Articles

https://doi.org/10.1038/s41563-019-0452-y

Programming shape using kirigami tessellations

Gary P. T. Choi   1, Levi H. Dudte1 and L. Mahadevan   1,2,3*

Kirigami tessellations, regular planar patterns formed by partially cutting flat, thin sheets, allow compact shapes to morph
into open structures with rich geometries and unusual material properties. However, geometric and topological constraints
make the design of such structures challenging. Here we pose and solve the inverse problem of determining the number, size
and orientation of cuts that enables the deployment of a closed, compact regular kirigami tessellation to conform approxi-
mately to any prescribed target shape in two or three dimensions. We first identify the constraints on the lengths and angles of
generalized kirigami tessellations that guarantee that their reconfigured face geometries can be contracted from a non-trivial
deployed shape to a compact, non-overlapping planar cut pattern. We then encode these conditions into a flexible constrained
optimization framework to obtain generalized kirigami patterns derived from various periodic tesselations of the plane that can
be deployed into a wide variety of prescribed shapes. A simple mechanical analysis of the resulting structure allows us to deter-
mine and control the stability of the deployed state and control the deployment path. Finally, we fabricate physical models that
deploy in two and three dimensions to validate this inverse design approach. Altogether, our approach, combining geometry,
topology and optimization, highlights the potential for generalized kirigami tessellations as building blocks for shape-morphing
mechanical metamaterials.

K
irigami, the creative art of paper cutting and folding, and state and still be able to tile a patch of the Euclidean plane compactly
the tessellations derived from it have recently emerged as when undeployed?
prototypical routes towards a new class of mechanical meta- To find a solution requires us to search for potentially admissible
materials. Indeed, various studies have focused on quantifying the results in the deployed space. A first step in this process is to quan-
geometry and kinematics of deployment of a given kirigami pat- tify the constraints that will allow the deployed initialization geom-
tern, and their potential as auxetic structures and shape-morphing etry to correspond to a valid kirigami pattern that can compactly
sheets1–18, the so-called forward problem. From a mathematical, close onto a Euclidean patch. The necessary and sufficient contract-
physical and a technological perspective, perhaps an even more ibility conditions imply that a valid deployed configuration must
interesting question is the inverse problem: can one design the kiri- be able to contract (undeploy) the configuration into a generalized
gami tessellations in a closed, compact subset of the plane, so that it kirigami pattern that closes consistently along the cuts without any
can be deployed into a prescribed final shape in two or three dimen- mismatch or overlap in lengths and angles. As illustrated in Fig. 1b,
sions? Here, we pose this puzzle as a constrained optimization prob- the contractibility constraints are: (1) every pair of edges with edge
lem and solve it in a range of situations. lengths a, b in the deployed space that correspond to the same cut
The simplest planar deployable kirigami patterns are based on must satisfy the condition
periodic tilings of the plane using triangles, squares and hexagons,
although other complex tessellations inspired by art and architec- a2 � b2 ¼ 0 ð1Þ
ture have been explored recently4,19. Here we focus on using the
quadrilateral kirigami pattern for ease of exposition; however, our (2) Every set of four angles in the deployed space that correspond
methods generalize to any periodic pattern (see Supplementary to an interior node must sum to 2π, so that
Information, Section 1). In Fig. 1a, we show the quad tessellation of
the plane in its compact and deployed state, with the cuts along the θ1 þ θ2 þ θ3 þ θ4 ¼ 2π ð2Þ
edges of the quads designed to allow for rotational in-plane deploy-
ment about a set of hinges. This pattern constitutes a one-degree-of- where θi are angles in the deployed space as illustrated in Fig. 1a (see
freedom mechanism whose planar deployment yields a continuous Supplementary Information, Section 2 for the analogous formula-
family of self-similar shapes that terminates at self-intersecting con- tion for other kirigami tessellations).
figurations. The basic unit cell underlying this pattern is also shown To be able to deploy a kirigami pattern to match a given shape,
in Fig. 1a in both its undeployed and deployed states and shows the we also need to ensure that for a given boundary curve ∂S we have a
mathematical constraints that define the system: pairs of deployed valid deployed configuration. This can be formulated asIboundary-
edges contract to the same edge in the undeployed state, and simul- shape matching constraints that force all nodes on the boundary of
taneously sets of deployed angles contract to the same single vertex the deployed configuration to lie exactly on ∂S. Mathematically, for
in the undeployed state. Given an initially periodic tiling of a patch every boundary node pi, this implies that I
of the plane with this unit cell, we can state our inverse design prob- I
lem thus (as shown in Fig. 1b): how should the unit cell be modu-
lated in space to approximate a given planar shape in its deployed p i k2 ¼ 0
kpi � e ð3Þ

