curvature. This method conserves volume ideally, but you might want to put on volume
constraints anyway due to numerical inaccuracies.
Warning: This method should only be used with a fixed g scale factor. And for stability, the factor should be
proportional to the fourth power of the shortest edge, since Laplacian of mean curvature is a fourth-derivative operator,
something like 0.001*length4 . This can make for very slow evolution for highly refined surfaces.
declaration:
area_method_name "null_area"
define facet attribute lmc_mobility real
define facet attribute lmc_mean_curvature real
quantity lmc energy method laplacian_mean_curvature global
15.6.5
General dimension methods
These methods are meant for the simplex model in any dimension.
simplex_vector_integral. Description: Integral of a vectorfield over a (n-1)-dimensional simplicial facet in nspace. Vectorfield is dotted with normal of facet; actually the side vectors of the simplex and the integrand vector are
formed into a determinant. Element: facet. Parameters: vector_integrand. Models: simplex. Ambient dimension: any.
Hessian: no. Orientable: yes. Example datafile declaration, for 4-volume under a 3D surface in 4D:
quantity xvint energy method simplex_vector_integral
vector_integrand:
q1: 0
q2: 0
q3: 0
q4: x4
simplex_k_vector_integral. Description: Integral of a simple (n-k)-vector over an oriented k-dimensional simplicial facet in n-space. The vector integrand lists the components of each of the k vectors sequentially. Evaluation is
done by forming a determinant whose first k rows are k vectors spanning the facet, and last (n-k) rows are vectors of
the integrand. Element: facet. Parameters: k_vector_order, vector_integrand. Models: simplex. Ambient dimension:
any. Hessian: yes. Orientable: yes. Example datafile declaration, for 3D surface in 5D:
quantity kvec energy method simplex_k_vector_integral
k_vector_order 3
vector_integrand:
q1: 0 // first vector
q2: 0
q3: 0
q4: 0
q5: x4
207
208
q6: 0 // second vector

q7: 0
q8: 0
q9: x3
q10: 0
edge_k_vector_integral. Description: Integral of a simple (n-k)-vector over an oriented k-dimensional simplicial

edge in n-space. The vector integrand lists the components of each of the k vectors sequentially. Evaluation is done
by forming a determinant whose first k rows are k vectors spanning the edge, and last (n-k) rows are vectors of the
integrand. Element: edge. Parameters: k_vector_order, vector_integrand. Models: linear, quadratic, simplex. Ambient
dimension: any. Hessian: yes. Orientable: yes. Example datafile declaration, for 3D edges of a 4D surface in 5D:
quantity kvec energy method edge_k_vector_integral
k_vector_order 3
vector_integrand:
q1: 0 // first vector
q2: 0
q3: 0
q4: 0
q5: x4
q6: 0 // second vector
q7: 0
q8: 0
q9: x3
q10: 0
15.6.6
Knot energy methods
These are a plethora of energies that various users have requested for knotted curves.
knot_energy. Description: An electrostatic energy in which vertices are endowed with equal charges. Inverse
power law of potential is adjustable via the global parameter knot_power, default value 2 (which is not electrostatic,
but the knot theorists like it). If the extra attribute node_charge is defined for vertices, then that value is used for the
vertex charge. Use of this energy is not restricted to knots; it has been used to embed complicated network graphs in
space. Element: vertex. Parameters: none. Models: linear. Ambient dimension: any. Hessian: yes. Example datafile
declaration:
parameter knot_power 2 // the default
quantity knotten energy method knot_energy global
uniform_knot_energy, edge_knot_energy. Description: A knot energy where vertex charge is proportional to

neighboring edge length. This simulates an electrostatic charge uniformly distributed along a wire. Inverse power law
of potential is adjustable via the global parameter knot_power (default 2). Element: vertex. Parameters: none. Models:
quantity knotten energy method uniform_knot_energy global
208
209
uniform_knot_energy_normalizer. Description: Supposed to approximate the part of uniform_knot_energy that

is singular in the continuous limit. Element: vertex. Parameters: Models: linear. Ambient dimension: any. Hessian:
quantity knottenorm energy method uniform_knot_energy global
method uniform_knot_energy_normalizer global
uniform_knot_normalizer1.
Description: Calculates internal knot energy to normalize singular divergence of integral of uniform_knot_energy.
Actually a synonym for uniform_knot_energy_normalizer. No gradient. Element: vertex. Parameters: none. Models:
linear. Ambient dimension: 3. Hessian: no. Example datafile declaration:
method uniform_knot_energy_normalizer1 global
uniform_knot_normalizer2. Description: Calculates internal knot energy to normalize singular divergence of

integral of uniform_knot_energy a different way from uniform_knot_energy_normalizer. Element: edge. Parameters:
none. Models: linear. Ambient dimension: 3. Hessian: no. Example datafile declaration:
method uniform_knot_energy_normalizer2 global
edge_edge_knot_energy, edge_knot_energy.
Description: Between pairs of edges, energy is inverse
square power of distance between midpoints of edges. Can also be called just edge_knot_energy. See also
edge_knot_energy_normalizer. (by John Sullivan) Element: edge. Parameters: none. Models: linear. Ambient
dimension: any. Hessian: no. Example datafile declaration:
quantity knotten energy method edge_edge_knot_energy global
edge_knot_energy_normalizer. Description: Calculates internal knot energy to normalize singular divergence

of integral of edge_edge_knot_energy. Element: edge. Parameters: none. Models: linear. Ambient dimension: 3.
quantity knotten energy method edge_edge_knot_energy global
method edge_knot_energy_normalizer global
simon_knot_energy_normalizer. Description: Another normalization of edge_knot_energy, which I dont feel

like deciphering right now. Element: edge. Parameters: none. Models: string linear. Ambient dimension: 3. Hessian:
quantity kenergy energy method edge_knotenergy global
method simon_knot_energy_normalizer global
209
210
facet_knot_energy.
Description: Charge on vertex is proportional to area of neighboring facets. Meant
for knotted surfaces in 4D. Power law of potential is adjustable via the global parameter knot_power. See also
facet_knot_energy_fix. Element: vertex. Parameters: none. Models: linear. Ambient dimension: any. Hessian:
quantity knotten energy method facet_knot_energy global
facet_knot_energy_fix. Description: Provides adjacent vertex correction to facet_knot_energy. Element: vertex.

Parameters: none. Models: linear. Ambient dimension: any. Hessian: no. Example datafile declaration:
quantity knotten energy method facet_knot_energy global
method facet_knot_energy_fix global
bi_surface. Description:
Named method of the knot energy family. Double integral over surface, i.e. all pairs of vertices weighted with
adjacent facet areas. Adapted from facet_knot_energy. Uses an arbitrary formula for energy, a function of the vector
between vertices, instead of just power rule. The formula is given by the scalar_integrand in datafile definition. The
vertex pairs it is evaluated over can be controlled. If the vertex integer attribute bi_surface_attr is defined, only those
with different values of bi_surface_attr will be included; otherwise all pairs are included. Element: vertex. Parameters:
none. Models: linear. Ambient dimension: any. Hessian: no. Example datafile declaration:
quantity qqqq energy method bi_surface
scalar_integrand x^2 + y^2 + z^2
buck_knot_energy. Description: Energy between pair of edges given by formula suggested by Greg Buck.
Power law of potential is adjustable via the global parameter knot_power. Element: edge. Parameters: none. Models:
quantity knotten energy method buck_knot_energy global
proj_knot_energy. Description: This energy is due to Gregory Buck. It tries to eliminate the need for a normalization term by projecting the energy to the normal to the curve. Its form is
Ee1 e2 =
L1 L2 cos p
|x1 x2 | p
(15.19)
where x1 , x2 are the midpoints of the edges and is the angle between the normal plane of edge e1 and the vector
x1 x2 . The default power is 2. Power law of potential is adjustable via the global parameter knot_power. Element:
edge. Parameters: none. Models: linear. Ambient dimension: any. Hessian: no. Example datafile declaration:
quantity knotten energy method proj_knot_energy global
210
211
circle_knot_energy. Description: This energy is due to Peter Doyle, who says it is equivalent in the continuous
case to the insulating wire with power 2. Its form is
Ee1 e2 =
L1 L2 (1 cos )2
,
|x1 x2 |2
(15.20)
where x1 , x2 are the midpoints of the edges and is the angle between edge 1 and the circle through x1 tangent to edge
2 at x2 . Only power 2 is implemented. Element: edge. Parameters: none. Models: linear. Ambient dimension: any.
quantity knotten energy method circle_knot_energy global
sphere_knot_energy.
Description: This is the 2D surface version of the circle energy. Its most general form is
E f1 f2 =
A1 A2 (1 cos ) p
,
|x1 x2 |q
(15.21)
where A1 , A2 are the facet areas, x1 , x2 are the barycenters of the facets, and is the angle between f1 and the sphere
through x1 tangent to f2 at x2 . The energy is conformally invariant for p = 1 and q = 4. For p = 0 and q = 1, one
gets electrostatic energy for a uniform charge density. Note that facet self-energies are not included. For electrostatic
energy, this is approximately 2.8A3/2 per facet.
The powers p and q are Evolver variables surface_knot_power and surface_cos_power respectively. The defaults
are p = 1 and q = 4. Element: facet. Parameters: none. Models: linear. Ambient dimension: any. Hessian: no.
parameter surface_knot_power 1 // the default
parameter surface_cos_power 4 // the default
quantity knotten energy method sphere_knot_energy global
sin_knot_energy. Description: Another weird way to calculate a nonsingular energy between midpoints of pairs
of edges. (by John Sullivan) Element: edge. Parameters: none. Models: linear. Ambient dimension: any. Hessian:
quantity knotten energy method sin_knot_energy global
curvature_binormal. Description: For string model. Evaluates to zero energy, but the force calculated is the mean
curvature vector rotated to the binormal direction. Element: vertex. Parameters: none. Models: linear. Ambient
quantity curbi energy method curvature_binormal global
ddd_gamma_sq. Description: Third derivative of curve position as function of arclength, squared. Element:
vertex. Parameters: none. Models: string, linear. Ambient dimension: 3. Hessian: no. Example datafile declaration:
quantity ddd energy method ddd_gamma_sq global
211
212
edge_min_knot_energy. Description: Between pairs of edges, energy is inverse square power of distance between
closest points of edges.

1
|e1 ||e2 |
(15.22)
d2
This should be roughly the same as edge_edge_knot_energy, but distances are calculated from edge midpoints there.
This is not a smooth function, so we dont try to compute a gradient. DO NOT use as an energy; use just for info_only
quantities. Element: edge. Parameters: none. Models: linear. Ambient dimension: 3. Hessian: no. Example datafile
declaration:
Energy =
quantity eminknot info_only method edge_min_knot_energy global
true_average_crossings. Description: Calculates the average crossing number of an edge with respect to all
other edges, averaged over all projections. Knot stuff. No gradient, so use just in info_only quantities. Element: edge.
Parameters: none. Models: linear. Ambient dimension: 3. Hessian: no. Example datafile declaration:
quantity true_cross info_only method true_average_crossings global
true_writhe. Description: For calculating the writhe of a link or knot. No gradient, so use just in info_only
quantities. Element: edge. Parameters: none. Models: linear. Ambient dimension: 3. Hessian: no. Example datafile
declaration:
quantity twrithe info_only method true_average_crossings global
twist. Description: Another average crossing number calculation. No gradient, so use just in info_only quantities.
Element: edge. Parameters: none. Models: linear. Ambient dimension: 3. Hessian: no. Example datafile declaration:
quantity twister info_only method twist global
writhe. Description: An average crossing number calculation. This one does have a gradient. Suggested by
Hermann Gluck. Programmed by John Sullivan. Between pairs of edges, energy is inverse cube power of distance
between midpoints of edges, times triple product of edge vectors and distance vector.
E = 1/d 3 (e1, e2, d)
(15.23)
Element: edge. Parameters: none. Models: linear. Ambient dimension: 3. Hessian: no. Example datafile declaration:
quantity writhy energy method writhe global
curvature_function. Description: Calculates forces as function of mean and Gaussian curvatures at vertices.
Function may be changed by user by altering teix.c. No energy, just forces. Element: vertex. Parameters: none.
Models: linear. Ambient dimension: any. Hessian: no. Example datafile declaration:
quantity curfun energy method curvature_function global
average_crossings. Description: To calculate the average crossing number in all projections of a knot. (by John
Sullivan) Element: edge. Parameters: none. Models: linear. Ambient dimension: 3. Hessian: no. Example datafile
declaration:
212
213
quantity across energy method average_crossings global
knot_thickness. Description: Calculates global radius of curvature at one vertex v, as the minimum radius of circle
containing the vertex and the endpoints of any non-adjacent edge. Because of the minimum, this has no gradient, so
should be used in info_only quantities. Element: vertex. Parameters: none. Models: linear. Ambient dimension: 3.
Gradient: no. Hessian: no. Example datafile declaration:
quantity kthick info_only method knot_thickness global
knot_thickness_0. Description: Calculates global radius of curvature at one vertex, as Lp integral of radius of
curvature of circle containing the vertex and the endpoints of edges not adjacent to the vertex. Integrand raised to -p
power. The power p is taken from the global variable knot_power. No factor of length in integral. This method has a
gradient. Element: vertex. Parameters: none. Models: linear. Ambient dimension: 3. Hessian: no. Example datafile
declaration:
quantity kthick info_only method knot_thickness_0 global
knot_thickness_p. Description: purpose: calculates global radius of curvature at one vertex v, as Lp integral of
radius of curvature of v and endpoints of nonadjacent edges. Includes factors of length at v and w. This method has
a gradient. The power p is taken from the global variable knot_power. Element: vertex. Parameters: none. Models:
linear. Ambient dimension: 3. Hessian: no. Example datafile declaration:
quantity kthick info_only method knot_thickness_p global
knot_thickness_p2. Description: Calculates the global radius of curvature at one vertex v, as Lp integral of
r(v,w1,w2) over all vertices w. Here w1 and w2 are the two neighbors of vertex w. Includes factors of length at v
and w. This has not been extended to allow open arcs (valence 1 vertices). This method does have a gradient. The
power p is taken from the global variable knot_power. Element: vertex. Parameters: none. Models: linear. Ambient
quantity kthick info_only method knot_thickness_p2 global
</pre>
knot_thickness2. Description: calculates global radius of curvature at one vertex v, as the minimum radius of
circle containing the vertex and the neighbor vertices of any non-adjacent vertex. Because of the minimum, this has
no gradient, so should be used in info_only quantities. Element: vertex. Parameters: none. Models: linear. Ambient
dimension: 3. Gradient: no. Hessian: no. Example datafile declaration:
quantity kthick info_only method knot_thickness2 global
knot_local_thickness. Description: Calculates the radius of curvature at a vertex of the circle containing the vertex and its two neighbor vertices. Meant to investigate the radius at individual vertices. Element: vertex. Parameters:
none. Models: linear. Ambient dimension: 3. Gradient: no. Hessian: no. Example datafile declaration:
quantity klocalthick info_only method knot_local_thickness global
213
15.6.7
214
Elastic stretching energy methods
These are various elastic energies for two-dimensional surfaces.

dirichlet_elastic. Description: Calculate the Dirichlet elastic strain energy for facets, minimization of which
gives conformal mapping. Let S be Gram matrix of unstrained facet (dots of sides). Let Q be the inverse of S. Let F
be Gram matrix of strained facet. Let C = FQ, the linear deformation matrix. Then energy density is Tr(CCT ).
Each facet has an extra attribute array form_factors[3] = s11,s12,s22, which are the entries in S. That is, s11 =
dot(v2-v1,v2-v1), s12 = dot(v2-v1,v3-v1), and s22 = dot(v3-v1,v3-v1). If form_factor is not defined by the user, it
will be created by Evolver, and the initial facet shape will be assumed to be unstrained. Element: facet. Parameters:
none. Models: linear. Ambient dimension: 3. Hessian: yes. Example datafile declaration:
quantity dirich energy method dirichlet_elastic global
general_linear_elastic. Description: Calculates the nonisotropic linear elastic strain energy energy for facets.
Let A be the linear transformation from the unstrained shape to the strained shape. Then the Cauchy-Green strain
tensor is C = (AT A I)/2. Let S1 and S2 be the sides of the unstrained facet. Let W1 and W2 be the transformed facet
sides. Let F be the Gram matrix of strained facet. Define
S
W
= [S1 S2 ],
Q = S 1
= [W1W2 ] = AS
= W T W = ST AT AS
Then
AT A = QT FQ
C
= (QT FQ I)/2
The energy density is

1
Ci j Ki jkl Ckl
2 i
jkl
where Ki jkl is the full tensor of elastic constants. By using symmetries, this can be reduced to
E1 E3 E4
C11
(1/2)[C11C22C12 ] E3 E2 E5 C22
E4 E5 E6
C12
Each facet has an extra attribute elastic_coeff of size 6 containing (E1 , E2 , E3 , E4 , E5 , E6 ), and an extra attribute array
elastic_basis of size 2x2 containing ((s11, s12), (s21, s22)), which are the two sides of the unstrained facet. Note that
the Ei are defined with respect to the original sides as defined by the form factors, so it is up to you to make sure
everything works out right. Test carefully!!! The elastic_coeff attribute must be created and initialized by the user.
Element: facet. Parameters: none. Models: linear. Ambient dimension: 3. Hessian: yes. Example datafile declaration:
linear_elastic. Description: To calculate the linear elastic strain energy energy for facets based on the CauchyGreen strain matrix. Let S be Gram matrix of unstrained facet (dots of sides). Let Q be the inverse of S. Let F be Gram
matrix of strained facet. Let C = (FQ I)/2, the Cauchy-Green strain tensor. Let v be Poisson ratio. Then energy
density is
(1/2/(1 + v))(Tr(C2 ) + v (TrC)2 /(1 (dim 1) v)).
(15.24)
Each facet has extra attribute poisson_ratio and extra attribute array form_factors[3] = s11,s12,s22, which are the
entries in S. That is, s11 = dot(v2-v1,v2-v1), s12 = dot(v2-v1,v3-v1), and s22 = dot(v3-v1,v3-v1). If form_factors
214
215
is not defined by the user, it will be created by Evolver, and the initial facet shape will be assumed to be unstrained.
For another variation, see linear_elastic_B. For a version of this method that gives compression zero energy, see
relaxed_elastic. Element: facet. Parameters: none. Models: linear. Ambient dimension: 3. Hessian: yes. Example
quantity lastic energy method linear_elastic global
linear_elastic_B. Description: To calculate the nonisotropic linear elastic strain energy for facets. Let A be
the linear transformation from the unstrained shape to the strained shape. Then the Cauchy-Green strain tensor is
C = (AT A I)/2. Let S1 and S2 be the sides of the unstrained facet. Let W1 and W2 be the transformed facet sides. Let
F be the Gram matrix of strained facet. Define
S
W
F
= [S1 S2 ], Q = S1
= [W1W2 ] = AS
= WT W = ST AT AS
Then
AT A = QT FQ
C = (QT > FQ I)/2
The energy density is
(1/2)C i j K i jkl C kl 
where Ki jkl is the full tensor of elastic constants. By using symmetries, this can be reduced to
E1 E3 E4
C11
(1/2)[C11C22C12 ] E3 E2 E5 C22
E4 E5 E6
C12
Each facet has extra attribute elastic_coeff of size 6 containing E1 , E2 , E3 , E4 , E5 , E6 , and extra attribute array elastic_basis of size 2x2 containing [[s11 , s12 ], [s21 , s22 ]], which are thech facet has extra attribute elastic_coeff of size 6
containing (E1 , E2 , E3 , E4 , E5 , E6 ), and extra attribute array elastic_basis of size 2x2 containing s11,s12,s21,s22, which
are the two sides of the unstrained facet. Note that the Ei are defined with respect to the original sides as defined by the
form factors, so it is up to you to make sure everything works out right. Test carefully!!! The elastic_coeff attribute
must be created and initialized by the user.
Element: facet. Parameters: none. Models: linear. Ambient dimension: 3. Hessian: yes. Example datafile
declaration:
define facet attribute elastic_basis real[2][2]
define facet attribute elastic_coeff real[6]
quantity genlastic energy method general_linear_elastic global
linear_elastic_B. Description: A variation of the linear_elastic method. To calculate the linear elastic strain
energy energy for facets based on the Cauchy-Green strain matrix. Let S be Gram matrix of unstrained facet (dots of
sides). Let Q be the inverse of S. Let F be Gram matrix of strained facet. Let C = (FQ I)/2, the Cauchy-Green
strain tensor. Let v be Poisson ratio. Then energy density is
(1/2/(1 + v))(Tr(C2 ) + v (TrC)2 /(1 (dim 1) v)).
(15.25)
Here each facet has extra attribute poisson_ratio and each vertex has two extra coordinates for the unstrained position.
Hence the entire surface must be set up as five dimensional. For a version of this method that gives compression zero
energy, see relaxed_elastic_A. Element: facet. Parameters: none. Models: linear. Ambient dimension: 3. Hessian:
215
216
space_dimension 5
quantity lastic energy method linear_elastic_B global
relaxed_elastic, relaxed_elastic1, relaxed_elastic2. Description: A variation of the linear_elastic method.

