Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

The critical state of granular media: Convergence, stationarity, and disorder

The critical state of granular media: Convergence, stationarity, and disorder Géotechnique 00, 1–9 [http://dx.doi.org/10.1680/geot.XX.XXXXX] MATTHEW R. KUHN Discrete element simulations are used to monitor several micro-scale characteristics within a granular material, demonstrating their convergence during loading toward the critical state, their stationarity at the critical state, and the evolution of their disorder toward the critical state. Convergence, stationarity, and disorder are studied in the context of the Shannon entropy and two forms of Kullback-Leibler relative entropy. Probability distributions of twenty aspects of micro-scale configuration, force, and movement are computed for three topological objects: particles, voids, and contacts. The probability distributions of these aspects are determined at numerous stages during quasi-static biaxial compression and unloading. Not only do stress and density converge to the critical state, but convergence and stationarity are manifested in all of the micro-scale aspects. The statistical disorder (entropy) of micro-scale movements and strains generally increases during loading until the critical state is reached. When the loading direction is reversed, order is briefly restored, but continued loading induces greater disorder in movements and strains until the critical state is reached again. KEYWORDS: constitutive relations; fabric/structure of soils; particle-scale behaviour; statistical analysis INTRODUCTION characteristics — persistent convergence toward stationary The critical state is a foundational concept in geomechanics, bulk attributes — resemble those of thermal systems that encompassing two characteristics of soils and other granular reliably approach a stable, equilibrium condition with materials (Schofield & Wroth, 1968). After sustained slow sufficient passing of time, even as the underlying molecular shearing from an initial packing, a granular material reaches motions remain dynamic and diverse. a critical (steady) state in which the ratio of shear stress In a pair of recent papers, the author proposed a third to mean stress remains nearly stationary during further characteristic of the critical state: at the critical state, the micro-scale landscape of particle arrangements and particle shearing. Not only does the stress reach a steady condition, but other bulk measures of internal fabric and structure interactions exhibits the greatest diversity and disorder that remain stationary as well (Thornton, 2000; Peña et al., 2009). is consistent with information known a priori — information Although the values of these characteristics will depend upon that constrains the available landscape (Kuhn, 2014, 2016b). the material’s composition and upon the form of loading, The two papers, based upon a conjecture of maximum for a given material and loading, a specific steady state disorder, demonstrated that observed micro-scale features condition is eventually attained. This critical state condition at the critical state are similar to those that derive from is most enduring at low confining stress, since particle the conjecture. These features include distributions of breakage at high stress can continue to alter the internal the contact forces, orientations, and movements and the fabric, even as the shearing stress remains constant (Luzzani topologic arrangements of particles. & Coop, 2002; Muir Wood & Maeda, 2007). Notably, the bulk Both works deduced this disordered nature by first steady condition occurs even while individual particles are considering the full extent (phase space) of potential undergoing seemingly erratic and highly varied motions and micro-states (particle arrangements, movements, forces, interactions, notwithstanding a slow, quasi-static loading etc.) and then applying available information in the form (Kuhn, 2016a). of reasonable bulk constraints on the space of micro-states. The second characteristic is the convergent quality of Among the many different micro-states that can satisfy the critical state: the eventual bulk attributes for a given the constraints, the author sought the specific collection material are insensitive to the initial particle arrangement of micro-states (called a “macro-state”) that exhibited and to prior stress conditions. For example, materials that are the greatest diversity of micro-states. This approach is initially either loose or dense or have different initial particle equivalent to finding the macro-state that encompasses arrangements will approach the same density after sufficient the greatest number of micro-states that satisfy the given shearing (Casagrande, 1936; Zhao & Guo, 2013). These constraints. Jaynes (1957) contended that this macro-state of maximum entropy satisfies the available information but is otherwise unbiased. As a simple illustration, consider micro-states of particle arrangement, with each micro-state Manuscript received. . . Prepared using GeotechAuth.cls [Version: 2015/01/21 v1.00] 1 arXiv:1812.07662v1 [cond-mat.soft] 18 Dec 2018 2 DISORDER AND THE CRITICAL STATE expressed as an ordered list of integers, between 1 and 10, the micro-state’s list. These frequencies can be extracted representing, say, the number of contacts of each particle for a particular micro-state, but other micro-states within in an assembly. Now impose an average of 7.0 for the the phase space can also share the same frequencies, most integers in a list — the average coordination number. simply as a reshuffling of the order of the N items  within Although it is certainly possible that every particle will the list. The mapping of micro-states and macro-states, have seven contacts, thus satisfying the bulk average, this therefore, is surjective: each micro-state corresponds to a condition is highly unlikely and has a low entropy, since unique macro-state, but many micro-states can comprise this macro-state is only realised by a single list of 7’s (a the same macro-state. In principle, constructing a specific singular micro-state) that corresponds to a nearly crystalized micro-state from a more general macro-state requires condition. Far more lists can be created by permuting equal additional information beyond the gross probabilities, and occurrences of the numbers 6 and 8, and the macro-state the amount of information that must supplement the macro- encompassing such micro-state lists would indicate a higher state, the Shannon “missing information,” is a measure of a entropy and a higher likelihood. In an analogous manner, macro-state’s diversity or disorder. The two terms are used the author applied principles of statistical mechanics (in interchangeably, and entropy will designate a quantified particular, the Jaynes formalism of maximising the Shannon measure of disorder. entropy) to derive characterisations of the critical state. Methods of quantifying the missing information must Starting with the assumption of maximum disorder, the accommodate, in part, characteristics that are discrete and two papers obtained maximally disordered distributions of countable (quantum states for molecules, or coordination various micro-quantities which compared favourably with numbers for a granular material); but a general method distributions measured in discrete element method (DEM) must also accommodate information that is continuous simulations at the critical state. (momenta for classical molecules, or forces, rates, etc. within The current work examines connections among the three a granular material). With discrete data, each item  can, it- characteristics of the critical state: its stationary character, self, be an m-list of discrete values n,  = fn ; n ; : : : ; n g , 1 2 m i i the convergence tendency, and the tendency toward maxi- belonging to the m-space of integers Z . For example, the mum disorder. Unlike the author’s previous works, which be- author showed that the topological arrangement of a two- gan with an assumption of maximum disorder at the critical dimensional assembly of N disks could be represented as a state, the current work directly measures disorder and tracks list (journal) of N integer pairs, each taken from Z (Kuhn, its evolution in DEM simulations. The simulations start from 2014). With continuous data, the items  can be lists of M the initial conditions of densely-packed and loosely-packed data points q,  = fq ; q ; : : : ; q g , taken from the contin- i 1 2 M i assemblies, continue through deviatoric loading in one di- uous M -space R . We will use both types of data — discrete rection, and finish with sustained loading in the reversed and continuous — to describe the evolution of disorder direction. With these simulations, the disorder in particle during the loading of granular media. movements is shown to generally increase during loading The disorder of a macro-state is measured by first until the critical state is reached. Once attained, all micro- expressing the underlying data in terms of probabilities p scale measures of fabric, force, and movement remain in a of the component categories  that appear within micro- steady condition. state lists. The Shannon entropy H is based upon these likelihoods — probabilities p for discrete systems and probability density p() for continuous systems — of the component characteristics  of a macro-state: INFORMATION AND DISORDER The intent is to quantify the diversity exhibited by the micro- H = p ln (p) or states of a granular material. For a thermal system, a micro- state might be presented as a snapshot of the positions and (1) H = p() ln (p()) d momenta of the N identifiable constituents (molecules) at a particular instant. In the current work, snapshots are taken of DEM simulations at particular strains, providing the local where phase space is the set of all possible  components, densities, coordination numbers, forces, movements, etc. A S P = , such that p = 1 and p() d = 1. The intrin- Boltzmann micro-state can be expounded as an ordered N - sic measure H quantifies the diversity of micro-states that list of such characteristics  that characterise a system’s N comprise a macro-state. Oddly, H does not depend upon the constituents,f ;  ; : : : ;  g, and the full set of all possible 1 2 values of the underlying data  but only on their probabilities. lists is the phase space of the system. Appropriate phase Note the lack of a Boltzmann-like constant in entropy H , spaces are later identified for depicting the slow shear-driven conventionally applied to the Shannon entropy as a scaling movements of particles and contacts. The macro-state to which a micro-state belongs is represented by the bulk relative frequencies (empirical probabilities) of items within Prepared using GeotechAuth.cls M. R. KUHN 3 factor to enforce coincidence with the thermodynamic en- of the measured posterior probability p is taken relative to tropy (Ben-Naim, 2008). Two variants of the Shannon mea- the most naive, uninformed prior: q is simply the uniform u u sure, described below, will be used to estimate the relative distribution, q or q (), spread evenly across phase space