John A. Paulson School of Engineering and Applied Sciences, Harvard University, Cambridge, MA, USA. 2Departments of Physics, and Organismic and
1

Evolutionary Biology, Harvard University, Cambridge, MA, USA. 3Kavli Institute for Nanobio Science and Technology, Harvard University, Cambridge, MA,
USA. *e-mail: [email protected]

Nature Materials | VOL 18 | SEPTEMBER 2019 | 999–1004 | www.nature.com/naturematerials 999

Articles NAture MAteriAls

a b Deployed space
Pattern space

Initial guess

Standard
kirigami pattern
Invalid deployed shape

Constrained
optimization
a b

θ1 θ2
a b
θ1 θ2

θ4 θ3
Contraction
r
θ4 θ3

Generalized
kirigami pattern
Valid deployed shape

Fig. 1 | Inverse design framework. a, A quad kirigami tessellation and its deployed configuration, with an enlargement of the unit cell of the quad kirigami
tessellation and its deployed configuration. Every pair of corresponding edges are connected by a red dashed line. The set of angles corresponding to the
same node are highlighted in blue. In a valid deployed configuration of a generalized kirigami pattern, every pair of edges should be equal in length, that
is a = b, and every set of corresponding angles should add up to 2π, that is, θ1 þ θ2 þ θ3 þ θ4 ¼ 2π. b, Our inverse design framework. Given a standard
kirigami tessellation, we start with an initial guess in the deployed space. Here I the initial guess shown is a conformal map from the standard deployed
configuration to the disc. The initial guess is usually invalid, violating either the edge length constraint or the angle constraint, or not exactly matching the
target boundary shape. We then solve a constrained optimization problem to morph the initial guess until it becomes a valid deployed shape, satisfying all
constraints. Finally, we use a simple contraction procedure to obtain the generalized kirigami pattern.

where e pi is the projection of pi onto ∂S and kk is the Euclidean and M is the total number of pairs of adjacent cells, subject to the
2-norm.I In addition to matching I
I the target boundary
I shape in the constraints (1)–(4).
deployed configuration, we can also control the shape of the kiri- Finally, we need a reasonable first approximation to the given
gami tessellations in their undeployed states by introducing bound- deployed configuration that matches the prescribed boundary curve.
ary angle constraints (see Supplementary Information, Section 2 This initial condition can be obtained via, for example, a conformal/
for details). quasi-conformal map20,21, although we could use any approximation
While the constraints described above ensure consistency that preserves the number and connectedness of the quads.
between corresponding edges and angles, they do not prevent the We solve the optimization problem numerically using MATLAB’s
faces from overlapping in the deployed state. To enforce this, we use built-in optimization routine fmincon (see Supplementary
the following non-overlapping constraint at every angle between Information, Section 3), noting that the problem is undercon-
two adjacent faces: strained with multiple admissible deployed kirigami patterns (see
Supplementary Information, Sections 2 and 4). Once we find a valid
hðb � aÞ ´ ðc � aÞ; n
^ i≥ 0 ð4Þ deployed structure, we can contract this into its compact pattern
form by rotating the faces and thus shrinking the entire structure.
where a, b and c are the nodes of adjacent faces, so that (b, a, c) We note here that our formulation of the optimization problem
forms a positive (right-hand ordered) angle between the two faces, shows some algorithmic similarities to our recent study of gener-
and n^ ¼ ð0; 0; 1Þ is the outward unit normal. alized origami tessellations22. This is both natural and inevitable,
ToI find an admissible deployed kirigami structure, we must sat- given the presence of geometrical constraints in both problems as
isfy the above interior and boundary conditions—this will yield a well as a condition of shape matching; indeed, many problems in
compact tiling that is related to the deployed state via a non-affine physical geometry requiring the ability to design shapes subject to
contraction. This can be framed as a constrained optimization constraints will also fall within this framework.
problem whose solution is sufficient to guarantee a valid deployed To illustrate the effectiveness of our approach, we first design
configuration of a generalized kirigami pattern that approximates a generalized kirigami patterns to approximately solve an ancient
prescribed shape. However, without a regularization procedure, the mathematical problem—that of circling the square—and a modern
solution is likely to be very rough with large gradients in the shapes one, making an egg shape from a square. In Fig. 2a, we show that by
of the quads. To produce a smooth kirigami tessellation, we there- distorting a periodic kirigami tiling of a square (see Supplementary
fore minimize the following objective function: Information), the deployed configurations can match either a circle
! or an egg. We can also use this method to generate novel general-
M
1X X 2
X 2 ized kirigami patterns that, when deployed, approximate different
ðαij � βij Þ þ ðaik � bik Þ ð5Þ
M i¼1 j
boundary shapes. Figure 2b shows two generalized kirigami pat-
k
terns; one deployed to approximate a domain with a boundary of
where αij ; βij are a pair of corresponding angles in two adjacent cells mixed curvature, and the other deployed to approximate a rectangle
and aik ;Ibik are the corresponding edge lengths in two adjacent cells, (see Supplementary Information, Section 4 for more generalized
I
1000 Nature Materials | VOL 18 | SEPTEMBER 2019 | 999–1004 | www.nature.com/naturematerials
NAture MAteriAls Articles
a b