Calculates the linear elastic strain energy energy for facets based on the Cauchy-Green strain matrix, with compression
counting for zero energy, simulating, say, plastic film. The effect is to permit wrinkling. Let S be Gram matrix of
unstrained facet (dots of sides). Let Q be the inverse of S. Let F be Gram matrix of strained facet. Let C = (FQ I)/2,
the Cauchy-Green strain tensor. Let v be Poisson ratio. Then energy density is
(1/2/(1 + v))(Tr(C2 ) + v (TrC)2 /(1 (dim 1) v)).
(15.26)
Here each facet has extra attribute poisson_ratio and each vertex has two extra coordinates for the unstrained position.
Hence the entire surface must be set up as five dimensional. The compression is detected by doing an eigenvalue
analysis of the strain tensor, and discarding any negative eigenvalues. The eigenvalues may be separately accessed by
the relaxed_elastic1_A (lower eigenvalue) and relaxed_elastic2_A (higher eigenvalue) methods, which are meant to
be used in info_only mode. For a sample datafile, see mylarcube.fe.For a version of this method that gives compression
zero energy, see relaxed_elastic_A. Element: facet. Parameters: none. Models: linear. Ambient dimension: 3.
space_dimension 5
quantity lastic energy method linear_elastic_B global
relaxed_elastic_A. Description: Calculates the linear elastic strain energy energy for facets based on the CauchyGreen strain matrix, with compression counting for zero energy, simulating, say, plastic film. The effect is to permit
wrinkling. Let S be the Gram matrix of unstrained facet (dots of sides). Let Q be the inverse of S. Let F be Gram
matrix of strained facet. Let C = (FQ I)/2, the Cauchy-Green strain tensor. Let v be Poisson ratio. Then the energy
is
(1/2/(1 + v))(Tr(C2 ) + v (TrC)2 /(1 (dim 1) v)).
(15.27)
Each facet has extra attribute poisson_ratio and extra attribute array form_factors[3] = s11,s12,s22, which are the entries in S. That is, s11 = dot(v2-v1,v2-v1), s12 = dot(v2-v1,v3-v1), and s22 = dot(v3-v1,v3-v1). If form_factors is
not defined by the user, it will be created by Evolver, and the initial facet shape will be assumed to be unstrained.
The compression is detected by doing an eigenvalue analysis of the strain tensor, and discarding any negative eigenvalues. The eigenvalues may Facets which are stressed in one or two dimensions can be separately counted by the
relaxed_elastic1_A (one stress direction, and one wrinkle direction) and relaxed_elastic2_A (two stressed directions)
methods, which are meant to be used in info_only mode. For a sample datafile, see mylarcube.fe. For a version of this
method that gives compression positive energy, see linear_elastic. Element: facet. Parameters: none. Models: linear.
Ambient dimension: 3. Hessian: yes. Example datafile declaration:
quantity lastic energy method relaxed_elastic_A global
SVK_elastic.
Description: SVK (Saint-Venant - Kirchhoff) potential. The facet energy is

/2 (tr(E))2 + (E : E) (3 + 2) ) tr(E)
where E = (C I)/2 is the Green-Lagrange Strain tensor, = T T 0 is the temperature deviation, and is the thermal
dilation coefficient. Needs real-valued facet attributes SVK_alpha, SVK_mu, SVK_lambda, and SVK_theta. Also
needs the facet attribute form_factors, decribed in the linear_elastic method. Written by Dr. Rabah Bouzidi. Element:
facet. Parameters: none. Models: linear. Ambient dimension: 3. Hessian: yes. Example datafile declaration:
216
217
define facet attribute SVK_alpha real

define facet attribute SVK_lambda real
define facet attribute SVK_mu real
define facet attribute SVK_theta real
define facet attribute form_factors real[3]
quantity svk energy method SVK_elastic global
15.6.8
Weird and miscellaneous methods
These methods dont really fit in any of the other categories.

wulff_energy. Description: Method version of Wulff energy. If Wulff filename is not given in top section of
datafile, then the user will be prompted for it. Element: facet. Parameters: none. Models: linear. Ambient dimension:
3. Hessian: no. Example datafile declaration:
wulff "crystal.wlf"
quantity wolf energy method wulff_energy global
area_square. Description: Energy of a facet is the square of the facet area. Element: facet. Parameters: none.
Models: linear. Ambient dimension: 3. Hessian: no. Example datafile declaration:
quantity asquare energy method area_square global
carter_energy. Description: Craig Carters energy. Given bodies B1 and B2 in R3 , define the energy
E=
B1 B2
This reduces to
E=
1
(3 p)(2 p) F
B
2
1
d 3 z2 d 3 z1
|z1 z2 | p
N1 N2
2 F1 B1

F2 F1
1
d 2 z1 d 2 z2 .
|z1 z2 | p2
(15.28)
(15.29)
And if we crudely approximate with centroids z1 and z2 ,

E=
1
(3 p)(2 p) F
B
2
2 F1 B1
A1 A2
,
|z1 z2 | p2
(15.30)
where A1 and A2 are unnormalized area vectors for the facets. The power p is set by the variable carter_power (default
6). Element: facet. Parameters: none. Models: linear. Ambient dimension: 3. Hessian: no. Example datafile
declaration:
parameter carter_power 6 // the default
quantity craig energy method carter_energy global
charge_gradient. Description: This energy is the gradient squared of the knot_energy method, assuming the
points are constrained to the unit sphere. Element: vertex. Parameters: none. Models: linear. Ambient dimension:
any. Hessian: no. Example datafile declaration:
217
218

quantity knotten energy method knot_energy global
johndust. Description: For all point pairs (meant to be on a sphere),
E = (pi asin(d/2))/d,
(15.31)
where d is chord distance. For point packing problems on the sphere. Element: vertex. Parameters: none. Models:
constraint 1 formula: x^2+y^2+z^2 = 1
quantity jms energy method johndust global
stress_integral. Description: Hmm. Looks like this one calculates integrals of components of a stress tensor.
The scalar_integrand value is set as an integer standing for which component to do. See the function stress_integral
in method3.c for details. Does not have a gradient, so should be used for just info_only quantities. Element: facet.
Parameters: scalar_integrand. Models: linear. Ambient dimension: 3. Hessian: no. Example datafile declaration:
quantity stressy info_only method stress_integral global
ackerman. Description: Not actually an energy, but a kludge to put inertia on vertices. Uses extra velocity
coordinates to represent vertex in phase space. Invocation actually transfers computed forces from space coordinates
to velocity coordinates, so forces become acceleration instead of velocity. Element: vertex. Parameters: none. Models:
quantity jeremy energy method ackerman global
218
Chapter 16
Technical Reference
This chapter explains the mathematics and algorithms used by the Evolver. You dont need to know this stuff to use
the Evolver, but its here for anybody who want to know what exactly is going on and for a programmers reference.
This chapter currently describes mostly the linear soapfilm model, although some of the named quantity methods
are done for quadratic model.
16.1 Notation
The term original orientation refers to the orientation of an element as recorded in its structure. Original orientations
begin with the orientations in the datafile, and are inherited thereafter during refinement and other operations.
Edges
In referring to a edge e:
t and h are the tail and head of the side.
s is the edge vector, s = h t.
ai and wi are the abscissas and weights for Gaussian quadrature on the interval [0,1].
Facets
In referring to a facet f :
T is the facet surface tension.
v0 , v1 , and v2 are the vertices around a facet in counterclockwise order.
s0 , s1 , and s2 are the edges in counterclockwise order around a facet, and v0 is the tail of s0 .
ai j and wi are the barycentric coordinates ( j is the vertex index) and weights for evaluation point i for Gaussian
quadrature on a triangle.
16.2 Surface representation

The basic geometric elements are vertices, edges, three-sided facets, and bodies. There is a structure type for each
(see skeleton.h for definitions). In addition, there is a facet-edge structure for each incident facet and edge pair. Edges,
facets, and facet-edges are oriented. Vertices and bodies could in principle be oriented, but they are not. Elements
can be referred to by identifiers which implicitly contain an orientation for the element. Currently these identifiers are
32-bit bitfields with one bit for an orientation bit, bits for element type, a bit for validity, and the rest for a list pointer.
Thus one structure is used for both orientations, and the programmer doesnt have to worry about the orientation of
the stored structure; all the orientation conversions are taken care of by the structure access functions and macros.
219
220
The topology of the surface is defined by the cross-references among the elements. Each facet-edge is in two
doubly-linked lists: one of facet-edges around the edge, and another of facet-edges around the facet. Each edge points
to its two endpoints (called head and tail) and to one facet-edge in the facet-edge loop around it. Each facet points
to one facet-edge in the facet-edge loop around it, and to the two bodies on its two sides. Each body points to one
facet-edge on its boundary. The upshot of all this is that it is easy to find all the facets around an edge or all the edges
around a facet. There are also various other linked lists connecting elements; see the Iterators section of the Operation
chapter.
16.3 Energies and forces

As described in the Energies section of the Model chapter, all energies are carried by facets or edges. (Volumes for
ideal gas energy are calculated from facets and edges also.) Hence it is straightforward to run through all the edges and
facets, adding up their energies and accumulating their energy gradients onto the total forces on their vertices. This
section gives the formulas used for all the types of energies and forces.
16.3.1
Surface tension
Energy of facet
E=
T
s0 s1 .
2
(16.1)
Force on vertex
F(v0 ) = T s1 (s0 s1 ) .
2 s0 s1
16.3.2
(16.2)
Crystalline integrand
is the Wulff vector corresponding to the normal of the facet. The facet has its original orientation.
W
Energy of facet
T
W s0 s1 .
2
(16.3)
F(v0 ) = T s1 W
.
2
(16.4)
E=
Force on vertex
16.3.3
Gravity
Let G be the acceleration of gravity. Let + be the density of the body from which the facet normal is outward, and let
be the density of the body with inward normal. Let zi be the z coordinate of vi . The following formulas are exact
for flat facets.
Energy of facet
E
= G(+ )
= G(+ )

f acet
1 2
z k dA
2
1
1 2 2 2
(z + z + z + z0 z1 + z0 z2 + z1 z2 ) (k s0 s1 ).
12 0 1 2
2
220
(16.5)
221
Force on vertex
F(v0 ) = G(+ ) ((2z0 + z1 + z2 )(k s0 s1 )k + (z20 + z21 + z22 + z0 z1 + z0 z2 + z1 z2 )(k s1 )).
24
16.3.4
(16.6)
Level set constraint integrals
This occurs for each constraint with energy integrand that an edge is deemed to satisfy. The edge orientation is the
original one. The points evaluated for Gaussian quadrature are
xi = aih + (1 ai )t.
Energy of edge
E=
edge
E ds
wiE(xi ) s.
(16.7)
(16.8)
Force on vertex
Component notation and the Einstein summation convention are used here, as vector notation gets confusing. Comma
subscript denotes partial derivative.
Fj (head) = wi (ai Ek, j (xi )sk + E j (xi )),
(16.9)
Fj (tail) = wi ((1 ai )Ek, j (xi )sk E j (xi )).
16.3.5
Gap areas
Every edge on a CONVEX constraint contributes here. Let qt be the projection of s on the constraint tangent space at
the tail, and qh likewise at the head. k is the gap constant.
Energy of edge
k
E=
12
(
)
2
2
2
2
2
2
(s qt )qt + (s qh )qh .
(16.10)
Force on vertex
This energy is not the direct derivative of the previous formula. Rather it comes directly from the geometric decrease
in gap area due to vertex motion. The formula is derived for the case of a flat gap surface, which should be close
enough to the truth to be useful.
2 2
2
F(tail) = k (s qt )qt qt ,
(16.11)
2q2
t
k (s2 q2h )q2h
F(head) =
qh .
2q2h
16.3.6
Ideal gas compressibility
The ideal gas mode is in effect when the keyword is listed in the first part of the datafile. Then prescribed volumes
become in essence moles of gas. Pamb is the ambient pressure. A body has prescribed volume V0 , actual volume V ,
and pressure P. The temperature is assumed constant, so PV = PambV0 .
221
222
Energy of body
P Pamb dV =
PambV0 /V Pamb dV
(16.12)
= PambV0 ln(V /V0 ) Pamb (V V0 )

Note the energy starts at 0 for V = V0 .
Force on vertex
Let gm be the gradient of the volume body m as a function of the coordinates of the vertex. Let Pm be the pressure of
body m. Then
F(v) = (Pm Pamb )gm .
(16.13)
m
16.3.7
Prescribed pressure
Ambient pressure is taken to be zero, and body pressures are constant.

Energy of body
E=
PdV = PV.
(16.14)
Force on vertex
Let gm be the gradient of the volume body m as a function of the coordinates of the vertex. Let Pm be the pressure of
body m. Then
F(v) = Pmgm .
(16.15)
m
16.3.8
Squared mean curvature
The integral of squared mean curvature in the soapfilm model is calculated as follows: Each vertex v has a star of
facets around it of area Av . The force due to surface tension on the vertex is
Fv =
Av
.
v
(16.16)
Since each facet has 3 vertices, the area associated with v is Av /3. Hence the average mean curvature at v is
hv =
1 Fv
,
2 Av /3
(16.17)
and this vertexs contribution to the total integral is

Ev = h2v Av /3 =
1 Fv2
.
4 Av /3
(16.18)
The corresponding calculation is done for the string model. Ev can be written as an exact function of the vertex
cooordinates, so the gradient of Ev can be fed into the total force calculation.
Philosophical note: The squared mean curvature on a triangulated surface is technically infinite, so some kind of
approximation scheme is needed. The alternative to locating curvature at vertices is to locate it on the edges, where
it really is, and average it over the neighboring facets. But this has the problem that a least area triangulated surface
would have nonzero squared curvature, whereas in the vertex formulation it would have zero squared curvature.
222
223
Practical note: The above definition of squared mean curvature seems in practice to be subject to instablities. One
is that sharp corners grow sharper rather than smoothing out. Another is that some facets want to get very large at the
expense of their neighbors. Hence a couple of alternate definitions have been added.
Effective area curvature: The area around a vertex is taken to be the magnitude of the gradient of the volume. This
is less than the true area, so makes a larger curvature. This also eliminates the spike instability, since a spike has more
area gradient but the same volume gradient. Letting Nv be the volume gradient at v,
hv =
1 Fv
,
2 ||Nv ||/3
(16.19)
and
3 Fv2
Av ,
(16.20)
4 ||Nv ||2
The facets of the surface must be consistently oriented for this to work, since the evolver needs an inside and
outside of the surface to calculate the volume gradient. This mode is toggled by the effective_area ON | OFF
command. There are still possible instabilities where some facets grow at the expense of others.
Normal curvature: To alleviate the instability of effective_area curvature, the normal_curvature mode considers
the area around the vertex to be the component of the volume gradient parallel to the mean curvature vector, rather
than the magnitude of the volume gradient. Thus
Ev = h2v Av /3 =
1 Fv ||Fv ||
,
(16.21)
2 Nv Fv /3
[
]
3 Fv Fv 2
Ev =
Av .
(16.22)
4 Nv Fv
This is still not perfect, but is a lot better. This mode is toggled by the normal_curvature ON | OFF command. If
you have effective area on, then normal curvature will supersede it, but effective area will be in effect when normal
curvature is toggled off.
Curvature at boundary: If the edge of the surface is a free boundary on a constraint, then the above calculation
gives the proper curvature under the assumption the surface is continued by reflection across the constraint. This
permits symmetric surfaces to be represented by one fundamental region. If the edge of the surface is a fixed edge or
on a 1-dimensional boundary, then there is no way to calculate the curvature on a boundary vertex from knowledge
of neighboring facets. For example, the rings of facets around the bases of a catenoid and a spherical cap may be
identical. Therefore curvature is calculated only at interior vertices, and when the surface integral is done, area along
the boundary is assigned to the nearest interior vertex.
WARNING: For some extreme shapes, effective_area and normal_curvature modes have problems detecting consistent local surface orientation. The assume_oriented toggle lets Evolver assume that the facets have been defined
with consistent local orientation.
hv =
16.3.9
Squared Gaussian curvature
The integral of squared Gaussian curvature over a soapfilm model surface only applies where the star of facets around
each vertex is planar. It is calculated as follows. The total Gaussian curvature at a vertex is the angle deficit around the
vertex, 2 i , where i are the vertex angles of the facets adjacent to the vertex. The average Gaussian curvature
is the total Gaussian curvature divided by one third of the area of the star. The average is then squared and integrated
over one third of the star.
At boundary: Treated the same way squared mean curvature is.
16.4 Named quantities and methods

This section gives details on the calculations of the methods used in named quantities. For those methods where exact
evaluation is impossible (due to integrals of functions or quadratic model), evaluation is done by Gaussian integration
223
224
with the number of evaluation points controlled by the integration_order variable. In the formulas below, edge are
parameterized by 0 u 1 and facets by 0 u1 1, 0 u2 1 u1 . S is the differential of the immersion into
space, Si j = xi /u j . J is the Jacobian, J = (det(ST S))1/2 . um are Gaussian integration points in the domain, xm are
the corresponding points in space, and wm are the Gaussian weights. For the linear model, S and J are constant, but
in the quadratic model they are functions of u: S(um ) and J(um ). Likewise for the facet normal N, which includes the
Jacobian factor.
16.4.1
Vertex value
Method name vertex_scalar_integral. A scalar function f (x) is evaluated at each vertex the method applies to.
16.4.2
Edge length
Method name edge_tension or edge_length. This method is entirely separate from the default energy calculation for
strings. For each edge, the value is the length in Euclidean coordinates. The edge density attribute, used in the default
energy calculation, is not used. No metric is used.
E = wm J(um )
(16.23)
However, the density_edge_length method does use the edge density :

E = wm J(um )
(16.24)
16.4.3
Facet area
Method name facet_tension or facet_area. This method is entirely separate from the default energy calculation for
surfaces. For each facet, the value is the area in Euclidean coordinates. The area density attribute, used in the default
energy calculation, is not used. No metric is used.
E = wm J(um )
(16.25)
However, the density_facet_area method does use the facet density :

E = wm J(um )
(16.26)
Method name facet_area_u. In the quadratic model, this gives an upper bound of area, for the paranoids who dont
trust the regular facet_area method. It uses the fact that
(
)1/2 (
)1/2

Area =
(det(ST S))1/2 du1 du2
det(ST S)du1 du2
du1 du2
(16.27)
The integrand on the right is a fourth degree polynomial, and so may be done exactly with 7-point Gaussian integration.
This method sets the integral order to at least 4.
Method name spherical_area. This is the area of the triangle projected out to the unit sphere, assuming the vertices
are all on the unit sphere. The calculation is done in terms of the edge lengths a, b, c (chord lengths, not arcs) by
calculating the angle deficit. The formula for angle C is
]
[
a2 b2 (1 a2 /4)(1 b2 /4)
.
(16.28)
acos
2(a2 + b2 c2 ab/2)
Note this formula avoids taking cross products (so it works in any ambient dimension), and avoids dot products of
vectors making a small angle.
224
16.4.4
225
Path integrals
Method name edge_scalar_integrand. A scalar integrand f (x) is evaluated on a edge by Gaussian quadrature:
E = wm f (xm )J(um ).
(16.29)
16.4.5
Line integrals
Method name edge_vector_integrand. A vector integrand f (x) is evaluated on a edge by Gaussian quadrature:
E = wm f (xm ) S(um )
(16.30)
16.4.6
Scalar surface integral
Method name facet_scalar_integral. A scalar integrand f (x) is evaluated on a facet by Gaussian quadrature:
E = wm f (xm )J(um )
(16.31)
16.4.7
Vector surface integral
Method name facet_vector_integral. A vector integrand f (x) is evaluated on a facet by Gaussian quadrature:
E = wm f (xi ) N(um )
(16.32)
For 3D only.
16.4.8
2-form surface integral
Method name facet_2form_integral. A 2-form integrand (xi ) is evaluated on a facet by Gaussian quadrature:
E = wm < A(um ), (xm ) >
(16.33)
where A is the 2-vector representing the facet, wi are the Gaussian weights, and xi are the evaluation points. For any
dimensional ambient space.
16.4.9
General edge integral
Method name edge_general_integral. A scalar function f (x,t ) of position and tangent direction is integrated over the
edge by Gaussian quadrature:
E = wm f (xm , S(um ))
(16.34)
m
For proper behavior, f should be homogeneous of degree 1 in t.
16.4.10
General facet integral
Method name facet_general_integral. A scalar function f (x,t) of position and normal direction is integrated over the
edge by Gaussian quadrature:
E = wm f (xm , N(um ))
(16.35)
m
For proper behavior, f should be homogeneous of degree 1 in t.

225
16.4.11
226
String area integral
Method name edge_area. This is the contribution of one edge to the area of a cell in the string model.
ydx
E=
(16.36)
It includes corrections for torus domains, along the same lines as explained below for facet_volume.
16.4.12
Volume integral
Method name facet_volume. This is the contribution of one facet to the volume of a body, using the Divergence
Theorem:

E=
zk N
(16.37)
This is done exactly, without the need for Gaussian integration. Note that this method has no automatic connection to
body volumes. If you want to use this to calculate body volumes, you must define a separate quantity for each body.
Further, it applies to the positive orientation of the facet, so if your body has negative facets, you must define a separate
method instance with modulus -1 for those facets. A little ugly, but maybe it will be cleaned up in a later version.
In a torus domain, wrap corrections are included. Assume the fundamental region is a unit cube. Otherwise, convert
to parallelpiped coordinates and multiply the resulting volume by the volume of the fundamenatal parallelpiped. For a
body B,

V=
1dxdydz
(16.38)
Confine the body to one fundamental region by introducing cut surfaces where the body wraps around. Let Mk be
the cut on the k period face of the fundamental region. Actually, one has two copies Mk and Mk+ , differing only by
translation by the kth period vector. By the Divergence Theorem,

zdxdy +
Mk
zdxdy +
zdxdy
+
Mz+
Mk
zdxdy
(16.39)
1dxdy.
The surface Mz+ is bounded by interior curves Mz+ and by cut lines Cx+ , Cx , Cy+ , and Cy . So

zdxdy +
Mz
zdxdy +
xdy
+
Mz+
Cx+
1dy
(16.40)
xdy + y+ y
where the positive and negative ys are at the endpoints of Cx+ . Note that all of the quantities in the last line can be
computed locally on each facet just from the edge wraps. The only ambiguity is in the values of the ys by units of 1.
So the final body volumes may be off by multiples of the torus volume. Careful set-up of the problem can avoid this.
16.4.13
Gravity
Method name gravity_method

This does the same calculation as the usual gravity, but not including incorporating body densities or the gravitational constant. The modulus should be set as the product of the body density and the gravitational constant. Let zi be
the z coordinate of vi . The following formulas are exact for flat facets.
E
f acet
1 2
z k dA
2
2 2 2
1
(z0 + z1 + z2 + z0 z1 + z0 z2 + z1 z2 ) (k s0 s1 ).
12
2
It is also exactly calculable in the quadratic model.
=
226
(16.41)
16.4.14
227
Hooke energy
Method name Hooke_energy. One would often like to require edges to have fixed length. The total length of some
set of edges may be constrained by defining a fixed quantity. This is used to fix the total length of an evolving knot,
for example. But to have one constraint for each edge would be impractical, since projecting to n constraints requires
inverting an n n matrix. Instead there is a Hookes Law energy available to encourage edges to have equal length. Its
form per edge is
E = |L L0 | p
(16.42)
where L is the edge length, L0 is the equilibrium length, embodied as an adjustable parameter hooke_length, and the
power p is an adjustable parameter hooke_power. The default power is p = 2, and the default equilibrium length
is the average edge length in the initial datafile. You will want to adjust this, especially if you have a total length
constaint. A high modulus will decrease the hooke component of the total energy, since the restoring force is linear in
displacement and the energy is quadratic (when p = 2). As an extra added bonus, a hooke_power of 0 will give
E = log |L L0 |.
(16.43)
To give each edge its own equilibrium length, use the hooke2_energy method. Each edge has an equilibrium length
extra attribute hooke_size.
To give each edge an energy according to an elastic model,
E=
1 L L0 )2
2 L0
(16.44)
use the hooke3_energy method. Each edge has an equilibrium length extra attribute hooke_size. The exponent can
be altered from 2 by setting the parameter hooke3_power.
16.4.15
Local Hooke energy
Method name local_hooke_energy. Energy of edges as springs with equilibrium length being average of lengths of
neighbor edges. Actually, the energy is calculated per vertex,
(
)
L1 L2 2
E=
(16.45)
L1 + L2
where L1 and L2 are the lengths of the edges adjacent to the vertex.
16.4.16
Integral of mean curvature
Method name mean_curvature_integral. This computes the integral of the signed scalar mean curvature (average of
sectional curvatures) over a surface. The computation is exact, in the sense that for a polyhedral surface the mean
curvature is concentrated on edges and singular there, but the total mean curvature for an edge is the edge length times
its dihedral angle. The contribution of one edge is
)
(
(a b)(a b) (a a)(b c)
,
(16.46)
E = L arccos
((a a)(b b) (a b)2 )1/2 ((a a)(c c) (a c)2 )1/2
where L is the edge length, a is the edge, and b,c are the two adjacent sides from the tail of a.
16.4.17
Integral of squared mean curvature
Method name sq_mean_curvature. This is the named method version of the integral of squared mean curvature
discussed earlier in this chapter. Including IGNORE_CONSTRAINTS in the method declaration will force the calculation of energy even at fixed points and ignoring constraints.
227
228
Method name eff_area_sq_mean_curvature. This is the named method version of the integral of squared mean
curvature discussed earlier in this chapter, with the effective area discretization of mean curvature.
Method name normal_sq_mean_curvature. This is the named method version of the integral of squared mean
curvature discussed earlier in this chapter, with the normal curvature discretization of mean curvature.
16.4.18
Integral of Gaussian curvature
Method name Gauss_curvature_integral. This computes the total Gaussian curvature of a surface with boundary. The
Gaussian curvature of a polyhedral surface may be defined at an interior vertex as the angle deficit of the adjacent
angles. But as is well-known, total Gaussian curvature can be computed simply in terms of the boundary vertices,
which is what is done here. The total Gaussian curvature is implemented as the total geodesic curvature around the
boundary of the surface. The contribution of a boundary vertex is
(
)
E=
(16.47)
The total over all boundary vertices is exactly equal to the total angle deficit of all interior vertices plus 2, where
is the Euler characteristic of the surface.
16.4.19
Average crossing number
Method name average_crossings. Between pairs of edges, energy is inverse cube power of distance between midpoints
of edges, times triple product of edge vectors and distance vector:
E = 1/d 3 (e1, e2, d).
16.4.20
(16.48)
Linear elastic energy
Method name linear_elastic.