disorder within particular micro-states collected from DEM . In this manner, one measures the disorder of a given snapshots. system relative to the most possibly disordered system, in Estimating probabilities p is straightforward for a micro- which probabilities are uniformly distributed, state of a system having a finite number of discrete possibilities : by simply counting the occurrences of each H = p ln or condition  and dividing by the length of the data list, 2 (2) N . For the example presented above, entropy H for lists p() H = p() ln d exclusively constructed from 7’s is1:0 ln(1:0) = 0; for lists q () of 6’s and 8’s in equal number, H = 2(0:50) ln(0:5) = 0:693. u u In general, the more uniform and diverse the distribution, the where q and q () are proper distributions, each with larger is its entropy; more biased, uneven, or concentrated a net expectation of 1. The negative Kullback-Leibler distributions (i.e., those having a more certain outcome) distance H is always non-positive. If a system’s density have lower entropies. Indeed, the basis of Jaynes’ maximum p equals the uniform density q , then the relative entropy entropy (MaxEnt) principle is that the most reasonable H has its largest value of zero, corresponding to the most probability distribution is the one that is consistent with disordered system. Continuing the previous example, if all available information, but only this information: the coordination numbers are between 1 and 10, then q = 1=10, remaining,“missing” information is maximised. For example, and the entropy H for lists exclusively constructed from when no information is available for the 10 possible 7’s is 1:0 ln(1:0=0:1) = 2:303; for lists of 6’s and 8’s in constituents of the previous list, not even an average value, equal number,H = 2(0:50) ln(0:5=0:1) = 1:609; and for each constituent should be assumed equally probable, for a uniform distribution,H = 0. Again, larger (less negative) which H = 10 (0:1) ln(0:1) = 2:30. When the average is values express greater diversity and disorder. known to be 7.0, but no other information is available, With the second relative entropy, one exploits information entropy H is maximised with a distribution of exponential gained at the critical state, using knowledge of its distribu- form that is biased toward larger numbers and with the lower tions of particle arrangement, force, and movement as prior entropy of 2.16 (see Ben-Naim, 2008 for other examples). information, against which a granular specimen is compared The entropy measure in Eq. (1) is problematic, however, as it is being loaded toward the critical state. This relative when used to characterise the disorder exhibited in DEM entropy simply measures the (negative) distance between a cs data. The Shannon entropy of a continuous system, as in distribution p and the reference, critical state distribution q , Eq. (1 ), suffers from a lack of scale invariance, such that testing convergence toward (and stationarity of ) the critical a scaling of variables  and the corresponding region state: (perhaps with a change in dimensional units) will alter cs the entropy H (note the logarithm of density p with H = p ln or cs its troublesome dimensional units, see Ben-Naim, 2008). m (3) A second shortcoming is that H describes the missing p() cs H = p() ln d cs information associated with posterior probabilities p but q () does not consider possible access to prior information, even imperfect information, in the form of preferential This second relative entropy is a measure of disorder only probabilities. if one accepts the critical state as the most disordered These problems are avoided with an alternative measure system, a question that is answered by applying the first u cs of disorder — the relative entropy or negative Kullback- relative entropyH to DEM data. The scalarH is similar to Leibler distance (Meléndez & Español, 2014). In this manner, the “state parameter” of Been & Jefferies (1985), which also one can account for possible access to a priori information characterises the distance of a current condition from the in the form of the prior probabilities q or q(), which are critical state: but rather than a difference in bulk void ratios, extracted from past evidence (such as information gathered cs H is the difference between entire probability distributions at the critical state). Two sets of prior probabilities are of the underlying micro-scale data. Although the critical state considered. cs distributions q could be estimated using methods of the With the first relative entropy, it is assumed that no prior author’s previous papers, these distributions will be directly information is available about the particle arrangements, extracted from DEM simulations. forces, or movements, so that one can test whether the disorder of these quantities increases as a granular material is loaded. For this case, the missing information (or disorder) Prepared using GeotechAuth.cls 4 DISORDER AND THE CRITICAL STATE 1.0 DEM SIMULATIONS Dense Numerical simulations were used for tracking convergence 1 0.5 toward the critical state and the evolution of disorder during Loose deviatoric loading. The simulations were conducted on 0.0 simple, two-dimensional assemblies of 676 bi-disperse disks that were contained within periodic boundaries. Previous -0.5 work by the author showed that assemblies with at least a few hundreds of particles reach nearly the same deviatoric -1.0 stress at the critical state Kuhn & Bagi (2009). Using two disk 0 5 10 15 20 25 30 varieties, with ratios of 1.5:1 in size and 1:2.25 in number, Strain, −ε , percent prevented the crystallisation of particle arrangements that (a) are otherwise observed with mono-disperse assemblies. 0.26 To accumulate large data sets, the author created 50 Loose 0.24 randomly generated assemblies (initial micro-states) of dense assemblies and 50 random loose assemblies, and 0.22 then slowly loaded each of the 100 assemblies under 0.20 identical biaxial compression conditions. The creation of Dense these initial assemblies is confirmation that many different 0.18 configurations, both dense and loose, are consistent with a 0.16 given assembly size and preparation method, as in the P - 0 5 10 15 20 25 30 system of Blumenfeld & Edwards (2009). The particles were Strain, −ε , percent isotropically compacted from a sparse two-dimensional (b) granular gas in which the disks were assigned random Dense velocities and an artificial friction coefficient ( = 0:20 for 0.2 the dense assemblies, and  = 0:50 for the loose assemblies), 0.1 Loose with compaction proceeding until the assemblies “seized”, at which the void ratios were 0.175 and 0.246 for the dense 0.0 and loose assemblies. A 0:50 friction coefficient was then -0.1 assigned for the subsequent loading sequence. In the first phase of biaxial loading, the horizontal width -0.2 of each assembly was reduced (compressed) at a constant 0 5 10 15 20 25 30 rate while the height was allowed to expand so that the mean Strain, −ε , percent stress within the assembly remained constant. Loading (c) proceeded slowly to a horizontal strain of 30%, bringing the assemblies fully into the critical state of constant Fig. 1. Stress, strain, void ratio, and contact fabric anisotropy (Satake, 1982) during biaxial loading and unloading, averaged stress, volume, and fabric (Fig.1). The biaxial loading was from 50 numerical simulations of dense and loose assemblies of then reversed by increasing the width (i.e. extension) at a 676 disks. constant rate while reducing the height and maintaining the original mean stress. Reversed loading was ceased when the strain returned to 1%. This reversal phase also Bulk convergence, stationarity, and ergodicity brought the assemblies to the critical state, but with the Figure 1 shows changes in the deviator stress and void ratio principal directions of stress, strain, and fabric rotated by during the compressive loading phase and the reversed 90 . Snapshots were taken of all micro-data at specific strains loading phase, as averaged across 50 specimens. For both to capture the evolving conditions: 22 snapshots during the dense and loose assemblies, the bulk measures of deviator initial phase and 21 snapshots during the reversal phase. By stress, void ratio, average fabric anisotropy, and average running simulations on 50 assemblies, the author collected coordination number reached a steady, critical state at the data lists on about N = 29000 particles, N = 18000 voids, strain 20% during forward loading (arrow 1, Fig. 1), and and N = 47000 contacts for each snapshot of the dense or these bulk quantities also reached the corresponding steady loose assemblies. conditions during reversed loading, after the strain was cs u With each snapshot, entropiesH andH were computed reduced from 30% to 10% (arrows 2 and 3). The dense for various micro-scale characteristics, described in a later and loose assemblies attained nearly identical deviator section, by using the Voronoi method of Learned-Miller stresses, void ratios, and fabric anisotropies at the critical (2004), kernel density estimates of Duong (2007), and direct state, evidence of the state’s convergent character. Note that binning. the dense and loose assemblies behaved almost identically Prepared using GeotechAuth.cls Fabric anisotropy, F − F Void ratio, e Deviator stress, (σ − σ )/p 11 22 11 22 M. R. KUHN 5 1.0 which had strength variances of 0.0035–0.0070. These results indicate that the critical state can be characterised either 0.5 by measuring multiple specimens at a given strain or by measuring a single specimen across a range of strains. In 0.0 the current study, multiple specimens were averaged among multiple strain snapshots at the critical state. -0.5 The primary focus of the remainder of this paper is statistical analyses of micro-scale (rather than macro-scale, -1.0 bulk) data, and as described below, several tens of thousands 0 5 10 15 20 25 30 of micro-scale samples were used for this purpose. It is no Strain, −ε , percent surprise that the critical state exhibits bulk stationarity, as Fig. 2. Stress and strain of 50 simulations of loose assemblies. the state is largely defined by this characteristic. As will be seen, however, stationarity and convergence are more deep- rooted, as they pervade all micro-scale aspects as well as during the reversed loading, with their averaged stress-strain being expressed in bulk, macro-scale characteristics. behaviours being almost indistinguishable. Figure 2 shows stress and strain of all 50 loose specimens, disclosing the erratic character of stress evolution (the dense MICRO-SCALE STATIONARITY, CONVERGENCE, AND specimens exhibited similar fluctuations). The fluctuations DISORDER are also apparent in averaged data of Fig. 1a as slight, raspy The evolving micro-scale quantities are broadly grouped as variations from the average trend. The 50 specimens were follows: sufficient in number, however, such that the mean deviator stress q=p has a 95% confidence interval of0:02. I. local configuration (i.e. packing) quantities, such as co- An analysis of all 50 specimens shows that after the critical ordination number, density, and contact orientation, state