log10(boundary layer area)

–0.3
–0.4
–0.5
–0.6
–0.7
–0.8
–0.9
1 1.5 2 2.5 3
log10(number of tiles)

Fig. 2 | Generalized planar kirigami patterns. a, Examples of generalized kirigami patterns produced by our method for getting a circle or an egg shape
from a square on deployment. b, Examples of generalized kirigami patterns produced by our method for achieving boundary shapes with mixed curvature
or zero curvature. It can be observed that our method is capable of producing generalized kirigami patterns that match boundary curves with different
curvature properties when deployed. c, Examples of circling the square with different resolutions (number of tiles = 8 × 8, 16 × 16, 20 × 20), together
with a log–log plot of the boundary layer area against the number of tiles. Here, the boundary layer area is defined as the total area of the gaps between
the circle and the boundary of the deployed kirigami patterns. The dots on the log–log plot represent kirigami patterns with different numbers of tiles
(4 × 4, 8 × 8, 12 × 12, 16 × 16, 20 × 20) and the straight line is the least-square regression line. The result shows that there is an accuracy–effort trade-off in
approximating a prescribed shape using generalized kirigami tessellations.

kirigami patterns with other base tessellations). These generalized gaps between the target boundary shape and the boundary of the
kirigami patterns with different base tessellations exhibit different deployed configuration. Figure 2c shows that the boundary layer
behaviours in terms of their effective porosity and overall magnifica- area (denoted by A) decreases as the number of tiles (denoted by n2)
tion relative to their compact state (see Supplementary Information, increases following the power law A / ðn2 Þ�1=2 ¼ n�1. To further
Section 5). These examples naturally raise the question: what are explain this, we approximate every boundary
I gap by a triangle and
the limits of kirigami design? One expects that these generalized measure the change in the average triangle base length ~l and aver-
kirigami patterns can only achieve a range of relative area changes, age triangle height ~h for different resolutions. We observe
I that
perimeter changes and curvature changes on deployment, and ~l / n�1 and ~h / n�1,I and hence the average area of the triangles
we can provide theorems on these bounds (see Supplementary ~aI / n�2. As the
I number of boundary gap triangles is approximately
Information, Section 6). However, since these bounds are a strong I we have A  4n~a / n�1.
4n,
function of cut topology, it is possible, nevertheless, to generate a While ourI inverse design approach guarantees that the end-
large variety of shapes. points of deployment, that is the deployed and undeployed states,
Increasing the accuracy of the approximation using kirigami tes- exist, our method so far is agnostic to the path of deployment. To
sellations can be achieved by using a larger number of smaller tiles, explore the deployment process of our patterns, we now extend our
suggesting that there is an accuracy–effort trade-off in matching a purely geometrical approach to a mechanical model by having lin-
prescribed shape. Figure 2c shows several generalized kirigami pat- ear springs along the edges and diagonals of the quads, and simple
terns for circling the square with increasing accuracy; with more torsional springs at the nodal hinges to model the ligaments there.
tiles, the boundary of the deployed pattern gets closer to a perfect Then, the total mechanical energy of the system is given by
circle (see Supplementary Information, Section 4 for multiresolu-
tion results for other patterns). To quantitatively assess the accu-   2
1 X xi � xj  � lij 1 X 2
racy of the patterns for the approximation, we define the boundary Eðx1 ; x2 ; :::; xN Þ ¼ þλ θ ð6Þ
Ns i;j lij Nc i i
layer area of a generalized kirigami pattern by the total area of the