Calculates a linear elastic strain energy for facets based on the Cauchy-Green strain matrix. Let S be Gram matrix
of the unstrained facet (dot productss of sides). Let Q be the inverse of S. Let F be Gram matrix of strained facet. Let
C = (FQ I)/2, the Cauchy-Green strain tensor. Let v be Poisson ratio. Then energy density is
(1/2/(1 + v))(Tr(C2 ) + v (TrC)2 /(1 (dim 1) v))
(16.49)
Each facet has extra attribute poisson_ratio and the extra attribute array form_factors[3] = { s11,s12,s22}. If
form_factor is not defined by the user, it will be created by Evolver, and the initial facet shape will be assumed to
be unstrained.
16.4.21
Knot energies
One way of smoothing a knotted curve is to put electric charge on it and let it seek its minimum energy position.
To prevent the curve from crossing itself and unknotting, the potential energy should have a high barrier (preferably
infinite) to curve crossing. This is often done by using a inverse power law potential with a higher power than the
standard inverse first power law of electrostatics. A length constraint on the curve is generally necessary to prevent
the charges from repelling each other off to infinity.
The Evolver implements several discrete approximations of these potentials. All use the quantity-method feature
described in the next section. That is, each type of energy is a method, and these methods can be attached to usernamed quantities. The power in the power law is an adjustable parameter knot_power, or perhaps some other variable
in specific cases. It may be changed interactively with the A command. The modulus is the multiple of the method
added to the quantity. A modulus of 0 inactivates the method. The value of the quantity can be seen with the A
command under the name of the quantity.
228
229
One of the general problems with discretization is that edges dont know when they are crossing each other. Edges
can cross without either their endpoints or midpoints getting close, especially if said edges get long. To keep edge
lengths close to equal, there is also an energy called hooke energy described in the Model chapter. Another energy
of interest is total squared curvature, which is elastic bending energy. However, it does not provide any barrier to
crossing.
All the knot energies proper correspond to double integrals in the continuous limit, so are double sums. Hence the
computation time is quadratic in the number of vertices. Hooke energy and square curvature are linear time.
Generally, the continuous integrals are divergent, and in the literature have various regularization terms subtracted
off. However, the Evolver discretizations do not have such regularization. The logic is that the discrete versions are
finite, and the discete regularization terms are practically constant, so there is no sense in wasting time computing
them.
A sample datafile incorporating some of these is knotty.fe.
These discrete knot energies are classified below by whether they are computed over all pairs of vertices, edges, or
facets.
I. Vertex pair energies. The total energy is the sum over all pairs of vertices v1 , v2 :
E=
1
Ev1 v2 .
2 v
=v
1
(16.50)
Ia. Conducting wire. There is a unit charge on each vertex, and edge lengths are not fixed.
Ev1 v2 =
1
|v1 v2 | p
This is approximating the continuous integral

)
(
1
1
E=
(s)(t) ds dt
|x(s) x(t)| p d(s,t) p
(16.51)
(16.52)
where s,t are arclength parameters, d(s,t) is the shortest arc distance between points, and is charge density. The
second term in this integral is a normalization term to prevent infinite energy. It is not present in the discrete version,
since the discrete energy is finite and the normalization term is approximately constant. Note that refining doubles the
total charge, so refining should be accompanied by reducing the modulus by a factor of 4.
Note that the discrete energy can apply to any dimension surface in any dimension ambient space since it does not
use the regularization. It can be used as a dust energy for sphere packing (high knot_power for a hard core potential)
or an energy for knotted 2-spheres in 4-space.
The default power is 2.
Datafile line:
quantity knot_energy ENERGY modulus 1 global_method knot_energy
Ib. Insulating wire.

Ev1 v2 =
L1 L2
|v1 v2 | p
(16.53)
where L1 is the average length of the two edges adjacent to v1 and L2 is the average length of the two edges adjacent
to v2 . This is approximating the continuous integral
)
(
1
1
E=
ds dt
(16.54)
|x(s) x(t)| p d(s,t) p
where s,t are arclength parameters, d(s,t) is the shortest arc distance between points. This corresponds to a uniform
charge density, which is why it is an insulating wire. The second term in the integral is a normalization term, which
is not included in the discrete version since it is approximately constant. However, a discrete normalization term can
be calculated as a separate quantity.
229
230
This energy assumes that each vertex has exactly 2 edges attached to it, so it is not suitable for surfaces.
Datafile line:
quantity knot_energy ENERGY modulus 1 global_method edge_knot_energy
The edge in the method name refers to the edge-weighting of the vertex charges. The default power is 2.
Datafile line for normalization term:
quantity norm_term INFO_ONLY modulus 1 global_method uniform_knot_energy_normalizer
Ic. Insulating surface.

Ev1 v2 =
A1 A2
|v1 v2 | p
(16.55)
where A1 is the area of the facets adjacent to v1 and A2 is the area of the facets adjacent to v2 . This corresponds to a
uniform charge density, which is why it is an insulating surface.
Datafile line:
quantity knot_energy ENERGY modulus 1 global_method facet_knot_energy
The facet in the method name refers to the edge-weighting of the vertex charges. The default power is 4.
II. Edge pair energies. The total energy is the sum over all pairs of edges e1 , e2 with endpoints v11 , v12 , v21 , v22 :
E=
1
Ee1 e2
2 e
=e
1
(16.56)
IIa. Box energy. This energy is due to Gregory Buck. It has the form
Ee1 e2 =
L1 L2
.
(d1 + d2 + d3 + d4 2(L1 + L2 )) p
(16.57)
Here L1 , L2 are the edge lengths and d1 , d2 , d3 , d4 are the distances between endpoints on different edges. This provides
an infinite barrier to crossing in the discrete case in a simple form involving only the edge endpoints. The denominator
becomes zero for parallel coincident lines or for perpendicular lines on opposite edges of a tetrahedron. This energy
should not be turned on until the curve is refined enough that the denominator is always positive. The default power is
2.
Datafile line:
quantity buck_energy ENERGY modulus 1 global_method buck_knot_energy
IIb. Normal projection energy. This energy is also due to Gregory Buck. It tries to eliminate the need for a
normalization term by projecting the energy to the normal to the curve. Its form is
Ee1 e2 =
L1 L2 cos p
|x1 x2 | p
(16.58)
where x1 , x2 are the midpoints of the edges and is the angle between the normal plane of edge e1 and the vector
x1 x2 . The default power is 2.
Datafile line:
quantity proj_energy ENERGY modulus 1 global_method proj_knot_energy
IIc. Conformal circle energy. This energy is due to Peter Doyle, who says it is equivalent in the continuous case
to the insulating wire with power 2. Its form is
Ee1 e2 =
L1 L2 (1 cos )2
,
|x1 x2 |2
(16.59)
where x1 , x2 are the midpoints of the edges and is the angle between edge 1 and the circle through x1 tangent to edge
2 at x2 . Only power 2 is implemented.
Datafile line:
230
231
quantity circle_energy ENERGY modulus 1 global_method circle_knot_energY
III. Facet pair energies. These energies are sums over all pairs of facets.
IIIa. Conformal sphere energy. This is the 2D surface version of the conformal circle energy. Its most general
form is
A1 A2 (1 cos ) p
E f1 f2 =
,
(16.60)
|x1 x2 |q
where A1 , A2 are the facet areas, x1 , x2 are the barycenters of the facets, and is the angle between f1 and the sphere
through x1 tangent to f 2 at x2 . The energy is conformally invariant for p = 1 and q = 4. For p = 0 and q = 1, one
gets electrostatic energy for a uniform charge density. Note that facet self-energies are not included. For electrostatic
energy, this is approximately 2.8A3/2 per facet.
The powers p and q are Evolver variables; see the datafile lines below. The defaults are p = 1 and q = 4.
There is a nice expression for the general dimensional version of the conformal version of the energy:
r r r s1 r s2
A1 A2
1
4E f1 f2 = 4 6 det t1 r t1 s1 t1 s2
(16.61)
r
r
t2 r t2 s1 t2 s2
where r = |x1 x2 | and s1 , s2 ,t1 ,t2 are the sides of the two facets.
Datafile lines: parameter surface_knot_power = 1 // this is q parameter surface_cos_power = 0 // this is p quantity
sphere_energy ENERGY modulus 1 global_method sphere_knot_energy
16.5 Volumes
A body can get volume from a surface integral over its boundary facets and from content integrals of boundary facets
edges on constraints. Also important are the gradients of body volumes at vertices. There are two choices built-in
surface integrals: default and SYMMETRIC_CONTENT.
16.5.1
Default facet integral
A facet oriented with normal outward from a body makes a contribution to its volume of

V=
= 1 (z0 + z1 + z2 )k s0 s1 .
zk dA
6
f acet
(16.62)
The gradient of the volume of the body as a function of the vertex v0 coordinates is
g =
16.5.2
)
1(
(k s0 s1 )k + (z0 + z1 + z2 )k s1 .
6
(16.63)
Symmetric content facet integral
A facet oriented with normal outward from a body makes a contribution to its volume of
V=
1
3
= 1v0 v1 v2 .
(xi + yj + zk) dA
6
f acet
(16.64)
This can be seen most easily by looking at the facet as the base of a tetrahedron with its fourth vertex at the origin. The
integral over the three new faces is zero, so the integral over the original facet must be the volume of the tetrahedron.
The gradient of the volume of the body as a function of the vertex v0 coordinates is
1
g = v1 v2 .
6
231
(16.65)
16.5.3
232
Edge content integrals
Then b gets
Let body b have facet f with outward normal. Suppose f has edge s deemed to have content integrand U.
volume
ds
xi )s.
V= U
wiU(
(16.66)
e
The volume gradients for the body are (summation convention again)
g j (head) =
g j (tail) =
16.5.4
wi (aiUk, j (xi )sk +U j (xi )),

wi ((1 ai )Uk, j (xi )sk U j (xi )).
(16.67)
Volume in torus domain
The wrapping of the edges across the faces of the unit cell makes the calculation of volumes a bit tricky. Suppose body
b has vertices vk . Ideally, we would like to adjust the vertices by multiples of the unit cell basis vectors to get a body
whose volume we could find with regular Euclidean methods. Unfortunately, all we know are the edge wraps, i.e. the
differences in the adjustments to endpoints of edges. But this turns out to be enough, if we are a little careful with the
initial volumes in the datafile.
Let Ak be the vertex adjustment for vertex k, and T j be the wrap vector (difference in endpoint adjustments) for
facet-edge j. Let m index facets. Then
V
=
=
1
6
(vm0 + Am0 ) (vm1 + Am1 ) (vm2 + Am2 )
(16.68)
f acets m
1
(S1 + S2 + S3 + S4 )
6
where
S1
vm0 vm1 vm2 ,
f acets m
S2
v j0 v j1 A j2 ,
vk Ak1 Ak2 ,
(16.69)
f acetedges j
S3
f acetverts k
S4
Am0 Am1 Am2 .
f acets m
The first of these sums is straightforward. The second sum can be regrouped, pairing the two oppositely oriented
facet-edges j and j for each edge together:
S2
v j0 v j1 A j2 +v j0 v j1 A j2
v j0 v j1 (A j2 A j2 )
edges j
(16.70)
edges j
1
v j0 v j1 (T j2 + Tj1 Tj2 T j1 )
2 edges
j
which can be regrouped into a sum over facet-edges, which can be done facet by facet:
S2 =
v j0 v j1 (T j1 T j2 ).
f acetedges j
232
(16.71)
233
The third sum we group terms with a common vertex together, with the inner facet sum over facets around vertex
k:
(
S3
(Ai1 Ak0 ) (Ai2 Ak )
vk
(16.72)
)
Ti0 Ti2
f acets i
vertices k
f acetsi
vk
vertices k
Ai1 Ai2
f acetsi
vk
vertices k
vk Tk2 Tk0 ,
f acetvertices k
which again can be done facet by facet.

The fourth sum is a constant, and so only needs to be figured once. Also, it is a multiple of the unit cell volume.
1
Therefore, if we assume the volume in the datafile is accurate to within 12
Vc , we can calculate the other sums and
figure out what the fourth should be.
The body volume gradient at v0 will be
(
)
1
1
g =
vm1 vm2 + vm1 (Tm1 Tm2 ) + Tm0 Tm2 .
(16.73)
6 f acetsm on v
2
16.6 Constraint projection

Suppose vertex v has constraints that are the zero sets of functions f1 , . . . , fn .
16.6.1
Projection of vertex to constraints
The solution will be by Newtons method. Several steps may be necessary to get within the tolerance desired of the
constraint. Inequality constraints are included if and only if the inequality is violated.
For one step, we seek a displacement v that is a linear combination of the constraint gradients and will cancel the
difference of the vertex from the constraints:
v = ci fi
such that
v f j = f j (v)
for each j.
(16.74)
Thus
ci fi f j = f j (v)
for each j.
(16.75)
This is a linear system that can be solved for the ci and thus v.
16.6.2
Projection of vector onto constraint tangent space
We want to project a vector F (say, a force vector) onto the intersection of all the tangent planes of the constraints at a
vertex. We project F by subtracting a linear combination of constraint gradients:
Fpro j = F ci fi
such that
Fpro j f j = 0
for each j.
(16.76)
This leads to the system of linear equations for the ci :
ci fi f j = F f j .
i
233
(16.77)
234
16.7 Volume and quantity constraints

This section explains how volume and quantity constraints are enforced. Henceforth in this section, everything referring to the volume of a body can also refer to the value of a quantity constraint. There are two parts to this enforcement
on each iteration: a volume restoring motion that corrects for any volume deviations, and a projection of the vertex
motions to the subspace of volume-preserving motions. Let gbv be the gradient of the volume of body b as a function
of the position of vertex v. We will also use B as a body index. If the bodies fill a torus, then one body is omitted to
prevent singular systems of equations.
16.7.1
Volume restoring motion
Let the current excess volume of body b be b (which may be negative, of course). We want a volume restoring motion
Rv
Rv = cBgBv
and
(16.78)
Rv gbv = b for each b
v
which leads to the linear system for the cB :
cB gBv gbv = b
16.7.2
for each b.
(16.79)
Motion projection in gradient mode
The motion in this mode is the scale factor times the force on a vertex. Let Fv be the total force at v. We want a
projected force Fvpro j such that
Fvpro j = Fv aBgBv
Fvpro j gbv = 0
and
for each b
(16.80)
which leads to the linear system for aB
aB gBv gbv = Fv gbv

v
for each b.
(16.81)
The coefficients aB are the pressures in the bodies when the surface is in equilibrium.
16.7.3
Force projection in mean curvature mode
The motion in this mode is the scale factor times the force on a vertex divided by the area of the vertex. Let Fv be the
total force at v and Av its area. We want a projected force Fvpro j such that
Fvpro j = Fv aBgBv
Fvpro j gbv /Av = 0
and
for each b
(16.82)
which leads to the linear system for ab
ab gBv gbv /Av = Fv gbv /Av

b
for each b.
(16.83)
The coefficients ab are the pressures in the bodies even when the surface is not in equilibrium. At equilibrium, the
pressures calculated with or without the mean curvature option are equal.
234
16.7.4
235
Pressure at z = 0
The pressure reported by the v command is actually the Lagrange multiplier for the volume constraint. The pressure
may vary in a body due to gravitational force, for example. In that case, the Lagrange multiplier is the pressure where
other potential energies are 0. For example, if gravitational force is acting, then the Lagrange multiplier is the pressure
at z = 0. Let be surface tension, H the scalar mean curvature, N the unit normal vector to the surface, the body
density, G the gravitational constant, and the lagrange multiplier. The gradient of area energy is 2H N, the gradient
of gravitational potential energy is GzN, and the gradient of volume is N. Therefore at equilibrium
2H N + GzN = N.
(16.84)
2H N = N.
(16.85)
At z = 0
So pressure(z = 0) = 2H = .
16.8 Iteration
This section explains what happens during each iteration (g command). t is the current scale factor.
16.8.1
Fixed scale motion
1. Do diffusion if called for. For each facet f , let A f be the area of the facet. Let m1 and P1 be the prescribed
volume (mass) of a body on one side of f , and m2 and P2 those of the body on the other side. Let be the
diffusion constant. If there is no body on one side, the pressure there is zero or ambient pressure for the ideal
gas model. The mass transfers are:
dm1 = (P1 P2 )A f t
and
dm2 = (P2 P1 )A f t.
(16.86)
2. All the forces on vertices are calculated, as detailed in the Energies and Forces section above. If the mean
curvature option is active, each force is divided by the vertexs associated area.
3. All forces at FIXED vertices are set to zero. Forces on constrained vertices are projected to the constraint tangent
spaces. Forces on vertices on boundaries are converted to forces in boundary parameter spaces.
4. The volume restoring motion is calculated, and the forces are projected according to body volume constraints.
5. Jiggle if jiggling is toggled on.
6. Move vertices by volume restoring motion and current scale times force.
7. Enforce constraints on vertices. Do up to 10 projection steps until vertex is within tolerance of constraints.
8. Recalculate volumes, pressures, areas, and energy.
16.8.2
Optimizing scale motion
The idea is to find three scale factors that bracket the energy minimum and then use quadratic interpolation to estimate
the optimum scale factor. The first four steps are the same as for fixed scale motion.
5. Save current coordinates.
6. Move vertices by volume restoring motion and current scale times force and enforce constraints. Recalculate
energy.
235
236
7. Restore coordinates, double scale factor, move surface, and recalculate energy. If this energy is higher, keep
cutting scale factor in half until energy increases again. If it is lower, keep doubling scale factor until energy
increases.
8. Eventually get three scale factors s1 , s2 , s3 with energies e1 , e2 , e3 bracketing a minimum. The scale factors are
successive doublings: s2 = 2s1 , s3 = 2s2 . Do quadratic interpolation to find approximate optimum:
sopt = 0.75s2 (4e1 5e2 + e3 )/(2e1 3e2 + e3 ).
(16.87)
9. Jiggle if jiggling is on.

10. Move vertices by volume restoring motion and optimum scale times force.
11. Project to constraints.
12. Recalculate volumes, pressures, areas, and energy.
16.8.3
Conjugate gradient mode
The U command toggles conjugate gradient mode. The conjugate gradient method does not follow the gradient downhill, but makes an adjustment using the past history of the minimization. It usually results in much faster convergence.
At iteration step i, let Si be the surface, Ei its energy, Fi (v) the force at vertex v, and hi (v) the history vector of v.
Then
hi (v) = Fi (v) + hi1
(16.88)
where
=
v Fi (v) Fi (v)
.
v Fi1 (v) Fi1 (v)
(16.89)
Then a one-dimensional minimization is performed in the direction of hi , as described in steps 6,7,8 above. It is
important that all volumes and constraints be enforced during the one-dimensional minimization, or else the method
can go crazy. The history vector is reset after every surface modification, such as refinement or equiangulation.
The above formula is the Fletcher-Reeves version of conjugate gradient. The version actually used is the PolakRibiere version, which differs only in having
=
v (Fi (v) Fi1 (v)) Fi (v)

.
v Fi1 (v) Fi1 (v)
(16.90)
Polak-Ribiere seems to recover much better from stalling. The ribiere command toggles between the two versions.
Conjugate gradient blowups: Sometimes conjugate gradient wants to move so far and fast that it loses contact
with volume constraints. Before, Evolver did only one or two Newton method projections back to volume constraints
each iteration; now it will do up to 10. Ordinary iteration does only one projection still, but you can get the extra
projections with the post_project toggle. If convergence fails after 10 iterations, you will get a warning message,
repeated iterations will stop, and the variable iteration_counter will be negative.
16.9 Hessian iteration

The hessian command constructs a quadratic approximation of the energy and solves for the minimum (or whatever
the critical point happens to be). See the Hessian section of the Model chapter for remarks on its use. The independent
variables used are the displacements of vertex coordinates. For vertices on level set constraints, the coordinates are
replaced by set of variables equal in number to the degrees of freedom. Fixed vertices are not represented. Let X
236
237
denote the vector of independent variables. Let B denote the energy gradient, and let H be the Hessian matrix of
second partial derivatives. Hence the energy change is
1
E = X T HX + BT X.
2
(16.91)
Global constraints like volumes and fixed named quantities are actually scalar functions of X with fixed target values.
Index these constraints by i, and let Ci be the gradient and Qi the Hessian of constraint i. Let Fi be the current value
and Fi0 be the target value of constraint i. Then we have the constraint conditions on X
1 T
X Qi X +CiT X = Fi0 Fi .
2
(16.92)
Let i be the Lagrange multiplier for constraint i. The Lagrange multiplier is shown in the pressure column of the
v command. At a critical point we have
X T H + BT = i (X T Qi +CiT ).
(16.93)
So we want to solve the system

(H i Qi )X iCi
i
= B
(16.94)
1 T
X Qi X +CiT X
2
= Fi0 Fi .
We do this by Newtons method. This means solve the augmented system

(
)( ) (
)
H i i Qi C
X
C B
=
d
F0 F
CT
0
(16.95)
where C is the matrix with columns Ci , is the vector of i , d is the change in , and F0 , F are the vectors of
constraint values. This matrix is typically sparse, and is factored with sparse matrix algorithms suitable for indefinite
matrices. The index of the Hessian proper turns out to be the index of the augmented Hessian minus the rank of C.
There is another approach to solving this, which was the default approach until version 2.17, and which can be
reinstated by the command augmented_hessian off. For convenience, denote A = H i i Qi . The solution of this is
d = (CT A1C)1 (CT A1 (C B) (F0 F))
X = A1 (C B) A1Cd.
(16.96)
I choose to solve for A1 first for two reasons: (1) A is sparse, while C is not, and (2) finding A1 and CT A1C can
tell us about the stability of the critical point. Actually, A1 is never found as such. Instead, a Cholesky factorization
A = U T DU is found and used to solve AY = C for Y = A1C and so forth. Here U is an upper triangular matrix, and
D is a diagonal matrix. The factoring is done with routines from the Yale Sparse Matrix Package, modified a bit since
the matrix A may not be positive definite. Negative entries in D are allowed, and the total number of them is the index
of A. Zero entries are also allowed, since many surfaces have degenerate minimums of energy (due to translational
invariance, for example). The index of the constrained Hessian is given by the relation
index(Hconstrained ) = index(A) index(CT A1C).
(16.97)
End of explanation of old method.

