Date: March 16, 2017

To: PFC 5 Documentation Set

From: David Potyondy
Re: Material-Modeling Support in PFC [fistPkg25]
Ref: ICG7766-L

This memo describes material-modeling support for PFC 5.0 as provided in the PFC 5.0 FISHTank
(or fistPkg), and this memo corresponds with fistPkg25.

The following capabilities are provided by fistPkg25:

 Material genesis of linear, contact-bonded, parallel-bonded, flat-jointed and user-defined (3D
hill) materials in polyaxial, cylindrical and spherical vessels. Material grains can be either
balls or clumps. Material tests are compression (confined, unconfined and uniaxial strain),
diametral-compression and direct-tension. Microstructural monitoring includes properties
(such as grain-size distribution) with microstructural plot sets and crack monitoring for
bonded materials.

Usage instructions are given in fistPkg-README.txt and summarized as:

 Create a 2D or 3D material (using the appropriate MatGen project).
 Perform compression, diametral-compression and direct-tension tests on the material (using
the CompTest, DiamCompTest and TenTest projects).

The following enhancements are planned:

 Provide direct-shear and fracture-toughness tests, smooth-jointed interface, and stress-
installation procedure.
Material-Modeling Support in PFC [fistPkg25] Page 2

TABLE OF CONTENTS

1.0 PFC MODEL......................................................................................................................... 5

2.0 PFC MATERIALS AND INTERFACE ............................................................................... 9
2.1 Common Material Properties .......................................................................................... 10
2.2 Linear Material ................................................................................................................ 11
2.3 Bonded Materials ............................................................................................................ 13
2.4 Contact-Bonded Material ................................................................................................ 13
2.5 Parallel-Bonded Material ................................................................................................ 16
2.6 Flat-Jointed Material ....................................................................................................... 19
2.6.1 High-Level Description of Flat-Jointed Material....................................................... 22
2.6.2 Flat-Jointed Material Microstructure ......................................................................... 24
2.7 User-Defined Material..................................................................................................... 26
2.8 Smooth-Jointed Interface ................................................................................................ 27
3.0 MICROSTRUCTURAL MONITORING........................................................................... 28
3.1 Microstructural Properties............................................................................................... 28
3.2 Microstructural Plot Sets ................................................................................................. 31
3.3 Force-Chain Fabric.......................................................................................................... 34
3.4 Crack Monitoring ............................................................................................................ 35
4.0 MATERIAL GENESIS....................................................................................................... 38
4.1 Material Vessels .............................................................................................................. 38
4.2 Material-Genesis Procedure ............................................................................................ 41
4.2.1 Packing Phase............................................................................................................. 41
4.2.2 Finalization Phase ...................................................................................................... 45
5.0 LABORATORY-TESTING PROCEDURES..................................................................... 49
5.1 Stress, Strain and Porosity in the Material ...................................................................... 50
5.2 Material Deformability.................................................................................................... 53
5.3 Servomechanism ............................................................................................................. 56
5.4 Loading Rate ................................................................................................................... 58
Material-Modeling Support in PFC [fistPkg25] Page 3

5.5 Compression Test............................................................................................................ 58

5.6 Diametral-Compression Test........................................................................................... 60
5.7 Direct-Tension Test......................................................................................................... 63
5.8 Fracture-Toughness Test ................................................................................................. 64
6.0 STRESS-INSTALLATION PROCEDURE ....................................................................... 64
7.0 REFERENCES.................................................................................................................... 65
Material-Modeling Support in PFC [fistPkg25] Page 4

The PFC model provides a synthetic material consisting of an assembly of rigid grains that interact
at contacts.1 This synthetic material encompasses a vast microstructural space, and only a small
portion of this space has been explored. For example, the bonded-particle modeling methodology
provides a rich variety of microstructural models in the form of bonded materials (Potyondy, 2015;
Potyondy and Cundall, 2004).2 The PFC model includes both granular and bonded materials as well
as an interface that can be inserted into the bonded materials. The support for material modeling that
is provided by PFC2D and PFC3D version 5.0 is described in this document. The material-modeling
support is provided in the form of a consistent set of FISH functions, which we call the PFC 5.0
FISHTank (or fistPkg).3 The PFC 5.0 FISHTank provides a state of the art embodiment of four
well-defined materials and a user-defined material to support: practical applications (via boundary-
value models made from these materials), and scientific inquiry (via further exploration of the
microstructural space described above).

The PFC model is described in Section 1.0. The synthetic materials and interface are described in
Section 2.0. The microstructural monitoring support is described in Section 3.0. The material-genesis
procedure that creates the material is described in Section 4.0. The testing procedures that perform
standard rock-mechanics laboratory tests upon the material are described in Section 5.0. The tests are
used to measure mechanical properties and observe microstructural behavior. The properties and
behaviors are typically compared with that of the physical material. A stress-installation procedure
that initializes stress in the material for subsequent use in boundary-value modeling is described in
Section 6.0. Examples of each synthetic material are provided in the three extensions of this memo,
which are named “Material-Modeling Support in PFC (Example Materials 1, 2 and 3).”

1
Particles in the PFC model interact at contacts by means of a generalized internal force. Contact mechanics is embodied
in particle-interaction laws that employ a soft-contact approach for which all deformation occurs at the contacts between
the rigid bodies. The particle-interaction laws are referred to as contact models. The PFC contact models are described in
Itasca (2016, PFC Model Components: Contacts and Contact Models).
2
A bonded material is defined as a packed assembly of rigid grains joined by deformable and breakable cement at grain-
grain contacts to which larger-scale joints can be added and whose mechanical behavior is simulated by the distinct-
element method using the two- and three-dimensional programs PFC2D and PFC3D.
3
The PFC 5.0 FISHTank replaces the PFC 4.0 FISHTank from PFC2D and PFC3D 4.0. This document replaces the
PFC FISHTank chapters of the PFC2D and PFC3D 4.0 manuals. Material-modeling support may be provided in future
versions of PFC by embedding the material-modeling capability directly into the command set of the PFC programs.
Material-Modeling Support in PFC [fistPkg25] Page 5

1.0 PFC MODEL

The PFC programs (PFC2D and PFC3D) provide a general purpose, distinct-element modeling
framework that includes a computational engine and a graphical user interface. A particular instance
of the distinct-element model is referred to as a PFC model, which refers to both the 2D and 3D
models. The PFC model simulates the movement and interaction of many finite-sized particles.4 The
particles are rigid bodies with finite mass that move independently of one another and can both
translate and rotate. Particles interact at pair-wise contacts by means of an internal force and
moment. Contact mechanics is embodied in particle-interaction laws that update the internal forces
and moments. The time evolution of this system is computed via the distinct-element method, which
provides an explicit dynamic solution to Newton’s laws of motion. The PFC model provides a
synthetic material consisting of an assembly of rigid grains that interact at contacts and includes both
granular and bonded materials.

We here generalize and expand upon the definition of the PFC model given above — refer to
Potyondy (2015) and Itasca (2016) for a comprehensive definition of the PFC model. Potyondy
(2015) summarizes the development of the bonded-particle modeling methodology, generalizes our
view of the BPM to consist of a base material (of rigid grains joined by deformable and breakable
cement at grain-grain contacts) to which larger-scale joints can be added, describes and classifies the
rich variety of microstructural models that can be produced, discusses current limitations and
suggests avenues for further development.

The PFC model simulates the movement of particles and their mechanical interaction at pair-wise
contacts. We denote each particle as a body to clarify that it is not a point mass, but instead, is a rigid
body with finite mass and a well-defined surface. The PFC model consists of bodies and contacts
(see Figure 1). There are three types of bodies: balls, clumps and walls. Bodies have surface
properties that are assigned to the pieces on the body surface. A ball consists of one piece, which is
the ball itself, while the pieces of a clump and wall are called pebbles and facets, respectively. A ball
is a rigid unit-thickness disk in 2D or sphere in 3D. A clump is a collection of pebbles that are rigid
unit-thickness disks in 2D or spheres in 3D. The 2D model consists of unit-thickness disks (see
Figure 2). Clumps model arbitrarily shaped rigid bodies. The pebbles comprising a clump can
overlap but contacts do not exist between them; instead, contacts form between the pebbles on the
boundary of a clump and other bodies. A wall is a collection of facets that are linear segments in 2D
or triangles in 3D and that form a manifold and orientable surface.

Contact mechanics is embodied in particle-interaction laws that employ a soft-contact approach for
which all deformation occurs at the contacts between the rigid bodies. The mechanical interaction

4
The thermal capability expands our definition of the PFC model to include the simulation of transient heat conduction
and storage as well as the development of thermally induced displacements and forces. The expanded PFC model
simulates the movement and both mechanical and thermal interaction of many finite-sized particles. This document
describes the simpler PFC model for which all interactions are mechanical and occur at mechanical contacts.
Material-Modeling Support in PFC [fistPkg25] Page 6

between the surfaces of two bodies occurs at one or more pair-wise mechanical contacts. Contacts
are created and deleted based on body proximity by the contact-detection logic. A contact provides
an interface between two pieces. 5 The interface consists of a contact plane with location  x c  ,

normal direction  nˆ c  and coordinate system  nst  . The contact plane is centered within the
interaction volume (either gap or overlap) of the two pieces, oriented tangential to the two pieces and
rotated to ensure that relative motion of the piece surfaces remains symmetric w.r.t. the contact
plane. Each contact stores a force  Fc  and moment Mc  that act at the contact location in an equal
and opposite sense on the two pieces. 6 The internal force and moment are updated by the particle-
interaction law, which takes the relative motion and surface properties of the two pieces as input. We
refer to the particle-interaction law as a contact model.

Figure 1 PFC model showing bodies and contacts (left) and contact plane with
internal force (right). (From Fig. 1 of Itasca [2016]7.)

5
Contacts are referred to by the types of the two pieces. For example, a contact between a ball and facet is called a ball-
facet contact and has a contact type of ball-facet. In the current PFC model, facet-facet contacts are not allowed and
piece 1 is a ball or pebble while piece 2 is a ball, pebble or facet to give six contact types: ball-ball, ball-pebble, ball-
facet, pebble-ball, pebble-pebble and pebble-facet. For most purposes, we do not differentiate between ball-pebble and
pebble-ball contacts so that there are five contact types.
6
Contact models that simulate interaction at a distance update the force and two moments that need not be equal and
opposite.
7
In documentation set at PFC Model Components: PFC Model Formulation: Model Components.
Material-Modeling Support in PFC [fistPkg25] Page 7

Figure 2 Global coordinate system and orientation of the 2D model, which

consists of unit-thickness disks and linear segments centered in the xy
plane. (From Fig. 1 of Itasca [2016]8.)
The modeling philosophy that underlies both the development and application of the PFC model
(and which includes the importance of having a well-defined model of a system, and the limitations
of any such model) is well stated in the following excerpt from the chapter “Models and Their
Limitations” in a book about the physics of baseball.

In his analysis of a real system, a physicist constructs a well-defined model of the system and
addresses the model. The system we address here is baseball. . . . We cannot calculate from first
principles the character of the collision of an ash bat with a sphere made up of layers of different
tightly wound yarns, nor do we have any precise understanding of the effect of the airstream on
the flight of that sphere, with its curious yin-yang pattern of stitches. What we can do is construct
plausible models of those interactions that play a part in baseball that do not violate basic
principles of mechanics. Though these basic principles . . . severely constrain such models, they
do not completely define them. It is necessary that the models touch the results of observations
― or the results of the controlled observations called experiments ― at some points so that the
model can be more precisely defined and used to interpolate between known results, or to
extrapolate from them. . . . If the model is well chosen, so as to represent the salient points of the
real system adequately, conclusions derived from an analysis of the model can apply to the
system to a useful degree. (Adair, 2002, pp. 1–2)

8
In documentation set at PFC Model Components: PFC Model Formulation: Conventions.
Material-Modeling Support in PFC [fistPkg25] Page 8

A modeling philosophy addresses the essential question: Why are we modeling this problem, and
what can we expect to learn from the model? This question must be answered by each modeler,
because no model is complete or fully verifiable (Oreskes et al., 1994); instead, the best that can be
done is to sanction the model. According to Winsberg (2010, p. 23):

The sanctioning of simulations does not cleanly divide into verification and validation. In fact,
simulation results are sanctioned all at once: simulationists try to maximize fidelity to theory, to
mathematical rigor, to physical intuition, and to known empirical results. But it is the
simultaneous confluence of these efforts, rather than the establishment of each one separately,
that ultimately gives us confidence in the results.

PFC models for intact rock have been sanctioned by demonstrating that they match the response
obtained during tension and compression tests of typical compact rocks. Further discussion of the
sanctioning of PFC models is provided in Potyondy (2015), and further discussion of modeling
philosophy is given in Starfield and Cundall (1988), Starfield (1997) and Nicolson et al. (2002).
Material-Modeling Support in PFC [fistPkg25] Page 9

2.0 PFC MATERIALS AND INTERFACE

A PFC material consists of rigid grains that interact at contacts (see Figure 3).9 The grains of the
PFC material can be either balls or clumps drawn from a general grain-size distribution. It is the type
of contact model at the grain-grain contacts that defines the material as being linear, contact-bonded,
parallel-bonded or flat-jointed. The linear material is a granular material, and the other materials are
bonded materials. A smooth-jointed interface can be inserted into the bonded materials. A user-
defined material can also be created. All materials are produced within a material vessel such that
they form a homogeneous, isotropic and well-connected grain assembly with a specified non-zero
material pressure. The PFC material is produced by the material-genesis procedures described in
Section 0.

The PFC materials and interface are described in the remaining subsections. Each material and
interface is defined by a set of material properties. These properties control the material-genesis
procedure, install the desired contact model at selected contacts and assign contact-model properties.
The properties common to all PFC materials are described in the first subsection. The linear material
is described in the second subsection. A general description of the bonded materials is provided in
the third subsection, and followed by descriptions of each bonded material, the user-defined
material, and the smooth-jointed interface.