is reached, these erratic variations in stress exhibited II. local force quantities, such as particle stresses and both stationarity and ergodicity. Although both terms are contact forces, and usually applied to time series data, in the current context, III. local movement quantities, such as particle transla- a monotonically increasing (or decreasing) strain serves tions and rotations, contact movements, and local the role of time, and the spatial domain is represented by strains. the stresses realised in different specimens at the same strain. Stationarity, meaning constant statistical measures of as presented in the columns of Table 1. The first two aspects stress across a range of strains, was tested by computing the are associated with the status or configuration of a granular mean and variance of stress for the specimens during two assembly, which are related to the configuration distribu- strain intervals: 20–25% and 25–30%. Stationarity was clearly tions of Edwards & Oakeshott (1989) or the configurational demonstrated at the critical state, as the means and variances entropy of Valanis et al. (1993). The third set of attributes of deviator stress (  )=p were nearly same for the two are transitional rate quantities that are driven by the bulk de- 11 22 intervals (means of 0.561 and 0.556, and variances of 0.0050 formation and will depend upon the loading direction. The and 0.0053). distinction between state (being) and transition (becoming) If a process exhibits the ergodicity property, multiple is explored further below. instances of the process taken at a given time yield the The micro-scale of a two-dimensional granular material same statistical characteristics as that of a single instance can be described with any of three topological objects: its tracked across a sufficiently extended time range. In the non-rattler grains, which can be represented as the nodes of current setting, an “instance” is a single simulation and a particle graph; the contacts between particles, represented “time” is strain. Ergodicity means that similar fluctuations as links between nodes; and the voids, represented as would be observed by tracking a single simulation across polygonal loops formed by the nodes and links (Satake, 1992). a range of strains as would be observed by noting the These objects must be distinguished, because configuration, results of many simulations at a single strain. Ergodicity force, and movement are expressed in different ways among was tested by computing the mean and variance of all 50 the objects, and the three objects’ associated micro-state lists specimens at strain 25% and comparing these values to f ;  ; : : : ;  g will contain different quantities and have 1 2 N those of individual specimens taken across a strain range different lengths N . The three objects are presented in the of 20–30%. The statistical measures were almost the same rows of Table 1. for the two cases, with nearly identical spatial and temporal Altogether, the table identifies twenty local micro-scale means, and spatial and temporal variances of 0.0051 and characteristics that were characterised in the form of their 0.0043. The statistical characteristics at strain 25% are nearly probability distributions p at various strain snapshots. the same as those at other strains in the range 20–30%, Three of the quantities are discrete (coordination number, Prepared using GeotechAuth.cls Deviator stress, (σ − σ )/p 11 22 6 DISORDER AND THE CRITICAL STATE Table 1. Matrix of micro-scale characteristics within a granular assembly. Local characteristics of a granular assembly 1 2 I) Configuration II) Force III) Movement 3 8 Particles Coordination no. Pressure Horizontal velocity Deviator stress Vertical velocity Shear stress Rotational velocity 4 9 Voids Valence — Dilation rate Contact journal Compression rate Density Transverse shear rate Rotation rate 7 10 Contacts Orientation Normal force Slip rate Tangential force Rolling rate Rigid rotation rate forces are normalised by dividing by mean stress p. movements are normalised by dividing by strain increment " . number of contacts of a particle. number of links (contacts) around a polygonal void. list of integer pairs for reconstructing an entire particle graph (see Kuhn, 2014, §2). void ratio of a polygonal void, ignoring any rattler particles by treating them as void space. angular orientation of a contact normal vector. particle stresses computed from the contact forces (see Bagi, 1996), expressed as the pressure ( +  ), deviator stress   , and shear stress  . 11 22 11 22 12 local deformation rate within a polygonal void, computed from the particle movements around the void’s perimeter (Kuhn, 1999, §2.1). The rate is a local velocity gradient L , ij expressed in four combinations: dilation L + L , deviatoric compression L L , 11 22 11 22 transverse shear L + L , and rotation L L . 12 21 12 21 relative movements of a particle pair, computed from the particles’ six component rates (two translation rates and a rotation rate for each particle). These six values are expressed in four combinations: three tangential combinations (slip, rolling, and rigid rotation, as described in Kuhn & Bagi, 2004) and the relative normal movement of the two particles. valence, and the contact journal); the other quantities are 5 2 0.0 continuous. -0.2 Convergence and stationarity cs MeasureH in Eq. (3) characterises the difference (negative distance) between the probability distributions of micro- Dense, loading -0.4 Dense, unloading scale quantities during loading and their eventual distribu- Loose, loading tions at the critical state. This measure was applied to each Loose, unloading of the twenty micro-scale characteristics in Table 1. For the -0.6 initial loading phase, distributions were compared with the cs reference distribution q at strain 30%; during the reversed 0 10 20 30 loading, the final critical state distributions at strain 1% were Strain, −ε , percent used. Fig. 3. Evolving distance between distributions of the deviator As an example of the twenty characteristics, Fig. 3 shows stresses within particles and the distribution at the critical state. cs A value of zero represents perfect conformance with the critical the evolution ofH for the particular distribution of the local state distribution. deviator stresses within particles, the difference 11 22 for each particle within the N=29,000 particle samples. The bulk stress was initially isotropic, such that any small deviator stresses (both positive and negative) within individual particles balanced to yield a bulk deviator stress of zero. the distribution of the particle deviator stresses had become This initial distribution of deviator stress was quite different closer to that of the critical state. During further loading than that at the critical state, as is shown in Fig. 4: upon (arrows 1 and 2, Fig. 3), the distribution of stress progressively the start of loading, the distribution becomes broader and approached the distribution of the critical state. At strains more diverse, exhibiting greater disorder. At the strain 0%, greater than 20%, the distributions became stationary and cs the (negative) distanceH was0:23 (Fig. 3). After loading to were nearly identical to the critical state distribution: that is, cs strain 0.4%, the negative distance was0:05, indicating that H attained a steady value of zero. The distributions at both Prepared using GeotechAuth.cls Object Particle deviator stresses, (σ − σ ): 11 22 cs Entropy relative to critical state, H M. R. KUHN 7 0.8 -6 Strain ε 0% 0.6 0.4% -8 22% and 30% 0.4 Dense, loading -10 0.2 Dense, unloading Loose, loading Loose, unloading 0.0 -12 -2 -1 0 1 2 3 Particle deviator stress, σ − σ 11 22 0 10 20 30 Fig. 4. Density distributions of the mean stress within particles Strain, −ε , percent for the dense specimens during loading at various strains. Fig. 5. Disorder in the joint probability distribution of particle velocities (horizontal, vertical, and rotational) as measured with the relative entropyH . 22% and 30% are plotted in Fig. 4, and any small differences can only be discerned by enlarging the plot. When the loading was reversed at strain 30%, the deforming assembly, but one can ask whether this static distribution of deviator stress was very different from its assembly is “at the critical state”? Although the question may eventual critical state distribution (arrow 3 in Fig. 3), seem as semantic nuance, any loading direction other than a since the previous loading had imparted a positive bias continued forward biaxial loading will divert the assembly to the particle deviator stresses, a direction counter to the from its current critical state of deformation. That is, the opposite bias that was progressively induced during further critical state is an active condition in which deformation reversed loading. The reversed loading quickly brought the and movement are essential ingredients. As will be seen, distributions to those of the eventual (reversed) critical state disorder in the micro-scale movements does increase while cs (arrows 4 and 5 in Fig. 3), with anH equal to zero. an assembly is approaching the critical state. On the other As with the local deviator stresses, the other nineteen hand, notable examples of reduced disorder occur among characteristics in Table 1 were also considered, and the joint the configuration and force quantities in Table 1 (columns I probability distributions of multiple characteristics in each and II). For example, the anisotropy that develops in the row of the table were also analysed. The relative entropies of contact orientations during loading (e.g., Thornton, 2000) all micro-scale characteristics exhibited similar trends: their is a form of increasing order, since a uniform, isotropic distributions converged to the corresponding distribution of distribution of orientations exhibits the least order. Another cs the critical state, such thatH equalled zero at a (forward) form of order develops during plastic deformation: when strain of about 20% or after a reversal of strain from 30% contact forces reach their frictional limit, they are constricted to about 10%. The distributions remained stationary during by the friction law, which reduces disorder in the tangential further loading. That is, convergence and stationarity at the components of the contact forces (the friction law can be critical state are manifested in all twenty micro-scale aspects. viewed as a form of constrictive information). An example of the tendency toward increasing disorder in Disorder the micro-scale movements is shown in Fig. 5. This plot is The evolution of micro-scale disorder was investigated by of the relative entropyH of the combined, joint probability computing the relative entropyH of Eq. (2) for the twenty distribution of all three components of particle movements: characteristics in Table 1. The author’s original hypothesis horizontal, vertical, and rotational velocities. EntropyH is was that the disorder of each characteristic would increase taken relative to the most disordered (uniform) distribution during loading toward the critical state. Instead, although q , and a value of zero represents the greatest