Degeneracy is detected during factoring of the Hessian by the appearance of a zero diagonal entry. If the zero
does represent a true degeneracy, then the whole row and column of that zero should also be zero at that stage of the
factoring. The factoring routine then puts a 1 on the diagonal to insure an invertible D. This has no effect on any of the
solutions. The cutoff value for considering a diagonal element to be zero is given by the variable hessian_epsilon
which may be set by the user with an ordinary assignment command. Its default value is 1e-8.
237
238
If the constrained Hessian index is positive, then a warning message is printed. Once the Hessian method has
converged, this provides a test for being a local minimum. At a saddle point, the constrained index is positive and it is
possible to find a downhill direction. The command saddle implements this.
Now a word on handling vertices on level-set constraints. Note that above, X was composed of independent
variables. For a vertex on k level-set constraints, there are n k independent variables, where n is the space dimension.
Let the constraints have the quadratic approximation (here X is original space variables)
1
KiT X + X T Gi X = 0,
1 i k.
(16.98)
2
and let K have columns Ki . Let P have for columns a basis for the nullspace of K T (which is the tangent space to all
the constraints), and let Y be coordinates in the P basis, so X = PY . Y is the set of independent coordinates. Now if
we have a quantity E with quadratic representation,
1
E = X T HX + BT X,
(16.99)
2
then we want to get an expression in Y , keeping in mind that Y really coordinatizes a tangent space orthogonally
projected back to a curved constraint. For a given Y , the projection back will be a linear combination of constraint
normals,
X = i Ki
(16.100)
i
so
1
KiT (PY + K) + (PY + K)T Gi (PY + K) = 0,
1 i k.
2
We keep only terms quadratic in Y , and since K T P = 0 and is quadratic in Y ,
1
KiT K + Y T PT Gi PY = 0,
2
So
1
T T
j = (K T K)1
ji Y P Gi PY,
2
i
1 i k.
1 j k.
Hence we can write E in terms of Y as

1
1
T T
T
E = Y T PT HPY + BT K j (K T K)1
ji Y P Gi PY + B PY.
2
2
j i
For each vertex, P and
T T
K j (K T K)1
ji Y P Gi PY
j
(16.101)
(16.102)
(16.103)
(16.104)
(16.105)
are calculated just once, and then used with whatever Hs and Bs turn up in the course of the calculation of E. Like
adjustments are made for all the Qi s above also.
Some notes for people rash enough to write hessian routines for their own named quantity methods: Hessians can
only be written for methods that are evaluated independently on elements. Two vertices are allocated a Hessian entry
only if they are joined by an edge. See q_edge_tension_hessian() in hessian.c for an example. For debugging, the
command hessian_menu brings up a little menu. You can turn debugging on, in which case you get a whole lot
of information dumped to the screen as the algorithm proceeds. Dont do that with more that 3 or 4 unfixed vertices.
The sequence of menu options for one iteration is 1,2,3,4 (stepsize 1). For checking Hessian matrix values, there is
a hessian_diff command at the main prompt which will toggle on the calculation of the Hessian matrix by finite
differences.
If hessian_normal is toggled on, then each vertex is constrained to move perpendicular to the surface. This
eliminates all the fiddly sideways movement of vertices that makes convergence difficult. Highly recommended.
Perpendicular is defined as the volume gradient, except at triple junctions and such, which are left with full degrees of
freedom. If you wish to run in hessian_normal mode but exempt particular vertices from the restriction, you can set
the vertices no_hessian_normal attribute, for example
set vertex no_hessian_normal where z > 1.2
238
239
16.10 Dirichlet and Sobolev approximate Hessians

This section describes features not necessary to understanding Evolver operation, but they are included because I think
they are interesting.
The problem in using quickly converging iteration methods such as Newtons method (as embodied in the hessian
command) is that the Hessian of the energy is usually indefinite except when extremely close to a minimum. The
idea here is to construct a positive definite quadratic form that is tangent to the area functional and whose Hessian is
an approximation of the possibly indefinite area Hessian. The approximation can be solved with standard numerical
linear algebra, and the solution will decrease area, if the area is not at a critical point. If the area is at a critical point,
then so is the approximation.
Consider an m-dimensional simplex in Rn . Let F be a mn matrix whose rows are the side vectors of emanating
from one vertex. Then the area of is
1
A (F) = (det FF T )1/2 .
(16.106)
m!
For convenience, let A = FF T .
For the first derivative, let G be a displacement of F, and let be a parameter. Then
A = (F + G)(F + G)T
and
DA (G) =
=
=

(
)
A 1
A (F + G)
1 1
Tr
A
(det A)1/2
=

2 m!
=0
1 1
Tr((GF T + FGT )A1 )(det A)1/2
2 m!
1
Tr(GF T A1 )(det A)1/2 .
m!
(16.107)
(16.108)
For the second derivative, let G1 and G2 be displacements. Then

2 A (F + G1 + G2 )
2
D A (G1 , G2 ) =

==0
=
=
=
=
2 1
(det((F + G1 + G2 )(F + G1 + G2 )T ))1/2
m!

2 1
1/2
(det A)
m!
==0

(
)

A 1
1 1
1/2
Tr
A
(det A)
m! 2
==0
[
( 2
)
1 1
A 1 A 1 A 1
Tr
A
A
A
m! 2

)
)]
(
(

1
A 1 1
A 1
1/2
(det A)
+ Tr
A
Tr
A
2
==0
[
1 1
Tr(2G1 GT2 A1 (G1 F T + FGT1 )A1 (G2 F T + FGT2 )A1 )
m! 2
]
1
+ Tr(2G1 F T A1 )Tr(2G2 F T A1 ) (det A)1/2
4
1 [
Tr(G1 GT2 A1 G1 F T A1 G2 F T A1 FGT1 A1 G2 F T A1 )
m!
]
+Tr(G1 F T A1 )Tr(G2 F T A1 ) (det A)1/2 .
239
(16.109)
240
Thus the true area Hessian can be written as the quadratic form
1 [
< G1 , G2 >H =
Tr(G1 GT2 A1 G1 F T A1 G2 F T A1 FGT1 A1 G2 F T A1 )
m!
]
+Tr(G1 F T A1 )Tr(G2 F T A1 ) (det A)1/2 .
(16.110)
Renka and Neuberger [RN] start with a Sobolev space inner product, and ultimately come down to taking the
approximate Hessian to be
]
1 [
< G1 , G2 >S =
Tr(G1 GT2 A1 FGT1 A1 G2 F T A1 ) + Tr(G1 F T A1 )Tr(G2 F T A1 ) (det A)1/2 .
(16.111)
m!
Note that compared to the true Hessian, this drops the term
Tr(G1 F T A1 G2 F T A1 )
(16.112)
which is responsible for the possible nonpositivity of the Hessian. The positive semidefiniteness may be seen by
writing the Sobolev form as
]
1 [
Tr(A1/2 G(I F T A1 F)GT A1/2 ) + Tr(GF T A1 )2 (det A)1/2 ,
< G, G >S =
(16.113)
m!
since I F T A1 F = I F T (FF T )1 F is an orthogonal projection.
There is a similar idea due to Polthier and Pinkall [PP]. Their scheme minimizes the Dirichlet integral of the image
simplex over the domain simplex:
E = Tr(F F T )da.
(16.114)
where F is the linear map from the old simplex to the new. Letting F being the old vectors and G being the new
vectors, we see that
F = GF 1
(16.115)
where F 1 is defined on the old simplex. Thus
E = Tr(GT (FF T )1 G)
1
(det(FF T ))1/2 .
m!
(16.116)
Or, with A = FF T ,
1
1
Tr(GT A1 G)(det A)1/2 =
Tr(GGT A1 )(det A)1/2 .
m!
m!
Hence the Dirichlet quadratic form is
E =
1
1
Tr(GT1 A1 G2 )(det A)1/2 =
Tr(G1 GT2 A1 )(det A)1/2 .
m!
m!
This drops even more terms of the true Hessian. So a priori, the Sobolev scheme should be a little better.
The iteration scheme is to find a perturbation G of the vertices that minimizes the energy
< G1 , G2 >D =
E(G) = A + DA (G) +
1
< G, G >
2
(16.117)
(16.118)
(16.119)
where < , > is the approximate Hessian of your choice on facet .

Use of these approximate hessian for the facet_area method is triggered by the dirichlet_mode and sobolev_mode
commands respectively.
The command dirichlet does one iteration minimizing the Dirichlet energy, and sobolev does one minimization of
the Sobolev energy. In practice, both Sobolev and Dirichlet iteration are good for restoring sanity to badly messed
up surfaces. However, as the surface approaches minimality, these iterations become just versions of gradient descent
since the true Hessian is not used. Hence they do not converge much faster than ordinary iteration, and slower than
the conjugate gradient method on ordinary iteration. The commands dirichlet_seek and sobolev_seek act as gradient
descent methods, doing a line search in the solution direction to find the minimum energy. These could be plugged in
to the conjugate gradient method, but I havent done that yet.
240
241
16.11 Calculating Forces and Torques on Rigid Bodies

It may be of interest to know the forces and torques a liquid exerts on a solid object, for example the forces exerted on
a microchip by liquid solder during the soldering phase of assembly of a circuit board. In general, the force exerted
on a rigid body in a system is the negative rate of change of energy of the system with respect to translation of the
rigid body while other constraints on the system remain in effect. Likewise, torque around a given axis is the negative
rate of change of energy with respect to rotating the body about the axis. Below are five ways of calculating force and
torque, expressed as a general rate of change of energy with respect to some parameter q. Other forces on the body
due to energies not included in the Evolver datafile, such as gravity, must be calculated separately.
16.11.1
Method 1. Finite differences
Algorithm: Evolve the system to a minimum of energy, change the parameter q to q + q, re-evolve to a minimum of
energy, and calculate the force as
E(q + q) E(q)
F =
,
(16.120)
q
where E(q) is the minimum energy for parameter value q.
Advantages:
Conceptually simple.
Often requires no special preparation in the datafile.
Disadvantages:
The system must be very close to a minimum, else energy reduction due to minimization will contaminate the
energy difference due to changing q.
Re-evolving can be time consuming.
The finite difference formula can be of limited precision due to nonlinearity of the energy as a function of q. A
more precise central difference formula could be used:
F =
E(q + q) E(q q)
.
2q
(16.121)
q needs to be chosen wisely to maximize precision. A rule of thumb is q should be the square root of the
accuracy of the energy for the one-sided difference, and the cube root for the central difference.
The datafile may need to be modified so that q is a parameter that can be changed with the desired effect on the
system.
The system is changed, and usually needs to be restored to its original condition.
16.11.2
Method 2. Principle of Virtual Work by Finite Differences
If the original system is at equilibrium, it is not necessary to re-evolve the perturbed system to equilibrium. The energy
difference from equilibrium of the perturbed system is only of order q2 in general, so it is only necessary to enforce
the constraints.
Algorithm: Evolve the system to a minimum of energy, change the parameter q to q + q, enforce any constraints
(say by doing one iteration with a scale factor of 0), and calculate the force as
F =
E(q + q) E(q)
.
q
Advantages:
241
(16.122)
242

The system need not be very close to a minimum, since there is no further evolution to contaminate the energy
difference.
No time consuming re-evolving.
Disadvantages:
F =
E(q + q) E(q q)
.
2q
(16.123)
system.
16.11.3
Method 3. Principle of Virtual Work using Lagrange Multipliers
This is the same as Method 2, except that adjustment to global constraints (such as volumes) is done using the corresponding Lagrange multipliers rather than doing a projection to the constraints. Pointwise constraints must still
be done, but those can be done with the recalc command rather than g. A brief review of Lagrange multipliers:
Suppose the system is at equilibrium, and let X be the perturbation vector. Let B be the energy gradient vector, so that
in matrix notation
E = BT X
(16.124)
to first order. Let Ci , i = 1, . . . , k, be the gradient vectors of the constraints. Then at equilibrium, the energy gradient
must be a linear combination of the constraint gradients
B = iCi
(16.125)
where the i are called the Lagrange multipliers. Now back to business. After the perturbation, let X be the deviation
from the new equilibrium. Then the deviations in the constraints will be
Vi = CiT X
(16.126)
so the deviation from the new equilibrium energy will be

E = BT X = iCi X = i Vi .
i
(16.127)
Hence we can calculate the force as

F =
E(q + q) i i Vi E(q)
.
q
(16.128)
Algorithm: Evolve the system to a minimum of energy, change the parameter q to q + q, enforce pointwise
constraints with recalc, doing one iteration with a scale factor of 0),
F =
E(q + q) i i Vi E(q)
.
q
Recall that the Lagrange multipliers for body volumes are available under the body attribute "pressure".
Advantages:
242
(16.129)
243

difference.
Avoids an actual projection to global constraints.
Disadvantages:
F =
E(q + q) E(q q)
2q
(16.130)
where the energies include the Lagrange multiplier corrections.

system.
16.11.4
Method 4. Explicit forces
The force on a body is exerted physically by pressure, surface tension, etc., so one could calculate the net force by
adding all those up. That is, calculate the area of liquid contact and multiply by the pressure, find the vector force that
is exerted by surface tension on each edge on the body, etc.
Algorithm: Identify all the forces acting on the body (surface tension, pressure, etc.) and calculate explicitly from
the equilibrium surface.
Advantages:
difference.
Does not change the system.
Full precision, since no differencing is done.
Disadvantages:
From tests, this method is extremely inaccurate and slow to converge at higher refinement.
Requires user to explicitly identify and calculate all forces.
It can be very difficult to do some forces explicitly, such as surface tension, in the quadratic and higher order
lagrange models.
243
16.11.5
244
Method 5. Variational formulation
This method calculates the rate of change of energy in the manner of the Calculus of Variations, setting up a vectorfield
perturbation and writing the derivatives of energy and constrained quantities as integrals over the surface. By the
Principle of Virtual Work, the force is
(16.131)
F = [E (q) iVi (q)].
i
Algorithm: For each component of the energy, create a named method whose value is the derivative of the component with respect to the parameter. Also do this for each global quantity. Use these derivatives to create a force
quantity following the Principle of Virtual Work formula.
Advantages:
difference.
Does not change the system.
Full precision, since no differencing is done.
Dont have to guess at forces, since one can systematically write variations for all energy and constraint components.
Disadvantages:
Requires user to include quantities for the variations of all energies and global constraints in the datafile.
Note on perturbations: When changing a parameter, it is best to deform the surface as uniformly as possible. If
you simply change a parameter governing the height of a boundary, for example, then all the deformation is inflicted
on just the adjacent facets. If the deformations are much smaller than the facet size, then this is not too significant. But
for elegance, a uniform deformation is nice. For example, suppose one has a catenoid whose top is at zmax and whose
bottom is at zmin . Suppose zmax is to be increased by dz. A uniform deformation would be
z z+
z zmin
dz.
zmax zmin
(16.132)
So a command sequence to do the perturbation would be

dz := 0.00001;
foreach vertex do set z z+(z-zmin)/(zmax-zmin)*dz;
zmax := zmax + dz;
recalc;
Note on choosing finite difference perturbation size: Let E be the energy as a function of parameter q. The
one-sided finite difference force calculation is
F =
E(q + dq) E(q) +

,
dq
(16.133)
where represents the error in the energy difference due to limited machine precision. Plugging in Taylor series terms,
F
E(q) + E (q)dq + E (q)dq2 /2 E(q) +

dq
= E (q) (E (q)dq/2 + /dq).
=
244
(16.134)
245
The error term is minimized for

dq =
2/E (q),
(16.135)
hence the rule of thumb dq = 1/2 with error 1/2 . For 15 digit machine precision, one could take dq 107 and
expect at most 7 or 8 digits of precision in the force.
The two-sided finite difference force calculation is
F =
E(q + dq) E(q dq) +

.
2dq
(16.136)
Plugging in Taylor series terms,

F
E(q) + E (q)dq + E (q)dq2 /2 + E (q)dq3 /6 (E(q) E (q)dq + E (q)dq2 /2 E (q)dq3 /6) +

2dq
= E (q) (E (q)dq2 /6 + /2dq).

The error term is minimized for
(16.137)
dq = (1.5/E (q))1/3 ,
1/3
hence the rule of thumb dq =

with error
expect at most 10 digits of precision in the force.
16.11.6
2/3 .
(16.138)
For 15 digit machine precision, one could take dq 105 and
Example of variational integrals
Here we compute the force of the column.fe example using variational integrals (method 5). The energy has surface
tension and gravitational components,

E=
z2
G k dA.
2
S
T dA +
S
(16.139)
where T is the surface tension of the free surface S, is the gravitational constant, and G is the acceleration of gravity.
Given a variation vectorfield g, the unconstrained variation of the energy is
)

(
z2
z2
E = T
(divg N Dg N)dA + G
div( k)g curl(g k) dA,
(16.140)
2
2
S
S
where N is the unit normal. This variation may be obtained by considering the two integrals separately. For the area
integral, the variation is the divergence of g in the tangent plane, and divg N Dg N is an invariant way to write that;
Dg is the matrix of first partials of g. The second term is obtained by noting that the original surface and the perturbed
surface almost form the boundary of a thin region. The variation is obtained by applying the divergence theorem to
the region, and adding in the missing boundary strip by means of Stokes Theorem.
Of course, we have to correct for any volume change, and since the pressure P is the Lagrange multiplier relating
volume and energy changes at a minimum,
E = PV = P
g dA.
(16.141)
Projecting volume back to the volume constraint subtracts this, so the net change is
)
(

z2
z2
P
(div k)g curl(g k) dA
(divg N Dg N)dA + G
E = T
g dA.
2
2
S
S
S
(16.142)
We shall take the perturbation to be g = (z + ZH)j/2ZH, which leaves the bottom plane fixed and gives a unit shift to
the top plane. It is important when working with piecewise linear surfaces that the perturbations be likewise piecewise
linear. The force, then, is
E = T
Nz Ny
dA + G
2ZH
z2
k dA P
4ZH
245
z + ZH
j dA.
2ZH
(16.143)
246
The Nz Ny term requires the use of the facet_general_integral method, which integrates an arbitrary scalar function
of position and normal vector over facets. This method requires that the integrand be homogeneous of degree 1 in the
unnormalized facet normal N, i.e. that the integrand be
1 Nz Ny
.
2ZH |N|
(16.144)
The third term of E is automatically zero by the Divergence Theorem, since the integrand is zero on the top and
bottom pads. The two remaining terms are defined as two method instances in the datafile below, and combined into
one quantity forcey.
Example
The datafile below calculates the vertical force exerted by a catenoid (without body), for which exact values can
be calculated analytically. All five methods are done. Following is a table of the errors in the force as a function of
refinement and the representation used. Lagrange order refers to the order of polynomial used to represent facets; 1
is linear, and 2 is quadratic.
// catforce.fe
// Evolver data for catenoid.
// For testing vertical force on upper ring calculated various
// ways and compared to exact solution.
PARAMETER
PARAMETER
PARAMETER
parameter
ZMAX = 0.7
ZMIN = -0.7
RMAX = cosh(zmax)
true_area =
pi*(zmax+sinh(2*zmax)/2-zmin-sinh(2*zmin)/2)
// for putting vertices exactly on catenoid

constraint 1
formula: sqrt(x^2+y^2) = cosh(z)
// for restoring after perturbation
define vertex attribute old_coord real[3]
// following method instances and quantity for vertical force
method_instance darea method facet_general_integral
scalar_integrand: 1/(zmax-zmin)*(x4^2+x5^2)/sqrt(x4^2+x5^2+x6^2)
quantity forcez info_only global_method darea
x1: RMAX * cos(p1)
x2: RMAX * sin(p1)
x3: ZMAX
x1: RMAX * cos(p1)
x2: RMAX * sin(p1)
x3: ZMIN
//
//
upper ring
lower ring
246
vertices
1
0.00
2
pi/3
3 2*pi/3
4
pi
5 4*pi/3
6 5*pi/3
7
0.00
8
pi/3
9 2*pi/3
10
pi
11 4*pi/3
12 5*pi/3
edges
1
1
2
2
3
3
4
4
5
5
6
6
7
7
8
8
9
9
10
10
11
11
12
12
13
1
14
2
15
3
16
4
17
5
18
6
faces
1
1
2
2
3
3
4
4
5
5
6
6
// given in terms of boundary parameter

boundary 1
fixed
boundary 1
fixed
boundary 1
fixed
boundary 1
fixed
boundary 1
fixed
boundary 1
fixed
boundary 2
fixed
boundary 2
fixed
boundary 2
fixed
boundary 2
fixed
boundary 2
fixed
boundary 2
fixed
2
3
4
5
6
1
8
9
10
11
12
7
7
8
9
10
11
12
14
15
16
17
18
13
247
boundary 1
boundary 1
boundary 1
boundary 1
boundary 1
boundary 1
boundary 2
boundary 2
boundary 2
boundary 2
boundary 2
boundary 2
fixed
fixed
fixed
fixed
fixed
fixed
fixed
fixed
fixed
fixed
fixed
fixed
-7 -13
-8 -14
-9 -15
-10 -16
-11 -17
-12 -18
read
hessian_normal
// For saving coordinates before perturbation
save_coords := { foreach vertex vv do
{ set vv.old_coord[1] x;
247
248
set vv.old_coord[2] y;
set vv.old_coord[3] z;
}
}
// For restoring coordinates after perturbation
restore_coords := { foreach vertex vv do
{ set vv.x old_coord[1];
set vv.y old_coord[2];
set vv.z old_coord[3];
}
}
// Force by central difference of energy minima
method1 := { save_coords;
dzmax := 0.00001;
zmax := zmax - dzmax;
optimize 1;
hessian; hessian;
lo_energy := total_energy;
restore_coords;
zmax := zmax + 2*dzmax;
hessian; hessian;
hi_energy := total_energy;
restore_coords;
force1 := -(hi_energy - lo_energy)/2/dzmax;
}
// Force by central difference and Principle of Virtual Work
old_scale := scale;
dzmax := 0.00001;
recalc; m 0; g;
lo_energy := total_energy;
restore_coords;
recalc; m 0; g;
hi_energy := total_energy;
restore_coords;
scale := old_scale; optimize 1;
}
// Force by central difference and Principle of Virtual Work and
// Lagrange multipliers
dzmax := 0.00001;
recalc;
248
249
lo_energy := total_energy ;
restore_coords;
recalc;
hi_energy := total_energy ;
restore_coords;
}
// Using same overall perturbation as forcez
method3a := { save_coords;
dzmax := 0.00001;
set vertex z z-(z-zmin)/(zmax-zmin)*dzmax;
recalc;
lo_energy := total_energy ;
restore_coords;
zmax := zmax + dzmax;
set vertex z z+(z-zmin)/(zmax-zmin)*dzmax;
zmax := zmax + dzmax;
recalc;
hi_energy := total_energy ;
restore_coords;
force3a:= -(hi_energy - lo_energy)/2/dzmax;
}
// Force by explicit calculation of forces. Surface tension force
// only here, since pressure not acting on chip in lateral direction.
method4 := {
force4 := sum(edge ee where on_boundary 1,
(ee.y*ee.facet[1].x - ee.x*ee.facet[1].y)/ee.facet[1].area);
}
// Force by variational integrals:
method5 := { force5 := -forcez.value;
}
// all methods at once
methods := {
quiet on; method1; method2; method3; method3a;
if (linear) then method4;
method5;
quiet off;
printf "\\n\\nSummary:\\n" ;
printf "True force:
%-#20.15g\\n",-2*pi*rmax/cosh(zmax);
printf "Force by method 1: %-#20.15g\\n",force1;
printf "Force by method 3a: %-#20.15g\\n",force3a;
if (linear) then printf "Force by method 4: %-#20.15g\\n",force4;
249
250
printf "Force by method 5:
%-#20.15g\\n",force5;
Table 1. Errors in calculated force by method 3 (uniform stretch with

central difference:
Lagrange order
refinement
1
2
3
4
5
6
0
8.5e-1 3.8e-3 6.2e-3 3.6e-3 1.7e-4 1.7e-5
1
2.0e-1 1.8e-3 3.0e-4 1.0e-5 1.0e-6 7.8e-8
2
5.1e-2 1.2e-4 2.1e-5 1.8e-7 1.3e-8 3.2e-10
3
1.3e-2 6.9e-6
4
4.3e-3 4.3e-7
5
8.2e-4
Lagrange multipliers) for dq = 105 and