Nature Materials | VOL 18 | SEPTEMBER 2019 | 999–1004 | www.nature.com/naturematerials 1001

Articles NAture MAteriAls
a
2.5
×10–3 One can easily note that the contractibility constraints for sur-
λ=0 face fitting are the same as in equations (1) and (2). For the non-
λ = 0.00001 overlapping constraints, to prevent adjacent faces in the deployed
2 λ = 0.0001
λ = 0.001 configuration from overlapping with or intersecting each other, we
replace the unit normal n^ in equation (4) by the normal computed
1.5 I This implies that we must enforce the
using one of the two faces.
following inequality constraints for every pair of adjacent faces in
E

1 the deployed configuration:

0.5 hðb � aÞ ´ ðc � aÞ; ðc � aÞ ´ ðd � aÞi≥ 0 ð8Þ

0
where a, b, c and d are the nodes of adjacent faces, so that (b, a, c)
0 0.5 1 1.5 2 form a positive (right-hand ordered) angle between the two faces
∆L / L0 and similarly (c, a, d) also form a positive (right-hand ordered)
b angle.
In addition to the above constraints, we need to enforce the con-
dition that the faces are planar so that the volume of the tetrahedron
associated with each face F vanishes, corresponding to the planarity
constraint:
5 cm
VolumeðFÞ ¼ 0 ð9Þ