disorder. In H did increase with all local movement characteristics Fig. 5, the movements of particles within the dense and loose (column III of Table 1), it did not necessarily increase with assemblies are seen to increase from their initially ordered the configuration or force characteristics in columns I condition (large negativeH ) at zero strain (arrow 1, Fig. 5) to and II. Upon further thought, one recognises that the critical a larger value at the critical state (arrow 2). When the loading state is not a fixed condition or status; instead it is a is reversed (arrow 3), the entropy of particle movements is transitional, active quality, which has meaning only in the immediately reduced, corresponding to more ordered (less context of a driving deformation. Consider, for example, diverse) movements. Further reversed loading causes greater the condition of an assembly being deformed in (forward) disorder (arrow 4) until the assembly reaches the reverse biaxial compression at strain 30% (see Fig. 1). One could critical state (arrow 5). The same trends were measured with separately build a static arrangement of disks having the all ten movement quantities in column III of Table 1: the local, same configuration and force characteristics as that of the micro-scale movements and strains of the particles, voids, Prepared using GeotechAuth.cls Probability density, p(κ) Particle velocities, v , v , ω : 1 2 Entropy relative to uniform distribution, H 8 DISORDER AND THE CRITICAL STATE 1.2 Although not rigorous confirmation, the current work and Strain ε 1.0 the author’s previous works give support to the concept 0% of maximum entropy at the critical state. Simulations of 0.8 0.4% 22% the biaxial loading of simple, two-dimensional disc assem- 0.6 30% blies show that micro-scale distributions of arrangement, 0.4 force, and movement converge to stationary distributions. 0.2 Furthermore, disorder in the movements, as characterised 0.0 by the entropy relative to a uniform distribution, tends to -4 -2 0 2 4 increase during loading until the critical state is reached. Void rotation rate, L − L 12 21 That is, starting from initial packings, the particle motions Fig. 6. Density distributions of the void rotations within the are relatively ordered, but as loading approaches the peak dense specimens during loading at various strains. stress and beyond, the distributions of motions become more diverse (i.e., disordered). Upon a reversal of loading — a temporary disturbance from the stationary, stable critical and contacts generally increased as deformation progressed state — the distributions of micro-scale arrangement, force, toward the critical state. and movement abruptly deviate from their stable condition, The increasing disorder is also seen in the probability but after sustained reversed loading, the distributions regain distributions of Fig. 6. This figure shows distributions of the the diverse critical state condition. rotations (vorticity) of the voids in the dense assemblies When one focuses on individual particles or contacts during the loading phase. A fairly peaked distribution at during deformation of a soil or other granular material, the strain 0% broadens and becomes more diverse during motions appear bewildering and complex, with seemingly loading, attaining a nearly stationary entropy at the critical erratic fits, stalls, and reversals in all directions (Kuhn, state snapshots of 22% and 30%. 2016a). Yet from this local ferment, bulk behavior at the In some cases, the assemblies temporarily reached a larger critical state maintains a steady monotony of stress and entropy H prior to the critical state (for example, during density. An unsolved and perplexing problem in granular loading at the small strains of 1%–3% in Fig. 5). Note thatH mechanics is the prediction of constitutive behavior with measures the comparison of a given probability distribution models that are based upon the complex local interactions with the most naive, disordered system having uniform of particles. In the current work, we can glimpse a possible distribution q . Comparisons to the uniform distribution resolution of the problem, by focusing upon the probability provide a gross measure of disorder, butq fails to account for distributions of micro-scale quantities, which naturally certain constraints that apply at all stages of loading, from the express their disordered character. If one accepts the critical start of loading through the critical state. These constraints state distribution as the eventual stationary and stable state include the consistency of particle motions with the applied and uses, as a generalised “state parameter”, the difference bulk deformation, consistency of contact forces with applied between a current distribution and that of the critical stresses, and energy conservation principles. Such imposed state, one may be able to predict the progress of both conditions induce biases in the distributions of micro-scale micro- and macro-quantities during a strain sequence. quantities that are not considered when comparisons are Viewing the critical state as a convergent, stationary, and made with a uniform distribution. Previous work has shown maximally disorder condition may enable fresh progress in that, when considered, these constraints can account for understanding and predicting soil behavior. many of the micro-scale trends that are observed at the critical state (Kuhn, 2016b). CONCLUSION In 1872, Boltzmann published his H-theorem, a rigorous derivation of an entropy that was based upon micro-scale analysis of particle collisions (Boltzmann, 2003). Signifi- cantly, this probabilistic approach to ideal diffuse gases demonstrated that any temporary disturbance from an equi- librium distribution is eventually corrected, as the particle assembly naturally returns to the most probable (i.e. equi- librium) state. An H-theorem has been proposed for force distributions within static dense granular assemblies (Met- zger, 2008), but no rigorous proof has yet been presented of a maximum-entropy principle for dense granular flow. Prepared using GeotechAuth.cls Probability density, p(κ) M. R. KUHN 9 NOTATION Kuhn, M. R. (2014). Dense granular flow at the critical state: cs critical state maximum entropy and topological disorder. Granul. Matter 16, e void ratio No. 4, 499–508. Kuhn, M. R. (2016a). Contact transience during slow loading of F components of fabric tensor ij dense granular materials. J. Eng. Mech. , C4015003doi:10.1061/ H Shannon entropy (ASCE)EM.1943- 7889.0000992. H relative entropy cs Kuhn, M. R. (2016b). Maximum disorder model for dense steady- H entropy relative to critical state distribution state flow of granular materials. Mech. of Mater. 93, 63–80. cs H entropy relative to uniform distribution Kuhn, M. R. & Bagi, K. (2004). Contact rolling and deformation in L components of velocity gradient ij granular media. Int. J. Solids Struct. 41, No. 21, 5793–5820. N number of particles or objects Kuhn, M. R. & Bagi, K. (2009). Specimen size effect in discrete p mean stress element simulations of granular assemblies. J. Eng. Mech. 135, p probability of discrete No. 6, 485–492. p() probability density of continuous Learned-Miller, E. G. (2004). Hyperspacings and the estimation of information theoretic quantities. Technical Report 04.104, q reference probability distribution cs University of Massachusetts - Amherst. q critical state probability distribution, discrete Luzzani, L. & Coop, M. R. (2002). On the relationship between q uniform probability distribution, discrete cs particle breakage and the critical state of sands. Soils and Found. q () critical state probability distribution, continuous 42, No. 2, 71–82. q () uniform probability distribution, continuous Meléndez, M. & Español, P. (2014). Gibbs-Jaynes entropy versus u uniform relative entropy. Journal of Statistical Physics 155, No. 1, 93–105, R m-dimension space, real numbers doi:10.1007/s10955- 014- 0954- 6, URL http://dx.doi.org/ Z m-dimension space, integers 10.1007/s10955-014-0954-6. " strain in x direction 11 1 Metzger, P. T. (2008). H theorem for contact forces in granular data values materials. Phys. Rev. E 77, No. 1, 011307. Muir Wood, D. & Maeda, K. (2007). Changing grading of soil: effect friction coefficient at particle contacts on critical states. Acta Geotechnica 3, No. 1, 3–14. stress components ij Peña, A. A., García-Rojo, R., Alonso-Marroquín, F. & Herrmann, H. J. phase space of possible  values (2009). Investigation of the critical state in soil mechanics using DEM. In Powders and Grains 2009 (Nakagawa, M. & Luding, S., eds.), Amer. Inst. of Phy., pp. 185–188. Satake, M. (1982). Fabric tensor in granular materials. In Proc. IUTAM Symp. on Deformation and Failure of Granular Materials REFERENCES ( Vermeer, P. A. & Luger, H. J., eds.), Rotterdam: A.A. Balkema, pp. Bagi, K. (1996). Stress and strain in granular assemblies. Mech. of 63–68. Mater. 22, No. 3, 165–177. Satake, M. (1992). A discrete-mechanical approach to granular Been, K. & Jefferies, M. G. (1985). A state parameter for sands. materials. Int. J. Engng. Sci. 30, No. 10, 1525–1533. Géotechnique 35, No. Volume 35, Issue 2, 99–112. Schofield, A. N. & Wroth, P. (1968). Critical state soil mechanics. New Ben-Naim, A. (2008). A farewell to entropy: statistical thermodynam- York: McGraw-Hill. ics based on information. World Scientific. Thornton, C. (2000). Numerical simulations of deviatoric shear Blumenfeld, R. & Edwards, S. F. (2009). On granular stress statistics: deformation of granular media. Géotechnique 50, No. 1, 43–53. Compactivity, angoricity, and some open issues. The Journal of Valanis, K. C., Peters, J. F. & Gill, J. (1993). Configurational entropy, Physical Chemistry B 113, No. 12, 3981–3987. non-associativity and uniqueness in granular media. Acta Boltzmann, L. (2003). Further studies on the thermal equilibrium Mechanica 100, 79–93, doi:10.1007/BF01176863. of gas molecules. In The kinetic theory of gases : an anthology Zhao, J. & Guo, N. (2013). Unique critical state characteristics in of classic papers with historical commentary (Hall, N. S., ed.), granular media considering fabric anisotropy. Géotechnique 63, London: Imperial College Press, pp. 262–349, (English translation No. 8, 695–704. by Brush, S. G.). Casagrande, A. (1936). Characteristics of cohesionless soils affecting the stability of slopes and earth fills. J. Boston Soc. Civil Eng. 23, No. 1, 13–32. Duong, T. (2007). ks: Kernel density estimation and kernel discriminant analysis for multivariate data in R. Journal of Statistical Software 21, No. 7, 1–16. Edwards, S. F. & Oakeshott, R. (1989). Theory of powders. Physica A 157, No. 3, 1080 – 1090. Jaynes, E. T. (1957). Information theory and statistical mechanics. Phys. Rev. 106, No. 4, 620–630, doi:10.1103/PhysRev.108.171. Kuhn, M. R. (1999). Structured deformation in granular materials. Mech. of Mater. 31, No. 6, 407–429. Prepared using GeotechAuth.cls http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Condensed Matter arXiv (Cornell University)

The critical state of granular media: Convergence, stationarity, and disorder

Condensed Matter , Volume 2018 (1812) – Dec 18, 2018

Loading next page...