Figure 3 Bonded material consisting of grains (with balls in yellow and clumps
in blue and red) and cement (drawn as pairs of black lines). (From
Fig. 6 of Potyondy [2015].)

9
The physical behavior at each contact is embodied in its contact model. For the bonded-particle materials, this physical
behavior mimics cement, and thus, a bonded material has been said to consist of rigid grains joined by cement at grain-
grain contacts. The PFC materials include a granular material along with the bonded materials; therefore, the present
definition of a PFC material replaces “rigid grains joined by cement” with “rigid grains that interact at contacts.”
Material-Modeling Support in PFC [fistPkg25] Page 10

2.1 Common Material Properties

The properties common to all PFC materials are described here and defined by the parameters listed
in Table 1. The material name is used both to create a model title that corresponds with the current
state of the model and to name the save states. The model title is displayed at the top of most views.
The material type may be either one of the four supported materials or a user-defined material.
Kinetic energy can be dissipated via local damping by specifying a non-zero value for the local-
damping factor of each grain — quasi-static conditions are modeled by setting the local-damping
factor to 0.7, and dynamic effects can be studied by choosing the local-damping factor to correspond
with the seismic quality factor of the rock (Potyondy and Cundall, 2004, Eq. 3). The grain density is
specified either directly or indirectly in terms of the bulk density. If the bulk density is given, then a
single value of grain density is computed such that the mass of synthetic material will equal the bulk
density times the vessel volume. The grain shape and size distribution are specified as follows. The
grain shape must be either all balls or all clumps. If clumps are chosen, then a clump template must
be provided for each clump shape. The size distribution is specified in terms of a finite number of
separate size distributions, each of which is either uniform or Gaussian. The volume fraction and
diameter range is given for each distribution. 10 The size distribution can be shifted by modifying the
diameter multiplier — synthetic materials that differ only in their average grain diameter can be
created by varying the diameter multiplier.

Table 1 Common Parameters

Parameter Type Range Default Description

Nm , cm_matName STR NA PFCmat material name

material-type code
 0, linear
 1, contact-bonded
Tm , cm_matType INT 0,4 0 
 2, parallel-bonded
 3, flat-jointed

 4, user-defined

contact-model name ( Tm  4,
N cm , cm_modName STR NA NA
also redefine ft_setMatBehavior)
local-damping factor
 , cm_localDampFac FLT 0.0,0.7 0 .0
(for local damping)
density code
C p , cm_densityCode INT 0,1 0  0, grain

 1, bulk

10
A given grain-size distribution (GSD) can be matched by specifying the volume fractions corresponding with the range
of grain sizes — i.e., by breaking the given GSD into a finite number of uniform distributions (see Figure 14).
Material-Modeling Support in PFC [fistPkg25] Page 11

density value (set grain density:

 v , C  0
g  
v , cm_densityVal FLT  0.0, NA  vVv Vg , C  1
Vv is volume of vessel, and
Vg is total volume of grains)
Grain shape & size distribution group:
grain-shape code
S g , cm_shape INT 0,1 0  0, all balls

 1, all clumps
nSD , cm_nSD INT nSD  1 NA number of size distributions
size-distribution type
TSD , cm_typeSD( nSD ) STR 0,1 0  0, uniform

 1, gaussian
Nct j , cm_ctName( n SD ) STR NA NA clump-template name ( Sg 1)

Dl  j  , cm_Dlo( n SD ) FLT 0.0, NA diameter range (lower)

diameter range (upper)
Du j  , cm_Dup( n SD ) FLT Du j  Dl  j  NA
(clumps: volume-equiv. sphere)

  j  , cm_Vfrac( n SD ) FLT 0.0,1.0 NA volume fraction       1.0 

j

Dmult , cm_Dmult FLT 0.0, 1.0

diameter multiplier
(shifts the size distribution)

2.2 Linear Material

A linear material is a granular assembly in which the linear contact model exists at all grain-grain
contacts at the end of the material-finalization phase; new grain-grain contacts that may form during
subsequent motion are also assigned the linear contact model. A behavior summary of the linear
contact model follows — refer to Itasca (2016)11, for a comprehensive description of the model.

The linear model provides the behavior of an infinitesimal interface that does not resist relative
rotation so that the contact moment equals zero ( Mc  0). The contact force is resolved into linear

and dashpot components ( Fc  Fl  Fd ). The linear component provides linear elastic (no-tension),
frictional behavior, while the dashpot component provides viscous behavior (see Figure 4). The

In documentation set at PFC Model Components: Contacts and Contact Models: Contact Models: Built-in Contact
11

Models: Linear Model.

Material-Modeling Support in PFC [fistPkg25] Page 12

linear force is produced by linear springs with constant normal and shear stiffnesses, kn and ks. The
linear springs cannot sustain tension and slip is accommodated by imposing a Coulomb limit on the
shear force using the friction coefficient,  . The dashpot force is produced by dashpots with
viscosity given in terms of the normal and shear critical-damping ratios, n and s . The linear
springs act in parallel with the dashpots. The linear model with inactive dashpots and a reference gap
of zero corresponds with the model of Cundall and Strack (1979).

Figure 4 Behavior and rheological components of the linear model. (From Fig.
1 of Itasca [2016]12.)
A linear material is defined by the parameters in Table 2. The properties in the linear material group
are used to set the relevant properties of the linear model during material finalization, and these same
properties are assigned to grain-grain contacts during packing and that may form subsequent to
material finalization.

The relevant properties of the linear model are as follows. The linear model provides linear and
dashpot components. Only the linear component is active, the reference gap is zero and the normal-
force update mode is absolute. The normal and shear stiffnesses are set based on a specified contact
deformability ( E * and  * of the deformability method), and the friction coefficient is set to  .

Table 2 Linear Material Parameters

Parameter Type Range Default Description

Material microproperties are listed via @mpListMicroProps.

In documentation set at PFC Model Components: Contacts and Contact Models: Contact Models: Built-in Contact
12

Models: Linear Model.

Material-Modeling Support in PFC [fistPkg25] Page 13

Common material parameters are listed in Table 1.

Packing parameters are listed in Table 7.
Linear material group:
E *
, lnm_emod FLT 0.0, 0 .0 effective modulus

 *
, lnm_krat FLT 0.0, 0 .0 stiffness ratio

 , lnm_fric FLT 0.0, 0 .0 friction coefficient

2.3 Bonded Materials

Bonded materials are analogous to intact rock, which can be viewed as an aggregate of crystals and
amorphous particles joined by varying amounts of cementing materials. The bonded materials
produced by the material-genesis procedure approximate an intact compact rock with average
homogeneous isotropic properties at a scale larger than the material granularity. The synthetic
material is analogous to brittle cookie dough that has been solidified by baking and for which grain
size corresponds with dough granularity. The granularity is quantified by means of its resolution (or
number of grains across a relevant dimension).

A rich variety of microstructures can be produced by modifying the bonded material itself. Such
microstructures are obtained either by modifying the properties of the grains and cement or by
modifying the packing fabric. The grain properties are size and shape. The cement properties are
deformability and strength as well as evolving damage. The cement properties are embodied in the
contact model, but the macroscopic material behavior is also sensitive to the ways in which new
contacts form and contacts deemed to be broken behave. The packing fabric is affected by the size
distribution and shapes of the grains as well as the material pressure. Structural features at a scale
larger than the material granularity can be overlaid on the base material. These features include
voids, material regions and joints (see Potyondy [2015]).

Damage in the bonded materials consists of bond-breakage events. In the contact- and parallel-
bonded materials, the entire interface breaks, whereas in the flat-jointed material, the elements break.
Each breakage event is denoted as a crack; thus, a fully broken interface in a 2D flat-joint contact
with four elements contains four cracks. Cracks in the flat-jointed material are shown in Figure 10.

2.4 Contact-Bonded Material

A contact-bonded material is a granular assembly in which the linear contact bond contact model
exists at all grain-grain contacts at the end of the material-finalization phase; new grain-grain
contacts that may form during subsequent motion are assigned the linear contact model. A behavior
Material-Modeling Support in PFC [fistPkg25] Page 14

summary of the linear contact bond contact model follows — refer to Itasca (2016)13 for a
comprehensive description of the model.

The linear contact bond model provides the behavior of an infinitesimal, linear elastic and either
bonded or frictional interface that carries a force (see Figure 5). The interface does not resist relative
rotation and is either bonded or unbonded. If bonded, the behavior is linear elastic until the strength
limit is exceeded and the bond breaks making the interface unbonded. If unbonded, the behavior is
linear elastic and frictional with slip being accommodated by imposing a Coulomb limit on the shear
force. The unbonded linear contact bond model is equivalent to the linear model. The linear contact
bond model with inactive dashpots and a reference gap of zero corresponds with the contact-bond
model of Potyondy and Cundall (2004).

Figure 5 Behavior and rheological components of the linear contact bond

model with inactive dashpots. (From Fig. 1 of Itasca [2016]14.)
A contact-bonded material is defined by the parameters in Table 3. The properties in the contact-
bonded material group are used to set the relevant properties of the linear contact bond model during
material finalization, and the properties in the linear material group are assigned to grain-grain
contacts during packing and that may form subsequent to material finalization.

13
In documentation set at PFC Model Components: Contacts and Contact Models: Built-in Contact Models: Contact
Models: Linear Contact Bond Model.
14
In documentation set at PFC Model Components: Contacts and Contact Models: Contact Models: Built-in Contact
Models: Linear Contact Bond Model.
Material-Modeling Support in PFC [fistPkg25] Page 15

The relevant properties of the linear contact bond model are as follows. There are linear and dashpot
components. Only the linear component is active. The reference gap is zero unless a bond is
installed. If a bond is installed, then the reference gap is set equal to the contact gap at the time of
bond installation (thereby establishing reference surfaces that are just touching and effectively
removing the overlap or extending the grain surfaces). The normal-force update mode is absolute.
The normal and shear stiffnesses are set based on a specified contact deformability ( E* and  * of the
deformability method), and the friction coefficient is set to  . Bonds are installed during material
finalization at the grain-grain contacts with a gap less than or equal to the installation gap  gi  . The
remaining properties correspond directly with properties in the contact-bond group of the linear
contact bond model.
The relevant properties of the linear model are as follows. The linear model provides linear and
dashpot components. Only the linear component is active, the reference gap is zero and the normal-
force update mode is absolute. The normal and shear stiffnesses are set based on a specified contact
deformability ( E n * and  n * of the deformability method), and the friction coefficient is set to n .

Table 3 Contact-Bonded Material Parameters

Parameter Type Range Default Description

Material microproperties are listed via @mpListMicroProps.

Common material parameters are listed in Table 1.
Packing parameters are listed in Table 7.
Contact-bonded material group:
Linear group:
E *
, cbm_emod FLT 0.0, 0 .0 effective modulus

 *
, cbm_krat FLT 0.0, 0 .0 stiffness ratio

 , cbm_fric FLT 0.0, 0 .0 friction coefficient

Contact-bond group:

gi , cbm_igap FLT 0.0, 0 .0 installation gap

T m,sd FLT 0.0, 0.0,0.0 tensile-strength dist. [stress]

(mean and std. deviation)
cbm_tens_{m,sd}

S m,sd FLT 0.0, 0.0,0.0 shear-strength dist. [stress]

(mean and std. deviation)
cbm_shears_{m,sd}
Linear material group (for grain-grain contacts during packing and that may form
subsequent to material finalization):
En* , lnm_emod FLT 0.0, 0 .0 effective modulus
Material-Modeling Support in PFC [fistPkg25] Page 16

 n * , lnm_krat FLT 0.0, 0.0 stiffness ratio

 n , lnm_fric FLT 0.0, 0.0 friction coefficient

2.5 Parallel-Bonded Material

A parallel-bonded material is a granular assembly in which the linear parallel bond contact model
exists at all grain-grain contacts at the end of the material-finalization phase; new grain-grain
contacts that may form during subsequent motion are assigned the linear contact model. A behavior
summary of the linear parallel bond contact model follows — refer to Itasca (2016)15 for a
comprehensive description of the model.

The linear parallel bond model provides the behavior of two interfaces: an infinitesimal, linear
elastic (no-tension) and frictional interface that carries a force and a finite-size, linear elastic and
bonded interface that carries a force and moment (see Figure 6). The first interface is equivalent to
the linear model: it does not resist relative rotation, and slip is accommodated by imposing a
Coulomb limit on the shear force. The second interface is called a parallel bond, because when
bonded, it acts in parallel with the first interface. When the second interface is bonded, it resists
relative rotation, and its behavior is linear elastic until the strength limit is exceeded and the bond
breaks making it unbonded. When the second interface is unbonded, it carries no load. The
unbonded linear parallel bond model is equivalent to the linear model. The linear parallel bond
model with inactive dashpots and a reference gap of zero corresponds with the parallel-bond model
of Potyondy and Cundall (2004).

In documentation set at PFC Model Components: Contacts and Contact Models: Built-in Contact Models: Contact
15

Models: Linear Parallel Bond Model.

Material-Modeling Support in PFC [fistPkg25] Page 17

Figure 6 Behavior and rheological components of the linear parallel bond

model with inactive dashpots. (From Fig. 1 of Itasca [2016]16.)
A parallel-bonded material is defined by the parameters in Table 4. The properties in the parallel-
bonded material group are used to set the relevant properties of the linear parallel bond model during
material finalization, and the properties in the linear material group are assigned to grain-grain
contacts during packing and that may form subsequent to material finalization.