Fig. 3 | Planar deployment of generalized kirigami tessellation. More explicitly, for quad tessellations the constraint becomes
a, Energetics of the deployment simulations of the square to circle example
with different choice of λ. Here ΔL is the average displacement of the hðb � aÞ ´ ðc � aÞ; d � ai ¼ 0 ð10Þ
pulling points and L0 is the average rest length of the extensional springs.
The insets show the initial, intermediate and final configurations of the where a, b, c and d are the four points of the quad F (see
generalized kirigami pattern under deployment. b, Snapshots of the Supplementary Information for generalizations to other base tes-
deployment of a monostable fabricated model. sellations). We note that this condition, which is automatically sat-
isfied for planar deployment, differentiates the three-dimensional
deployment problem from the two-dimensional one.
Finally, to get smooth solutions, we extend the objective function
where xi are the locations of the nodes, θi are the angles between (5) and the contraction process from our planar design approach
every pair of edges created under the cuts, lij are the rest lengths to three-dimensional surface fitting. This allows us to determine a
of the extensional springs, Ns is the total number of extensional valid generalized kirigami pattern that can be deployed to approx-
springs, Nc is the total number of torsional springs and λ is the ratio imate a prescribed surface in three dimensions by solving a con-
of the torsional spring constant to the extensional spring constant. strained optimization problem using fmincon in MATLAB, with
A larger λ corresponds to a thicker ligament, which has a stronger equations (1), (2) and (7)–(9) to be satisfied.
tendency to close up. By iteratively moving the boundary nodes In Fig. 4 we show several generalized kirigami patterns that can
towards the target boundary shape and solving for the intermediate be deployed to fit surfaces of varying complexity, such as those
deployed configurations by minimizing (6), we obtain a continuous with monotonic positive and negative gauss curvature (Fig. 4a,b).
deployment path. Figure 3a shows the energetics of the deployment Additionally, just as for planar deployments, we can impose extra
simulations as a function of λ: as λ → 0, we see the appearance of boundary angle constraints to produce different pattern design
bistability, while for sufficiently large λ, monostability is observed. effects such as using rectangular quad patterns that can be deployed
To test our predictions experimentally, we fabricated a physi- to fit either a hyperbolic paraboloid as shown in Fig. 4a or an ellip-
cal model by laser cutting a sheet of super-stretchable abrasion- tic paraboloid as shown in Fig. 4b. Our approach also allows us to
resistant natural rubber. Figure 3b shows the deployment snapshots design surfaces with complex non-monotonic Gauss curvature,
of a fabricated model with monostability, demonstrating that the for example a periodic patch of an egg-carton shape (Fig. 4c) and
simulated path and real deployment have similar behaviours (see a bivariate Gaussian (Fig. 4d). In all these examples, we see that
Supplementary Information, Section 4 for another fabricated model our kirigami algorithm provides locally planar tessellations that
of a generalized kagome kirigami pattern). approximately tile surfaces in R3 that have non-zero curvature in
Our inverse design approach has so far focused on approximat- general. This suggests that theI effective surface curvature of the
ing planar shapes. However, we now show that it can be extended holes between the piecewise planar tilings must be non-zero. To
to fit surfaces in three dimensions as well, so that the undeployed quantify this, we fit every hole in the deployed configurations of
pattern space is in R2 while the deployed space is in R3. To fit a sur- our generalized quad kirigami patterns by a bicubic Bézier sur-
I
face S in R3, we replace the boundary shape matching I constraints face, and compute the mean curvature and the Gaussian curvature
(3) by the Isurface matching constraints so that every node xi in the of the interpolant (see Supplementary Information, Section 7 for
deployed configuration satisfies the condition the details); the rightmost columns in Fig. 4 show that the holes
between the planar tilings are indeed curved.
x i k2 ¼ 0
kx i � e ð7Þ To simulate the physical process of deployment, we extend the
planar energetic model into three dimensions, with an additional
where exi is the projection of xi onto S and kk is the Euclidean planarity constraint enforced to ensure that all quads remain pla-
2-norm.I The extra constraints for controlling the
I boundary shape nar through the simulations. Figure 5a shows the deployment
of the undeployed configuration in the planar case (analogous to simulations with the four boundaries of a generalized kirigami
equation (3)) can be directly extended to the surface fitting problem pattern pulled towards the target positions for fitting a hyperbolic
(see Supplementary Information, Section 7). paraboloid. While the intermediate states are warped, the final

1002 Nature Materials | VOL 18 | SEPTEMBER 2019 | 999–1004 | www.nature.com/naturematerials

NAture MAteriAls Articles
a H K
0.03 0.05
0.02
0.025
0.01
0 0
–0.01
–0.025
–0.02
–0.03 –0.05

b
H K
30 100
20
50
10
0 0
–10
–50
–20
–30 –100

c
H K
0.6 2
0.4
1
0.2
0 0
–0.2
–1
–0.4
–0.6 –2

d
H K
0.6 2
0.4
1
0.2
0 0
–0.2
–1
–0.4
–0.6 –2

Fig. 4 | Generalized kirigami patterns for three-dimensional surface fitting. a–d, The target surfaces are a hyperbolic paraboloid (with negative curvature)
(a), a paraboloid (with positive curvature) (b), a periodic patch of an egg-carton shape (c) and a bivariate Gaussian (d). Columns: the target surfaces
(leftmost), the generalized kirigami patterns, the deployed configurations of the patterns that fit the target surfaces, the top views of the deployed patterns
with the holes coloured with the approximated mean curvature H and the top views of the deployed patterns with the holes coloured with the approximated
Gauss curvature K (rightmost).