 
/lp/arxiv-cornell-university/the-critical-state-of-granular-media-convergence-stationarity-and-mtynV5mew0
ISSN
0016-8505
eISSN
ARCH-3331
DOI
10.1680/jgeot.16.P.008
Publisher site
See Article on Publisher Site

Abstract

Géotechnique 00, 1–9 [http://dx.doi.org/10.1680/geot.XX.XXXXX] MATTHEW R. KUHN Discrete element simulations are used to monitor several micro-scale characteristics within a granular material, demonstrating their convergence during loading toward the critical state, their stationarity at the critical state, and the evolution of their disorder toward the critical state. Convergence, stationarity, and disorder are studied in the context of the Shannon entropy and two forms of Kullback-Leibler relative entropy. Probability distributions of twenty aspects of micro-scale configuration, force, and movement are computed for three topological objects: particles, voids, and contacts. The probability distributions of these aspects are determined at numerous stages during quasi-static biaxial compression and unloading. Not only do stress and density converge to the critical state, but convergence and stationarity are manifested in all of the micro-scale aspects. The statistical disorder (entropy) of micro-scale movements and strains generally increases during loading until the critical state is reached. When the loading direction is reversed, order is briefly restored, but continued loading induces greater disorder in movements and strains until the critical state is reached again. KEYWORDS: constitutive relations; fabric/structure of soils; particle-scale behaviour; statistical analysis INTRODUCTION characteristics — persistent convergence toward stationary The critical state is a foundational concept in geomechanics, bulk attributes — resemble those of thermal systems that encompassing two characteristics of soils and other granular reliably approach a stable, equilibrium condition with materials (Schofield & Wroth, 1968). After sustained slow sufficient passing of time, even as the underlying molecular shearing from an initial packing, a granular material reaches motions remain dynamic and diverse. a critical (steady) state in which the ratio of shear stress In a pair of recent papers, the author proposed a third to mean stress remains nearly stationary during further characteristic of the critical state: at the critical state, the micro-scale landscape of particle arrangements and particle shearing. Not only does the stress reach a steady condition, but other bulk measures of internal fabric and structure interactions exhibits the greatest diversity and disorder that remain stationary as well (Thornton, 2000; Peña et al., 2009). is consistent with information known a priori — information Although the values of these characteristics will depend upon that constrains the available landscape (Kuhn, 2014, 2016b). the material’s composition and upon the form of loading, The two papers, based upon a conjecture of maximum for a given material and loading, a specific steady state disorder, demonstrated that observed micro-scale features condition is eventually attained. This critical state condition at the critical state are similar to those that derive from is most enduring at low confining stress, since particle the conjecture. These features include distributions of breakage at high stress can continue to alter the internal the contact forces, orientations, and movements and the fabric, even as the shearing stress remains constant (Luzzani topologic arrangements of particles. & Coop, 2002; Muir Wood & Maeda, 2007). Notably, the bulk Both works deduced this disordered nature by first steady condition occurs even while individual particles are considering the full extent (phase space) of potential undergoing seemingly erratic and highly varied motions and micro-states (particle arrangements, movements, forces, interactions, notwithstanding a slow, quasi-static loading etc.) and then applying available information in the form (Kuhn, 2016a). of reasonable bulk constraints on the space of micro-states. The second characteristic is the convergent quality of Among the many different micro-states that can satisfy the critical state: the eventual bulk attributes for a given the constraints, the author sought the specific collection material are insensitive to the initial particle arrangement of micro-states (called a “macro-state”) that exhibited and to prior stress conditions. For example, materials that are the greatest diversity of micro-states. This approach is initially either loose or dense or have different initial particle equivalent to finding the macro-state that encompasses arrangements will approach the same density after sufficient the greatest number of micro-states that satisfy the given shearing (Casagrande, 1936; Zhao & Guo, 2013). These constraints. Jaynes (1957) contended that this macro-state of maximum entropy satisfies the available information but is otherwise unbiased. As a simple illustration, consider micro-states of particle arrangement, with each micro-state Manuscript received. . . Prepared using GeotechAuth.cls [Version: 2015/01/21 v1.00] 1 arXiv:1812.07662v1 [cond-mat.soft] 18 Dec 2018 2 DISORDER AND THE CRITICAL STATE expressed as an ordered list of integers, between 1 and 10, the micro-state’s list. These frequencies can be extracted representing, say, the number of contacts of each particle for a particular micro-state, but other micro-states within in an assembly. Now impose an average of 7.0 for the the phase space can also share the same frequencies, most integers in a list — the average coordination number. simply as a reshuffling of the order of the N items  within Although it is certainly possible that every particle will the list. The mapping of micro-states and macro-states, have seven contacts, thus satisfying the bulk average, this therefore, is surjective: each micro-state corresponds to a condition is highly unlikely and has a low entropy, since unique macro-state, but many micro-states can comprise this macro-state is only realised by a single list of 7’s (a the same macro-state. In principle, constructing a specific singular micro-state) that corresponds to a nearly crystalized micro-state from a more general macro-state requires condition. Far more lists can be created by permuting equal additional information beyond the gross probabilities, and occurrences of the numbers 6 and 8, and the macro-state the amount of information that must supplement the macro- encompassing such micro-state lists would indicate a higher state, the Shannon “missing information,” is a measure of a entropy and a higher likelihood. In an analogous manner, macro-state’s diversity or disorder. The two terms are used the author applied principles of statistical mechanics (in interchangeably, and entropy will designate a quantified particular, the Jaynes formalism of maximising the Shannon measure of disorder. entropy) to derive characterisations of the critical state. Methods of quantifying the missing information must Starting with the assumption of maximum disorder, the accommodate, in part, characteristics that are discrete and two papers obtained maximally disordered distributions of countable (quantum states for molecules, or coordination various micro-quantities which compared favourably with numbers for a granular material); but a general method distributions measured in discrete element method (DEM) must also accommodate information that is continuous simulations at the critical state. (momenta for classical molecules, or forces, rates, etc. within The current work examines connections among the three a granular material). With discrete data, each item  can, it- characteristics of the critical state: its stationary character, self, be an m-list of discrete values n,  = fn ; n ; : : : ; n g , 1 2 m i i the convergence tendency, and the tendency toward maxi- belonging to the m-space of integers Z . For example, the mum disorder. Unlike the author’s previous works, which be- author showed that the topological arrangement of a two- gan with an assumption of maximum disorder at the critical dimensional assembly of N disks could be represented as a state, the current work directly measures disorder and tracks list (journal) of N integer pairs, each taken from Z (Kuhn, its evolution in DEM simulations. The simulations start from 2014). With continuous data, the items  can be lists of M the initial conditions of densely-packed and loosely-packed data points q,  = fq ; q ; : : : ; q g , taken from the contin- i 1 2 M i assemblies, continue through deviatoric loading in one di- uous M -space R . We will use both types of data — discrete rection, and finish with sustained loading in the reversed and continuous — to describe the evolution of disorder direction. With these simulations, the disorder in particle during the loading of granular media. movements is shown to generally increase during loading The disorder of a macro-state is measured by first until the critical state is reached. Once attained, all micro- expressing the underlying data in terms of probabilities p scale measures of fabric, force, and movement remain in a of the component categories  that appear within micro- steady condition. state lists. The Shannon entropy H is based upon these likelihoods — probabilities p for discrete systems and probability density p() for continuous systems — of the component characteristics  of a macro-state: INFORMATION AND DISORDER The intent is to quantify the diversity exhibited by the micro- H = p ln (p) or states of a granular material. For a thermal system, a micro- state might be presented as a snapshot of the positions and (1) H = p() ln (p()) d momenta of the N identifiable constituents (molecules) at a particular instant. In the current work, snapshots are taken of DEM simulations at particular strains, providing the local where phase space is the set of all possible  components, densities, coordination numbers, forces, movements, etc. A S P = , such that p = 1 and p() d = 1. The intrin- Boltzmann micro-state can be expounded as an ordered N - sic measure H quantifies the diversity of micro-states that list of such characteristics  that characterise a system’s N comprise a macro-state. Oddly, H does not depend upon the constituents,f ;  ; : : : ;  g, and the full set of all possible 1 2 values of the underlying data  but only on their probabilities. lists is the phase space of the system. Appropriate phase Note the lack of a Boltzmann-like constant in entropy H , spaces are later identified for depicting the slow shear-driven conventionally applied to the Shannon entropy as a scaling movements of particles and contacts. The macro-state to which a micro-state belongs is represented by the bulk relative frequencies (empirical probabilities) of items within Prepared using GeotechAuth.cls M. R. KUHN 3 factor to enforce coincidence with the thermodynamic en- of the measured posterior probability p is taken relative to tropy (Ben-Naim, 2008). Two variants of the Shannon mea- the most naive, uninformed prior: q is simply the uniform u u sure, described below, will be used to estimate the relative distribution, q