The relevant properties of the linear parallel bond model are as follows. The linear parallel bond
model provides the behavior of two interfaces. The first interface is equivalent to the linear model,
and the second interface is called a parallel bond.
 The relevant properties of the first interface are listed in the linear group. The first interface
provides linear and dashpot components. Only the linear component is active. The reference
gap is zero but may be altered when a bond is installed. If a bond is installed and the contact
gap is negative, then the reference gap is set equal to the contact gap at the time of bond
installation (thereby establishing reference surfaces that are just touching and effectively
removing the overlap). The normal-force update mode is absolute. The normal and shear
stiffnesses of the first interface are set based on a specified contact deformability ( E * and  *
of the deformability method), and the friction coefficient is set to  .

In documentation set at PFC Model Components: Contacts and Contact Models: Contact Models: Built-in Contact
16

Models: Linear Parallel Bond Model.

Material-Modeling Support in PFC [fistPkg25] Page 18

 The relevant properties of the second interface are listed in the parallel-bond group. The
second interface carries load only when it is bonded. Bonds are installed during material
finalization at the grain-grain contacts with a gap less than or equal to the installation gap
 g i  . The normal and shear stiffnesses of the second interface are set based on a specified
deformability ( E * and  * of the pb_deformability method). The remaining properties of
the second interface correspond directly with properties in the parallel-bond group of the
linear parallel bond model.

The relevant properties of the linear model are as follows. The linear model provides linear and
dashpot components. Only the linear component is active, the reference gap is zero and the normal-
force update mode is absolute. The normal and shear stiffnesses are set based on a specified contact
deformability ( E n * and  n * of the deformability method), and the friction coefficient is set to n.

Table 4 Parallel-Bonded Material Parameters

Parameter Type Range Default Description

Material microproperties are listed via @mpListMicroProps.

Common material parameters are listed in Table 1.
Packing parameters are listed in Table 7.
Parallel-bonded material group:
Linear group:

E * , pbm_emod FLT 0.0, 0 .0 effective modulus

 * , pbm_krat FLT 0.0, 0.0 stiffness ratio

 , pbm_fric FLT 0.0, 0 .0 friction coefficient

Parallel-bond group:

gi , pbm_igap FLT 0.0, 0 .0 installation gap

 , pbm_rmul FLT  0.0, 1.0 radius multiplier

E *
, pbm_bemod FLT 0.0, 0 .0 bond effective modulus

 * , pbm_bkrat FLT 0.0, 1.0 bond stiffness ratio

, pbm_mcf FLT  0.0,1.0  0.0 moment-contribution factor

c m,sd FLT 0.0, 0.0,0.0 tensile-strength dist. [stress]

(mean and std. deviation)
pbm_ten_{m,sd}

c m,sd FLT 0.0, 0.0,0.0 cohesion dist. [stress]

(mean and std. deviation)
Material-Modeling Support in PFC [fistPkg25] Page 19

pbm_coh_{m,sd}

 , pbm_fa FLT 0.0,90.0  0 .0 friction angle [degrees]

Linear material group (for grain-grain contacts during packing and that may form
subsequent to material finalization):
E n * , lnm_emod FLT 0.0, 0 .0 effective modulus

 n * , lnm_krat FLT 0.0, 0.0 stiffness ratio

 n , lnm_fric FLT 0.0, 0 .0 friction coefficient

2.6 Flat-Jointed Material

A flat-jointed material is a granular assembly in which the flat-joint contact model exists at all grain-
grain contacts with a gap less than or equal to the installation gap at the end of the material-
finalization phase; all other grain-grain contacts as well as new grain-grain contacts that may form
during subsequent motion are assigned the linear contact model. A behavior summary of the flat-
joint contact model follows — refer to Itasca (2016)17 for a comprehensive description of the model.
A high-level description of a flat-jointed material is provided in Section 2.6.1.

The flat-joint model provides the macroscopic behavior of a finite-size, linear elastic and either
bonded or frictional interface that may sustain partial damage (see Figure 7). The interface is
discretized into elements. Each element is either bonded or unbonded, and the breakage of each
bonded element contributes partial damage to the interface. The behavior of a bonded element is
linear elastic until the strength limit is exceeded and the bond breaks making the element unbonded,
while the behavior of an unbonded element is linear elastic and frictional with slip accommodated by
imposing a Coulomb limit on the shear force. Each element carries a force and moment that obey the
force-displacement law described below, while the force-displacement response of the flat-joint
interface is an emergent behavior that includes evolving from a fully bonded state to a fully
unbonded and frictional state.

In documentation set at PFC Model Components: Contacts and Contact Models: Contact Models: Built-in Contact
17

Models: Flat-Joint Model.

Material-Modeling Support in PFC [fistPkg25] Page 20

Figure 7 Behavior and rheological components of the flat-joint model. (From

Fig. 8 of Itasca [2016]18.)
A flat-jointed material is defined by the parameters in Table 5. The properties in the flat-jointed
material group are used to set the relevant properties of the flat-joint model during material
finalization, and the properties in the linear material group are assigned to grain-grain contacts
during packing and that may form subsequent to material finalization.

The relevant properties of the flat-joint model are as follows. The flat-joint contact model is installed
at all grain-grain contacts with a gap less than or equal to the installation gap ( gi ). The normal and
shear stiffnesses are set based on a specified deformability ( E * and  * of the deformability method).
 
The property set B ,G ,  go m,sd , C , v is described in Section 2.6.2. The remaining properties
correspond directly with properties of the flat-joint model

The relevant properties of the linear model are as follows. The linear model provides linear and
dashpot components. Only the linear component is active, the reference gap is zero and the normal-
force update mode is absolute. The normal and shear stiffnesses are set based on a specified contact
deformability ( E n * and  n * of the deformability method), and the friction coefficient is set to  n .

In documentation set at PFC Model Components: Contacts and Contact Models: Contact Models: Built-in Contact
18

Models: Flat-Joint Contact Model.

Material-Modeling Support in PFC [fistPkg25] Page 21

Table 5 Flat-Jointed Material Parameters

Parameter Type Range Default Description

Material microproperties are listed via @mpListMicroProps.

Common material parameters are listed in Table 1.
Packing parameters are listed in Table 7.
Flat-jointed material group:
microstructure-tracking flag
CMS , fjm_trackMS BOOL true, false false (draw microstructure with
the faced grain plot set)
gi , fjm_igap FLT 0.0, 0 .0 installation gap

B  , fjm_B_frac FLT 0.0,1.0 NA bonded fraction

G  , fjm_G_frac FLT 0.0,1.0 NA gapped fraction

 go m,sd , fjm_G_{m,sd} FLT 0.0, 0.0,0.0 initial surface-gap distribution

(mean and std. deviation)
elements in radial direc.
Nr , fjm_Nr INT 1, 2
(2D model: total elements)
elements in circumf. direc.
N , fjm_Nal INT 3,  4
(3D model only)
radius-multiplier code
C , fjm_rmulCode INT 0,1 0  0, fixed

 1, varying
radius-multiplier value
 f , C  0
  v  
 v , fjm_rmulVal FLT  0.0, 1.0  o , C  1
 f is fixed value, and
o is starting value

E *
, fjm_emod FLT 0.0, 0 .0 effective modulus

 * , fjm_krat FLT 0.0, 0.0 stiffness ratio

 , fjm_fric FLT 0.0, 0 .0 friction coefficient

c m,sd FLT 0.0, 0.0,0.0 tensile-strength dist. [stress]

(mean and std. deviation)
fjm_ten_{m,sd}

cm,sd FLT 0.0, 0.0,0.0 cohesion dist. [stress]

(mean and std. deviation)
fjm_coh_{m,sd}
 , fjm_fa FLT 0.0,90.0  0 .0 friction angle [degrees]
Material-Modeling Support in PFC [fistPkg25] Page 22

Linear material group (for grain-grain contacts during packing, that are not flat-jointed, and
that may form subsequent to material finalization):
E n * , lnm_emod FLT 0.0, 0 .0 effective modulus

 n * , lnm_krat FLT 0.0, 0.0 stiffness ratio

 n , lnm_fric FLT 0.0, 0 .0 friction coefficient

+ Slit fraction:  S  1   B   G  0   S  1  .

2.6.1 High-Level Description of Flat-Jointed Material

A flat-joint contact and its corresponding flat-jointed material are shown in Figure 8. A flat-joint
contact simulates the behavior of an interface between two notional surfaces, each of which is
connected rigidly to a piece of a body. A flat-jointed material consists of bodies (balls, clumps or
walls) joined by flat-joint contacts such that the effective surface of each body is defined by the
notional surfaces of its pieces, which interact at each flat-joint contact with the notional surface of
the contacting piece. The notional surfaces are called faces, which are lines in 2D and disks in 3D.

The following description of a flat-jointed material applies to the case in which bodies are balls;
however, the flat-joint model can be installed at both ball-ball and ball-wall contacts. We refer to the
balls of a flat-jointed material as faced grains, each of which is depicted as a circular or spherical
core and a number of skirted faces. The faced grains are created when the flat-joint model is installed
at the ball-ball contacts of a packed ball assembly (see Figure 9). An interface exists between each
set of adjoining faces and is discretized into elements with each element being either bonded or
unbonded. The breakage of each bonded element contributes partial damage to the interface, and
each breakage event is denoted as a crack (see Figure 10).19 If the relative displacement at a flat-joint
contact becomes larger than the flat-joint diameter, then the adjoining faces may be removed
(because the contact may be deleted) making the associated balls locally circular or spherical; if
these balls come back into contact, the behavior will be that of an interface between circular or
spherical surfaces (if the linear contact model is assigned to the new contact).

19
Cracks in the 2D flat-jointed material are depicted as colored lines lying on the interface between the two disk-shaped
pebbles with color denoting breakage mode and with line thickness proportional to the element gap. Cracks in the 3D
flat-jointed material are depicted as colored octagons lying on the interface between the two spherical pebbles with color
denoting breakage mode.
Material-Modeling Support in PFC [fistPkg25] Page 23

Figure 8 Flat-joint contact (left) and flat-jointed material (right). (From Fig. 1
of Itasca [2016]20.)

In documentation set at PFC Model Components: Contacts and Contact Models: Contact Models: Built-in Contact
20

Models: Flat-Joint Contact Model.

Material-Modeling Support in PFC [fistPkg25] Page 24

Figure 9 Creation of faced grains showing packed ball assembly (left) and
initial faced-grain assembly (right). (From Fig. 2 of Itasca [2016]21.)

Figure 10 Partially damaged flat-jointed material showing faced grains with

cracks colored red/blue for tensile/shear failure. (From Fig. 3 of Itasca
[2016]22.)
2.6.2 Flat-Jointed Material Microstructure
The microstructure of a flat-jointed material (see Figure 11) is such that each flat-joint contact is
initially either bonded ( go  0 , bonded), gapped ( g o  0 , unbonded) or slit ( go  0 , unbonded). This

21
In documentation set at PFC Model Components: Contact and Contact Models: Contact Models: Built-in Contact
Models: Flat-Joint Contact Model.
22
In documentation set at PFC Model Components: Contact and Contact Models: Contact Models: Built-in Contact
Models: Flat-Joint Contact Model.
Material-Modeling Support in PFC [fistPkg25] Page 25

information is stored in the property ‘fj_mtype’ of each flat-jointed contact. The material
microstructural parameters are the fraction of bonded and gapped contacts ( B and G), and the mean
and standard deviation of the initial gap for the gapped contacts (  go m,sd ). When the flat-jointed
material is created, each flat-joint contact is designated as being of type B, G or S (as described
below), and the type G contacts are assigned an initial gap drawn from a normal distribution with the
specified mean and standard deviation. The initial microstructural types of the flat-jointed material
are computed and listed by mpListMicroStrucProps, which returns the number of flat-joints
initially bonded (mp_nFJiB), gapped (mp_nFJiG), slit (mp_nFJiS) or undefined (mp_nFJiU).

The following type-designation process is applied to the set of flat-joint contacts. Denote the
fractions of bonded, gapped and slit contacts as B, G and S, respectively. These fractions are related
to the number of bonded, gapped and slit contacts ( nB , nG and nS , respectively) as well as the total
number of flat-joint contacts ( n FJ ) by the relations:

nB n n
B  , G  G ,  S  S
nFJ nFJ nFJ (1)
 B  G   S  1 and nFJ  nB  nG  nS .

The microstructural types are assigned by looping through the set of flat-joint contacts. The first nB
of these contacts are designated as type B, the next nG of these contacts are designated as type G and
the remaining nS of these contacts are designated as type S. The type designations are spatially
random, because the contact list is spatially random (which is assumed to be the case).

Microstructural validity describes whether or not the grain facets overlap one another; a flat-jointed
material has a valid microstructure if and only if the facets of each grain can be connected to the
grain center with no overlap.23 Microstructural validity is controlled as follows (see Figure 12). The
radius multipliers of a flat-jointed material can either be fixed or allowed to vary. If the multipliers
are fixed ( C  0 ), then all flat-joint radius multipliers are set equal to the specified value. If the
multipliers are allowed to vary ( C  1 ), then all flat-joint radius multipliers are initially set equal to
the specified value and then reduced (in a non-uniform fashion) to obtain a valid microstructure.

23
A microstructurally invalid model may produce useful behavior, and as such, its use can be justified — parallel-
bonded materials with radius multipliers of unity have invalid microstructures.
Material-Modeling Support in PFC [fistPkg25] Page 26

Figure 11 Initial microstructure of flat-jointed material: each flat-joint contact

is bonded (type B), gapped (type G) or slit (type S).

Figure 12 Initial radius multipliers of flat-jointed material: each flat-joint radius

multiplier is fixed, which may produce an invalid microstructure (left)
or each flat-joint radius may be reduced from an initial maximum
value to obtain a valid microstructure (right). {DP: This image was
produced by PFC2D 4.0; it will be reproduced using the faced-grain
plotting capability in PFC2D 5.0.}
2.7 User-Defined Material
A user-defined material is a granular assembly in which a user-defined contact model exists at all
grain-grain contacts. The user-defined contact model is referred to in commands and FISH by the
contact-model name ( N cm in Table 1), and is provided as a dynamic link library (DLL) file that is
loaded into PFC at runtime. The user-defined material example creates a 3D hill material that uses
Material-Modeling Support in PFC [fistPkg25] Page 27

the hill contact model by specifying the contact-model name as hill. The user-defined material is
defined by a set of parameters that are used to set the relevant properties of the contact model during
material genesis and subsequent modeling — the properties are specified in the redefinition of the
ft_setMatBehavior FISH function. The following additional FISH functions must also be
provided for a user-defined material: udm_checkParams, udm_listProps,
udm_computeMicroStrucProps, and udm_listMicroStrucProps. Refer to the user-
defined material example for more information about how to create a user-defined material.