7
1.1e-5
1.5e-8
5.5e-10
Table 2. Time of calculation (r; g5; hessian; hessian; hessian) in seconds:

Lagrange order
refinement
1
2
3
4
5
6
0
1.272
3.094
9.204
20.289
1
0.340
3.184
4.777
12.127
36.212
80.155
2
1.291
12.277
19.328 49.190
145.850 326.570
3
6.479
51.384
83.740 209.982
4
19.839 227.046
5
98.441
8
6.1e-6
1.7e-9
3.8e-10
7
32.598
131.710
523.593
8
106.563
425.142
1789.710
16.12 Equiangulation
Triangulations work best when the facets are close to equilateral (that is, equiangular). Given a set of vertices, how
do you make them into a triangulation that has triangles as nearly as possible equilateral? In the plane, the answer is
the Delaunay triangulation, in which the circumcircle of each triangle contains no other vertex. See [S]. It is almost
always unique. It can be constructed by local operations beginning with any triangulation. Consider any edge as the
diagonal of the quadrilateral formed by its adjacent triangles. If the angles of the two vertices off of the edge add to
more than , then the circumcircle criterion is violated, and the diagonal should be switched to form a replacement
pair of facets.
Suppose now we have a triangulation of a curved surface in space. For any edge with two adjacent facets, we
switch the edge to the other diagonal of the skew quadrilateral if the sum of the angles at the off vertices is more than
. Let a be the length of the common edge, b and c the lengths of the other sides of one triangle, and d and e the
lengths of the other sides of the other triangle. Let 1 and 2 be the off angles. Then by the law of cosines
a2 = b2 + c2 2bc cos 1 ,
a2 = d 2 + e2 2de cos 2 .
(16.145)
The condition 1 + 2 > is equivalent to cos 1 + cos 2 < 0. So we switch if

b2 + c2 a2 d 2 + e2 a2
+
< 0.
bc
de
(16.146)
This is guaranteed to terminate since a switch reduces the radii of the circumcircles and there are a finite number of
triangulations on a finite number of vertices. When using the u command to equiangulate, you should use it several
times until no changes happen. If you do get a genuine infinite loop, report it as a bug.
250
251
16.13 Dihedral angle

The dihedral angle attribute of an edge is defined in the soap film model for any edge with two adjacent facets. (Other
edges return a dihedral angle of 0.) Letting the common edge be a, and letting b and c be the other edge vectors from
the base of a, we use the inner product of 2-vectors:

a c a a

b c b a
< a b,c a >
(16.147)
cos =
=
.

||a b||||c a||
a a a b1/2 c c c a1/2

b a b b a c a a
If the denominator is 0 or cos > 1, then = 0. If cos < 1, then = . Note that this definition works in any
ambient dimension.
In the string model, this is the angle from straightness of two edges at a vertex. If there are less than two edges,
the value is 0. If two or more edges, the value is 2*asin(F/2), where F is the magnitude of the net force on the vertex,
assuming each edge has tension 1. Upper limit clamped to pi.
16.14 Area normalization

By default, the gradients of quantities like energy and volume are calculated simply as the gradient of the quantity as
a function of vertex position. This gives the force on a vertex, for example. But sometimes it is desirable to have force
per area instead, for example in directly estimating mean curvature. The mean curvature vectorfield h of a surface is
defined to be the gradient of the area of the surface in the sense that if the surface is deformed with an instantaneous
velocity u then the rate of change of surface area is
dA
=
dt

sur f ace
u hdA.
(16.148)
According to the triangulation formulation, with Fv being the force on vertex v, the rate of change is
dA
= u(v) Fv .
dt
v
(16.149)
h = Fv /dAv
(16.150)
Hence as approximation to h, we take

and take the area dAv associated with a vertex to be one-third of the total areas of the facets surrounding the vertex.
Since each facet has three vertices, this allocates all area. Command a can be used to toggle area normalization.
16.15 Hidden surfaces

This section describes the algorithm used by the painter graphics interface to hide hidden surfaces.
1. Each facet has the viewing transformation applied to it.
2. Each facet has its maximum and minimum in each dimension recorded (bounding box).
3. Facets are sorted on their maximum depth (rear of bounding box).
4. List is traversed in depth order, back to front. If a facets bounding box does not overlap any others, it is drawn.
If it does overlap, a linear program tries to find some point where one facet is directly in front of the other. If there is
such a point and the order there is backward, the facets are switched on the list. The traversal then restarts there. If it
makes over 10 switches without drawing a facet, it gives up and draws the current facet anyway.
251
252
16.16 Extrapolation
The final energy value at refinement step n is saved as En . Assuming the difference from minimum energy follows a
power law of unknown power, three successive values are used to extrapolate to the minimum energy E0 :
= E0 + acb ,
= E0 + a(2c)b ,
En
En1
(16.151)
= E0 + a(4c) .
En2
(16.152)
Hence
E0 = En
(En En1 )2
.
En 2En1 + En2
(16.153)
16.17 Curvature test

This only applies to the quadratic model.
16.18 Annealing (jiggling)

This happens at each command j , or at each iteration if the jiggling option (command J )is activated. Each coordinate
of each non-FIXED vertex is moved by
x = gT L
(16.154)
where g is a random value from the standard Gaussian distribution (calculated from the sum of five random values
from the uniform distribution on [0,1]), T is the current temperature, and L is a characteristic length that starts as the
diameter of the surface and is cut in half at each refinement.
16.19 Long wavelength perturbations (long jiggling)

This is command longj. An amplitude A and a wavelength L are chosen at random from a unit sphere by picking
random vectors in a cube of side 2 centered at the origin and rejecting attempts that are outside the unit sphere. A
and L are then multiplied by the size of the surface to give them the proper scale. The wavelength is inverted to a
wavevector w. The amplitude is multiplied by the temperature. A random phase is picked. Then each vertex v is
moved by A sin(v w + ).
16.20 Homothety
This scales the total volume of all bodies to 1 by multiplying each coordinate by
(
)1/3
Vbody
(16.155)
(16.156)
bodies
In the string model, the factor is
)1/2
Vbody
bodies
252
253
16.21 Popping non-minimal edges

All facets are assumed to have equal tension. This looks for edges with more that three facets that are not fixed or on
boundaries or constraints. When found, such an edge is split with a new facet in between. The two closest old facets
are attached to the new edge. This is repeated until only three facets are on the original edge. Each split is propagated
along the multiple junction line as far as possible. If it is impossible to propagate beyond either endpoint, the edge is
subdivided to provide a vertex which can be split.
16.22 Popping non-minimal vertex cones

String model
Any vertex with four or more edges is tested. The edges into the vertex are separated into two contiguous groups. The
endpoints of one group are transfered to a new vertex, and a short edge is introduced between the new and old vertex.
All groupings are tested to see which one pulls apart the most. This method works when the edges have non-equal
tensions.
Soapfilm model
All facets are assumed to have equal tension. This assumes that all edges have at most three facets adjoining, so edge
popping should be done first. Here the facet and edge structure around each vertex is analyzed to find which have the
wrong topology. The analysis is done by looking at the network formed by the intersection of the facets and edges
containing the vertex with the surface of a small sphere around the vertex. The numbers of sides of the cells of the
network are counted. A simple plane vertex has two cells of one side each. A triple edge vertex has three cells of two
sides each. A tetrahedral point has four cells with three sides each. Any other configuration is popped.
The popping itself is done by replacing the vertex with a hollow formed by truncating each cell-cone except the
cell with the largest solid angle. In case the network is disconnected, the solid angles of all the cells will add up to
over 4. Then the vertex is duplicated and the different components are assigned to different vertices. This lets necks
shrink to zero thickness and pull apart.
16.23 Refining
First, all edges are subdivided by inserting a midpoint. Hence all facet temporarily have six sides. For edges on
constraints, the midpoints get the same set of constraints, and are projected to them. For edges on boundaries, the
parameters of the midpoint are calculated by projecting the vector from the edge tail to the midpoint back into the
parameter space and adding that to the tail parameters. This avoids averaging parameters of endpoints, which gives
bad results when done with boundaries that wrap around themselves, such as circles. The new edges and the midpoint
inherits their attributes from the edge.
Second, each facet is subdivided into four facets by connecting the new midpoints into a central facet. The new
edges inherit their attributes from the facet.
16.24 Refining in the simplex model

The goal is to divide an n-dimensional simplex into 2n equal volume simplices, although these cannot be congruent
for n > 2. First, each 1-dimensional edge is subdivided by inserting a new vertex. This vertex inherits the intersection
of the attributes of the edge endpoints (i.e. fixity and constraints). Second, the n + 1 tips of the old simplex are lopped
off, leaving a core. This core is recursively subdivided. An edge midpoint is picked as a base point. Faces of the
core not adjacent to the base point are either simplices (where tips got lopped off) or cores of one lower dimension.
The core is given a conical subdivision from the base point to simplicial subdivisions of the nonadjacent faces. The
base point is always picked to have the lowest id number, in order that refinements of adjacent simplices are consistent.
253
254
16.25 Removing tiny edges

The edge list is scanned, looking for edges shorter than the cutoff length. It is NOT removed if any of the following
are true:
1. Both endpoints are FIXED.
2. The edge is FIXED.
3. The boundary and constraint attributes of the endpoints and the edge are not identical.
For an edge that passes these tests, one endpoint is eliminated. If one endpoint is FIXED, that is the one kept. All the
facets around the edge are eliminated (collapsing their other two edges to one), and the edge itself is eliminated.
16.26 Weeding small triangles

This is meant to help eliminate narrow triangles that tiny edge removal cant because they dont have short edges. This
procedure finds the shortest edge of a triangle and eliminates it by the same process as the regular tiny edge removal.
If that fails, it tries the other edges.
16.27 Vertex averaging

For each vertex, this computes a new position as the area-weighted average of the centroids of the facets adjoining
the vertex. FIXED vertices or vertices on boundaries are not moved. Vertices on triple lines are only averaged with
neighboring vertices along the triple line. Tetrahedral points do not get moved. For normal vertices, the scheme is as
follows: If vertex v is connected by edges to vertices v1 , . . . , vn , then the new position is calculated as
vavg =
f acets area xcentroid

.
f acets area
(16.157)
The volume one one side of all the facets around the vertex calculated as a cone from the vertex is
V=
v N f
(16.158)
f acets f
where N f is the facet normal representing its area. The total normal N is
N =
N f .
(16.159)
f acets f
To preserve volume, we subtract a multiple of the total normal to the average position:
(vavg N) N =v N,
so
(16.160)
vavg N v N
.
N N
(16.161)
vnew = (vavg N).
(16.162)
=
Then the new vertex position is
Constrained vertices are then projected to their constraints.

The rawv command will do vertex averaging as just described, but without conserving volumes. The rawestv
command will also ignore all restrictions about like constraints and triple edges, but it will not move fixed points or
points on boundaries. Points on constraints will be projected back to them.
254
255
16.28 Zooming in on vertex

First, all vertices beyond the cutoff distance from the given vertex are deleted. Then all edges and facets containing
any deleted vertices are deleted. Any remaining edge that had a facet deleted from it is made FIXED.
16.29 Mobility and approximate curvature

This section derives the resistance and mobility matrices used in approximate curvature. The derivation is done in the
full generality of vertex constraints, boundaries, volume and quantity constraints, and effective area.
be a
First, we must define the inner product or metric on global vectors, which is the resistance matrix. Let W
global vectorfield, that is, W has a component vector Wv at each vertex v. At each vertex v, let Uv,i be a basis for the
degrees of freedom of motion of vertex v due to level set constraints or boundaries. Consider one edge e in the string
model. Let P be the projection operator onto the normal of e if effective area is in effect, and let P be the identity map
if not. Then the inner product of two vector fields is
,V > = L
<W
1
0
a + tW
b ) P((1 t)Va + tVb )dt
P((1 t)W
1
1
1
1
= L( PW
Va + PW
Vb + PW
Va + PW
Vb ).
a P
a P
b P
b P
3
6
6
3
(16.163)
(16.164)
If the vectors are expressed in terms of the bases Uv with coefficient column matrices Wv and Vv , then
,V >= L( 1 WaiU
a,i PU
a, jVa j
<W
3 i, j
+
1
b,i PU
a, jVa j
WbiU
6
i, j
a,i PU
b, jVb j
+ 16 i, j WaiU
b,i PU
b, jVb j ).
+ 13 i, j WbiU
(16.165)
So the components of the resistance matrix S are

Sai,a j
Sai,b j
Sai,a j
Sai,b j
L
a, j ,
Ua,i PU
3
L
b, j .
Ua,i PU
6
(16.166)
In the soapfilm model, for a facet of area A,

A
a, j ,
Ua,i PU
6
A
b, j .
Ua,i PU
12
(16.167)
The total matrix S is formed by adding the contributions from all edges. Note that S has a block on its diagonal for
each vertex and a block for each pair of edges joined by an edge. Hence it is reasonably sparse. S is a symmetric
positive semidefinite matrix, positive definite if effective area is not in effect.
Now suppose we have a form H to convert to a vector. H is represented as a covector Hv at each vertex v. The
mobility matrix M = (Mb j,ai ) is the inverse of S. The first step is to find the coefficients in the dual basis of U by
contracting:
a,i , Hv > .
ha,i =< U
(16.168)
The U basis components of the vector dual to H are then
cb, j = Mb j,ai ha,i .
255
(16.169)
256
Actually, the components are found by solving the sparse system

Sai,b j ca,i = ha,i .
(16.170)
The foregoing ignores volume and quantity constraints. Suppose Bk are gradient forms for the volume and quantity
constraints. We need to project the motion in such a way as to guarantee that when the velocity is zero, the energy
gradient is a linear combination of constraint gradients.
pro j(H) = H k Bk .
(16.171)
V = M(H k Bk ).
(16.172)
The projected velocity is

The condition for motion tangent to the constraints is
Bm M(H k Bk ) = 0
for each m.
(16.173)
Therefore
and so
Bm MBk k = Bm MH,
(16.174)
k = (Bm MBk )1 Bm MH.
(16.175)
256
Chapter 17
Miscellaneous
This chapter contains some miscellaneous topics that are not of interest to the ordinary user.
17.1 Customizing graphics

This section is provided for those people who need to write their own graphics interface module. All device-specific
graphics output has been collected into a few basic routines. There are two styles of interface. One provides facets
in random order for routines that can do their own hidden surface removal. The other provides facets in back to front
order (painter algorithm).
17.1.1
Random-order interface
The random-order interface has several global pointers to functions which should be set to the users functions. This
permits several sets of graphics routines in one program. The facets are generated by graphgen(), which call the user
functions. The function display() is called as the overall display function; it should set the function pointers and call
graphgen(). Example: iriszgraph.c
The function pointers are:
void (*graph_start)(void);
This is called at the start of each display. It should do whatever device initialization and screen clearing is
needed.
void (*graph_facet)(struct graphdata *,facet_id)}
This is called to graph one triangular facet. See extern.h for the definition of graphdata. The second
argument is a facet identifier for routines that are not happy with just the information in graphdata; it is
for Evolver gurus only. Facets are presented in random order. The coordinates are not transformed; the
current transformation matrix is in view[][], which is in homogeneous coordinates.
void (*graph_edge)(struct graphdata *)}
This function is called to graph one edge in the string model.
void (*graph_end)(void)}
This function is called after all data has been given.
257
17.1.2
258
Painter interface
The painter model is similar, except there is an extra layer of functions. Example: xgraph.c, psgraph.c, gnugraph.c. In
display(), the user should set
graph_start = painter_start;
graph_facet = painter_facet;
graph_end
= painter_end;
The user should also set these function pointers:
void (*init_graphics)(void);
Called by painter_start() to do device initialization.
void (*display_facet)(struct tsort *);

Called by painter_end() to graph sorted facets one at a time. See extern.h for struct tsort.
void (*finish_graphics)(void);
Called at the end of all facets.
The user should set graph_edge as in the random interface.
17.2 Dynamic load libraries

Many Evolver features, such as level set constraints, parametric boundaries, named method integrands, and Riemannian metrics require user-defined functions of a set of arguments. The expressions for these functions are ordinarily
stored as a parse tree and interpreted each time needed, which can be much slower that evaluating compiled expressions. There is a way to use a set of compiled functions specific to a datafile through a mechanism known as dynamic
loading. Here a library of functions for a datafile is separately compiled, and then loaded at runtime when a the datafile
is loaded. Currently, the Evolver only implements a dynamic loading mechanism found on many unix systems, whose
presence can be tested by looking for the existence of the file /usr/include/dlfcn.h. If it exists, you can enable dynamic
loading by including -DENABLE_DLL in the CFLAGS line in the Makefile. On some systems, you may need to include
-ldl on the GRAPHLIB line also, to link Evolver with functions such as dlopen().
To create the library for a datafile, write a source file containing C code for the desired functions, compile it, and
link it into a shared library. The function should be able to compute the value and the partial derivatives of the function,
and its second partials if you are going to use any Hessian features. A sample source file for a 2-dimensional datafile:
#define FUNC_VALUE 1
#define FUNC_DERIV 2
#define FUNC_SECOND 3
#define MAXCOORD 4
/* must be same as in Evolver!! */
#define REAL double /* long double if Evolver compiled with -DLONGDOUBLE */
struct dstack \lbrace REAL value;
REAL deriv[2*MAXCOORD];
REAL second[2*MAXCOORD][2*MAXCOORD]; \rbrace;
258
259
void func1 ( mode, x, s )

int mode; /* FUNC_VALUE, FUNC_DERIV, FUNC_SECOND */
REAL *x; /* pointer to list of arguments */
struct dstack *s; /* for return values */
\lbrace REAL value;
s->value = x[0] + x[1]*x[1];
if ( mode == FUNC_VALUE ) return;
/* first partials */
s->deriv[0] = 1.0;
s->deriv[1] = 2*x[1];
if ( mode == FUNC_DERIV ) return;
/* second partials */
s->second[0][0] = 0.0;
s->second[0][1] = 0.0;
s->second[1][0] = 0.0;
s->second[1][1] = 2.0;
return;
\rbrace
Supposing the sourcefile name to be foo.c, compile and link on SGI systems (IRIX 5.0.1 or above) with
cc -c foo.c
ld -shared foo.o -o foo.so
Sun systems are the same, but with -s in place of -shared. For other systems, consult the ld documentation for the
option to make a shared library or dynamic load library.
To use the functions in a datafile, include a line at the top of the datafile before any of the functions are used:
load_library "foo.so"
Up to 10 libraries may be loaded. Afterwards, any of the functions may be invoked just by using their name, without
an explicit argument list because the argument list is always implicit where these functions are legal. Examples,
supposing func2 is also defined with one argument:
constraint 1
formula: func1
x1: func2
x2: 3*func2 + sin(p1)
It is up to you to make sure the number of arguments your function expects is the same as the number implicit in the
use of the function. You do not need to explicitly declare your functions in the datafile. Any undefined identifier is
checked to see if it is a dynamically loaded function.
NOTE: This implementation of dynamic loading is experimental, and the interface described here may change in
the future.
259
Chapter 18
Helpful hints and notes