deployed configuration resembles the shape of a smooth hyper- Our inverse design approach allows us to create non-periodic
bolic paraboloid very well. Furthermore, just as planar deployment compact kirigami patterns that when deployed can approximate
can be either bistable or monostable depending on the value of the any given shape in two or three dimensions. Simple fabrica-
torsional spring constant λ, we see the same effect in three-dimen- tion methods using cutting and moulding allow us to verify our
sional deployment as well; when λ is large enough, the deployment designs for a few planar and three-dimensional deployable shapes.
becomes monostable. When our geometry-based constrained optimization framework
Our inverse design framework is agnostic to the range of possi- is generalized to account for the mechanical response of the til-
ble deployment trajectories aimed towards the target three-dimen- ings and hinges, we see that the response of the generalized kiri-
sional shape, for example by applying displacements to multiple gami patterns can be tuned to switch between monostability and
parts of the boundary, as well as fabrication methods of physical bistability. Altogether, harnessing the underlying topological and
kirigami structures. Figure 5b shows the manual deployment of a geometrical complexity of kirigami in a constrained optimiza-
physical model of a hyperbolic paraboloid fabricated by laser cut- tion framework opens the path for the use of generalized kirigami
ting a natural rubber sheet model. We can also create generalized tesselations as the building blocks of shape-morphing mechani-
kirigami patterns by creating a mould into which one can pour a liq- cal metamaterials. Simultaneously, the present work on inverse
uid polymer and cross-linking it (see Supplementary Information, design of kirigami together with our ability to solve the inverse
Section 7 for a fabricated model produced using polydimethylsilox- design problem in origami22 suggests a follow-up question: can
ane (PDMS) and generalized kagome kirigami patterns for surface we combine origami and kirigami, coupling geometry, topology
fitting obtained by our method). and mechanics, to create structures that morph from any shape

Nature Materials | VOL 18 | SEPTEMBER 2019 | 999–1004 | www.nature.com/naturematerials 1003

Articles NAture MAteriAls
a
×10–3 6. Blees, M. K. et al. Graphene kirigami. Nature 524, 204–207 (2015).
2.5
7. Zhang, Y. et al. A mechanically driven form of Kirigami as a route to 3D
λ=0
λ = 0.00001 mesostructures in micro/nanomembranes. Proc. Natl Acad. Sci. USA 112,
λ = 0.0001 11757–11764 (2015).
2
λ = 0.001 8. Shyu, T. C. et al. A kirigami approach to engineering elasticity in
nanocomposites through patterned defects. Nat. Mater. 14, 785 (2015).
9. Rafsanjani, A. & Bertoldi, K. Buckling-induced kirigami. Phys. Rev. Lett. 118,
1.5
084301 (2017).
10. Konaković, M. et al. Beyond developable: computational design and
E

fabrication with auxetic materials. ACM Trans. Graph. 35, 89 (2016).

1
11. Tang, Y. & Yin, J. Design of cut unit geometry in hierarchical kirigami-based
auxetic metamaterials for high stretchability and compressibility. Extreme
Mech. Lett. 12, 77–85 (2017).
0.5
12. Gatt, R. et al. Hierarchical auxetic mechanical metamaterials. Sci. Rep. 5,
8395 (2015).
13. Kolken, H. M. & Zadpoor, A. A. Auxetic mechanical metamaterials. RSC Adv.
0
0 0.5 1 1.5 2 7, 5111–5129 (2017).
∆L / L0 14. Mitschke, H., Robins, V., Mecke, K. & Schröder-Turk, G. E. Finite
auxetic deformations of plane tessellations. Proc. R. Soc. Lond. A 469,
b 20120465 (2013).
15. Shan, S., Kang, S. H., Zhao, Z., Fang, L. & Bertoldi, K. Design of planar
isotropic negative Poissons ratio structures. Extreme Mech. Lett. 4,
96–102 (2015).
16. Isobe, M. & Okumura, K. Initial rigid response and softening transition of
5 cm
highly stretchable kirigami sheet materials. Sci. Rep. 6, 24758 (2016).
17. Neville, R. M., Scarpa, F. & Pirrera, A. Shape morphing kirigami mechanical
Fig. 5 | Three-dimensional deployment of generalized kirigami metamaterials. Sci. Rep. 6, 31067 (2016).
18. Celli, P. et al. Shape-morphing architected sheets with non-periodic cut
tessellation. a, Energetics of the three-dimensional deployment simulations
patterns. Soft Matter 14, 9744–9749 (2018).
of the pattern in Fig. 4a with different choices of λ. Here ΔL is the average 19. Grünbaum, B. & Shephard, G. C. Tilings and Patterns (Freeman, 1987).
displacement of the pulling points and L0 is the average rest length of 20. Choi, G. P.-T. & Lui, L. M. A linear formulation for disk conformal
the extensional springs. The insets show the initial, intermediate and parameterization of simply-connected open surfaces. Adv. Comput. Math. 44,
final configurations of the pattern under deployment. b, Snapshots of the 87–114 (2018).
21. Meng, T. W., Choi, G. P.-T. & Lui, L. M. TEMPO: feature-endowed
deployment of a monostable fabricated model, with thin threads used Teichmüller extremal mappings of point clouds. SIAM J. Imaging Sci. 9,
for pulling the four sides. Both the numerical simulation and physical 1922–1962 (2016).
deployment results fit the hyperbolic paraboloid very well. 22. Dudte, L. H., Vouga, E., Tachi, T. & Mahadevan, L. Programming curvature
using origami tessellations. Nat. Mater. 15, 583–588 (2016).