2.8 Smooth-Jointed Interface

A smooth-jointed interface can be inserted into the bonded materials by identifying the contacts near
the interface, and replacing their contact models with the smooth-joint contact model. New grain-
grain contacts that may form during subsequent motion are assigned the linear contact model, with
the exception of the new contacts associated with the interface, which are assigned the smooth-joint
contact model and aligned with the interface direction. {DP: To be continued…}
Material-Modeling Support in PFC [fistPkg25] Page 28

3.0 MICROSTRUCTURAL MONITORING

The macroscopic behavior of the synthetic material arises from the evolution of the microstructural
features. Microstructural monitoring includes computation of quantitative microstructural properties,
microstructural plot sets, force-chain fabric and crack monitoring. These items are described in the
following four subsections.

3.1 Microstructural Properties

The microstructural properties of the material are computed and listed by
mpListMicroStrucProps and include the following items.
 Grain Size and Packing Information. Number of grains in the model, grain-size
distribution (discussed below), average and median grain diameters, vessel resolutions w.r.t.
the average and median grain diameters,24 measurement-based porosity (defined in Section
5.1), and overlap ratios.25
 Contact Information. The number of active linear-based contacts along with the number of
such contacts that are grain-grain and grain-wall.
 Bonded-Material Information. The bonded materials provide this information. Bond
coordination number  cb  .26 Number of contact-bonded bonds, parallel-bonded bonds, flat-
jointed contacts, flat-jointed elements, and flat-jointed bonds. The initial microstructural
types of the flat-jointed material (defined in Section 2.6.2).
 User-Defined Material Information. A user-defined material may provide this information
via udm_{compute,list}MicroStrucProps.

The cumulative distribution of grain size is most often used to characterize particulate media
(Santamarina et al., 2001). The typical grain size distribution curve is a cumulative distribution of
mass according to particle size — it is NOT a cumulative distribution of particle counts. The curve
expresses the mass percentage of particles that can pass through a given sieve size. The horizontal
axis is particle size, and the vertical axis is “mass percent passing.” A typical grain size distribution
curve is shown in Figure 13. If the vertical axis is labelled “percent finer than,” then it is understood
that this denotes the percentage by mass, not the percentage by number. The equivalent particle sizes
corresponding to selected percentiles are frequently quoted. For example, d 50 corresponds to the
sieve size that 50 percent of the particles can pass through (and these particles comprise 50 percent
of the total mass), and d10 corresponds to the sieve size that 10 percent of the particles can pass

24
Vessel resolution is defined as the number of grains across the minimum vessel dimension.
25
The overlap ratios (ORs) of all active contacts are computed, and the maximum, minimum and average values are
returned. The OR is the overlap divided by the piece diameter. For grain-grain contacts, there are two ORs, while for
grain-wall contacts, there is one OR.
26
The bond coordination number is defined for the bonded materials as the average number of intact bonds per grain.
Material-Modeling Support in PFC [fistPkg25] Page 29

through (and these particles comprise 10 percent of the total mass). The quantity d 50 is called the
median grain size. The slope of the cumulative distribution and its convexity or concavity are
characterized by the coefficients of uniformity C u and curvature C c :

d 60 d 302
Cu  and Cc  . (2)
d10 d10 d 60

A well-graded material with Cu  4 indicates that there are large particles forming the soil frame
and that voids are filled with smaller particles, potentially forming a dense soil mass.

Figure 13 Typical grain size distribution curve showing the cumulative

distribution of mass according to particle size. (From Fig. 2.10 of
Santamarina et al. [2001].)
The grain size distribution (GSD) for a sample of the synthetic material is computed via
@gsdMeasure(n), where n is the number of bins over which the GSD is sampled, and the
sampled grains are defined by @gsdMark (see Figure 14). The GSD is returned in a table named
‘GSD’ for which the x-axis is the effective grain diameter and the y-axis is the cumulative mass
percent passing. A table named ‘GSD-retained’ is also created for which the x-axis is the effective
grain diameter and the y-axis is the incremental mass percent retained. The cumulative mass percent
passing curve corresponds with the typical grain size distribution curve shown in Figure 13. The
incremental mass percent retained histogram corresponds with the material retained on each sieve
during a sieve analysis. The grain size for percentile pct is extracted from the table named ‘GSD’
Material-Modeling Support in PFC [fistPkg25] Page 30

via @gsdGetSizeOfPercentile(pct). For circular and spherical grains, the grain mass is
given by

  Dg 
2

   t , 2D  t  1
  2  (3)
M g   gV g , V g   3
 4  Dg 
 3   2  , 3D
  

where  g is grain density, Vg is grain volume and Dg is grain diameter.27 Some of the grain size
distribution information can be obtained by the measurement logic. 28

Figure 14 Grain size distribution information produced by the FISH function

gsdMeasure(19) showing cumulative mass percent passing (in black)
and incremental mass percent retained (in red).

27
The current fistPkg supports circular and spherical grain shapes by considering such grains as balls. The fistPkg will
support general grain shapes by considering such grains as clumps (by setting Sg  1 in Table 1, and providing a clump
template for each shape).
28
The cumulative volume percent passing curve can be obtained with the measurement logic (by specifying the number
of bins and the grain-diameter range for the size-distribution computation when creating a measurement sphere with the
measure create command, and extracting the information into a table with the measure dump command). If all
sampled grains have the same density, then the cumulative volume percent passing curve is equal to the cumulative mass
percent passing curve.
Material-Modeling Support in PFC [fistPkg25] Page 31

3.2 Microstructural Plot Sets

Microstructural plot sets are provided for the bonded materials to display the material microstructure
and thereby reveal how the evolution of the microstructure influences the macroscopic behavior. The
microstructural plot sets include depictions of the grains and the grain-grain interfaces, and when
used with the crack-monitoring package, include the interface damage in the form of bond
breakages.

A plot set is defined as a plot in which the plot information is stored as a geometry set, and displayed
by the geometry plot item by specifying: {ColorBy: Set}, and {Sets: <geometry set name>}. The
microstructural plot sets are regenerated at a specified rate (msUpdateRate) during cycling, and
the regeneration can be forced to occur via @msForceUpdate. An axis-aligned microstructural
box (defined by @msBoxDefine) is associated with the microstructural plot sets. The box is
displayed by the geometry plot item by specifying: {ColorBy: Set}, and {Sets: msBox}. The grains
with centers inside the box can be displayed with either: (a) the ball and clump plot items by
specifying: {Color By: Text Val: msBoxIN}, or (b) the grain and faced grain plot sets. The contact-
based plot sets display only the entities joining the grains in the box. The faced grain plot set
displays only the faces joining grains with centers inside the box. The following microstructural plot
sets are available (see Figure 15).

Grains. A contact- or parallel-bonded material mimics the microstructure of rigid ball- or

clump-shaped grains cemented with either spot welds or epoxy. The grain plot set (geometry
set name: “grains”) displays each grain as a sphere — only ball-shaped grains are displayed,
clump-shaped grains are not supported. The sphere is centered at the ball location, and has
radius equal to the scale factor (msGN_sfacRad) times the ball radius. For the 2D model,
the grains are displayed as circles.

Contact bonds. A contact-bonded material mimics the microstructure of rigid grains

cemented with spot welds. The contact bond plot set (geometry set name: “contact bonds”)
displays each intact grain-grain contact bond as a line. The line is centered at the contact
location, and has length equal to the scale factor (msCB_sfacLen) times the distance
between the centers of the two contacting pieces.

Parallel bonds. A parallel-bonded material mimics the microstructure of rigid grains

cemented with epoxy. There are two plots sets. The first plot set depicts the parallel bond as
an interface, while the second plot set depicts the parallel bond as cement.

The parallel-bond interfaces plot set (geometry set name: “pbond interfaces”) displays each
intact grain-grain parallel bond as a disk with radius equal to the scale factor (  ,
msPBi_sfacRad) times the parallel-bond radius  R  . The disk is centered at the contact
Material-Modeling Support in PFC [fistPkg25] Page 32

location and lies on the contact plane. For the 2D model, the parallel-bond interfaces are
displayed as lines (of length 2 R , and unit-thickness depth).

The parallel-bond cement plot set (geometry set name: “pbond cement”) displays each
intact grain-grain parallel bond as a cylinder with radius equal to the scale factor (  ,
msPBc_sfacRad) times the parallel-bond radius  R  . The cylinder is centered at the
contact location, and has a length  L  equal to the scale factor (msPBc_sfacLen) times
the distance between the centers of the two contacting pieces. For the 2D model, the parallel-
bond cement is displayed as a thick line (of length L , thickness 2 R , and unit-thickness
depth).

Flat Jointed Material. A flat-jointed material mimics the microstructure of rigid faced
grains with bonded and frictional interfaces. There are two plots sets. The first plot set
depicts the interfaces, while the second plot set depicts the faced grains.

The flat-joint interfaces plot set (geometry set name: “FJ interfaces”) displays the interface
of each grain-grain flat-jointed contact. The interface, coordinate systems and discretization
can be displayed by the geometry plot item by specifying {ColorBy: Group} and {Sets: FJ
interfaces}, with {Colors:} used to turn each entity on/off and specify its color. The
visualization scheme for each entity is as follows.
a. Interface (group ‘interface’). The interface is drawn as a disk with radius equal to
the scale factor (  , msFJi_sfacRad) times the flat-joint radius  R  . The disk is
centered at the contact location and lies on the contact plane. For the 2D model, the
flat-joint interfaces are displayed as lines (of length 2 R , and unit-thickness depth).
b. Coordinate systems (groups ‘s-tSys’ and ‘xi-etaSys’). The two coordinate systems
associated with the interface ( nst ) and ( n ) are drawn as lines from interface
center extending just beyond the interface edge, with the lines in the s and 
directions being shorter than the lines in the t and  directions. For the 2D model,
the group is ‘n-tSys’, and only the n and t directions of the nst coordinate system
are drawn.
c. Discretization (group ‘elements’). The interface discretization shows elements and
centroid locations. The outline of each element is drawn as a set of edges, and there
is a short line (aligned with the s direction) at each element centroid. For the 2D
model, each element is outlined by edges oriented perpendicular to the interface, with
a shorter perpendicular line at each element centroid.
Material-Modeling Support in PFC [fistPkg25] Page 33

The faced grain plot set (geometry set name: “faced grains”) displays each grain as a faced
grain — only ball-shaped grains are displayed, clump-shaped grains are not supported. The
faced grains are created when the flat-joint model is installed at the ball-ball contacts of a
packed ball assembly such that each face remains rigidly connected to its corresponding
grain during subsequent motion. The faced grain is drawn as a spherical core with a skirted
face at the associated notional surface of each flat-jointed contact. The core and skirted faces
of each grain can be displayed by the geometry plot item by specifying {ColorBy: Group}
and {Sets: faced grains}, with {Colors:} used to turn each entity on/off and specify its color.
For the 2D model, two geometry plot items should be used. The first geometry plot item
turns on the entities ‘core’ and ‘grain in skirt’, with {Edges: Off}, {Polys: checked} and
{Polys: Wireframe: unchecked}. The second geometry plot item turns on the entity ‘skirted
face’, with {Edges: On} and {Polys: unchecked}. The visualization scheme for each entity is
as follows.
a. Core (group ‘core’). The core is drawn as a sphere with radius equal to the minimum
of all facet radial distances. For the 2D model, the core is drawn as a circle.
b. Skirted Face (group ‘skirted face’, also ‘grain in skirt’ for the 2D model). The
skirted face is drawn as a disk (line for the 2D model) surrounded by a skirt that
extends toward the core center until it intersects the core surface.
Material-Modeling Support in PFC [fistPkg25] Page 34

Figure 15 Microstructural plot sets for bonded materials with the same initial
packing showing (clockwise from upper left): microstructural box and
grains in the box (grey); contact-bonded material with contact bonds
in the box; parallel-bonded material with parallel-bond cement (gold,
50% size) and parallel-bond interfaces (gold, 50% size); and flat-
jointed material with flat-jointed interfaces (blue, 50% size).
3.3 Force-Chain Fabric
The PFC material carries load via force chains that propagate from one grain to the next across
grain-grain contacts. The force-chain fabric is depicted as a network of cylinders, with a cylinder at
each contact. Force magnitude corresponds with cylinder thickness, and force direction corresponds
with cylinder orientation. The force-chain fabric can be displayed as either bi-colored to denote
compression and tension or scale-colored to denote magnitude (see Figure 16). Both displays are
provided by the contact plot item. The bi-colored display is obtained by specifying: {Shape:
Cylinder}, {Color By: Text Val: force chain}, {Color Opt: Named}, {Scale by Force: checked}, and
{Directed: checked}. The scale-colored display is provided by specifying: {Shape: Cylinder},
Material-Modeling Support in PFC [fistPkg25] Page 35

{Color By: Vector Qty: force}, {Color Opt: Scaled}, {Scale by Force: checked}, and {Directed:
checked}.