18.1 Hints
This is a collection helpful hints gained from my own experience with Evolver and from helping others.
Evolver works in dimensionless units, and the default settings work best when size, surface tension, volume, etc.
are near 1. If you decide to work in units that give very large or small numbers, you may have to adjust parameters
such as scale_limit, target_tolerance, and constraint_tolerance.
When drawing a sketch for constructing the initial datafile, make it as big as you can. You will have lots of notation
to put on it. Number all vertices, edges, and facets. Put orientation arrows on the edges, and indicate the orientation
of facets (I like to use curved arrows around the facet numbers).
Initial faces should be convex. Although Evolver handles nonconvex faces, the triangulation algorithm is very
simple-minded, and the triangulation of a nonconvex face can be ugly. Just put in an extra edge or two to divide the
face into a couple of convex faces.
Make separate constraints for edges with constraint energy or content integrals, and for edges without. Even if the
other edges are fixed, it is much easier to check that the integrands are correct when only the precisely needed edges
are on constraints with integrals.
If you dont have all your elements numbered consecutively (which usually happens due to numbering schemes
you use, or adding or deleting elements), run Evolver with the -i command line option so mouse-picking reports the
same element numbers as in your datafile. You can instead put keep_originals in the top of your datafile for the same
effect.
Make sure all your body facets are oriented properly. Evolver will complain if there are mismatched facet orientations on an ordinary edge, but fixed edges, constrained edges, etc. are exempt from this checking. A good way to
check is by coloring, for example:
set body[1].facet color green
Make sure vertices, edges, and facets are on their proper constraints. You can check visually by coloring, e.g.
set edge color red where on\_constraint 1
set facet color green where on\_constraint 1
You cant color vertices directly, but you can get close to the same effect by refining a couple of times and coloring
edges adjacent to vertices:
foreach vertex vv where on\_constraint 1 do set vv.edge color blue
Check that all the energies, volumes, quantities, etc. in your initial datafile are correct. See the section on reasonable scale factors below for more details on how to check in great detail.
If you are doing liquids with contact lines on solid walls, I suggest making the first datafile with all the boundary
surfaces of the liquid represented explicitly as facets, and then make a second version of the datafile using constraint
260
261
energy and content integrals to replace the facets on the fixed walls. It is far easier to get the energies and volumes
right in the first version, but it is also far more prone to problems during evolution. Use the first version to check the
correctness of the second version, and use the second version for serious work.
If your edges on curved constraints try to short-cut the curve, there are several ways to discourage that:
1. Make a second guide constraint, so that the intersection of the two constraints define guiderails for vertices to
run along. By using vertex attributes to customize the guide constraint, you only need one guide constraint. For
example:
define vertex attribute guides real[2]
constraint 1
formula: x\^2 + y\^2 = rad\^2
// curved constraint
constraint 2
formula: guides[1]*x + guides[2]*y = 0 // radial guide planes
Then you can set the guide coefficients at runtime with
set vertex.guides[1] -y where on\_constraint 1
set vertex.guides[1] x where on\_constraint 1
2. If you understand exactly what energy or volume condition is encouraging the short-cutting, you can adjust the
energy or content integrand on the curved constraint to compensate enough to eliminate the encouragement.
This basically means calculating the surface area of the gap between the edge and the curved constraint, or the
volume bounded by the gap.
3. Declare the curved constraint CONVEX. This adds an energy roughly proportional to the gap area. This is simple
to do, and works if you set the gap_constant high enough (you should leave the gap constant as low as will work,
however), but you cannot use any Hessian commands if you use convex constraints.
Run at low resolution before refining. A good evolution script usually winds up having alternating refining and
evolultion. Having many triangles not only takes a long time to calculate, but motion can propagate only one triangle
per iteration. Dont over-evolve at a particular refinement. Remember its an approximation. There is not much point
in evolving to 12 digits precision an approximation that is only accurate to 4 digits.
Groom your surface triangulation with V (vertex averaging), u (equiangulation), l (long edge division), and t (tiny
edge deletion). It may take some experimenting to get the right sequence, along with refinements. It may be better to
divide certain long edges than simply refine the whole surface. However, overdoing it may be counterproductive to
convergence; sometimes the converged surface doesnt want to be entirely equiangulated or averaged, and you can get
into an endless loop of iteration and grooming. Once you work out a good script, write it down in a handy command
at the end of the datafile for easy use.
Use the dump or d commands to save your evolved surface regularly. Remember that Evolver has no undo feature
to roll back disastrous commands.
Use conjugate gradient mode for faster gradient descent, but not too soon. Use regular gradient descent to adjust
to volume or constraint changes. Conjugate gradient should be used only when regular motion has settled down.
Conjugate gradient assumes a quadratic energy function, and may get confused when its not. Conjugate gradient may
need to be toggled off and on to make it forget its history.
During gradient descent (including conjugate gradient), keep an eye on the scale factor. The scale factor should
remain fairly steady. A scale factor going to 0 does NOT mean convergence; it means the surface is having trouble.
However, a good scale factor may depend on refinement and other considerations. See the section on reasonable scale
factors below.
Second-order Hessian convergence is much faster than first-order gradient descent, when Hessian works. So my
advice is to use gradient descent just to get to where its safe to use hessian or hessian_seek. Actually, hessian_seek
is pretty much always safe to use, since it makes sure energy is decreasing. I have found circumstances where hessian_seek does an amazingly good job as an iteration step, even though the surface is nowhere near convergence.
261
262
Beware saddle points of energy. A symmetric surface, e.g. a blob of solder on a pad or around a wire, may seem to
converge with gradient descent, but just have reached a saddle point. Use the eigenprobe command to test for stability,
and if not stable, use the saddle command to get off the saddle point.
Judging convergence in gradient descent is tough. If iterations run at a more or less constant scale factor and
energy isnt changing much, and running in conjugate gradient mode for a long time doesnt change much, then youre
probably in good shape. But use the eigenprobe command to make sure, and hessian to finish off convergence.
If you intend to use quadratic mode or Lagrange mode for higher precision, evolve in linear model first until the
final stage, since it is much quicker and there are more triangulation grooming commands available.
18.2 Checking your datafile

You should always check your initial datafile to be sure it is doing exactly what you want. It is easy to get signs on
integrands wrong, or apply quantities to the wrong elements. When you load the initial datafile, the initial energy,
body volumes, and quantities values should be exactly what you expect, either from hand calculation or from another
datafile you trust. In particular, when using constraint integrals to replace omitted facets, I suggest you make a separate
datafile with facets instead of integrals just for checking the agreement between the two.
With the named methods and quantities feature, it is possible to get very detailed information on where numbers
are coming from. If you give the convert_to_quantities command, every energy, volume, and constraint integrand will
be internally converted to named methods and quantities (although the user interface for all remains the same). These
internal quantities are ordinarily not displayed by the v or Q commands, but if you do show_all_quantities then
they will be displayed. Further, Q will show all the component method instances also. For an example, consider the
following output:
Enter command: convert\_to\_quantities
Enter command: show\_all\_quantities
Enter command: Q
Quantities and instances:
(showing internal quantities also; to suppress, do "show\_all\_quantities off")
1. default\_length
64.2842712474619 info\_only quantity
modulus
1.00000000000000
2. default\_area
4.00000000000000 energy quantity
modulus
1.00000000000000
3. constraint\_1\_energy
-0.342020143325669 energy quantity
modulus
1.00000000000000
4. constraint\_2\_energy
-0.342020143325669 energy quantity
modulus
1.00000000000000
5. body\_1\_vol
1.00000000000000 fixed quantity
target
1.00000000000000
modulus
1.00000000000000
body\_1\_vol\_meth
0.000000000000000 method instance
modulus
1.00000000000000
body\_1\_con\_2\_meth
1.00000000000000 method instance
modulus
1.00000000000000
6. gravity\_quant
0.000000000000000 energy quantity
modulus
0.000000000000000
Heres a detailed explanation of the output of the Q command above:
default_length - total edge length, using the edge_length method. This would be the default energy in the string
model, and I guess it really doesnt need to exist in a soapfilm model. But its an info_only quantity, which means it is
only evaluated when somebody asks to know its value.
default_area - the default energy in the soapfilm model, and included in the energy here, as indicated by "energy
quantity" at the right.
262
263
constraint_1_energy - the energy integral of constraint 1, using the edge_vector_integral method applied to all
edges on constraint 1.
constraint_2_energy - the energy integral of constraint 2, using the edge_vector_integral method applied to all
edges on constraint 2.
body_1_vol - the volume of body 1, as a sum of several method instances. body_1_vol_meth is the
facet_vector_integral of (0,0,z) over all the facets on the body. body_con_2_meth is the integral of the constraint
2 content integrand over all edges on facets of body 1 which are edges on constraint 2.
gravity_quant - the total gravitational energy of all bodies with assigned densities. This quantity is always present
even if you dont have any bodies, or dont have any body densities. But youll notice the modulus is 0, which means
its evaluation is skipped, so the presence of this quantity doesnt harm anything.
You can find the quantity or method contribution of single elements by using the quantity or method name as an
attribute of elements. Using a quantity name really means summing over all its constituent methods that apply to the
element. For example,
Enter command: foreach edge ee where on\_constraint 2 do printf "%d %f\n",id, ee.body\_1\_con\_2\_meth
5 0.000000
6 0.000000
7 1.000000
8 0.000000
Enter command: foreach edge where constraint\_1\_energy != 0 do print constraint\_1\_energy
-0.342020143325669
18.3 Reasonable scale factors

Trouble in evolving is usually signaled by a small scale, which means there is some obstacle to evolution. Of course,
that means you have to know what a reasonable scale is, and that depends on the type of energy you are using and how
refined your surface si. In normal evolution, the size of the scale is set by the development of small-scale roughness
in the surface. Combined with a little dimensional analysis, that leads to the conclusion that the scale should vary as
L2q , where L is the typical edge length and the units of energy are lengthq . The dimensional analysis goes like this:
Let D be the perturbation of one vertex away from an equilibrium surface. In general, energy is quadratic around an
equibrium, so
E = D2 Lq2
So the gradient of energy at the vertex is
E = 2DLq2
The motion is the scale times the gradient, which we want proportional to D, so
scale E = scale 2DLq2 = D
So scale is on the order of L2q . Some examples:
Energy
Energy dimension Scale Example file
Area of soapfilm
L2
L0
quad.fe
1
Length of string
L
L1
flower.fe
Squared curvature of string
L1
L3
elastic8.fe
0
Squared mean curvature of soapfilm L
L2
sqcube.fe
In particular, the scale for area evolution is independent of refinement, but for most other energies the scale decreases with refinement.
Another common influence on the scale for area evolution is the surface tension. Doing a liquid solder simulation
in a system of units where the surface tension of facets is assigned a value 470, say, means that all calculated gradients
are multiplied by 470, so the scale decreases by a factor of 470 to get the same geometric motion. Thus you should set
scale_limit to be the inverse of the surface tension.
263
Chapter 19
Bugs
There are no known outright bugs at present. but they undoubtably exist. When you run across one, I would like to
hear about it. Bug reports should be submitted by email to [email protected]. Please include the Evolver version
number, a description of the problem, the initial data file, and the sequence of commands necessary to reproduce the
problem.
There are a few shortcomings, however:
Not all features are implemented in all models.
Vertex and edge popping is not elegant or complete for vertices on boundaries or constraints.
Zero length edges and zero area triangles stall things.
Surfaces can intersect each other without knowing it.
Convergence to a minimum energy can be difficult to judge.
264
Chapter 20
Version history
Version 1.0 August 4, 1989
First public version.
Various bug fixes.
Constraint integral specification changed to have just components, not density.
Facet-edge specification made obsolete (but still legal).
Vertex motion adjustment for quadratic model.
Zoom factor changed from 2 to 1.2.
Initial datafile reading revamped to use lex and yacc. Has simple macros. Comments must be delimited. Expressions
now in algebraic form, not RPN.
Version 1.10 October 18, 1989
More bug fixes.
Graphics driver for HP98731 contributed by Eric Haines.
Datafile reading:
Equations permitted as constraints.
Backslash line splicing.
Constant folding in expressions.
Version 1.11 May 18, 1990
More bug fixes.
x,y,z accepted as synonyms for x1,x2,x3
Datafile coordinates may be given as expressions.
Constraints revamped. Number of constraints per element raised to sizeof(int). One-sided constraints may be applied
selectively. GLOBAL constraints automatically apply to all vertices (they count in number limit).
265
266
Version 1.2 June 17, 1990

Comments removed from input stream before lex analyzer. So comments in macros safe now.
Scale factor upper bound adjustable when toggling with m command.
Adjustable constants implemented. Syntax in data file: PARAMETER name = constant-expression These are treated
as constants in expressions, but can be changed with the A command. Useful for dynamically changing contact angles
and other stuff in constraint expressions.
Symmetric content evaluation added for linear model. Volumes evaluated as surface integral of (xi + yj + zk)/3 rather
than of zk. Use keyword SYMMETRIC CONTENT in datafile. Permits more accurate evaluation of lunes.
Return value of sprintf not used, as this varies among systems.
Arbitrary surface energy integrands added. Vector field integrated over designated facets. Syntax
surface energy {\sl n}
e1:
{\sl expression}
e2:
{\sl expression}
e3:
{\sl expression}
Designate facet by following facet definition with energy n, in same manner as giving constraints. Linear model
only. This useful for changing direction of gravity, by putting in Divergence Theorem equivlent surface integrand for
gravitational potential energy and putting adjustable constant in for direction.
The q or x exit command gives you a chance to load another datafile, continue with the current configuration, or
exit.
You can change the initial upper limit on the scale factor by putting a line in the data file:
SCALE LIMIT value
value must be a number (no expression).
The V command (for Vertex averaging) will replace each unconstrained vertex by the average of its neighboring
vertices (connected by edges). Good for uncramping skinny triangles sometimes.
TENSION allowed as synonym for DENSITY in setting facet surface tension.
Binary save/reload disabled, since it is out of date and a binary file turns out twice as large as the ascii dump file.
Shared memory interface added for MinneView (which is a public domain 3D graphics viewer for Irises written by
the Geometry Project at the Minnesota Supercomputer Institute.)
Histogram added for ridge notcher (option n).
Ported to NeXT (no screen graphics, but PostScript output files can be displayed).
Automatic energy recalculation added after all options that change energy.
Version 1.3 July 30, 1990
Reptition counts before letters in graphics commands.
Clockwise (c) and counterclockwise (C) rotations added to graphics.
Printout during quadratic search for optimum scale suppressed.
i command added for information. v reports just volumes.
Extrapolation command e fixed.
Constant expressions permitted wherever real value needed.
F command to log commands to file.
-ffilename command line option to read commands from file. Take commands from stdin afterwards.
Dump file real values printed to 15 decimal places so accuracy not lost.
Adjustable accuracy on edge integrals via INTEGRAL_ORDER keyword in datafile for setting order of Gaussian
quadrature.
Datafile now allows specification of constraint tolerance with CONSTRAINT_TOLERANCE.
All two-word keywords made into single words by connecting the words with _.
266
267
Parsing of datafile continues after errors.

b command permits editing body volumes.
v command prints none for prescribed volume of those bodies with no prescribed volume.
Manual in TEXformat with PostScript figures.
CONVEX gap energies and forces fixed up. k = 1 best approximation to area.
Vertex popping fixed to handle disjoint components of tangent cones. Also not to screw up its data structures after a
modification.
Version 1.31
Added long-wavelength random jiggle (command jj).
Fixed simultaneous Minneview to handle torus display options.
Simultaneous MinneView can be stopped and restarted.
Fixed bug in Gaussian quadrature of content integrals.
Fixed equiangulation to work on arbitrarily small triangles.
Histograms adjusted to current length and area order of magnitude.
Zero area triangle test area cutoff adjusted to scale of surface.
Vertex averaging improved to preserve volumes.
Small edge removal bug fixed. In collapsing a facet, will preferentially keep an edge on a constraint.
VOLCONST adjustment to body volumes added to datafile.
Quantities added. A quantity is a sum of vector integrals over facets and edges. Can be simply tallied for informational purposes, or can be used as mathematical constraints with fixed values. Good for center of mass, moment of
inertia, magnetic flux, etc.
Conjugate gradient energy minimization added. Use the U command to toggle between gradient descent and this.
Version 1.4 posted August 20, 1990
Version 1.41 September 22, 1990
Bug fixes on vertex popping and element list management.
Long jiggle command jj lets user put in own numbers, use random numbers, or use previous numbers.
Version 1.42
Bug fixes:
segment violation on error message at end of data file.
STRING model not graphing third dimension
On SOAPFILM model, edges without facets are displayed. Useful for showing wire boundaries beyond surface. Such
edges will generate warnings when datafile is read, however.
2 dimensional surfaces can live in N-dimensional space. See SPACE_DIMENSION.
Simplex model added to represent k-dimensional surfaces in N-dimensional space.
Embryonic query language added. Does list and show. Also set and unset attributes.
Background metric on space added for string model only.
Commands can be read from a file with read "filename"
267
268

Riemannian metric extended to all dimensional surfaces.
Refine, delete added to query language. List query has same format output as datafile dump.
Quotient spaces added.
Piping output added to commands and queries.
SET FACET COLOR query added. Colors include CLEAR. Also SET FACET TRANSPARENCY for Iris and like.
Squared mean curvature added as possible energy.
Queries can refer to elements by original number of parent in datafile.
Adjustable constants can take values from a file according to ID number of element applied to.
Revised notch command (n) to subdivide adjacent facets instead of edge itself. Supposed to be followed with equiangulation.
Put in K command to subdivide longest edges of skinny triangles (as judged by their smallest angle).
Added VALENCE attribute for edges for queries. Is the number of facets on an edge.
Added minimization by using Newtons method on Hessian matrix of energy. Only for no-constraint area minimization
with no other energies. Command hessian.
NeXT version given graphical interface.
User-defined functions of coordinates added. See userfunc.c.
Version 1.76 March 20, 1992
Autopopping and autochopping in string model for automatic evolution.
Phase-dependent grain boundary energies.
Approximate polyhedral curvature.
Stability test for approximate curvature.
Squared Gaussian curvature as part of energy only, not force.
system command to execute shell commands.
check_increase to guard against blowup during interation.
effective_area to count only resistance to motion normal to surface.
Runge-Kutta iteration.
Command and query language much extended.
geomview interface added.
Fixed area added as a constraint.
Multiple viewing transforms can be specified in the datafile so one fundamental region of a symmetric surface can be
displayed as the whole surface.
Commands can be included at the end of the datafile, introduced by the keyword READ.
Some alternate definitions of squared curvature added. Invoked by effective_area ON | OFF or normal_curvature
ON | OFF.
268
269

Shaded colors added to xgraph and cheygraph.
Restriction of motion to surface normal added. Toggle tt normal_motion.
Squared mean curvature, Gaussian curvature, and squared Gaussian curvature extended to surfaces with boundary.
Datafile element attribute bare added for vertices and edges in soapfilm model so they wont generate erroneous
warnings.
Force calculation added for squared Gaussian curvature, so it can be used in the energy.
All prompts that require real value responses now accept arbitrary expressions.
User-defined mobility added, both scalar and tensor forms.
Default squared curvature works for 2-surfaces in Rn .
close_show command added to close show window (the native graphics window, not geomview).
Graphics command checks string for illegal characters before doing any transformations.
Dihedral angles now work for 2-surfaces in any dimension.
Permanent variable assignments may be made with ::= instead of :=. Such assignments will not be forgotten when
a new surface is begun.
Conditional expressions of C form added: expr ? expr : expr. Useful for patching constraints together.
Version 1.88 December 16, 1992
SET BACKGROUND color command added for native graphics.
View transformation generators and expressions added.
Exact bounding box calculated for PostScript files.
Version 1.88a January 6, 1993
Default constraint_tolerance lowered from 1e-5 to 1e-12.
Fixed bug in volume constraint calculation introduced in 1.88.
Version 1.89 February 18, 1993
Postscript draws fixed and boundary edges in interior of surface. All internal graphics should be consistent in the
special edges they draw (bare edges, triple edges, etc.).
Viewing matrix can be read from datafile and will be dumped. Keyword VIEW_MATRIX
Mod operator % added, and functions floor(), ceil().
rebody command added to recalculate connected bodies after neck pinching and any other body disruption.
If squared mean curvature part of energy, then squared mean curvature at a vertex is available as a query attribute as
"sqcurve".
These quantities can now be used in command expressions: vertex_count, edge_count, facet_count, body_count,
total_energy, total_area, total_length, scale.
Dump file records defined procedures in read section at end.
Knot energies added, both conducting and insulating wire.
dissolve command added to erase elements and leave gaps in surface, unlike delete command, which closes gaps.
Can only dissolve elements not neede by higher dimensional elements.
269
270
Command repeat numbers have been restricted to just three types of commands: 1. single letter commands that dont
have optional arguments (l,t,j,m,n,w have optional arguments) 2. command list in braces 3. user-defined procedure
names This is to prevent disasters like list vertex 1293 which before would produce 1293 full lists of all vertices.
dump without argument will dump to default file name, which is datafile name with .dmp extension.
SIGTERM is caught and causes dump to default dump file and exit. Useful for interrupting scripts running in the
background with kill -TERM. Likewise for SIGHUP.
Version 1.90 April 2, 1993
Conjugate gradient ON/OFF state saved in dump file. Note that conjugate gradient history vector is not saved.
Notching and dihedral attribute apply to vertices in the string model.
FOREACH iterator added. Syntax: FOREACH element [ name ] [WHERE expr] DO command
LOAD command added. Syntax: LOAD filename. Useful for starting new surfaces, especially in scripts.
PRINTF command added for formatted printing. Syntax: PRINTF "format string",expr,expr,... expr is floating point,
so use %f or %g formats.
String variables have been added. Can be used where quoted strings needed. Can be assigned to. SPRINTF is version
of PRINTF giving string output.
A view transformation matrix in the datafile may be preceded by "color n" to give that transform a color (overrides
any facet color).
In queries, element attribute oid added, which returns a signed version of id.
Many knot energies added. Also a hooke energy that keeps edges near a uniform length.
PostScript output optionally includes color.
Two sides of facets can have different colors. COLOR applies to both sides, FRONTCOLOR and BACKCOLOR
to different sides.
Attributes of individual named elements can be set inside loops, i.e. foreach facet ff do set ff color red
Every time a command changed a global variable, the surface was being recalculated. This slowed down scripts
immensely. So now the only variables that cause recalculation are 1) adjustable parameters defined in the datafile 2)
quantity moduli and parameters
History commands now echoed.
Surface area can be minimized by minimizing Dirichlet integral, according to scheme of Polthier and Pinkall. Command dirichlet.
To reduce need for explicit line-splicing on long commands, the parser now keeps track of depth of brace and parenthesis nesting, and will call for more input if a line ends inside nest. So if you want to type a multiline command, start
with and end with many lines later. Also does auto line-splicing if certain tokens are last token in line (such as
+).
facet_knot_energy_fix method added.
Command assignment fixed to assign only one command. so ggg := g; g will be the same as { ggg := g}; g and not
ggg := { g;g}.
Queries can run through edges and facets adjacent to a vertex, and facets of a body, as in list vertices vv where
max(vv.facet,color==red) > 0
Improved NeXT terminal interface. -u option for no graphics, -t for terminal and graphics.
view_4d command to toggle sending full 4D coordinates to geomview. Default is OFF.
SGI parallelism enabled for named quantity calculations.
method-instance scheme introduced.
Torus periods can be specified with expressions using adjustable parameters.
Verbs (list,refine,delete,dissolve) may apply to single elements: foreach edge ee do refine ee
270
271
FIX and UNFIX may be used as verbs:

fix vertices where on_constraint 1
Toggle command names may now be used as boolean values in expressions in commands. Also new boolean
read-only variables: torus, torus_filled, symmetry_group, simplex_representation. New numeric read-only variables:
space_dimension, surface_dimension, integration_order.
Mac version repeated commands interruptable with control-..
Macintosh and Dos versions pipe to a file instead of a command:
Enter command: list vertices | "filename"
For a torus domain, the torus periods may be specified using expressions with parameters, so the fundamental cell may
be changed interactively. Do a recalc after changing such a parameter to update the torus periods.
gv_binary toggle for binary/ascii data to geomview. Default ON for binary, which is faster. Ascii mode useful for
debugging.
history command added to print command history. Single-letter commands now included in history for convenience.
But history does not record responses to prompts commands may issue.
command repeat counts can now be expressions
More internal variables for counters on command events: equi_count, delete_count, notch_count, dissolve_count,
pop_count, where_count.
Quantities for scalar and vector integrands over edges and facets added.
Quantity names may now be used as element attributes. Value of total quantity must be referred to as total quantityname.
DOS version has improved graphics. Recognizes higher resolution and more colors.
Can apply named quantities to elements with SET command.
Undocumented user_attr.
Version 1.94 January 24, 1994
New named quantity methods: vertex_scalar_integrand, facet_2form_integral.
Hessians for named quantity methods: edge_length, facet_area, vertex_scalar_integral, edge_scalar_integral,
edge_vector_integral, facet_scalar_integral, facet_vector_integral, facet_2form_integral, gravity_method.
Added edge wrap as readable attribute.
Added coordinate attributes for edges and facets. Interpreted as edge vector components and facet normal components.
Commands are added to history list after being successfully parsed, rather than after successful execution.
Unfound files are treated as errors rather than prompting for new name, except for datafiles.
New arithmetic operators: mod (synonym for %), imod, idiv. New arithmetic function: atan2(y,x).
Show conditions for edges and facets are saved in read section of datafile.
Total energy is in a comment at the top of a dump file.
PostScript output in case of string model in 3D has option for doing bordered crossings.
w is synonym for coordinate x4.
Named quantity methods for squared curvature: sq_mean_curvature, eff_area_sq_mean_curvature, normal_sq_mean_curvature. All work with h_zero.
New quantities edge_general_integral, facet_volume, facet_torus_volume, facet_general, stress_integral, edge_area,
edge_torus_area. Also quadratic model versions of basic quantities. Also hessian.
Quadratic midpt for edges included in dump. Optional in datafile.
midv attribute for edges in quadratic model.
iteration_counter variable for printing current repetition number.
New math functions: tanh, asinh, acosh, atanh.
271
272
Extra attributes, indexable. Can have non-dumping extra attributes.