or q (), spread evenly across phase space disorder within particular micro-states collected from DEM . In this manner, one measures the disorder of a given snapshots. system relative to the most possibly disordered system, in Estimating probabilities p is straightforward for a micro- which probabilities are uniformly distributed, state of a system having a finite number of discrete possibilities : by simply counting the occurrences of each H = p ln or condition  and dividing by the length of the data list, 2 (2) N . For the example presented above, entropy H for lists p() H = p() ln d exclusively constructed from 7’s is1:0 ln(1:0) = 0; for lists q () of 6’s and 8’s in equal number, H = 2(0:50) ln(0:5) = 0:693. u u In general, the more uniform and diverse the distribution, the where q and q () are proper distributions, each with larger is its entropy; more biased, uneven, or concentrated a net expectation of 1. The negative Kullback-Leibler distributions (i.e., those having a more certain outcome) distance H is always non-positive. If a system’s density have lower entropies. Indeed, the basis of Jaynes’ maximum p equals the uniform density q , then the relative entropy entropy (MaxEnt) principle is that the most reasonable H has its largest value of zero, corresponding to the most probability distribution is the one that is consistent with disordered system. Continuing the previous example, if all available information, but only this information: the coordination numbers are between 1 and 10, then q = 1=10, remaining,“missing” information is maximised. For example, and the entropy H for lists exclusively constructed from when no information is available for the 10 possible 7’s is 1:0 ln(1:0=0:1) = 2:303; for lists of 6’s and 8’s in constituents of the previous list, not even an average value, equal number,H = 2(0:50) ln(0:5=0:1) = 1:609; and for each constituent should be assumed equally probable, for a uniform distribution,H = 0. Again, larger (less negative) which H = 10 (0:1) ln(0:1) = 2:30. When the average is values express greater diversity and disorder. known to be 7.0, but no other information is available, With the second relative entropy, one exploits information entropy H is maximised with a distribution of exponential gained at the critical state, using knowledge of its distribu- form that is biased toward larger numbers and with the lower tions of particle arrangement, force, and movement as prior entropy of 2.16 (see Ben-Naim, 2008 for other examples). information, against which a granular specimen is compared The entropy measure in Eq. (1) is problematic, however, as it is being loaded toward the critical state. This relative when used to characterise the disorder exhibited in DEM entropy simply measures the (negative) distance between a cs data. The Shannon entropy of a continuous system, as in distribution p and the reference, critical state distribution q , Eq. (1 ), suffers from a lack of scale invariance, such that testing convergence toward (and stationarity of ) the critical a scaling of variables  and the corresponding region state: (perhaps with a change in dimensional units) will alter cs the entropy H (note the logarithm of density p with H = p ln or cs its troublesome dimensional units, see Ben-Naim, 2008). m (3) A second shortcoming is that H describes the missing p() cs H = p() ln d cs information associated with posterior probabilities p but q () does not consider possible access to prior information, even imperfect information, in the form of preferential This second relative entropy is a measure of disorder only probabilities. if one accepts the critical state as the most disordered These problems are avoided with an alternative measure system, a question that is answered by applying the first u cs of disorder — the relative entropy or negative Kullback- relative entropyH to DEM data. The scalarH is similar to Leibler distance (Meléndez & Español, 2014). In this manner, the “state parameter” of Been & Jefferies (1985), which also one can account for possible access to a priori information characterises the distance of a current condition from the in the form of the prior probabilities q or q(), which are critical state: but rather than a difference in bulk void ratios, extracted from past evidence (such as information gathered cs H is the difference between entire probability distributions at the critical state). Two sets of prior probabilities are of the underlying micro-scale data. Although the critical state considered. cs distributions q could be estimated using methods of the With the first relative entropy, it is assumed that no prior author’s previous papers, these distributions will be directly information is available about the particle arrangements, extracted from DEM simulations. forces, or movements, so that one can test whether the disorder of these quantities increases as a granular material is loaded. For this case, the missing information (or disorder) Prepared using GeotechAuth.cls 4 DISORDER AND THE CRITICAL STATE 1.0 DEM SIMULATIONS Dense Numerical simulations were used for tracking convergence 1 0.5 toward the critical state and the evolution of disorder during Loose deviatoric loading. The simulations were conducted on 0.0 simple, two-dimensional assemblies of 676 bi-disperse disks that were contained within periodic boundaries. Previous -0.5 work by the author showed that assemblies with at least a few hundreds of particles reach nearly the same deviatoric -1.0 stress at the critical state Kuhn & Bagi (2009). Using two disk 0 5 10 15 20 25 30 varieties, with ratios of 1.5:1 in size and 1:2.25 in number, Strain, −ε , percent prevented the crystallisation of particle arrangements that (a) are otherwise observed with mono-disperse assemblies. 0.26 To accumulate large data sets, the author created 50 Loose 0.24 randomly generated assemblies (initial micro-states) of dense assemblies and 50 random loose assemblies, and 0.22 then slowly loaded each of the 100 assemblies under 0.20 identical biaxial compression conditions. The creation of Dense these initial assemblies is confirmation that many different 0.18 configurations, both dense and loose, are consistent with a 0.16 given assembly size and preparation method, as in the P - 0 5 10 15 20 25 30 system of Blumenfeld & Edwards (2009). The particles were Strain, −ε , percent isotropically compacted from a sparse two-dimensional (b) granular gas in which the disks were assigned random Dense velocities and an artificial friction coefficient ( = 0:20 for 0.2 the dense assemblies, and  = 0:50 for the loose assemblies), 0.1 Loose with compaction proceeding until the assemblies “seized”, at which the void ratios were 0.175 and 0.246 for the dense 0.0 and loose assemblies. A 0:50 friction coefficient was then -0.1 assigned for the subsequent loading sequence. In the first phase of biaxial loading, the horizontal width -0.2 of each assembly was reduced (compressed) at a constant 0 5 10 15 20 25 30 rate while the height was allowed to expand so that the mean Strain, −ε , percent stress within the assembly remained constant. Loading (c) proceeded slowly to a horizontal strain of 30%, bringing the assemblies fully into the critical state of constant Fig. 1. Stress, strain, void ratio, and contact fabric anisotropy (Satake, 1982) during biaxial loading and unloading, averaged stress, volume, and fabric (Fig.1). The biaxial loading was from 50 numerical simulations of dense and loose assemblies of then reversed by increasing the width (i.e. extension) at a 676 disks. constant rate while reducing the height and maintaining the original mean stress. Reversed loading was ceased when the strain returned to 1%. This reversal phase also Bulk convergence, stationarity, and ergodicity brought the assemblies to the critical state, but with the Figure 1 shows changes in the deviator stress and void ratio principal directions of stress, strain, and fabric rotated by during the compressive loading phase and the reversed 90 . Snapshots were taken of all micro-data at specific strains loading phase, as averaged across 50 specimens. For both to capture the evolving conditions: 22 snapshots during the dense and loose assemblies, the bulk measures of deviator initial phase and 21 snapshots during the reversal phase. By stress, void ratio, average fabric anisotropy, and average running simulations on 50 assemblies, the author collected coordination number reached a steady, critical state at the data lists on about N = 29000 particles, N = 18000 voids, strain 20% during forward loading (arrow 1, Fig. 1), and and N = 47000 contacts for each snapshot of the dense or these bulk quantities also reached the corresponding steady loose assemblies. conditions during reversed loading, after the strain was cs u With each snapshot, entropiesH andH were computed reduced from 30% to 10% (arrows 2 and 3). The dense for various micro-scale characteristics, described in a later and loose assemblies attained nearly identical deviator section, by using the Voronoi method of Learned-Miller stresses, void ratios, and fabric anisotropies at the critical (2004), kernel density estimates of Duong (2007), and direct state, evidence of the state’s convergent character. Note that binning. the dense and loose assemblies behaved almost identically Prepared using GeotechAuth.cls Fabric anisotropy, F − F Void ratio, e Deviator stress, (σ − σ )/p 11 22 11 22 M. R. KUHN 5 1.0 which had strength variances of 0.0035–0.0070. These results indicate that the critical state can be characterised either 0.5 by measuring multiple specimens at a given strain or by measuring a single specimen across a range of strains. In 0.0 the current study, multiple specimens were averaged among multiple strain snapshots at the critical state. -0.5 The primary focus of the remainder of this paper is statistical analyses of micro-scale (rather than macro-scale, -1.0 bulk) data, and as described below, several tens of thousands 0 5 10 15 20 25 30 of micro-scale samples were used for this purpose. It is no Strain, −ε , percent surprise that the critical state exhibits bulk stationarity, as Fig. 2. Stress and strain of 50 simulations of loose assemblies. the state is largely defined by this characteristic. As will be seen, however, stationarity and convergence are more deep- rooted, as they pervade all micro-scale aspects as well as during the reversed loading, with their averaged stress-strain being expressed in bulk, macro-scale characteristics. behaviours being almost indistinguishable. Figure 2 shows stress and strain of all 50 loose specimens, disclosing the erratic character of stress evolution (the dense MICRO-SCALE STATIONARITY, CONVERGENCE, AND specimens exhibited similar fluctuations). The fluctuations DISORDER are also apparent in averaged data of Fig. 1a as slight, raspy The evolving micro-scale quantities are broadly grouped as variations from the average trend. The 50 specimens were follows: sufficient in number, however, such that the mean deviator stress q=p has a 95% confidence interval of0:02. I. local configuration (i.e. packing) quantities, such as co- An analysis of all 50 specimens shows that