Figure 16 Force-chain fabric in the linear material packed at 150 kPa material
pressure showing: bi-colored (left) and scale-colored (right) displays.
The linear material cannot sustain tension, therefore the bi-colored
display shows only compression.
3.4 Crack Monitoring
Damage in the bonded materials consists of bond breakages, which we denote as cracks. Each crack
has: type (parallel bonded, contact bonded, flat jointed or smooth jointed); failure mode (tensile or
shear); geometric information (size, position, normal direction and gap); step number at which it
formed; and orphan indicator. The numbers of all cracks (partitioned by type and failure mode) are
returned by the ck_nX functions. Crack data is stored as a Discrete Fracture Network (DFN). The
DFN plot item supports visualization of the cracks.29 Crack filtering occurs by calling the FISH
function ckFilter( g ), which selects cracks with a gap less than g . The ckFilter function
can be redefined to provide a user-defined filtering criterion. The DFN plot item displays cracks by
specifying: {Shape: Fracture}, {Color By: Text Val: group (slot 1 or 2, for all or filtered cracks,
respectively)}, and {Color Opt: Colors: select on/off and color for each type and failure mode}. The
size of the displayed cracks is scaled by the shrink factor in the Shape attribute of the DFN plot item:
{Shape: Fracture… Shrink Fac}.

29
Gap-based plots are not yet supported. Gap-based plots add a thickness to each crack icon that is proportional to the
current gap.
Material-Modeling Support in PFC [fistPkg25] Page 36

Geometric Information:
A crack is a disk for the 3D model and a linear segment of unit-thickness depth for the 2D
model. A crack has a size (diameter for the 3D model and length for the 2D model), position,
normal direction and gap.
 The size is set at creation and then frozen. The size for each type of crack is:
o contact-bonded crack: two times the radius of the deformability method.
o parallel-bonded crack: two times the parallel-bond radius.
o flat-jointed crack: two times the element radius (which is the radius giving the same
area as the element), or the element length (for the 2D model).
o smooth-jointed crack: two times the smooth-joint radius.
 The position, normal direction and gap are updated to correspond with material motion
subsequent to bond breakage (see ckForceUpdate). If the originating contact exists, then
these quantities are obtained from the contact and its associated contact model as described
below. If the originating contact no longer exists, then these quantities are set equal to the
contact-plane location, contact-plane normal direction and contact gap, respectively, that
would be assigned to a contact between the two parent pieces. 30 However, if at least one of
the parent pieces no longer exists, then these quantities are frozen at their last updated values,
and such cracks are called orphans. The crack position is given by

 x   , flat-jointed crack
e
x (4)
 x c , otherwise

where x  is the centroid location of element  e  , and xc is the contact-plane location. The
e

crack normal direction is given by

 nˆ j , smooth-jointed crack
nˆ   (5)
 nˆ c , otherwise

where nˆ j is the joint normal direction, and nˆ c is the contact-plane normal direction. The
crack gap is given by

 g s  g c  g r , contact- or parallel-bonded crack


g g  e  , flat-jointed crack (6)
 g j , smooth-jointed crack


The three geometric quantities are given by Eq. (2) in Itasca (2016, PFC Model Components: Contacts and Contact
30

Models: Contact Resolution).

Material-Modeling Support in PFC [fistPkg25] Page 37

where g s is the surface gap, gc is the contact gap, g r is the reference gap, g   is the gap at
e

the centroid of element  e  , and g j is the joint gap.

Global Functions:
 ckInit, ckOn, ckOff, ckForceUpdate, ckListData, ck_nOrphans,
ck_nFiltered. The crack-monitoring package is initialized and turned on by calling
ckInit. If the package is reinitialized, then crack data is reset. The package is activated by
ckOn, and deactivated by ckOff. Cracks are only monitored when the package is on. The
crack data can be listed to the screen by ckListData.

Global Variables:
 ck_nAll, ck_n{CB,PB,FJ,SJ}{t,s} : crack counts (used with history logic to
monitor cracking during a simulation)
 ckUpdateRate : crack geometry update rate (number of cycles)
 ckFilterGap : display cracks with gap less than this value (used by ckFilter)

Crack Data (stored for each crack, which is a DFN fracture):

 The size, position and normal direction are stored as DFN data.
 group (slot 1) : name indicating type & failure mode ({CB,PB,FJ,SJ}-
{ten,shear}Fail) of all cracks
group (slot 2) : name indicating type, failure mode and gap-based filtering criterion
({CB,PB,FJ,SJ}-{ten,shear}Fail(gap < [ckFilterGap])) of filtered
cracks
 extra 1 : originating contact
extra 2 : originating element (for flat-jointed cracks only)
extra 3 : parent piece-1 (ball or pebble)
extra 4 : parent piece-2 (ball, pebble or facet)
extra 5 : gap
 property ‘failStep’ : step number at which the crack formed
property ‘orphan’ : orphan indicator (1: orphaned crack, 0: not an orphaned crack)
Material-Modeling Support in PFC [fistPkg25] Page 38

4.0 MATERIAL GENESIS

Material behavior is embodied in the contact model at the grain-grain contacts. The material-genesis
procedure creates the following PFC materials: linear, contact-bonded, parallel-bonded, flat-jointed
and user-defined.31 There is a MatGen project for each material type, and these projects are in
directories: MatGen-X, where X = {Linear, ContactBonded, ParallelBonded,
FlatJointed, Hill}. The materials consist of a homogeneous, isotropic and well-connected
grain assembly with a specified non-zero material pressure. Material genesis occurs within a material
vessel that may be either a physical vessel or a periodic space. Specimens formed in a physical
vessel may be removed from the vessel for subsequent boundary-value modeling, while specimens
formed in periodic space are converted into periodic bricks for subsequent boundary-value modeling
(via the brick.assembly FISH function). The boundary-value modeling may involve trimming
the material into a desired shape and applying boundary conditions. The material vessels are
described in the first subsection. The material-genesis procedure is described in the second
subsection.

4.1 Material Vessels

Material genesis occurs within a material vessel that may be either a physical vessel or a periodic
space. The physical vessels include polyaxial, cylindrical and spherical cells (see Figure 17). A
polyaxial cell consists of planar walls that form a rectangular cuboid (for which each pair of adjacent
faces meets in a right angle). A cylindrical cell consists of two planar walls and a cylinder wall. A
spherical cell consists of a sphere wall. Walls of the physical vessels are expanded to prevent grains
from escaping if the walls are moved outwards during subsequent compression testing — the wall-
wall overlap is ignored, because walls interact only with grains. For materials formed in periodic
space, the periodic cell is a rectangular cuboid. Three spherical measurement regions are placed
symmetrically along the axis of the largest vessel dimension (with a spanning length of l times the
largest vessel dimension, and with a diameter of d times the smallest vessel dimension). For
materials formed in a physical vessel, the linear contact model is installed at the grain-wall contacts.
The walls are frictionless, and grain-wall contact stiffness is set based on a specified contact
deformability ( E * of the deformability method of the linear model). The material vessel parameters
are listed in Table 6.

The material-genesis procedure in the PFC 5.0 FISHTank creates a homogeneous material and does not include the
31

material and refinement regions of the PFC 4.0 FISHTank.

Material-Modeling Support in PFC [fistPkg25] Page 39

Table 6 Material-Vessel Parameters

Parameter, FISH Type Range Default Description

Material-vessel properties (including current vessel dimensions) are listed

via @mvListProps.
vessel-type code
Tv , mv_type INT 0,1 0  0, physical

 1, periodic
vessel-shape code
 0, rectangular cuboid

Sv , mv_shape INT 0,1, 2 0  1, cylinder
 2, sphere

(2D model: S v  0 )
height, width and depth
(sphere diameter is H ;
H,W, D , mv_{H,W,D} FLT  0.0, NA
2D model: D  1 ,
see Figure 2)
expansion factor
 , mv_expandFac FLT 1.0,  1 .2
of physical vessel
l ,d  , inset factors
FLT  0.0,1.0  0 . 8 , 0 . 8 
mv_inset{L,D}Fac of measurement regions
effective modulus
Ev* , mv_emod FLT  0.0, NA
of physical vessel
Material-Modeling Support in PFC [fistPkg25] Page 40

Figure 17 Material vessels with global coordinate system centered in each cell
and associated axial direction.
Material-Modeling Support in PFC [fistPkg25] Page 41

4.2 Material-Genesis Procedure

The material-genesis procedure occurs within a physical material vessel and produces a specimen
consisting of a homogeneous, isotropic and well-connected grain assembly with a specified non-zero
material pressure. The material-genesis procedure consists of a packing phase followed by a
finalization phase. During the packing phase, the grain assembly is subjected to a specified packing
procedure. During the finalization phase, the final material properties are assigned to the grain-grain
contacts, and additional material properties are specified that will be assigned to new contacts that
may form during subsequent motion. The two phases are described in the following two subsections.

4.2.1 Packing Phase

During the packing phase, the grain assembly is subjected to a specified packing procedure during
which the effective behavior at all grain-grain contacts is that of either the linear contact model or a
user-defined contact model. For the linear model, the properties are deformability (lnm_emod and
lnm_krat); for the user-defined model, the properties are specified in the redefined
ft_setMatBehavior FISH function. The friction coefficient is a packing parameter (either zero
or  CA in Table 7).

The packing procedures include boundary contraction and grain scaling. The boundary-contraction
procedure produces a dense or loose packing of a granular material, while the grain-scaling
procedure produces a dense packing of a granular material that will subsequently become a bonded
material during the material-finalization phase. Both procedures generate a cloud of grains, and then
allow them to rearrange into a packed state under conditions of zero friction. The desired material
pressure is obtained differently for each procedure. For the boundary-contraction procedure,
confinement is applied by moving the vessel walls under control of the servomechanism described in
Section 5.3, whereas for the grain-scaling procedure, the grain sizes are scaled iteratively to modify
the mean stress of the assembly. For the boundary-contraction procedure, the material friction
coefficient is chosen during confinement application to achieve a dense or loose packing, whereas
for the grain-scaling procedure, the material friction coefficient remains equal to zero throughout the
entire packing process to achieve a dense packing. The packing parameters are listed in Table 7. A
series of material instances can be created with the same statistical properties but different packing
arrangements by varying the seed of the random-number generator.

Table 7 Packing Parameters

Parameter Type Range Default Description

seed of random-number
S RN , pk_seed INT SRN 10,000 10, 000 generator
(affects packing)
Pm , pk_Pm FLT  0.0, NA material pressure
Material-Modeling Support in PFC [fistPkg25] Page 42

pressure tolerance
 P  Pm 
 P , pk_PTol FLT  0.0, 1  10  2   P 
 Pm 
where P is current pressure
 lim , equilibrium-ratio limit
pk_ARatLimit
FLT  0.0, 8  10  3
(parameter of ft_eq)
step limit
n lim , pk_stepLimit INT 1, 2  10 6
(parameter of ft_eq)
packing-procedure code
C p , pk_procCode INT 0,1 0  0, boundary contraction

 1, grain scaling
 0.58, 3D
 , Cp  0
 0.25, 2D
nc , pk_nc FLT 0.0,1.0  grain-cloud porosity
 0.35, 3D
 , Cp  1
 0.08, 2D
Boundary-contraction group ( C p  0 ):
material friction coefficient
 C A , pk_fricCA FLT 0.0, 0 .0
during confinement application
v l i m , pk_vLimit FLT  0.0, NA servo velocity limit (see Table 9)

Static-equilibrium is determined by successful return from the ft_eq FISH function:

 ft_eq(  lim , n lim ) cycles the model until a state of static equilibrium is obtained.
Cycling continues until either the mechanical aratio (  arat ) falls below the equilibrium-ratio
limit lim  :
arat  lim command: solve aratio lim 
or the number of steps ( ns ) exceeds the step limit  nlim  :
ns  n lim  command: solve max_cycles n lim  .

The aratio (average ratio) and mratio (maximum ratio) are defined as
Material-Modeling Support in PFC [fistPkg25] Page 43

 
F F  Fb ng
avg  Fub
c a
ng
 arat 
avg  F    F  F
int c a  Fb  ng
ng

max  Fc  Fa  Fb
max  Fub  ng
 mrat 
avg  Fint   Fng
c  Fa  Fb  ng

The aratio is the average (over all grains) unbalanced force magnitude divided by the average
(over all grains) force intensity, while the mratio is the maximum (over all grains)
unbalanced force magnitude divided by the average (over all grains) force intensity. The
unbalanced force is the vector sum of all forces acting on the grain (contact, applied and
body), and the force intensity is the sum of the force magnitudes. In the above expressions,
2 2 2
vector magnitude is defined by v  vx  vy  vz , and only free translational degrees-
of-freedom are used in the computations.

The boundary-contraction procedure is adapted from McDowell et al. (2006) and consists of the
following three steps (see Figure 18).

1. Generate a cloud of grains with porosity of nc . This porosity does not account for overlaps:
nc  Vv  Vg  Vg , where Vv is the volume of the material vessel and Vg is the total volume
of grains. The grains are drawn from the specified size distribution, and then placed at
arbitrarily chosen positions that lie fully within the material vessel such that there may be
large grain-grain overlaps. Typically, nc is chosen equal to nl , where nl corresponds with
the loose state for which grains are just in contact at zero mean stress. For a linear material of
equal-sized spheres, nl  0.58 . For a linear material of equal-sized disks, nl  0.25 .
2. Set the material friction coefficient to zero, and then allow the grains to rearrange until either
the mean stress is near zero (within 0.1% of P m ) or static-equilibrium is obtained.32 This
step eliminates the large overlaps and should provide an isotropic state.
3. Set the material friction coefficient to  CA , and then apply confinement of P m . The
confinement is applied by moving the vessel walls under control of the servomechanism

32
During the first set of cycles, the model is calmed (by setting all grain translational and rotational velocities to zero)
every five steps, and the calming continues until the maximum overlap ratio (defined at the active contacts as the contact
gap divided by the piece diameter) is small (less than _pkORmaxLimit, which defaults to 0.25). The calming process
prevents most grains from passing through the walls of the material vessel; those grains with centers lying outside of the
material vessel at the end of step 2 are deleted, and for clumped materials, those pebbles with centers lying outside of the
material vessel are also deleted.
Material-Modeling Support in PFC [fistPkg25] Page 44

until the wall pressures are within the specified pressure tolerance of the material pressure
and static-equilibrium has been obtained. Setting CA  0 gives the densest packing, and
progressively looser packings are obtained by increasing  CA .