Acknowledgements
This work was supported in part by the Croucher Foundation (G.P.T.C.), National
to a given shape with a given mechanical response in two or Science Foundation grant no. DMR 14-20570 (L.M.), DMREF grant no. 15-33985
three dimensions? (L.M.) and EFRI grant no. 18-30901 (L.M.). We thank M. Goldberg for contributing
to preliminary numerical work, M. Gazzola for helpful initial discussions, and A.
Nagarkar and the Whitesides Group for help with fabrication of some of the models
Online content using PDMS.
Any methods, additional references, Nature Research reporting
summaries, source data, statements of code and data availability and
Author contributions
associated accession codes are available at https://doi.org/10.1038/ L.H.D. and L.M. conceived the research. G.P.T.C., L.H.D. and L.M designed the research.
s41563-019-0452-y. G.P.T.C. and L.H.D. conducted the simulations and built the models. G.P.T.C., L.H.D.
and L.M. analysed the results and wrote the manuscript.
Received: 24 October 2018; Accepted: 4 July 2019;
Published online: 19 August 2019 Competing interests
We have filed a patent on our algorithms for kirigami design.
References
1. Grima, J. N. & Evans, K. E. Auxetic behavior from rotating squares. J. Mater.
Sci. Lett. 19, 1563–1565 (2000). Additional information
2. Grima, J. N., Alderson, A. & Evans, K. E. Negative Poisson’s ratios from Supplementary information is available for this paper at https://doi.org/10.1038/
rotating rectangles. Comp. Methods Sci. Technol. 10, 137–145 (2004). s41563-019-0452-y.
3. Grima, J. N., Alderson, A. & Evans, K. E. Auxetic behaviour from rotating Reprints and permissions information is available at www.nature.com/reprints.
rigid units. Phys. Status Solidi B 242, 561–575 (2005). Correspondence and requests for materials should be addressed to L.M.
4. Rafsanjani, A. & Pasini, D. Bistable auxetic mechanical metamaterials inspired
by ancient geometric motifs. Extreme Mech. Lett. 9, 291–296 (2016). Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in
5. Sussman, D. M. et al. Algorithmic lattice kirigami: a route to pluripotent published maps and institutional affiliations.
materials. Proc. Natl Acad. Sci. USA 112, 7449–7453 (2013). © The Author(s), under exclusive licence to Springer Nature Limited 2019

1004 Nature Materials | VOL 18 | SEPTEMBER 2019 | 999–1004 | www.nature.com/naturematerials

NAture MAteriAls Articles
Methods environment. See Supplementary Information for further details on the constraints,
Experiment. The physical models shown in Figs. 3 and 5 were fabricated by objective functions and initial conditions.
perforating patterns on super-stretchable abrasion-resistant natural rubber sheets
with a laser cutter. See Supplementary Information for further details and another Data availability
physical model fabricated using PDMS. The data that support the findings of this study are available from the
corresponding author on reasonable request.
Numerical computations. The numerical computations (solving the constrained
optimization problem and analysing energetics) were conducted with custom Code availability
MATLAB code. Analytic gradients of the constraints and objective functions are Computer codes used in this study are available from the corresponding author
provided to the fmincon constrained optimization routine within the MATLAB on request.

Nature Materials | www.nature.com/naturematerials