Mechanical and microscopic properties of the reversible plastic regime in a 2D jammed material

Nathan C. Keim∗ and Paulo E. Arratia† Department of Mechanical Engineering and Applied Mechanics, University of Pennsylvania, Philadelphia, PA 19104 (Dated: August 31, 2018) At the microscopic level, plastic flow of a jammed, disordered material consists of a series of particle rearrangements that cannot be reversed by subsequent deformation. An infinitesimal de- formation of the same material has no rearrangements. Yet between these limits, there may be a self-organized plastic regime with rearrangements, but with no net change upon reversing a defor- mation. We measure the oscillatory response of a jammed interfacial material, and directly observe rearrangements that couple to bulk stress and dissipate energy, but do not always give rise to global irreversibility.

PACS numbers: 83.60.La,63.50.Lm,62.20.F-,05.65.+b

The mechanical properties of disordered (amorphous) than any global timescale of deformation (e.g. a period materials far from equilibrium — from sand, to plas- of driving or the inverse strain rateγ ˙ −1), we can describe tics, to ice cream — continue to elude comprehensive changes to microstructure in terms of discrete, local plas- understanding [1–3]. These materials typically feature tic rearrangements, which are a key feature of the shear many particles (e.g. droplets, atoms, or grains) that are transformation zone (STZ) picture of plasticity [8, 9]. crowded together in close contact, and are both jammed Under steady shear, the piling-on of these events, each of so that each particle is fully constrained by its neighbors, which traverses a barrier between two local minima in po- and disordered so that these constraints vary greatly tential energy, ensures that the initial microstructure can among particles, and crystalline order rarely extends be- never be recovered out of a vast landscape of metastable yond several particle diameters [4]. A sufficiently large states. However, it is believed that individually and in imposed stress may cause these materials to flow plasti- isolation, many if not all plastic rearrangements can be cally as would a viscous liquid, permanently changing the reverted by applying a reverse stress [8–11]. Further- equilibrium arrangement — the microstructure — of the more, when a material is deformed cyclically with suffi- particles. Plastic flow, and the process of yielding that ciently small amplitude, recent simulations and experi- initiates it, are governed by local structural relaxations ments have observed that reversing the deformation may in which one particle squeezes past another, relieving reverse virtually all changes, returning the entire mate- nearby stresses and dissipating energy. These relaxations rial to its original state [12–18]. Viewed stroboscopically and many other behaviors are common to materials on a (once per cycle), the microstructure is static. This poses wide range of length scales and with varying microscopic new questions for a complete description of material re- physics, but the way specific microscopic processes orga- sponse: When rearrangements are stroboscopically invis- nize and give rise to macroscopic behaviors — the mate- ible, do they meaningfully affect bulk rheology? What rial’s bulk rheology — is still not well-understood [5–8]. are their characteristics? Could they clarify the yielding If the timescale of structural relaxation is much shorter transition, when bulk properties change and the material becomes stroboscopically dynamic? [15, 16, 18, 19] Here, we examine in detail the rearrangements in a a) b) Glass-oil-water contact line cyclically-sheared jammed material, in experiments in which it self-organizes to a steady state that is strobo- scopically static [15]. The material is a monolayer of arXiv:1308.6806v2 [cond-mat.soft] 20 Nov 2013 Oil Water repulsive microspheres adsorbed at an oil-water inter- face, for which we simultaneously measure mechanical 4 Needle response (rheology) and image many (5.6 × 10 ) individ- 100 µm ual particles. We find that even when the deformation is globally reversible, local rearrangements are plastic, dis- FIG. 1: (color online) Material and apparatus. (a) View playing hysteresis and altering rheology. The former is from above of bulk material: mutually repulsive polystyrene microspheres adsorbed at oil-water interface. (b) Interfacial a sign that the self-organized steady state is in fact a stress rheometer apparatus. The interfacial material is pinned limit cycle, as found in many other nonlinear dynamical on glass walls; a needle is embedded in the material between systems [16, 17, 20]. This reversible plasticity vanishes them, and is magnetically forced. Velocity profile is sketched. at small strain amplitude, and is gradually overwhelmed by irreversibility as the yielding transition is surpassed. 2