Figure 18 Boundary-contraction packing procedure: (a) initial grain cloud at

end of step 1, (b) relaxed grain cloud at end of step 2, and (c)
compacted granular assembly at end of step 3. The images show the
AG_Linear material of the linear material example.
The grain-scaling procedure is similar to what is referred to as the material-genesis procedure in
Potyondy and Cundall (2004) and Itasca (2008a&b, Section 3.6.2 Material-Genesis Procedure, in
FISH in PFC3D/PFC2D volumes) and consists of the following three steps (see Figure 19).

1. Generate a cloud of grains with porosity of nc . This porosity does not account for overlaps:
nc  Vv  Vg  Vg , where Vv is the volume of the material vessel and Vg is the total volume
of grains. The grains are drawn from the specified size distribution, and then placed at
arbitrarily chosen positions that lie fully within the material vessel such that there may be
large grain-grain overlaps. Typically, nc is chosen equal to nd , where nd corresponds with
the dense state for which grains are well packed and in good contact with one another at a
large mean stress but with relatively small overlaps. For a linear material of equal-sized
spheres, nd  0.35 . For a linear material of equal-sized disks, nd  0.08 .
2. Set the material friction coefficient to zero, and then allow the grains to rearrange until either
the mean stress is near zero (within 0.1% of P m ) or static-equilibrium is obtained (see
Material-Modeling Support in PFC [fistPkg25] Page 45

footnote 32). This step eliminates the large overlaps and should provide an isotropic and
close-packed state.
3. Scale the grain size iteratively to modify the mean stress of the assembly until the mean
stress is within the specified pressure tolerance of the material pressure and static-
equilibrium has been obtained. The grain-scaling procedure is described in Itasca (2008a&b,
Section 3.10.3 Isotropic Stress Installation Procedure, in FISH in PFC3D/PFC2D volumes).
For a bonded material, the material pressure is typically set to a low value relative to the
material strength.

Figure 19 Grain-scaling packing procedure: (a) initial grain cloud at end of step
1, (b) relaxed grain cloud at end of step 2, and (c) compacted granular
assembly at end of step 3. The images show the SS_ContactBonded
material of the contact-bonded material example.
4.2.2 Finalization Phase
During the finalization phase, the final material properties are assigned to the grain-grain contacts,
and additional material properties are specified that will be assigned to new contacts that may form
during subsequent motion. The first step of the finalization phase occurs only for the bonded
materials. We ensure the existence of contacts between all grains with a gap less than or equal to the
installation gap by specifying the installation gap as the proximity in the Contact Model Assignment
Table and executing the CLEAN command; this operation may create new grain-grain contacts in
the grain assembly. The grain connectivity is controlled by the material pressure and the installation
gap, with the installation gap being of primary importance (see Figure 20). The next step of the
material-finalization phase differs for each material as follows.
Material-Modeling Support in PFC [fistPkg25] Page 46

Figure 20 Grain assembly at end of packing phase with the bonded or flat-
jointed contacts of a typical grain when the installation gap is zero
(left) and greater than zero (right). Increasing the installation gap
increases the grain connectivity.
 For a linear material, the linear material properties (from Table 2) are assigned to all grain-
grain contacts. New grain-grain contacts that may form during subsequent motion are
assigned the linear model with properties derived from the linear material properties.
 For a contact-bonded material, the contact-bonded material properties (from Table 3) are
assigned to all grain-grain contacts, but only those grain-grain contacts with a gap less than
or equal to the installation gap are bonded. New grain-grain contacts that may form during
subsequent motion are assigned the linear contact bond model with properties derived from
the linear material group of the contact-bonded material properties, and the new contacts are
unbonded.
 For a parallel-bonded material, the parallel-bonded material properties (from Table 4) are
assigned to all grain-grain contacts, but only those grain-grain contacts with a gap less than
or equal to the installation gap are bonded. New grain-grain contacts that may form during
subsequent motion are assigned the linear parallel bond model with properties derived from
the linear material group of the parallel-bonded material properties, and the new contacts are
unbonded.
 For a flat-jointed material, the flat-joint contact model is installed at all grain-grain contacts
with a gap less than or equal to the installation gap, and the flat-jointed material properties
(from Table 5) are assigned to these flat-jointed contacts. The remaining grain-grain contacts
as well as new grain-grain contacts that may form during subsequent motion are assigned the
linear model with properties derived from the linear material group of the flat-jointed
material properties.
Material-Modeling Support in PFC [fistPkg25] Page 47

 For a user-defined material, the user-defined material properties are assigned to all grain-
grain contacts. New grain-grain contacts that may form during subsequent motion are
assigned the user-defined model with properties derived from the user-defined material
properties.
For the bonded materials, the final material properties are set to establish reference surfaces that do
not overlap (see Figure 21); therefore, there are no forces or moments in the bonded material. For the
bonded materials, the grain-vessel interface is smoothed (see Figure 22). The smoothing operation
consists of setting the reference gap equal to the contact gap for all grain-wall contacts with a
negative contact gap (thereby establishing reference surfaces that are just touching and effectively
removing the overlap). After the smoothing operation and the assignment of the final material
properties, there are no forces at the grain-wall interface and no forces or moments in the bonded
material. Subsequent compression testing will proceed as expected, because the specimen will
become loaded during the seating phase as the vessel walls move inward. New grain-wall contacts
will be created naturally, because only the grain-wall contacts protruding from the vessel at the time
of smoothing were affected by the smoothing operation.

The system is brought to static equilibrium while keeping the walls fixed. The specimen remains
within the material vessel, and the model state is saved. This model state can be restored for
subsequent materials testing. If a bonded material is being created, then the specimen is removed
from the material vessel and brought to static equilibrium, and the model state is saved. This model
state can be restored for subsequent boundary-value simulation.

Figure 21 Setting final material properties of bonded materials to establish

reference surfaces that do not overlap (contact-bonded, parallel-
bonded and flat-jointed materials denoted by cbond, pbond and fjoint,
respectively).
Material-Modeling Support in PFC [fistPkg25] Page 48

Figure 22 The effective grain-vessel interface before (left) and after (right) the
smoothing operation, which effectively cuts off the part of each ball or
pebble that protrudes from the material vessel.
Material-Modeling Support in PFC [fistPkg25] Page 49

5.0 LABORATORY-TESTING PROCEDURES

The following standard rock-mechanics laboratory tests are supported: compression (confined,
unconfined and uniaxial strain), diametral-compression, direct tension and fracture toughness (see
Figure 23). The tests are performed upon materials that were created within the physical material
vessel; however, the diametral-compression test can be performed upon any specimen that is
centered at the origin. The loading axis corresponds with the axial direction shown in Figure 17. The
specimen is loaded by walls in the compression and diametral-compression tests, and by surface
grains in the direct-tension and fracture-toughness tests (after having been removed from the
material vessel and allowed to relax). In the compression tests, the specimen is loaded in either a
polyaxial or cylindrical cell in which the axial walls act as loading platens. For the confined test, the
velocities of the radial walls are controlled by a servomechanism to maintain a constant confining
pressure ( Pc ); for the unconfined test, the radial walls do not contact the specimen; and for the
uniaxial-strain test, the radial walls are kept motionless. In the diametral-compression test, the
specimen (typically of a cylindrical or spherical shape) is loaded by the axial walls. In the direct-
tension and fracture-toughness tests, the velocities of selected surface grains are fixed in the axial
direction, thereby pulling apart the opposing specimen surfaces.

Figure 23 Loading conditions of laboratory-testing procedures.

The techniques used to measure stress, strain and porosity of the specimen within the material vessel
are described in the first subsection. The measurement of material deformability is discussed in the
second subsection. The servomechanism that controls wall velocities of the polyaxial, cylindrical
and spherical cells is described in the third subsection. A discussion of how loading rate affects
material response is given in the fourth subsection. The laboratory-test procedures are described in
the remaining subsections.
Material-Modeling Support in PFC [fistPkg25] Page 50

5.1 Stress, Strain and Porosity in the Material

Stress and strain of the specimen within the material vessel is measured in three different ways via
measurement regions, walls and gauge grains. The measurement-based quantities are taken as the
average values from three spherical measurement regions that are placed symmetrically along the
axis of the largest vessel dimension. The measurement logic is described in Itasca (2016)33. The
wall-based quantities provide stress as wall force divided by cross-sectional area of specimen side
(using either the current or initial vessel dimensions), and strain based on the change of distance
between opposing walls. The gauge-based quantities are obtained from gauge grains that mimic
LVDTs touching the specimen surface. Each gauge grain is the ball or clump that lies closest to the
center of the corresponding specimen surface. The gauge-based quantities provide strain based on
the change of distance between opposing gauge grains. Damage formation in a bonded material may
perturb the gauge grains, and thus, wall-based quantities provide a more uniform averaged response
over the entire specimen surface.

Porosity of the specimen within the material vessel is measured in two different ways via the above
measurement regions and walls. The measurement-based porosity  n  accounts for grain-grain

overlap via the procedure described in Itasca (2016)34. The wall-based porosity  nw  does not
account for overlaps: nw  Vv  Vg  Vg , where Vv is the current volume of the material vessel and

Vg is the total volume of grains.

We denote stress and strain by

stress:  xx ,  yy ,  zz ,  xy ,  xz ,  yz 
(7)
strain:  xx ,  yy ,  zz ,  xy ,  xz ,  yz 

where  ii  0 is tension and  ii  0 is extension. For the 2D model, the out-of-plane stress and
strain components are equal to zero so that stress is  xx ,  yy ,  xy  and strain is  xx ,  yy ,  xy  . An
axial direction is associated with each material vessel: the axial direction of the polyaxial and
cylindrical cells for the 3D model corresponds with the z-direction, while the axial direction of the
polyaxial cell for the 2D model corresponds with the y-direction (see Figure 17). The radial direction
is perpendicular to the axial direction. We define the following stress and strain quantities.

33
In documentation set at General Components: Measure.
34
In documentation set at General Components: Measure: Measured Quantities: Porosity.
Material-Modeling Support in PFC [fistPkg25] Page 51

 Axial and radial stress and strain (3D model):

1
 a   zz ,  r 
2
 xx   yy 
(8)
1
 a   zz ,  r    xx   yy 
2

 Axial and radial stress and strain (2D model,  zz   zz  0 ):

 a   yy ,  r   xx
(9)
 a   yy ,  r   xx

 Deviator stress and strain:35

d  a r , d  a  r (10)

 Mean stress and volumetric strain (3D model):

1 1
m 
3
  xx   yy   zz    a  2 r  ,  v   xx   yy   zz   a  2 r
3
(11)

 Mean stress36 and volumetric strain (2D model,  zz  0 ):

1 1
m 
2
  xx   yy    a   r  ,  v   xx   yy   zz   a   r
2
(12)

The material-vessel stress, strain and porosity quantities are listed in Table 8. The FISH variables for
the measurement-based quantities are denoted by mv_mX, while the wall- and gauge-based quantities
are denoted by mv_wX and mv_gX, respectively.

Table 8 Material-Vessel Stress, Strain and Porosity Quantities

Quantity, FISH Description

 xx , yy ,zz ,xy ,xz , yz  , material stress

mv_ms{xx,yy,zz,xy,xz,yz} (2D model: zz  xz  yz  0 )

35
The deviator stress should not be confused with the deviatoric stress (Engelder, 1994). The amount by which the axial
stress component departs from the confining pressure in a triaxial test is commonly called the differential stress in the
rock mechanics and geological literature, and is used by Paterson and Wong (2005, p. 6) in the direct reporting of triaxial
test results.
36
The mean stress for the 2D model is defined as the average of the in-plane stresses. It is not obtained by setting
 zz  0 in the expression for mean stress of the 3D model.
Material-Modeling Support in PFC [fistPkg25] Page 52

 xx , yy ,zz , xy , xz , yz  , material strain

mv_me{xx,yy,zz,xy,xz,yz} (2D model: zz  xz   yz  0 )

a ,r  , mv_ms{a,r} axial & radial stress

 a ,  r  , mv_me{a,r} axial & radial strain

 d , mv_msd deviator stress
 m , mv_msm mean stress
d , mv_med deviator strain
v , mv_mev volumetric strain
n , mv_mn measurement-based porosity
nw , mv_wn wall-based porosity

The three measurement techniques measure the following quantities:

 ijm  ijm measurement-based (6 terms each, symmetric)

 kw  kw wall-based (3 terms each) (13)

g guage-based (3 terms)
k
i , j  x , y , z
k  x , y , z , r

For the 2D model, the above equation satisfies the condition: i, j, k  x, y . The wall- and gauge-
based techniques measure only the direct stress and strain components, which we denote by a single
subscript (e.g.,  xx is denoted by  x ). The cylindrical and spherical cells measure the radial
components (  r and  r ).