after the critical ordination number, density, and contact orientation, state is reached, these erratic variations in stress exhibited II. local force quantities, such as particle stresses and both stationarity and ergodicity. Although both terms are contact forces, and usually applied to time series data, in the current context, III. local movement quantities, such as particle transla- a monotonically increasing (or decreasing) strain serves tions and rotations, contact movements, and local the role of time, and the spatial domain is represented by strains. the stresses realised in different specimens at the same strain. Stationarity, meaning constant statistical measures of as presented in the columns of Table 1. The first two aspects stress across a range of strains, was tested by computing the are associated with the status or configuration of a granular mean and variance of stress for the specimens during two assembly, which are related to the configuration distribu- strain intervals: 20–25% and 25–30%. Stationarity was clearly tions of Edwards & Oakeshott (1989) or the configurational demonstrated at the critical state, as the means and variances entropy of Valanis et al. (1993). The third set of attributes of deviator stress (  )=p were nearly same for the two are transitional rate quantities that are driven by the bulk de- 11 22 intervals (means of 0.561 and 0.556, and variances of 0.0050 formation and will depend upon the loading direction. The and 0.0053). distinction between state (being) and transition (becoming) If a process exhibits the ergodicity property, multiple is explored further below. instances of the process taken at a given time yield the The micro-scale of a two-dimensional granular material same statistical characteristics as that of a single instance can be described with any of three topological objects: its tracked across a sufficiently extended time range. In the non-rattler grains, which can be represented as the nodes of current setting, an “instance” is a single simulation and a particle graph; the contacts between particles, represented “time” is strain. Ergodicity means that similar fluctuations as links between nodes; and the voids, represented as would be observed by tracking a single simulation across polygonal loops formed by the nodes and links (Satake, 1992). a range of strains as would be observed by noting the These objects must be distinguished, because configuration, results of many simulations at a single strain. Ergodicity force, and movement are expressed in different ways among was tested by computing the mean and variance of all 50 the objects, and the three objects’ associated micro-state lists specimens at strain 25% and comparing these values to f ;  ; : : : ;  g will contain different quantities and have 1 2 N those of individual specimens taken across a strain range different lengths N . The three objects are presented in the of 20–30%. The statistical measures were almost the same rows of Table 1. for the two cases, with nearly identical spatial and temporal Altogether, the table identifies twenty local micro-scale means, and spatial and temporal variances of 0.0051 and characteristics that were characterised in the form of their 0.0043. The statistical characteristics at strain 25% are nearly probability distributions p at various strain snapshots. the same as those at other strains in the range 20–30%, Three of the quantities are discrete (coordination number, Prepared using GeotechAuth.cls Deviator stress, (σ − σ )/p 11 22 6 DISORDER AND THE CRITICAL STATE Table 1. Matrix of micro-scale characteristics within a granular assembly. Local characteristics of a granular assembly 1 2 I) Configuration II) Force III) Movement 3 8 Particles Coordination no. Pressure Horizontal velocity Deviator stress Vertical velocity Shear stress Rotational velocity 4 9 Voids Valence — Dilation rate Contact journal Compression rate Density Transverse shear rate Rotation rate 7 10 Contacts Orientation Normal force Slip rate Tangential force Rolling rate Rigid rotation rate forces are normalised by dividing by mean stress p. movements are normalised by dividing by strain increment " . number of contacts of a particle. number of links (contacts) around a polygonal void. list of integer pairs for reconstructing an entire particle graph (see Kuhn, 2014, §2). void ratio of a polygonal void, ignoring any rattler particles by treating them as void space. angular orientation of a contact normal vector. particle stresses computed from the contact forces (see Bagi, 1996), expressed as the pressure ( +  ), deviator stress   , and shear stress  . 11 22 11 22 12 local deformation rate within a polygonal void, computed from the particle movements around the void’s perimeter (Kuhn, 1999, §2.1). The rate is a local velocity gradient L , ij expressed in four combinations: dilation L + L , deviatoric compression L L , 11 22 11 22 transverse shear L + L , and rotation L L . 12 21 12 21 relative movements of a particle pair, computed from the particles’ six component rates (two translation rates and a rotation rate for each particle). These six values are expressed in four combinations: three tangential combinations (slip, rolling, and rigid rotation, as described in Kuhn & Bagi, 2004) and the relative normal movement of the two particles. valence, and the contact journal); the other quantities are 5 2 0.0 continuous. -0.2 Convergence and stationarity cs MeasureH in Eq. (3) characterises the difference (negative distance) between the probability distributions of micro- Dense, loading -0.4 Dense, unloading scale quantities during loading and their eventual distribu- Loose, loading tions at the critical state. This measure was applied to each Loose, unloading of the twenty micro-scale characteristics in Table 1. For the -0.6 initial loading phase, distributions were compared with the cs reference distribution q at strain 30%; during the reversed 0 10 20 30 loading, the final critical state distributions at strain 1% were Strain, −ε , percent used. Fig. 3. Evolving distance between distributions of the deviator As an example of the twenty characteristics, Fig. 3 shows stresses within particles and the distribution at the critical state. cs A value of zero represents perfect conformance with the critical the evolution ofH for the particular distribution of the local state distribution. deviator stresses within particles, the difference 11 22 for each particle within the N=29,000 particle samples. The bulk stress was initially isotropic, such that any small deviator stresses (both positive and negative) within individual particles balanced to yield a bulk deviator stress of zero. the distribution of the particle deviator stresses had become This initial distribution of deviator stress was quite different closer to that of the critical state. During further loading than that at the critical state, as is shown in Fig. 4: upon (arrows 1 and 2, Fig. 3), the distribution of stress progressively the start of loading, the distribution becomes broader and approached the distribution of the critical state. At strains more diverse, exhibiting greater disorder. At the strain 0%, greater than 20%, the distributions became stationary and cs the (negative) distanceH was0:23 (Fig. 3). After loading to were nearly identical to the critical state distribution: that is, cs strain 0.4%, the negative distance was0:05, indicating that H attained a steady value of zero. The distributions at both Prepared using GeotechAuth.cls Object Particle deviator stresses, (σ − σ ): 11 22 cs Entropy relative to critical state, H M. R. KUHN 7 0.8 -6 Strain ε 0% 0.6 0.4% -8 22% and 30% 0.4 Dense, loading -10 0.2 Dense, unloading Loose, loading Loose, unloading 0.0 -12 -2 -1 0 1 2 3 Particle deviator stress, σ − σ 11 22 0 10 20 30 Fig. 4. Density distributions of the mean stress within particles Strain, −ε , percent for the dense specimens during loading at various strains. Fig. 5. Disorder in the joint probability distribution of particle velocities (horizontal, vertical, and rotational) as measured with the relative entropyH . 22% and 30% are plotted in Fig. 4, and any small differences can only be discerned by enlarging the plot. When the loading was reversed at strain 30%, the deforming assembly, but one can ask whether this static distribution of deviator stress was very different from its assembly is “at the critical state”? Although the question may eventual critical state distribution (arrow 3 in Fig. 3), seem as semantic nuance, any loading direction other than a since the previous loading had imparted a positive bias continued forward biaxial loading will divert the assembly to the particle deviator stresses, a direction counter to the from its current critical state of deformation. That is, the opposite bias that was progressively induced during further critical state is an active condition in which deformation reversed loading. The reversed loading quickly brought the and movement are essential ingredients. As will be seen, distributions to those of the eventual (reversed) critical state disorder in the micro-scale movements does increase while cs (arrows 4 and 5 in Fig. 3), with anH equal to zero. an assembly is approaching the critical state. On the other As with the local deviator stresses, the other nineteen hand, notable examples of reduced disorder occur among characteristics in Table 1 were also considered, and the joint the configuration and force quantities in Table 1 (columns I probability distributions of multiple characteristics in each and II). For example, the anisotropy that develops in the row of the table were also analysed. The relative entropies of contact orientations during loading (e.g., Thornton, 2000) all micro-scale characteristics exhibited similar trends: their is a form of increasing order, since a uniform, isotropic distributions converged to the corresponding distribution of distribution of orientations exhibits the least order. Another cs the critical state, such thatH equalled zero at a (forward) form of order develops during plastic deformation: when strain of about 20% or after a reversal of strain from 30% contact forces reach their frictional limit, they are constricted to about 10%. The distributions remained stationary during by the friction law, which reduces disorder in the tangential further loading. That is, convergence and stationarity at the components of the contact forces (the friction law can be critical state are manifested in all twenty micro-scale aspects. viewed as a form of constrictive information). An example of the tendency toward increasing disorder in Disorder the micro-scale movements is shown in Fig. 5. This plot is The evolution of micro-scale disorder was investigated by of the relative entropyH of the combined, joint probability computing the relative entropyH of Eq. (2) for the twenty distribution of all three components of particle movements: characteristics in Table 1. The author’s original hypothesis horizontal, vertical, and rotational velocities. EntropyH is was that the disorder of each characteristic would increase taken relative to the most disordered (uniform) distribution