quietgo toggle to suppress output of just the g command.
ribiere toggle for conjugate gradient. Does not initiate CG. Also extra projections to constraints in conjugate gradient
or post-project mode, with iteration_counter set to -1 if nonconvergence in 10 projections.
assume_oriented toggle for square mean curvature.
Label option added to PostScript output.
Short-circuit evaluation of AND and OR.
Delete facet more aggressive; tries to eliminate all edges of facet in increasing length order until it succeeds.
G command takes numerical argument instead of repetition count.
hessian_menu command has experimental stuff for hessian, eigenvalues, and eigenvectors. Saddle will seek along
lowest eigenvector without the need to go into the menu.
sprintf, printf now accept string args, but they are pushed as 8 bytes, so %s in format string should be followed by half.
Extra attributes now inherited for same type element.
Added jiggle toggle, and jiggle_temperature internal variable.
Added total_time internal variable. Settable also.
Compound quantities allowing quantity energy to be a function of method instances.
Indexing on element generators.
Permitted element attributes in datafile expressions for quantities, etc, altho only works in named quantities.
hessian_normal flag for hessian motion constrained to surface normal at non-singular points.
Made ribiere default mode for conjugate gradient.
geomview string command added to let scripts send commands to geomview.
The print command also accepts strings. Example: print datafilename
The name of the current datafile can be referred to in commands as datafilename wherever a string can be used.
Named quantities can refer to qqq.value, qqq.target, qqq.pressure, and qqq.modulus.
Alternative Hessian factoring (via ysmp toggle), with Bunch-Kauffman option.
lanczos and eigenprobe commands added.
Dump saves states of toggles.
Lanczos(t,n) and ritz(t,n) commands added. Linear metric for eigenvalues.
-q option to turn everything into quantities.
geomview picking.
break and continue commands.
P, M commands take arguments
random_seed variable added for user control of random numbers.
Added 12-pt degree 6 and 28-pt degree 11 integration rules for facets. Made default facet cubature 12-pt 5th degree.
Much more stable for area. Separate integral_order_1D and integral_order_2D variables.
target, volconst attributes for bodies and quantities.
Higher degree cubature formulas.
linear_metric for hessian.
edgeswap command.
Torus translations added automatically to view transform generators.
272
273
Element structures only allocate needed storage.

convert_to_quantities command.
Constraint limit raised to 127.
Toggles print previous values.
Hessian operations save direction of motion. Eigenvalue saved in last_eigenvalue, stepsize in last_hessian_scale.
Eigenprobe(, n) finds eigenvector. Hessian menu option G minimizes squared gradient instead of energy for saddle
and hessian_seek. Can pick Ritz vector in hessian menu. Saddle and hessian_seek commands take stepsize limit
arguments.
move command.
-DSDIMn compiler option to hardwire dimension for optimization.
Hessian calculation parallelized for SGI_MULTI mode.
Bottominfo command.
Redefinition of single-letter commands.
geompipe toggle.
P command takes argument to short-circuit menu.
Version 2.00 April 30, 1996
HTML version of documentation, also used by help command.
Lagrange model.
Quad precision when compiled with -DLONGDOUBLE.
Windows NT version.
X command prints extra attribute dictionary.
optimizing_variable introduced.
postscript command with toggles instead of interactive (as in P 3)
return command for ending current command.
Mac 68K and Power PC versions.
"node_charge" vertex attribute for knot_energy. Useful for spreading graphs out.
new_vertex, etc.
V modified; vertex_average; works better on constrained vertices and quadratic edge midpoints.
Simplex equiangulation in 3D.
no_refine attribute for edges and vertices.
>> redirection
dynamic link libraries for functions.
DOS, Windows version leaves alphanumeric escape sequences alone.
Can print transform_expr, transform_count.
Window NT/95/98 version using OpenGL has much better graphics. Can rotate, translate, or zoom with left mouse
button, pick elements with right button, even do cross-eyed stereo. Type h in the graphics window for a command
summary. Also made catenoid icon for the program.
C-style assignment operators +=, -=, *=, /= work where reasonable.
Parameterized boundaries can have energy and content integrals, in the same way level set constraints have had.
Command output redirection to a file using >> for append, >>> for overwrite.
A variable can be toggled between optimizing or non-optimizing at run time with "unfix varname" and "fix varname"
respectively.
"keep_macros" in datafile header keeps macros active after datafile.
273
274
Command line option -i will keep element ids the same as in the datafile, rather than renumbering consecutively as is
the default.
If you want to reorder elements in the internal lists (the way elements are listed by, say, list vertices, you can define the
extra attributes vertex_order_key, edge_order_key, facet_order_key, body_order_key, facetedge_order_key, give them
all appropriate values, and then give the command reorder_storage. See reorder.cmd in the distribution fe directory.
renumber_all renumbers elements in internal list order.
Read-only internal variable random for random numbers uniformly distributed between 0 and 1.
Added warnings about using keywords as identifiers in the datafile. The datafile will still run, but your commands will
misinterpret those identifiers.
All keywords are in the on-line help. Do help "keyword" if you want to check on a potential keyword. help
also recognizes identifiers defined by the user in the current surface. For testing in scripts, there is a function
is_defined(stringexpr) that returns 1 if a name is already in use, 0 if not.
In the datafile, making an edge FIXED no longer fixes its endpoints. This is compatible with how fixed facets and
fixing edges at run time have always worked.
For named methods that logically depend on the orientation of the element (i.e. facet_vector_integral, etc.), the relative
orientation of the element when the method is applied is recorded. Default is positive in the datafile, unless a - is
added after the name of a method or quantity applied to an individual element.
Added "verbose" flag for action messages from pop edge, pop vertex, delete, notch, refine, dissolve, edgeswap, unstar.
IGNORE_FIXED and IGNORE_CONSTRAINT flags for sq_mean_curvature methods.
Old fixed_area finally gotten rid of.
Assignment statements permissible at start of expressions; useful for common subexpressions in constraint and quantity integrands.
hessian_slant_cutoff variable for controlling hessian on constraints.
-e option to echo input; useful for piped input.
Automatic conversion to named quantities when needed; suppressed by -a- option.
"scale" attribute for optimizing parameters for impedance matching of scale factors.
hessian_normal is now defaults to ON.
Put in automatic convert_to_quantities; still a place or two where cant convert on the fly. Command line option -adisables.
Added "verbose" toggle command for action messages from pop edge, pop vertex, delete, notch, refine, dissolve, and
edgeswap.
The "dissolve" command will now dissolve facets on bodies in the soapfilm model, and edges on facets in the string
model.
Edges have frontbody and backbody attributes in the string model.
There is a chdir command to change the working directory.
Element extra attributes can be declared with code to evaluate their values.
GLUT OpenGL graphics, with pull-down menu and multiple windows.
Mac OSX version.
Multiply dimensioned arrays.
Postscript can do visibility check to cut down output size.
274
275

Multi-dimensional arrays for variables and element extra attributes.
Functions and procedures with arguments.
Local variables.
FOR loop control construct.
Augmented hessian.
Sparse constraints.
Runtime defines of named quantities, method instances, constraints, and boundaries.
More popping commands: t1_edgeswap, pop_quad_to_quad, pop_tri_to_edge, pop_edge_to_tri; and toggles
pop_disjoin, pop_to_face, pop_to_edge.
Vertex_merge(), edge_merge(), and facet_merge().
Spherical arc methods.
star_perp_sq_mean_curvature method, best yet I think. And the star methods now work on partial stars.
cpu_counter variable for really high-resolution timing.
PDF version of manual with bookmarks and links.
binary_printf, reverse_orientation, quietload commands.
eigenvalues array.
Concatenation of quoted strings.
Version 2.l30 January 1, 2008
Clipping and slicing planes in graphics.
Whole-array operations.
Evmovie display program and binary_off_file.
Addload for multiple file loading.
Simple graphics text.
Showing string facets.
Subcommand prompt.
Debugging breakpoints.
275
Chapter 21
Bibliography
[A] F. Almgren, Minimal surface forms, The Mathematical Intelligencer, vol.4, no. 4 (1982), 164172.
[AV] V. I. Arnold, Catastrophe Theory, 3rd ed., Springer-Verlag, 1992.
[AT] F. Almgren and J. Taylor, The geometry of soap films and soap bubbles, Scientific American, July 1976,
8293.
[AYC] J. Ambrose, B. Yendler, and S. H. Collicot, Modeling to evaluate a spacecraft propellant guaging system,
Spacecraft adn Rockets 37 (2000) 833-835.
[B1] K. Brakke, The motion of a surface by its mean curvature, Princeton University Press, Princeton, NJ (1977).
[B2] K. Brakke, The surface evolver, Experimental Mathematics, vol. 1, no. 2 (1992), 141165.
[B3] K. A. Brakke, Minimal surfaces, corners, and wires, Journal of Geometric Analysis 2 (1992) 11-36.
[B4] K. A. Brakke, The opaque cube problem video, Computing Optimal Geometries, J. E. Taylor, ed., American
Mathematical Society, Providence RI, 1991.
[B5] K. A. Brakke, Grain growth with the Surface Evolver, Video Proceedings of the Workshop on Computational
Crystal Growing, J. E. Taylor, ed., American Mathematical Society, Providence RI, 1992.
[B6] K. A. Brakke, The opaque cube problem,, Am. Math. Monthly 99 (Nov. 1992), 866-871.
[BB] K. A. Brakke and F. Baginski, Modeling ascent configurations of strained high-altitude balloons, AIAA Journal 36 (1998) 1901-1920.
[B7] K. A. Brakke and F. Morgan, Instabilities of cylindrical bubble clusters,, Eur. Phys. J. E 9 (2002) 453-460.
[BS] K. A. Brakke and J. M. Sullivan, Using Symmetry Features of the Surface Evolver to Study Foams, in Mathematics and Visualization, ed. K. Polthier and H. Hege, Springer Verlag, Berlin, (1997).
[C] M. Callahan, P. Concus and R. Finn, Energy minimizing capillary surfaces for exotic containers, Computing
Optimal Geometries, American Mathematical Society, Providence RI, 1991.
276
277
[CS] S. Collicot, Convergence behavior of Surface Evolver applied to a generic propellant-management device, J.
Propulsion and Power 17 (2001) 845-851.
[DL] D. Dobkin and M. Laszlo, Primitives for the manipulation of three-dimensional subdivisions, technical report
CS-TR-089-87, Department of Computer Science, Princeton University, Princeton, New Jersey. (April 1987).
[FT] H. J. Frost and C. V. Thompson, Computer Simulation of Grain Growth, Current Opinion in Solid State and
Materials Science 3 (1996), 361.
[FHW] M. Freedman, X. He, and Z. Wang, On the energy of knots and unknots, Annals of Mathematics 139 no. 1
(1994) 150.
[FS] G. Francis, J. M. Sullivan, R. B. Kusner, K. A. Brakke, C. Hartman, and G. Chappell, The Minimax Sphere
Eversion, in Mathematics and Visualization, ed. K. Polthier and H. Hege, Springer Verlag, Berlin, (1997).
[HK] J. Hale and H. Kocak, Dynamics and Bifurcations, Springer-Verlag, 1991.
[HKS] L. Hsu, R. Kusner, and J. Sullivan, Minimizing the squared mean curvature integral for surfaces in space
forms, Experimental Mathematics 1 (1991) 191208.
[KR1] A. Kraynik and D. Reinelt, The linear elastic behaviour of a bidisperse Weiare-Phelan soap foam, submitted
to Chem. Eng. Comm.
[KR2] A. Kraynik and D. Reinelt, Simple shearing flow of a dry Kelvin soap foam, J. Fluid Mech. 311 (1996) 327.
[KRS] A. Kraynik, D. Reinelt, and F. van Swol, Structure of random monodisperse foam, Phys. Rev. E 67 (2003)
031403.
[KS1] R. Kusner and J. M. Sullivan, Comparing the Weaire-Phelan Equal-Volume Foam to Kelvins Foam, Forma
(The Society for Science on Form, Japan) ed. R. Takaki and D. Weaire, KTK Scientific Publishers. As book by
Taylor and Francis Ltd (1996).
[KS2] R. Kusner and J. M. Sullivan, Mobius Energies for Knots and Links, Surfaces and Submanifolds in Geometric Topology, Proceedings of the Georgia International Topology Conference, August 1993, ed. W. Kazez,
AMS/IP Studies in Advanced Mathematics, vol. 2, part 1 (1997) 570604.
[MB] X. Michalet and D. Bensimon, Observation of stable shapes and conformal diffusion in genus 2 vesicles,
Science 269 (4 Aug 1995) 666668.
[MH] H. Mittelmann, Symmetric capillary surfaces in a cube, Math. Comp. Simulation 35 (1993) 139152.
[MF1] F. Morgan, Harnack-type mass bounds and Bernstein theorems for area-minimizing flat chains modulo ,
Comm. Part. Diff. Eq. 11 (1986) 12571283.
[MT] F. Morgan and J. E. Taylor, Destabilization of the tetrahedral point junction by positive triple junction line
energy, Scripta Metall. Mater. 25 (1991) 1907-1910.
[PWB] R. Phelan, D. Weaire, and K. Brakke, Computation of equilibrium foam structures using the Surface
Evolver, Experimental Mathematics 4 (1995) 181-192.
277
278
[P] W. Press, B. Flannery, S. Teukolsky, and W. Vetterling, Numerical Recipes in C, Cambridge University Press,
New York, 1988.
[PP] K. Polthier and U. Pinkall,Computing discrete minimal surfaces and their conjugates, Experimental Math. 49
(1993) 1536.
[RN] R. J. Renka and J. W. Neuberger, Minimal surfaces and Sobolev gradients, SIAM J. Sci. Comput. 16 (1995)
14121427.
[RSB] L. M. Racz, J. Szekely, and K. A. Brakke, A General Statement of the Problem and Description of a Proposed
Method of Calculation for Some Meniscus Problems in Materials Processing, ISIJ International, Vol. 33, No.
2, (February 1993) 328335.
[S] R. Sibson, Locally equiangular triangulations, Comput. J. 21 (1978) 243245.
[SG] G. Strang, Linear Algebra and its Applications, Academic Press (1988).
[SJ] J. M. Sullivan, Sphere Packings Give an Explicit Bound for the Besicovitch Covering Theorem, J. Geometric
Analysis 4 (1993) 219231.
[T1] J. E. Taylor, The structure of singularities in soap-bubble-like and soap-film-like minimal surfaces, Ann. Math.
103 (1976), 489-539.
[T2] J. E. Taylor, Constructing crystalline minimal surfaces, Seminar on Minimal Submanifolds, E. Bombieri, ed.,
Annals of Math. Studies 105 (1983), 271288.
[T3] J. E. Taylor, Constructions and conjectures in crystalline nondifferentiable geometry, Proceedings of the Conference in Differential Geometry, Rio de Janiero, 1988, Pittman Publishing Ltd.
[Te] J. Tegart, Three-dimensional fluid interfaces in cylindrical containers, AIAA paper AIAA-91-2174, 27th Joint
Propulsion Conference, Sacramento, CA, June 1991.
[TW] W. Thompson (Lord Kelvin), On the division of space with minimum partitional area, Acta Math. 11(1887),
121134.
[U] A. Underwood, Constructing barriers to minimal surfaces from polyhedral data, Ph.D. thesis, Princeton, 1993.
[WM] D. Weaire and S. McMurry, Some Fundamentals of Grain Growth, in Solid State Physics: Advances in
Research and Applications (eds. H. Ehrenreich and F. Spaepen), Vol. 50, Academic Press, Boston (1997).
[WP] D. Weaire and R. Phelan, A counter-example to Kelvins conjecture on minimal surfaces, Phil. Mag. Letters
69 (1994), 107-110.
278
Index
+ graphics command, 180
- graphics command, 180
? graphics command, 180
#include, 92
__x, 54
show, edge, 63
0, 132
A, 21, 36, 42, 90, 129
a, 130, 164
abort, 115, 148
abs, 93
ackerman, 218
acos, 93
acosh, 93
actual_volume, 112
actual_volume, 72, 112
addition, 118
addload, 148
adjustable constants, 19, 35, 41

aggregate expressions, 119
alice, 149
ambient pressure, 83, 84, 101, 133
ambient_pressure, 135
ambient_pressure_value, 120
and, 118
annealing, 102, 252
approx_curv, 135
approx_curvature, 135
approximate curvature, 88
approximate_curvature, 135
approximate_curvature, 102
area, 65
area normalization, 85, 88, 130, 132, 251
area_fixed, 109
area_method_name, 97
area_normalization, 135
area_square, 217
area_fixed, 120
area_normalization, 101
areaweed, 149
array, 117
array assignment, 145
array operations, 148
array operations, 95
arrays, 95
arrays, 150
asin, 93
asinh, 93
assignment, 145
assume_oriented, 135
atan, 93
atan2, 93
atanh, 93
attributes, 17, 169
attributes, 158
attributes, common, 51
augmented_hessian, 135
autochop, 107
autochop, 135
autochop_length, 135
autochop_length, 120
autodisplay, 135
autopop, 107
autopop, 136
autopop_quartic, 136
autorecalc, 136
average_crossings, 212, 228
avg, 119
axial_point, 77
axial_point, 59
b, 130
B graphics command, 180
b graphics command, 180
backbody, 64, 66
backcolor, 169
backcolor, 68, 111
backcull, 136
background, 169
background, 120, 182
bad_error_count, 130
bad_next_prev_count, 130
279

bare, 55, 61
bezier_basis, 136
bi_surface, 210
bi_surface_attr, 210
big_endian, 136
binary numbers, 92
binary_off_file, 149
binary_printf, 149
black, 93
blas_flag, 136
blowup, 134, 252
blue, 93
bodies, 15
bodies, 17, 112
bodies_predicted, 110
body, 112
body element, 51
body facets, 70
body_count, 120
body_dissolve_count, 120
body_metis, 149
bottominfo, 159
boundaries, 21, 76, 80, 81, 90, 106, 180
boundaries, 151
boundary, 57
boundary, 110, 111
boundary, edge, 62
boundary, facet, 67
boundary, vertex, 57
boundary_curvature, 136
bounding box, 180
bounding_box_color, 182
break, 115
break_after_warning, 136
break_on_warning, 136
breakflag, 120
breakpoint, 149
brightness, 120
brown, 93
buck_knot_energy, 210
bug reports, 10
bugs, 10, 264
bunch_kauffman, 136
bunch_kaufman, 136
burchard, 150
bye, 165
C, 130
c, 130
C graphics command, 180
c graphics command, 180
280
calculate_in_3d, 137
carter_energy, 217
carter_power, 217
case, 17, 92
catenoid, 21
ceil, 93
centerofmass, 72, 137
central_symmetry, 79
charge_gradient, 217
chdir, 150
Chebyshev, 156
check, 150
check_count, 120
check_increase, 137
checking datafile, 262
circle_knot_energy, 211
circle_willmore, 206
circular_arc_area, 197
circular_arc_draw, 137
circular_arc_length, 197
clear, 169
clip_coeff, 137
clip_view, 137
clipped, 180
clipped, 24, 76, 137
clipped_cells, 137
clock, 120
close_show, 150
color, 169
color, 62, 111
color, edge, 62
color, facet, 67
colorfile, 121, 137
colormap, 137
colors, 93, 132
command line, 176
command line options, 176
command repitition, 114
commands, 15
commands, graphics, 179
comments, 17, 91
comparison, 118
compiling, 13
compound commands, 114
compressibility, 83
compressible, 101
conditional expression, 118
conducting_knot_energy, 101
conf_edge, 137
conformal_metric, 107
conj_grad, 137
280

conjugate gradient, 85, 133, 142, 236
connected, 180
connected, 24, 76, 137
connected_cells, 137
conserved, 103
consistency checks, 130
console_title, 121
constant expressions, 94
constraint, 169, 233
constraint, 104, 110, 111
constraint energy, 19, 80
constraint energy integral, 221
constraint energy integrand, 82
constraint normal, 56
constraint volume integral, 232
constraint volume integrand, 34, 40, 80, 84
constraint_tolerance, 121
constraint_tolerance, 86, 106, 107
constraints, 55, 76, 80, 82, 85, 90, 104, 180
constraints, 111, 151
constraints, edge, 61
constraints, facet, 66
contact angle, 19, 32, 38, 82
content, 84
content, 104, 106
content_rank, 104, 106
continue, 116
convert_to_quantities, 83, 150
convex, 34, 40, 80, 82, 104, 106, 221
coordinates, 54, 169
cos, 93
cosh, 93
count, 119
counts, 130
cpu_counter, 121
crossingflag, 141
crystalline integrand, 44, 83, 100, 220
cube, 16
cubelat symmetry group, 78
cubocta symmetry group, 78
curvature test, 134, 252
curvature_binormal, 211
curvature_function, 212
curvature_power, 121
cyan, 93
D, 130
d, 18, 130
d graphics command, 179
darkgray, 93
datafile, 15, 91, 176
281
datafilename, 121
date_and_time, 121
ddd_gamma_sq, 211
debug, 138
define, 95, 150
delete, 152
delete_count, 121
delete_text, 152
density, 169
density, 65, 82, 112
density, body, 70
density_facet_area, 198
density_facet_area_u, 200
density_edge_length, 192
detorus, 152
detorus_epsilon, 121, 137, 152
detorus_sticky, 137, 152
deturck, 138
diffusion, 85, 130
diffusion, 85, 102, 138
diffusion_coeff, 121
diffusion_constant, 85
dihedral, 64
dihedral angle, 251
dihedral, vertex, 58
dihedral_hooke, 194
dimension, 13, 74, 75
dirichlet, 152, 239
dirichlet_area, 200
dirichlet_elastic, 214
dirichlet_mode, 138
dirichlet_seek, 239
display, 130, 133
display_text, 152
display_origin, 76
display_periods, 76
displayable, 90, 112
dissolve, 152
dissolve_count, 152
div_normal_curvature, 138
Divergence Theorem, 45
division, 118
do, 115
dodecahedron symmetry group, 79
dot_product, 95
dump, 130
dump, 153
dump_memlist, 153
dynamic load libraries, 99, 258
e, 117, 130
281