The wall-based stress quantities are the average force on opposing walls divided by the
corresponding cross-sectional area:

 kw 
1
2  F k  F 
w 
k
w 
, k   x , y , z  2D model: k   x, y  (14)
Ak

where F 
k
w 
and F  k
w 
are the total force on the walls in the negative and positive k  -
directions, respectively, and A k is the specimen cross-sectional area perpendicular to direction-
 k  . For the cylindrical and spherical cells, the radial stress acting on the corresponding wall is
Material-Modeling Support in PFC [fistPkg25] Page 53

  Frw
 rw   cylindrical and spherical cells  (15)
Ar

where the numerator is the sum of the radial components of the forces on the wall, and Ar is the
surface area of the wall. The surface areas are determined (by mv_wAreaMode, M a ) using either
the current ( M a  0 ) or initial ( M a  1 ) vessel dimensions, and expressed in terms of the distances
w
between opposing walls ( d k ). For the polyaxial cell: Ax  dy dz , Ay  dx dz and Az  dx dy ; for
w w w w w w

  w 2
4 and Ar   drwdzw ; and for the spherical cell: Ar   drw  . For
2
the cylindrical cell: Az   dr

the polyaxial cell of the 2D model: Ax  dy t t  1 and Ay  dx t t  1 .

w w

The wall-based strain quantities are based on the change of distance between opposing walls:

d kw   d kw 
 
w o
, k   x, y , z , r  2D model: k   x, y  (16)
k
 d kw  o

where d kw is the distance between opposing walls in direction-  k  , and  d kw  o is the initial
distance between these walls. The gauge-based strain quantities are based on the change of distance
between opposing gauge grains:

d kg   d kg 
 
g o
, k   x, y , z , r  2D model: k   x, y  (17)
k
d  g
k o

where d kg is the distance between opposing gauge grains in direction-  k  , and  d kg  o is the initial
distance between these grains. The initial distances provide a reference for the strain measurement
and can be reset by the FISH functions mv_{w,g}StrainZero.

5.2 Material Deformability

The deformability of a bonded material can be quantified by the effective isotropic elastic constants,
and the deformability of a granular material can be quantified by the resilient modulus. The
discussion begins by considering the stiffness tensor of PFC material. Then the definition and
measurement of the effective isotropic elastic constants and resilient modulus are presented.

The stiffness tensor relates stresses to strains. The two fundamental assumptions of linear elastic
theory are that (a) stress-strain relations are linear, and (b) deformations are reversible. The first
Material-Modeling Support in PFC [fistPkg25] Page 54

assumption is expressed by Hooke’s law, giving 36 independent elastic constants. The existence of a
unique strain-energy potential (from assumption b) reduces the number of independent elastic
constants to 21. By assuming different levels of material symmetry, the number of independent
elastic constants can be reduced significantly. For example, an isotropic system has the property that
all directions are elastically equivalent and principal, which reduces the number of independent
elastic constants to two. Therefore, a linear elastic isotropic material model has two independent
elastic constants. A hexagonal system has a plane of elastic isotropy at each point, and these planes
are parallel at all points. Such a body is said to be transversely isotropic, and there are five
independent elastic constants. The stiffness tensor of the PFC material is affected by stress and
damage so that, in general, the material response does not correspond with that of a linear elastic
isotropic material model — e.g., the material response becomes transversely isotropic in response to
both applied compression and compression-induced cracking (Potyondy and Hazzard, 2008).

The effective isotropic elastic constants are defined as the Young’s modulus  E  and Poisson’s ratio
  of a linear elastic isotropic material model. It is useful to measure the effective isotropic elastic
constants of the PFC material — e.g., when calibrating the bonded material, E and  are compared
with the elastic constants measured during triaxial tests on real rock specimens. The effective
isotropic elastic constants are measured by performing a compression test on the PFC material and
interpreting the macroscopic force-displacement response. The interpretation differs for the 3D and
2D models. The 3D model is an assembly of rigid three-dimensional grains for which stress and
strain is defined in a straightforward fashion. The effective isotropic elastic constants of the 3D
model are computed as

 a
E
 a
 3D model  (18)

  r
 a

where the stress and strain components are defined in Section 5.1 such that  a is the axial stress,  a
is the axial strain and  r is the radial strain measured during a compression test in which a constant
confining pressure has been maintained during axial-strain application.

The 2D model is an assembly of rigid unit-thickness disks. The unit-thickness disks are rigid so that
they do not expand in the out-of-plane direction and no stress acts in the out-of-plane direction;
therefore, the out-of-plane stress and strain components are equal to zero. Note that this condition is
neither plane strain nor plane stress. A material subjected to plane-strain conditions will be stiffer in
its plane of loading than the same material subjected to plane-stress conditions, but plane-strain or
plane-stress conditions cannot be imposed on the 2D model. The 2D model boundary conditions are
Material-Modeling Support in PFC [fistPkg25] Page 55

neither plane strain nor plane stress, because the corresponding stress-strain constitutive relations are
not employed — i.e., the out-of-plane forces and displacements do not enter into the force-
displacement law at the contact level.

In order to measure effective isotropic elastic constants of the 2D material, we perform a

compression test on the material and interpret the macroscopic force-displacement response
assuming either plane-strain or plane-stress conditions. For the same 2D model, one macroscopic
force-displacement response is measured, but two sets of effective isotropic elastic constants,
corresponding with plane-stress and plane-strain conditions, are computed as follows.

The plane-stress elastic constants are computed as

 a
E 
 a
 2D model, plane stress  (19)

   r
 a

where the stress and strain components are defined in Section 5.1 such that  a is the axial stress,  a
is the axial strain and  r is the radial strain (defined in Section 5.1) measured during a compression
test in which a constant confining pressure has been maintained during axial-strain application. The
above equation is valid for the case of plane stress   z  0 and constant radial stress   r  0
during the stress-strain increments. Then, the plane-strain elastic constants are obtained via


 
1    2D model, plane strain  (20)
E  E  1   2

which is the general relation between plane-stress and plane-strain conditions (Ugural and Fenster,
1987, p. 71).

How are the effective isotropic elastic constants of the 2D model used? When calibrating the 2D
bonded material, E and  (not E  and   ) are compared with the elastic constants measured during
triaxial tests on real rock specimens. To what do the effective elastic constants correspond? If a
region of a 2D isotropic elastic continuum were extracted and replaced with the 2D model material,
and its corresponding set of elastic constants were assigned to the remaining continuum, then the
deformation state of the model material would match that of the continuum — i.e., there would be
full displacement compatibility along the extraction boundary. This procedure is used in the
construction of 2D boundary-value models of excavation damage in which the model material is
inserted into a pre-defined region within a larger continuum model. For a tunnel simulation, the
Material-Modeling Support in PFC [fistPkg25] Page 56

continuum model is run in plane-strain mode and assigned the plane-strain elastic constants E and
.

The resilient modulus  M R  is defined as the ratio of applied deviator stress to recoverable or
“resilient” strain (see Figure 24). It is a granular material characterization parameter that is stress
dependent. The resilient modulus is used to perform layered elastic analysis (LEA). LEA is utilized
extensively for pavement system evaluation and is a means of calculating pavement response under
loading. Each pavement layer is defined by its resilient modulus and Poisson’s ratio, even though
granular bases exhibit nonlinear elastoplastic behavior in laboratory and field applications. LEA is
used because it is a relatively simple analysis procedure and, more importantly, pavement loading is
generally of low enough magnitude that a linear-elastic approximation of pavement material
behavior is deemed suitable (Potyondy et al., 2016). According to Han and Vanapalli (2016), the
resilient modulus is “…the key soil property in the mechanistic pavement design methods to
rationally characterize the resilient behavior of the pavement materials, analyze the fatigue failure of
the surface layer, and dimension the multi-layer system of the pavement structure.”

Figure 24 Response of granular material subjected to triaxial loading, and

measurement of resilient modulus. (From Fig. 1 of Buchanan [2007].)
5.3 Servomechanism
The velocities of the walls of the polyaxial and cylindrical cells are controlled by a servomechanism.
The servo-control procedure is described in Itasca (2008a&b, Section 3.10.5 Controlling Wall
Velocity to Maintain Specified Stress, in FISH in PFC3D/PFC2D volumes). The servomechanism
parameters are listed in Table 9. The polyaxial cell consists of three pairs of opposing walls for the
3D model and two pairs of opposing walls for the 2D model. The cylindrical cell consists of a pair of
opposing axial walls and a cylinder wall. The spherical cell consists of a sphere wall. The boundary
condition for each pair of walls as well as the cylinder and sphere walls can be either velocity or
pressure. A velocity boundary condition sets the velocity of the wall pair to be equal and opposite
and the velocity of the cylinder- and sphere-wall vertices to be radial (with a positive or negative
value indicating opening or closing motion, respectively). A pressure boundary condition activates a
Material-Modeling Support in PFC [fistPkg25] Page 57

servomechanism that controls the velocities to maintain the specified pressure. The pressure acting
on the opposing walls in the  k  -direction is given by

Pkw   kw , P k
w
 0 is compression 
  x , y , z , polyaxial cell  2D model: k   x, y 
 (21)
with k    z , r , cylindrical cell

 r , spherical cell.

The servomechanism is turned on and off by mvs_on and mvs_off, respectively. When the
servomechanism is on, the boundary conditions (set by mvs_setBCs) are enforced during
subsequent cycling. Application of the specified pressures is determined by successful return from
the mvs_eqP FISH function:

 mvs_eqP(  P , n lim ) cycles the model until either the pressures of all servo-controlled
walls  Pkw  are within the equilibrium-pressure tolerance (  P ) of the target values ( Pk ):
t

 Pkw  Pk t 

 Pk t
  P  , k   x, y , z , r

 2D model: k   x, y 
 

or the number of steps ( ns ) exceeds the step limit  nlim  :

ns  n lim  command: solve max_cycles n lim  .

Table 9 Servomechanism Parameters

Parameter Type Range Def. Description

boundary-condition code

 Bc k , INT 0,1 NA
 0, velocity
 , k   x, y, z or  z , r
mvs_BC{x,y,z,r}  1, pressure
(2D model: k  x, y )
boundary-condition value
 vk , velocity   Bc   0 
 Bv k , FLT 0,1 NA

 t
k
, k   x , y , z or  z , r 
mvs_BC{x,y,z,r}Val  Pk , pressure   Bc  k  1

(2D model: k  x, y )

ngain , FLT 1, 25 servo gain update rate

Material-Modeling Support in PFC [fistPkg25] Page 58

mvs_gainUpdateRate
v l i m , mvs_vLimit FLT  0.0, NA velocity limit ( vk  vlim , vlim  0 )

5.4 Loading Rate

The response of a real material and a DEM model are sensitive to loading rate. In most cases, we
choose a loading rate that is slow enough to insure a quasi-static response. The quasi-static response
is the response obtained when the test has been conducted under quasi-static conditions, which
means that the loading has been slow enough that the system has time to adjust to the force
redistribution that accompanies each nonlinear event (slip or bond breakage). This response could be
obtained by performing a strain-controlled test, whereby the loading velocity is set to zero after each
nonlinear event until the system reaches a new state of static equilibrium. In this way, we would
trace out the stress-strain curve up to the peak, and the peak value would correspond with the quasi-
static response. One can ensure that such a response has been obtained by conducting a strain-
controlled test at a series of constant loading velocities, and demonstrating that the response is the
same for all loading velocities below some critical velocity.

When calibrating a DEM model, we might compare the quasi-static response with the quasi-static
response obtained for the real material (from a test that may have been performed at a different strain
rate). This allows us to compare the results of the DEM model and the real material. If dynamic
effects are also affecting the real system response, then these effects would also need to be included
in the DEM model, which involves a more complex calibration process, whereby we match the
dissipative mechanisms occurring in the real material. If we only compare quasi-static responses,
then the dissipative mechanisms need not be replicated in the DEM model (e.g., the DEM model can
be heavily damped by using local damping with a default damping coefficient of 0.7 to approximate
quasi-static conditions).

The loading rate for the compression, diametral-compression, and direct-tension tests is the axial
strain rate  a  . The axial strain rate should be chosen based on the previous considerations. Note
that an axial strain rate that is sufficient to produce quasi-static loading on a specimen of a given
length will most likely not produce quasi-static loading on a specimen of a different length. A
reasonable estimate of the strain rate needed for the different-length specimen is found by equating
the loading velocities for the two cases.

5.5 Compression Test

Compression tests (confined, unconfined and uniaxial strain) are performed in either the polyaxial or
cylindrical cell of the physical material vessel. The polyaxial cell provides a polyaxial loading
condition while the cylindrical cell provides a triaxial loading condition. The loading axis
corresponds with the axial direction shown in Figure 17, and the loading conditions for the three test
Material-Modeling Support in PFC [fistPkg25] Page 59

types are shown in Figure 23. The axial walls act as loading platens. For a confined test, the
velocities of the radial walls are controlled by a servomechanism to maintain a constant confining
pressure. For an unconfined test, the radial walls are moved away from the specimen and kept
motionless.37 For a uniaxial-strain test, the radial walls are kept motionless. The compression-test
parameters are listed in Table 10. Compression testing is provided by the CompTest project, which
is a subdirectory of the MatGen-X directory.

Table 10 Compression-Test Parameters

Parameter Type Range Default Description

Material-vessel parameters are listed in Table 6.

test-type code
 0, confined
Tt , ct_testType INT 0,1, 2 0 
 1, unconfined
 2, uniaxial strain

confining pressure
Pc , ct_Pc FLT  0.0, NA
( Pc  0 is compression)
axial strain rate
a , ct_eRate FLT  0.0, NA ( v  12 a ho , a  0 , see Figure 23
and Section 5.4)
loading-phase code
Cl , ct_loadCode INT 0,1 0  0, single stage

 1, multiple stages
load-termination factor ( Cl  0 ) for
 , ct_loadFac FLT 0.0,1.0 0 .9
termination criterion: d   d
w w
max

Servo-control group:
 P  Pc 
pressure tolerance   P 
P , ct_PTol FLT  0.0, pk_PTol
 Pc 
where P is current pressure
equilibrium-ratio limit
 lim , ct_ARatLimit FLT  0.0, 1  1 0 5
(parameter of ft_eq)
step limit
nlim , ct_stepLimit INT 1, pk_stepLimit
(parameter of ft_eq)

37
The wall-based axial stress is computed using the axial cross-sectional area of the initial vessel. The wall-based radial
stress, radial strain and volumetric strain are zero; the corresponding measurement-based quantities remain valid and are
used by the hidden views pl-mer_wea and pl-mev_wea. The radial walls can be removed from model views by
adding these walls to the filter of the wall plot item.
Material-Modeling Support in PFC [fistPkg25] Page 60

vlim , ct_vLimit FLT  0.0, 10H a servo velocity limit (see Table 9)

The compression test consists of a seating phase followed by a loading phase. During the seating
phase, the strains are reset to zero. For a confined test, confining pressure is applied by activating the
servomechanism with a pressure boundary condition in all directions. For an unconfined test, the
radial walls are first moved away from the specimen, whereas for a uniaxial-strain test, the radial
walls are kept in their initial position. Then for both unconfined and uniaxial-strain tests, an axial
pressure (equal to Pc ) is applied by activating the servomechanism with a pressure boundary
condition in the axial direction and a velocity boundary condition of zero in the radial direction.
Seating is successful if both mvs_eqP() and ft_eq() indicate that static-equilibrium has been
obtained. The model state is saved at the end of the seating phase.