during loading toward the critical state. Instead, although q , and a value of zero represents the greatest disorder. In H did increase with all local movement characteristics Fig. 5, the movements of particles within the dense and loose (column III of Table 1), it did not necessarily increase with assemblies are seen to increase from their initially ordered the configuration or force characteristics in columns I condition (large negativeH ) at zero strain (arrow 1, Fig. 5) to and II. Upon further thought, one recognises that the critical a larger value at the critical state (arrow 2). When the loading state is not a fixed condition or status; instead it is a is reversed (arrow 3), the entropy of particle movements is transitional, active quality, which has meaning only in the immediately reduced, corresponding to more ordered (less context of a driving deformation. Consider, for example, diverse) movements. Further reversed loading causes greater the condition of an assembly being deformed in (forward) disorder (arrow 4) until the assembly reaches the reverse biaxial compression at strain 30% (see Fig. 1). One could critical state (arrow 5). The same trends were measured with separately build a static arrangement of disks having the all ten movement quantities in column III of Table 1: the local, same configuration and force characteristics as that of the micro-scale movements and strains of the particles, voids, Prepared using GeotechAuth.cls Probability density, p(κ) Particle velocities, v , v , ω : 1 2 Entropy relative to uniform distribution, H 8 DISORDER AND THE CRITICAL STATE 1.2 Although not rigorous confirmation, the current work and Strain ε 1.0 the author’s previous works give support to the concept 0% of maximum entropy at the critical state. Simulations of 0.8 0.4% 22% the biaxial loading of simple, two-dimensional disc assem- 0.6 30% blies show that micro-scale distributions of arrangement, 0.4 force, and movement converge to stationary distributions. 0.2 Furthermore, disorder in the movements, as characterised 0.0 by the entropy relative to a uniform distribution, tends to -4 -2 0 2 4 increase during loading until the critical state is reached. Void rotation rate, L − L 12 21 That is, starting from initial packings, the particle motions Fig. 6. Density distributions of the void rotations within the are relatively ordered, but as loading approaches the peak dense specimens during loading at various strains. stress and beyond, the distributions of motions become more diverse (i.e., disordered). Upon a reversal of loading — a temporary disturbance from the stationary, stable critical and contacts generally increased as deformation progressed state — the distributions of micro-scale arrangement, force, toward the critical state. and movement abruptly deviate from their stable condition, The increasing disorder is also seen in the probability but after sustained reversed loading, the distributions regain distributions of Fig. 6. This figure shows distributions of the the diverse critical state condition. rotations (vorticity) of the voids in the dense assemblies When one focuses on individual particles or contacts during the loading phase. A fairly peaked distribution at during deformation of a soil or other granular material, the strain 0% broadens and becomes more diverse during motions appear bewildering and complex, with seemingly loading, attaining a nearly stationary entropy at the critical erratic fits, stalls, and reversals in all directions (Kuhn, state snapshots of 22% and 30%. 2016a). Yet from this local ferment, bulk behavior at the In some cases, the assemblies temporarily reached a larger critical state maintains a steady monotony of stress and entropy H prior to the critical state (for example, during density. An unsolved and perplexing problem in granular loading at the small strains of 1%–3% in Fig. 5). Note thatH mechanics is the prediction of constitutive behavior with measures the comparison of a given probability distribution models that are based upon the complex local interactions with the most naive, disordered system having uniform of particles. In the current work, we can glimpse a possible distribution q . Comparisons to the uniform distribution resolution of the problem, by focusing upon the probability provide a gross measure of disorder, butq fails to account for distributions of micro-scale quantities, which naturally certain constraints that apply at all stages of loading, from the express their disordered character. If one accepts the critical start of loading through the critical state. These constraints state distribution as the eventual stationary and stable state include the consistency of particle motions with the applied and uses, as a generalised “state parameter”, the difference bulk deformation, consistency of contact forces with applied between a current distribution and that of the critical stresses, and energy conservation principles. Such imposed state, one may be able to predict the progress of both conditions induce biases in the distributions of micro-scale micro- and macro-quantities during a strain sequence. quantities that are not considered when comparisons are Viewing the critical state as a convergent, stationary, and made with a uniform distribution. Previous work has shown maximally disorder condition may enable fresh progress in that, when considered, these constraints can account for understanding and predicting soil behavior. many of the micro-scale trends that are observed at the critical state (Kuhn, 2016b). CONCLUSION In 1872, Boltzmann published his H-theorem, a rigorous derivation of an entropy that was based upon micro-scale analysis of particle collisions (Boltzmann, 2003). Signifi- cantly, this probabilistic approach to ideal diffuse gases demonstrated that any temporary disturbance from an equi- librium distribution is eventually corrected, as the particle assembly naturally returns to the most probable (i.e. equi- librium) state. An H-theorem has been proposed for force distributions within static dense granular assemblies (Met- zger, 2008), but no rigorous proof has yet been presented of a maximum-entropy principle for dense granular flow. Prepared using GeotechAuth.cls Probability density, p(κ) M. R. KUHN 9 NOTATION Kuhn, M. R. (2014). Dense granular flow at the critical state: cs critical state maximum entropy and topological disorder. Granul. Matter 16, e void ratio No. 4, 499–508. Kuhn, M. R. (2016a). Contact transience during slow loading of F components of fabric tensor ij dense granular materials. J. Eng. Mech. , C4015003doi:10.1061/ H Shannon entropy (ASCE)EM.1943- 7889.0000992. H relative entropy cs Kuhn, M. R. (2016b). Maximum disorder model for dense steady- H entropy relative to critical state distribution state flow of granular materials. Mech. of Mater. 93, 63–80. cs H entropy relative to uniform distribution Kuhn, M. R. & Bagi, K. (2004). Contact rolling and deformation in L components of velocity gradient ij granular media. Int. J. Solids Struct. 41, No. 21, 5793–5820. N number of particles or objects Kuhn, M. R. & Bagi, K. (2009). Specimen size effect in discrete p mean stress element simulations of granular assemblies. J. Eng. Mech. 135, p probability of discrete No. 6, 485–492. p() probability density of continuous Learned-Miller, E. G. (2004). Hyperspacings and the estimation of information theoretic quantities. Technical Report 04.104, q reference probability distribution cs University of Massachusetts - Amherst. q critical state probability distribution, discrete Luzzani, L. & Coop, M. R. (2002). On the relationship between q uniform probability distribution, discrete cs particle breakage and the critical state of sands. Soils and Found. q () critical state probability distribution, continuous 42, No. 2, 71–82. q () uniform probability distribution, continuous Meléndez, M. & Español, P. (2014). Gibbs-Jaynes entropy versus u uniform relative entropy. Journal of Statistical Physics 155, No. 1, 93–105, R m-dimension space, real numbers doi:10.1007/s10955- 014- 0954- 6, URL http://dx.doi.org/ Z m-dimension space, integers 10.1007/s10955-014-0954-6. " strain in x direction 11 1 Metzger, P. T. (2008). H theorem for contact forces in granular data values materials. Phys. Rev. E 77, No. 1, 011307. Muir Wood, D. & Maeda, K. (2007). Changing grading of soil: effect friction coefficient at particle contacts on critical states. Acta Geotechnica 3, No. 1, 3–14. stress components ij Peña, A. A., García-Rojo, R., Alonso-Marroquín, F. & Herrmann, H. J. phase space of possible  values (2009). Investigation of the critical state in soil mechanics using DEM. In Powders and Grains 2009 (Nakagawa, M. & Luding, S., eds.), Amer. Inst. of Phy., pp. 185–188. Satake, M. (1982). Fabric tensor in granular materials. In Proc. IUTAM Symp. on Deformation and Failure of Granular Materials REFERENCES ( Vermeer, P. A. & Luger, H. J., eds.), Rotterdam: A.A. Balkema, pp. Bagi, K. (1996). Stress and strain in granular assemblies. Mech. of 63–68. Mater. 22, No. 3, 165–177. Satake, M. (1992). A discrete-mechanical approach to granular Been, K. & Jefferies, M. G. (1985). A state parameter for sands. materials. Int. J. Engng. Sci. 30, No. 10, 1525–1533. Géotechnique 35, No. Volume 35, Issue 2, 99–112. Schofield, A. N. & Wroth, P. (1968). Critical state soil mechanics. New Ben-Naim, A. (2008). A farewell to entropy: statistical thermodynam- York: McGraw-Hill. ics based on information. World Scientific. Thornton, C. (2000). Numerical simulations of deviatoric shear Blumenfeld, R. & Edwards, S. F. (2009). On granular stress statistics: deformation of granular media. Géotechnique 50, No. 1, 43–53. Compactivity, angoricity, and some open issues. The Journal of Valanis, K. C., Peters, J. F. & Gill, J. (1993). Configurational entropy, Physical Chemistry B 113, No. 12, 3981–3987. non-associativity and uniqueness in granular media. Acta Boltzmann, L. (2003). Further studies on the thermal equilibrium Mechanica 100, 79–93, doi:10.1007/BF01176863. of gas molecules. In The kinetic theory of gases : an anthology Zhao, J. & Guo, N. (2013). Unique critical state characteristics in of classic papers with historical commentary (Hall, N. S., ed.), granular media considering fabric anisotropy. Géotechnique 63, London: Imperial College Press, pp. 262–349, (English translation No. 8, 695–704. by Brush, S. G.). Casagrande, A. (1936). Characteristics of cohesionless soils affecting the stability of slopes and earth fills. J. Boston Soc. Civil Eng. 23, No. 1, 13–32. Duong, T. (2007). ks: Kernel density estimation and kernel discriminant analysis for multivariate data in R. Journal of Statistical Software 21, No. 7, 1–16. Edwards, S. F. & Oakeshott, R. (1989). Theory of powders. Physica A 157, No. 3, 1080 – 1090. Jaynes, E. T. (1957). Information theory and statistical mechanics. Phys. Rev. 106, No. 4, 620–630, doi:10.1103/PhysRev.108.171. Kuhn, M. R. (1999). Structured deformation in granular materials. Mech. of Mater. 31, No. 6, 407–429. Prepared using GeotechAuth.cls

Journal

Condensed MatterarXiv (Cornell University)

Published: Dec 18, 2018

There are no references for this article.