e_boundary, 62
e_constraint_list, 62
e_method_list, 64
edge, 111
edge attributes, 50
edge constraints, 61
edge element, 50
edge facets, 61
edge popping, 132, 253
edge tension, 60
edge valence, 61
edge wraparound, 76
edge_count, 121
edge_delete_count, 121
edge_diffusion, 85
edge_dissolve_count, 121
edge_divide, 132
edge_edge_knot_energy, 209
edge_k_vector_integral, 208
edge_knot_energy, 208, 209
edge_knot_energy_normalizer, 209
edge_merge, 153
edge_min_knot_energy, 212
edge_pop_count, 121
edge_refine_count, 122
edge_area, 193, 226
edge_general_integral, 193, 225
edge_length, 192, 224
edge_loop_count, 130
edge_pop_attribute, 140
edge_scalar_integral, 192, 225
edge_tension, 192, 224
edge_torus_area, 193
edge_vector, 62
edge_vector_integral, 193, 225
edges, 15, 111

edges, 17, 111
edges_predicted, 110
edges_same_vertices_count, 130
edgeswap, 153
edgeswap_count, 122
edgeweed, 153
eff_area_sq_mean_curvature, 203
eff_area_sq_mean_curvature, 227
effective_area, 88
effective_area, 138
effective_area, 102
efixed, 111
eigen_neg, 122
eigen_pos, 122
eigen_zero, 122
282
eigenneg, 122
eigenpos, 122
eigenprobe, 153
eigenvalue, 86, 156

eigenvalues, 153, 168
eigenvalues, 122
eigenvector, 86, 156, 168
eigenzero, 122
elastic_basis, 214
elastic_coeff, 214
element_modulus, 102, 187
ellipticE, 93
ellipticK, 93
else, 115
energy, 15, 82, 85, 220
energy, 104, 106, 111
eprint, 153
equal to, 118
equi_count, 122
equiangulate, 154
equiangulation, 133, 250
ergb, 142
errors, 177
errprintf, 153
estimate, 138
estimated_change, 122
everything_quantities, 104
evolver_version, 94
EVOLVERPATH, 12, 176
exec, 154
exit, 19, 134, 180
exp, 93
expressions, 93, 117
exprint, 154
extra attributes, 53, 90, 99, 110, 134, 150, 169
extra_boundary, 81
extra_boundary_param, 81
extrapolate, 130
extrapolation, 252
F, 130
f, 130
f_body_list, 66
f_boundary, 67
f_constraint_list, 67
f_method_list, 70
f_next_bfacet, 70
f_next_vfacet, 70
face, 111
faces, 111
faces, 17, 111
282
283
facet attributes, 50
facet bodies, 66
facet edges, 65
facet element, 50
facet vertices, 65
facet-edges, 15, 219
fixed, facet, 65
fixed_area, 109
facet_2form_integral, 199
facet_2form_sq_integral, 199
facet_area, 198
facet_area_u, 200
facet_colors, 138
facet_count, 122
facet_crosscut, 154
facet_delete_count, 122
facet_diffusion, 85
facet_dissolve_count, 122
facet_general_integral, 199
facet_knot_energy, 210
facet_knot_energy_fix, 210
facet_merge, 154
facet_refine_count, 122
facet_scalar_integral, 198
facet_tension, 198
facet_torus_volume, 199
facet_vector_integral, 198
facet_volume, 198
facet_2form_integral, 225
facet_area, 224
facet_area_u, 224
facet_general_integral, 225
facet_normal, 68
facet_scalar_integral, 225
facet_tension, 224
facet_vector_integral, 225
facet_volume, 226
force, 220
facetedge attributes, 51
facetedge edge, 73
facetedge element, 51
facetedge facet, 73
facetedge_count, 122
facetedges_predicted, 110
facets, 15
facets_predicted, 110
facets_same_vertices_count, 130
fbrgb, 142
file formats, 184

fix, 154
fix_count, 122
FIXED, 82, 90
fixed, 52, 55, 103, 106, 110, 111, 154
fixed, edge, 60
flip_rotate symmetry group, 78

floor, 93
flush_counts, 155
for, 115
force_deletion, 138
force_edgeswap, 138
force_pos_def, 138
forces, 241
foreach, 155
form_integrand, 102
form_factors, 214, 216
formula, 104
free_discards, 155
frgb, 142
frontbody, 64, 66
frontcolor, 169
frontcolor, 67, 111
full_bounding_box, 138, 183
full_gravity_method, 200
function, 99, 103
function_quantity_sparse, 138
functions, 117
G, 21, 117, 131
g, 18, 131, 235
gap constant, 132

gap energy, 80, 221
gap_energy, 200
gap_constant, 101
gaps, 80, 82, 85, 101, 132, 221
gauss_bdry_e, 205
gauss_bdry_v, 205
gauss_curvature, 100
gauss_curvature_integral, 205
Gauss_curvature_integral, 228
Gaussian curvature, 83
Gaussian quadrature, 80, 106
general_linear_elastic, 214
generator_power, 77
GENERIC makefile option, 13
genus2 symmetry group, 79
geompipe, 182
geomview, 13, 14, 18, 133, 139, 182, 184
global, 80, 103, 104
global_method, 103
go, 131
graphics commands, 179
graphics interfaces, 257
283

graphics_title3, 122
graphics_title, 122
graphics_title2, 122
gravity, 19, 32, 38, 76, 82, 101, 131, 220

gravity, 138
gravity_constant, 122
gravity_method, 200
gravity_constant, 101
gravity_method, 226
greater than, 118
green, 93
Greens Theorem, 45
gridflag, 142
gv_binary, 139
H, 131
h, 131
h graphics command, 180
h_inverse_metric, 139
h_zero, 195, 196, 202205
help, 131, 180

help, 155
hemisphere, 100
Hessian, 86, 156, 157, 236
hessian, 155
hessian_diff, 139
hessian_double_normal, 139
hessian_epsilon, 123
hessian_menu, 155
hessian_normal, 139
hessian_normal_one, 139
hessian_normal_perp, 139
hessian_quiet, 139
hessian_seek, 157
hessian_slant_cutoff, 123
hessian_special_normal, 139
hessian_normal, 238
hessian_special_normal_vector, 98
hexadecimal numbers, 92
hidden surfaces, 251, 257
high_boundary, 123
high_constraint, 123
hints, 260
histogram, 157
history, 113
history, 157
hit_constraint, 56
homogeneous coordinates, 257
homothety, 108, 134, 252
homothety, 139
Hooke energy, 227
284
hooke2_energy, 194, 227
hooke2_power, 194
hooke3_energy, 194, 227
hooke3_power, 194
hooke_size, 194
hooke_energy, 193
hooke_length, 193
hooke_power, 193
HTML version, 11
i, 131
id, 51
id numbers, 94
ideal gas model, 83, 84, 101, 112, 133, 221
identifiers, 92
idiv, 93
if, 115
ignore_constraints, 202
ignore_fixed, 202
ignore_constraints, 227
immediate_autopop, 139
imod, 93
incompleteEllipticE, 93
incompleteEllipticF, 93
inconsistent_bodies_count, 130
inequality constraint, 105
info_only, 103
information, 131
initiallization, 177
installation, 11
insulating_knot_energy, 101
integer, 99
integral_order, 123
integral_order_1d, 123
integral_order_2d, 123
integral_order, 80, 106
integration_order, 123
integration_order_1d, 123
integration_order_2d, 123
interp_bdry_param, 139
interp_normals, 140
interrupts, 131, 178
inverse_periods, 123
is_defined, 126, 157
itdebug, 140
iteration, 16, 18, 85, 131, 235
iteration_counter, 123
J, 131, 252
j, 131, 252
jiggle, 85, 131, 252

284

jiggle, 102, 140
jiggle_temperature, 123
jiggling, 44, 102

johndust, 218
K, 131
k, 132
k_vector_order, 207
k_altitude_mode, 140
keep_macros, 92
keep_originals, 94
Kelvin tetrakaidecahedron, 24
keylogfile, 157
keywords, 17, 93
klein_area, 200
KLEIN_METRIC, 75
klein_length, 197
Klein_metric, 107
kmetis, 160
knot energy, 101, 228
knot_energy, 208
knot_local_thickness, 213
knot_power, 208211, 217
knot_thickness, 213
knot_thickness2, 213
knot_thickness_0, 213
knot_thickness_p, 213
knot_thickness_p2, 213
kraynikpopedge, 140
kraynikpopvertex, 140
kusner, 140
l, 18, 132
l graphics command, 180
labelflag, 142
labels, 185
Lagrange, 158
lagrange, 98
Lagrange model, 74
Lagrange multiplier, 235
lagrange_order, 123
lagrange_multiplier, 103
lanczos, 158
laplacian_mean_curvature, 206
last_eigenvalue, 123
last_error, 123
last_hessian_scale, 123
length, 60
length_method_name, 97
lens, 100
less than, 118
285
lightblue, 93
lightcyan, 93
lightgray, 93
lightgreen, 93
lightmagenta, 93
lightred, 93
line splicing, 91
linear, 158
linear_elastic, 214
linear_elastic_B, 215
linear_metric, 87, 140
linear_metric_mix, 87, 123
linear_elastic, 228
lines, 17
list, 158
little_endian, 140
load, 159
load_library, 99, 258
local, 114
Local Hooke energy, 227
local variables, 114
local_hooke_energy, 194
local_hooke_flag, 194
log, 93
logfile, 159
logging, 130
loghistogram, 157, 159
long edges, 132
longj, 159, 252
M, 74, 132
m, 44, 85, 132
Macintosh, 12
macros, 19, 92
magenta, 93
makefile, 13
manual, 11
matrix_determinant, 159
matrix_inverse, 160
max, 119
MAXCOORD compile option, 75

maximum, 93
mean curvature, 83, 84, 88, 234
mean curvature integral, 100
mean_curvature_integral, 83
mean_curvature_integral_a, 202
mean_curvature, 59
mean_sq_mix, 204
memdebug, 140
285

memory, 130
memory_arena, 123
memory_used, 123
merit_factor, 109
method, 186
method instances, 102

method instances, 151
method_instance, 102
method_instances_predicted, 110
methods, 223
METIS, 13
metis, 160
metis_factor, 140
metric, 75, 107
metric, 107
metric_conversion, 140
metric_convert, 140
metric_facet_area, 200
metric_edge_length, 196
Microsoft Windows, 11
mid_edge, 58
mid_facet, 58
midv, 111
midv, 61
min, 119
minimum, 93
286
no_hessian_normal, 238
no_refine, 63, 111
no_refine, facet, 69
no_transform, 63
no_transform, facet, 69
node_charge, 208
nodisplay, 111
noncontent, 64, 69, 111
nonnegative, 80, 104
nonpositive, 80, 104
nonwall, 104
normal interpolation, 132
normal_curvature, 141
normal_motion, 141
normal_sq_mean_curvature, 203
normal_sq_mean_curvature, 228
not, 118
not equal, 118
notch, 132
notch, 161
notch_count, 123
nulgraph, 13
null_area, 97
null_length, 97
numbers, 92
mix_sq_mean_curvature, 204
o, 132
mobility, 87, 170

mobility, 107
mobility_tensor, 107
mod, 93
modulus, 102, 118
motion, 85
move, 160
multiplication, 118
octahedron, 44
off, 135
oid, 52
old_area, 141
ometis, 161
omitting faces, 19, 84
on, 135
on_assign_call, 95
on_boundary, 52, 57
on_constraint, 52
on_method_instance, 53
on_quantity, 53
one_sided_lagrange, 80
ooglfile, 182
opacity, 169
opacity, 68
OpenGL, 180
optimise, 162
optimising_parameter, 95
optimization, 13, 90
optimize, 162
optimizing parameter, 154
optimizing scale factor, 132
optimizing_parameter, 90
N, 132
n, 132
named methods, 186
named quantities, 103, 169, 186, 223

named quantity, 155, 171
named quantity attribute, 53
new_body, 161
new_edge, 160
new_facet, 161
new_vertex, 160
newsletter, 10
nextedge, 73
nextfacet, 73
no_dump, 140
no_display, 68, 111
286

optimizing_parameter, 95
or, 118
orientation, 15, 17, 82, 111, 112, 169, 219

orientation, 63
orientation, facet, 69
original, 53
P, 132, 184
p, 133, 164
p1, 57, 93
p2, 57, 93
p_force, 95
p_velocity, 95
painter algorithm, 251, 257

parameter, 95
parameter_file, 109
parameter_1, 195, 196
parameters, 90
parameters, vertex, 57
pause, 162
pdelta, 95
periodic surfaces, 75
periods, 97
permload, 162
perturbations, 159, 252
phase, 111
phase, body, 72
phase, facet, 69
phasefile, 100
phases, 100
pi, 93, 117
pickenum, 124, 182
pickfnum, 124, 182
picking, 179
pickvnum, 124, 182
pinning, 141
pipe, 114
Pixar, 132, 184
poisson_ratio, 214216
pop, 162
pop_count, 124
pop_disjoin, 141
pop_edge_to_tri, 163
pop_edge_to_tri_count, 124
pop_enjoin, 141
pop_quad_to_quad, 163
pop_quad_to_quad_count, 124
pop_to_edge, 141
pop_to_face, 141
pop_tri_to_edge, 163
pop_tri_to_edge_count, 124
287
pos_area_hess, 201
post_project, 142
post_project, 236
PostScript, 132, 185

postscript, 182
pow, 93
power, 118
Power User, 32, 38
prescribed mean curvature, 83, 84
prescribed pressure, 222
pressure, 8385, 130, 134, 169, 235
pressure, 84, 101, 112, 221
pressure, body, 72
prevedge, 73
prevfacet, 73
print, 163
printf, 164
procedure, 159
procedures, 116
procedures, 159
proj_knot_energy, 210
ps_bareedgewidth, 185
ps_bareedwidth, 124
ps_conedgewidth, 185
ps_conedwidth, 124
ps_fixededgewidth, 185
ps_fixedwidth, 124
ps_gridedgewidth, 185
ps_gridedwidth, 124
ps_labelsize, 124, 185
ps_linewidth, 185
ps_stringwidth, 124, 185
ps_tripleedgewidth, 124
ps_colorflag, 141
ps_crossingflag, 141
ps_gridflag, 142
ps_labelflag, 142
pscale, 95
pscolorflag, 141
Q, 133
q, 18, 19, 133
q graphics command, 180
quadratic, 98, 165
quadratic model, 74, 76, 132

quadratic_metric_mix, 124
quantities, 76, 83, 133, 223
quantities, 151
quantities_only, 142
quantities_predicted, 110
quantity, 110
287

quantity, 103
quarter_turn, 79
quarter_turn_period, 79
quiet, 142
quietgo, 142
quietload, 142
quit, 19, 133, 134, 180

quit, 165
quotient space, 77
quotient space, 97
R, 18
r, 18, 133
R graphics command, 180
r graphics command, 180
random, 124
random_seed, 125
raw_cells, 76, 142
raw_vertex_average, 166
raw_cells, 24
raw_velocity, 60
rawest_vertex_average, 166
rawestv, 165
rawv, 166
read, 166
real, 99
rebody, 166
recalc, 166
red, 93
redirecting input, 176

redirection, 114
refine, 90, 133, 253
refine, 166
refine_count, 125
refining, 112
relaxed_elastic, 216
relaxed_elastic1, 216
relaxed_elastic1_A, 216
relaxed_elastic2, 216
relaxed_elastic2_A, 216
relaxed_elastic_A, 216
remainder, 118
renumber_all, 166
reorder_storage, 167
replace_load, 167
reset, 180
reset_counts, 168
reset_profiling, 168
return, 116
reverse_orientation, 168
rgb_colors, 142
288
ribiere, 23, 142, 236
Riemannian metric, 75
ritz, 168
rotate, 179
rotate symmetry group, 77
rotate_lights, 142
rotation_order, 77
Runge_Kutta, 108
runge_kutta, 143
s, 133
s graphics command, 180
saddle, 168
save, 130
scalar_integrand, 102
scale, 263
scale, 106, 125
scale factor, 16, 85, 106, 132
scale optimizing, 131, 235
scale optimizing, 106
scale_scale, 85, 131
scale_limit, 106
scope, 114
screw_angle, 79
screw_height, 79
screw_symmetry, 79
scrollbuffersize, 125
self, 151
self_sim_coeff, 125
self_similar, 143
semicolon, 113
septum_flag, 143
set, 154, 171
shading, 143
shell, 169
show, 133
show, 63, 183
show_all_edges, 143
show_all_quantities, 143
show_bounding_box, 143
show_expr, 183
show_inner, 143
show_off, 150
show_outer, 143
show_trans, 183
show_vol, 134
showq, 183
shrink, 180
signals, 178
simon_knot_energy_normalizer, 209
simplex model, 75
288

simplex refine, 253
simplex_k_vector_integral, 207
simplex_representation, 125
simplex_to_fe, 170
simplex_vector_integral, 207
simplex_representation, 98
sin, 93
sin_knot_energy, 211
sinh, 93
sizeof, 126
skinny triangles, 131

slice_coeff, 143
slice_view, 143
small facets, 134
small triangles, 254
smooth_graph, 143
soapfilm, 96
soapfilm model, 74, 96
sobolev, 170, 239
sobolev_area, 201
sobolev_mode, 143
sobolev_seek, 239
SoftImage, 132, 185
space_dimension, 125
sparse_constraints, 143
sphere, 16
sphere example, 38
sphere_knot_energy, 211
spherical_arc_area_n, 197
spherical_arc_length, 197
spherical_area, 201
spherical_arc_area_s, 197
spherical_area, 224
spring_constant, 101
sprintf, 170
sq_gauss_curvature, 206
sq_mean_curvature, 202
sq_gaussian_curv_cyl, 195
sq_mean_curv_cyl, 196
sq_mean_curvature, 227
sq_torsion, 195
sqcurve, 59
sqcurve2_string, 195
sqcurve3_string, 195
sqcurve_string, 195
sqcurve_string_mark, 196
sqcurve_string_marked, 196
sqgauss, 101
sqr, 93
sqrt, 93
square gradient, 157
289
square_curvature, 100
square_gaussian_curvature, 101
square_gauss_modulus, 101
squared Gaussian curvature, 223

squared mean curvature, 83, 100, 222
squared_curvature, 100
squared_gaussian_curvature, 101
squared_gradient, 143
stability, 89
stability_test, 170
star_eff_area_sq_mean_curvature, 204
star_finagle, 144
star_gauss_curvature, 206
star_normal_sq_mean_curvature, 205
star_perp_sq_mean_curvature, 205
star_sq_mean_curvature, 204
Stokes Theorem, 45
stokes2d, 201
stokes2d_laplacian, 201
stokes_type, 201
stokes_velocity, 201
stress_integral, 218
string, 96
string model, 74, 90, 96, 105, 112
string_curve_tolerance, 125
string_gravity, 193
strings, 92
subcommand, 170
subtraction, 118
sum, 119
suppress_warning, 110
surface, 15
surface energy, 33, 39
surface tension, 82, 90, 100, 220
surface_dimension, 125
surface_cos_power, 211
surface_dimension, 96
surface_knot_power, 211
SVK_alpha, 216
SVK_elastic, 216
SVK_lambda, 216
SVK_mu, 216
SVK_theta, 216
swap_colors, 109
symmetric_content, 85, 97, 231
symmetry, 77, 80
symmetry_group, 125
symmetry_group, 97
system, 170
t, 36, 42, 133
289

t graphics command, 180
t1_edgeswap, 170
t1_edgeswap_count, 125
tag, 111
tan, 93
tanh, 93
tank example, 32
target, 169
target, body, 71
target_tolerance, 125
temperature, 85, 131
temperature, 102
tension, 60, 65, 82, 111
tetra_point, 59
tetrakaidecahedron, 24
then, 115
thicken, 144
thickening, 132
thickness, 125, 182
time, 107
tiny edges, 133, 254
toggles, 135
tolerance, 103
topinfo, 159
topology, 15, 85, 90, 132, 220
torques, 241
torus, 24, 75, 132, 232
torus, 75, 96, 125
torus duplication, 134
torus symmetry group, 77
torus wrapping, 76, 90, 111
torus_filled, 125
torus_filled, 76, 96
torus_periods, 75
total_area, 125
total_energy, 125
total_length, 125
total_time, 126
transform_count, 126
transform_depth, 183
transform_expr, 183
transforms, 183
translation, 180
transparent, 169
triangulation, 90
triple junction, 15
triple_point, 59
tripleedgewidth, 185
true_average_crossings, 212
true_writhe, 212
twist, 212
290
U, 23, 133, 236
u, 18, 36, 42, 133, 250
u graphics command, 179
ulong, 99
unary minus, 118
unfix, 171
unfix_count, 126
uniform_knot_energy, 208
uniform_knot_energy_normalizer, 209
uniform_knot_normalizer1, 209
uniform_knot_normalizer2, 209
unit cell, 97
units, 15
unset, 171
unsuppress_warning, 110
user-defined functions, 94
utest, 171
v, 18, 84, 134
v graphics command, 180
v_boundary, 57
v_constraint_list, 56
v_force, 59
v_method_list, 58
v_oldx, 60
v_velocity, 59
valence, 55, 61
valence, facet, 66
valid_element, 127
valid_boundary, 127
valid_constraint, 127
valleys, 180
value_of_constraint, 56
variables, 90, 95
vector_integrand, 102
verbose, 144
vertex, 110
vertex attributes, 49
vertex averaging, 254
vertex boundary, 57
vertex constraints, 55
vertex edges, 54
vertex element, 49
vertex facets, 55
vertex mean curvature, 59
vertex parameters, 57
vertex popping, 21, 132, 253
vertex_average, 171
vertex_count, 126
vertex_dissolve_count, 126
vertex_merge, 171
290
291
vertex_pop_count, 126
vertex_normal, 58
vertex_scalar_integral, 224
vertex_scalar_integral , 192
vertexnormal, 58
x, 54, 134
x graphics command, 180
x1, 93
x2, 93
x3, 93
vertices, 15, 110

vertices, 17
xyz symmetry group, 78
vertices_predicted, 110
view_4d, 184
view_matrix, 108
view_transform_generators, 109
view_transforms, 108
view_transforms_unique_point, 126
view_transforms_use_unique_point, 144
view_matrix, 126
y, 134
y, 54
viewing matrix, 108

visibility_test, 144
volconst, 41, 112
volconst, body, 71
volfixed, 71
volgrads_every, 144
volume, 84, 130, 132, 134
volume, 17, 71, 112
volume constraint, 19, 32, 38, 84, 112, 234
volume constraints, 85
volume_method_name, 98, 109
volumes, 231
yellow, 93
ysmp, 144
Z, 134
z, 54, 134
z graphics command, 180
zener_coeff, 144
zener_drag, 144
zoom, 134, 180, 255

zoom, 172
zoom_radius, 109
zoom_vertex, 109
W, 134
w, 36, 42, 134
w graphics command, 180
warning_messages, 164
where, 147
where_count, 126
whereami, 171
while, 115
white, 93
whitespace, 92
window_aspect_ratio, 182
Windows, 11
Windows 95/NT, 180
wrap, 63
wrap_compose, 77
wrap_inverse, 77
wrap_vertex, 172
wrap_list, 63
writhe, 212
wulff, 100
Wulff shape, 44
Wulff vectors, 44, 83
wulff_energy, 217
291

Surface Evolver Manual

Uploaded by

Copyright:

Available Formats

Surface Evolver Manual

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Surface Evolver Manual

Uploaded by

Copyright:

Available Formats

What are the different platforms supported for installation?

What are the different platforms supported for installation?

What are some examples of surfaces that can be modeled?

What are some examples of surfaces that can be modeled?

Surface Evolver Manual