Grain displacements are reset to zero after the seating phase. At the start of the loading phase, strains
are reset to zero. Then axial strain is applied by moving the axial walls at the specified strain rate
while keeping the confining pressure constant for a confined test or keeping the radial walls
motionless for either an unconfined or uniaxial-strain test. The loading phase may consist either of a
single stage that ends when the applied deviator stress falls below a specified fraction of its peak
value or multiple stages during which the axial-strain increments ( a , a  0 is opening) are
specified in the function ctPerformStages. At the end of each stage, the axial wall velocities are
set to zero and the model state is saved.

During the compression test, the crack-monitoring package is on (for bonded materials), and the
specimen behavior is monitored using the history mechanism to sample and store relevant quantities.
The monitored quantities include average stress and strain within the specimen and numbers of
cracks (for bonded materials).

5.6 Diametral-Compression Test

Diametral-compression tests are performed upon specimens that are centered w.r.t. the origin. Axial
walls that act as loading platens are created and centered w.r.t. the origin such that the loading axis
corresponds with the y-direction (see Figure 25). The platens are frictionless, and the grain-wall
contact stiffness is set based on a specified effective modulus. The diametral-compression test
parameters are listed in Table 11. Diametral-compression testing is provided by the
DiamCompTest project, which is a subdirectory of the MatGen-X directory.
Material-Modeling Support in PFC [fistPkg25] Page 61

Figure 25 Loading configuration of diametral-compression test.

Table 11 Diametral-Compression Test Parameters

Parameter Type Range Default Description

platen width and depth

 w, d  , dc_{w,d} FLT  0.0, NA
(2D model: d  1 )
go , dc_g0 FLT  0.0, NA initial platen gap
mv_emod platen effective modulus
E *p , dc_emod FLT  0.0, or NA (used by linear contact model)
axial strain rate
a , dc_eRate FLT  0.0, NA ( v  12 a g o , a  0 , see Figure 25
and Section 5.4)
loading-phase code
Cl , dc_loadCode INT 0,1 0  0, single stage

 1, multiple stages
load-termination factor ( Cl  0 ) for
 , dc_loadFac FLT 0.0,1.0 0 .9
termination criterion: Fa   Fa max

Static-equilibrium group:
equilibrium-ratio limit
 lim , dc_ARatLimit FLT  0.0, 1  1 0 5
(parameter of ft_eq)
pk_stepLimit step limit
 , dc_stepLimit INT 1.0, or 2  10
6
(parameter of ft_eq)
Material-Modeling Support in PFC [fistPkg25] Page 62

The diametral-compression test consists of a set-up phase followed by a loading phase. During the
set-up phase, the walls are created and the system is brought to static equilibrium (via the ft_eq
FISH function). The model state is saved at the end of the set-up phase.

Grain displacements are reset to zero after the set-up phase. During the loading phase, axial strain is
applied by moving the axial walls at the specified strain rate ( a ). The loading phase may consist of
either a single stage that ends when the axial-force magnitude falls below a specified fraction of its
peak value, or multiple stages during which the axial-strain increments ( a , a  0 is opening) are
specified in the function dcPerformStages. At the end of each stage, the wall velocities are set
to zero and the model state is saved.

During the diametral-compression test, the crack-monitoring package is on, and the specimen
behavior is monitored using the history mechanism to sample and store relevant quantities. The
monitored quantities include axial force and axial strain. The axial force is the average force on the
opposing walls:

Fa 
1
2
 Fy    Fy  ,
 
  Fa  0 is tension  (22)

   
 
where Fy and Fy are the total force on the bottom and top walls, respectively. The axial
strain is based on the change of distance between the opposing walls:

g  go
a  ,  a  0 is extension  (23)
go

where g is the wall gap, and g o is the initial wall gap.

When the diametral-compression test is performed upon a cylindrical specimen oriented such that
the loading axis coincides with the cylinder radial direction, then the Brazilian tensile strength is
given by

Fa
B  max
,  2D model: t  1 (24)
 Rt

where Fa max
is the peak axial force, and R and t are the radius and thickness, respectively, of the
Brazilian disk (Goodman, 1980).
Material-Modeling Support in PFC [fistPkg25] Page 63

5.7 Direct-Tension Test

Direct-tension tests are performed upon materials that have been removed from the physical material
vessel and brought to static equilibrium (which is the final step of the material finalization phase for
bonded materials). The loading axis corresponds with the axial direction shown in Figure 17. During
the direct-tension test, the specimen is gripped at its ends and pulled apart slowly while monitoring
the axial stress and strain (using the measurement-based quantities of  am and  am listed in Table 8).
The direct-tension test parameters are listed in Table 12. Direct-tension testing is provided by the
TenTest project, which is a subdirectory of the MatGen-X directory.

The specimen is gripped by identifying a thin layer of surface grains that will be used to load the
specimen. The surface grains are referred to as grip grains, and identified based on the specified grip
thickness ( tg ). The velocities of the grip grains are fixed in the axial direction, thereby pulling apart
the opposing specimen surfaces. The grip grains are not allowed to translate or spin during the test,
which is equivalent to the opposing surfaces being glued to flat and rigid platens. The velocity of the
grip grains is specified via the axial strain rate ( a ).

Table 12 Direct-Tension Test Parameters

Parameter Type Range Default Description

Material-vessel parameters are listed in Table 6.

tg , tt_tg FLT  0.0, 0.1H grip thickness
axial strain rate
a , tt_eRate FLT  0.0, NA ( v  12 a ho , a  0 , see Figure 23
and Section 5.4)
loading-phase code
Cl , tt_loadCode INT 0,1 0  0, single stage

 1, multiple stages
load-termination factor ( Cl  0 ) for
 , tt_loadFac FLT 0.0,1.0 0 .9
termination criterion:  a    a
m m
max

The direct-tension test consists of a set-up phase followed by a loading phase. During the set-up
phase, the grip grains are identified (by assigning the names ttGripTop and ttGripBottom to
the first slot of the group data for these grains), the motion of these grains is specified and the strains
are reset to zero. The model state is saved at the end of the set-up phase.

Grain displacements are reset to zero after the set-up phase. During the loading phase, axial strain is
applied by moving the grip grains at the specified strain rate. The loading phase may consist either of
Material-Modeling Support in PFC [fistPkg25] Page 64

a single stage that ends when the axial stress falls below a specified fraction of its peak value or
multiple stages during which the axial-strain increments ( a , a  0 is opening) are specified in
the function ttPerformStages. At the end of each stage, the grip-grain velocities are set to zero
and the model state is saved.

During the direct-tension test, the crack-monitoring package is on, and the specimen behavior is
monitored using the history mechanism to sample and store relevant quantities. The monitored
quantities include axial stress and axial strain within the specimen and numbers of cracks.

5.8 Fracture-Toughness Test

{DP: To be developed and described here. The implementation for the 3D model will follow that in
Potyondy and Emam (2014a&b, Sections 16.0&12.0 Fracture Toughness (KIC) Measurement).}

6.0 STRESS-INSTALLATION PROCEDURE

{DP: To be developed and described here. The implementation will follow that in Itasca (2008a&b,
Section 3.9 Stress-Installation Procedure, in FISH in PFC3D/PFC2D volumes).}
Material-Modeling Support in PFC [fistPkg25] Page 65

7.0 REFERENCES
Adair, R.K. (2002) The Physics of Baseball, Third Edition, New York: HarperCollins.

Buchanan, S. (2007) “Resilient Modulus: What, Why, and How?” Vulcan Materials Company,
August 31, 2007. [Document 2-Resilient-Modulus-Buchanan.pdf accessed from
www.vulcaninnovations.com/public/pdf on November 6, 2015.]

Cundall, P.A, and O.D.L. Strack. (1979) “A Discrete Numerical Model for Granular Assemblies,”
Géotechnique, 29, 47–65.

Engelder, T. (1994) “Deviatoric Stressitis: A Virus Infecting the Earth Science Community,” EOS
Trans. AGU, 75, 209.

Goodman, R.E. (1980) Introduction to Rock Mechanics, New York: John Wiley & Sons.

Han, Z., and S.K. Vanapalli. (2016) “State-of-the-Art: Prediction of Resilient Modulus of
Unsaturated Subgrade Soils,” Int. J. Geomechanics, 10.1061/(ASCE)GM.1943-5622.0000631,
04015104.

Holt, R.M., J. Kjolaas, I. Larsen, L. Li, A. Gotusso Pillitteri and E.F. Sonstebo. (2005) “Comparison
Between Controlled Laboratory Experiments and Discrete Particle Simulations of the Mechanical
Behavior of Rock,” Int. J. Rock Mech. & Min. Sci., 42, 985–995.

Itasca Consulting Group, Inc. (2016) PFC — Particle Flow Code in 2 and 3 Dimensions, Version
5.0, Documentation Set of version 5.00.30 [March 2, 2017]. Minneapolis: Itasca.

Itasca Consulting Group, Inc. (2008a) PFC3D — Particle Flow Code in 3 Dimensions, Version 4.0,
User’s Manual. Minneapolis: Itasca.

Itasca Consulting Group, Inc. (2008b) PFC2D — Particle Flow Code in 2 Dimensions, Version 4.0,
User’s Manual. Minneapolis: Itasca.

McDowell, G.R., O. Harireche, M. Konietzky, S.F. Brown and N.H. Thom. (2006) “Discrete
Element Modelling of Geogrid-Reinforced Aggregates,” Proceeding of the Institution of Civil
Engineers — Geotechnical Engineering, 159(1), 35–48.

Nicolson, C.R., A.M. Starfield, G.P. Kofinas and J.A. Kruse. (2002) “Ten Heuristics for
Interdisciplinary Modeling Projects,” Ecosystems, 5, 376–384.

Oreskes, N., K. Shrader-Frechette and K. Belitz. (1994) “Verification, Validation, and Confirmation
of Numerical Models in the Earth Sciences,” Science, 263, 641–646.
Material-Modeling Support in PFC [fistPkg25] Page 66

Paterson, M.S., and T.-f. Wong (1995) Experimental Rock Deformation — The Brittle Field,
Second Edition, Berlin: Springer-Verlag.

Potyondy, D. (2016) “Hill Contact Model [version 4],” Itasca Consulting Group, Inc., Technical
Memorandum ICG7795-L (October 12, 2016), Minneapolis, Minnesota.

Potyondy, D. O. (2015) “The Bonded-Particle Model as a Tool for Rock Mechanics Research and
Application: Current Trends and Future Directions,” Geosystem Engineering, 18(1), 1–28.

Potyondy, D.O., and P.A. Cundall. (2004) “A Bonded-Particle Model for Rock,” Int. J. Rock Mech.
& Min. Sci., 41(8), 1329–1364.

Potyondy, D., and S. Emam. (2014a) “PFC3D 4.0 Manual Modifications,” Itasca Consulting Group,
Inc., Technical Memorandum distributed with PFC3D 4.0 executable (February 6, 2014),
Minneapolis, Minnesota.

Potyondy, D., and S. Emam. (2014b) “PFC2D 4.0 Manual Modifications,” Itasca Consulting Group,
Inc., Technical Memorandum distributed with PFC2D 4.0 executable (February 6, 2014),
Minneapolis, Minnesota.

Potyondy, D.O., and J.F. Hazzard. (2008) “Effects of Stress and Induced Cracking on the Static and
Dynamic Moduli of Rock,” in Continuum and Distinct Element Numerical Modeling in Geo-
Engineering 2008 (Proc. 1st Int. FLAC/DEM Symposium, Minneapolis, Minnesota, USA, August
25–27, 2008), pp. 147–156, R. Hart, C. Detournay and P. Cundall, Eds., Minneapolis, MN: Itasca
Consulting Group.

Potyondy, D., J. Siekmeier, and L. Petersen. (2016) “Aggregate-Geogrid Interaction Model

Incorporating Moisture Effects,” in Transportation Research Board 2016 Annual Meeting
Compendium of Papers.

Santamarina, J.C., K.A. Klein and M.A. Fam. (2001) Soils and Waves: Particulate Materials
Behavior, Characterization and Process Monitoring, New York: John Wiley & Sons, Ltd.

Siekmeier, J., J. Bittmann, D. Potyondy, and L. Petersen. (2016) “Introducing a Geogrid Gain Factor
for Flexible Pavement Design,” in Proceedings University of Minnesota 64th Annual Geotechnical
Engineering Conference (Saint Paul, MN, March 4, 2016).

Starfield, A.M. (1997) “A Pragmatic Approach to Modeling for Wildlife Management,” J. Wildlife
Management, 61(2), 261–270.

Starfield, A.M., and P.A. Cundall. (1988) “Towards a Methodology for Rock Mechanics
Modelling,” Int. J. Rock Mech. Min. Sci. & Geomech. Abstr., 25(3), 99–106.
Material-Modeling Support in PFC [fistPkg25] Page 67

Ugural, A.C., and S.K. Fenster. (1987) Advanced Strength and Applied Elasticity, Second SI
Edition, New York: Elsevier Science Publishing Co., Inc.

Winsberg, E. (2010) Science in the Age of Computer Simulation, Chicago: University of Chicago
Press.