Our findings strongly suggest that microscopic rearrange- ized by the square of the typical interparticle spacing, ments and bulk plasticity are necessary but not sufficient a ' 6.8 µm; details are given in the Supplemental Materi- for irreversibility. als [22]. Figures 2(a,b) illustrate evolution to a reversible Our model material is a mixture (equal parts by steady state in which rearrangements occur, but are al- number) of 4.1 and 5.6 µm-diameter sulfate latex ways reversed by the end of each cycle; movies SM1–3 (polystyrene) particles (Invitrogen; nominal diameters 4 show the full evolution at 3 strain amplitudes [22]. We 2 and 6 µm) adsorbed at a water-decane interface with set a threshold D0 = 0.015, corresponding to a distur- area fraction φ ∼ 0.43. The particles do not touch, but bance ∼ 0.1a ' 1 pixel, and comparable with a value their electrostatic dipole-dipole repulsion [21] results in used for simulations of disordered solids [9]. Most par- 2 −3 a stable, soft (i.e. readily deformable) jammed material ticles in Fig. 2b have Dmin . 10 , while those with 2 2 (Fig. 1a); the particles’ large sizes and strong repulsion Dmin ≥ D0 are in clusters of . 20 particles, with median make thermal motion negligible. This material is sub- size ∼ 5 particles. jected to a linear shear deformation in an interfacial stress We may also measure change to microstructure as the rheometer (ISR) [22–24]. As shown in Fig. 1b, a magne- displacement of a particle relative to the material around tized needle is placed on the material to be studied, in an it (Fig. 2d). The resulting computed streamlines (Fig. 2e) open channel formed by 2 vertical glass walls. An electro- resemble the flow at a hyperbolic point in an incompress- magnet forces the needle, creating a uniform shear stress ible fluid, consistent with the geometry of a single plastic σ(t) on the material between the needle and the walls. event measured in sheared dry foams by Kabla and De- We measure material rheology by observing the needle’s bregas [25], and modeled by Picard et al. [26] for an other- motion [expressed as strain γ(t)] under oscillatory stress. wise elastic incompressible medium. Finally, rearranging Deformation in experiments is quasistatic and rear- particles lose and gain nearest neighbors, a process dis- rangements are discrete, insofar as the timescale for a cretized as T1 events [27] in Fig. 2f, and Movie SM4 [22]. rearrangement to complete (∼ 0.5 s) is much shorter Details of these computations are in the Supplemental than the shortest driving period (5 s) or largest inverse Materials [22]. strain rate (γ ˙ −1 = 20 s). We also require that the We find that the locations of rearrangements are not boundary conditions in the 3rd dimension be approxi- predicted by static material structure, such as local num- mately stress-free — that typical forces in the plane of ber density, presence of anomalously large or small parti- the material are much stronger than viscous drag from cles, or number of neighbors. However, we do see a differ- the liquid bath [24]. This ratio is the Boussinesq num- ence between more- and less-ordered regions. The bond ∗ ∗ ber Bq = |η |a/ηl, where η is the material’s observed order parameter magnitude |ψ6| measures the degree to complex viscosity, a = 230 µm is the needle diameter, which each particle’s neighbors are spaced 60◦ apart (de- −3 and ηl ' 10 Pa s is the oil and water viscosity. Here tails in Supplemental Materials [22]); Fig. 3 shows that Bq ∼ 102 and so our experiments are nearly 2D. Further the material has of regions of crystalline order with scale details of the material and apparatus are found in the ∼ 5a, and thick interstitial “grain boundaries.” Parti- Supplemental Materials [22]. For each experiment, we cles involved in plasticity are disproportionately in the prepare the material with 6 cycles of shearing at large latter, strongly suggesting that the material’s response is amplitude (γ0 ∼ 0.5), then stop. Resuming at smaller γ0 dominated by disorder. starts a transient relaxation to a steady state. We now verify that these rearrangements are a form of At each cycle of driving during the experiment, we plasticity, dissipating energy and coupling to bulk stress; can measure total (peak-to-peak) change in microstruc- this does not necessarily follow from non-affine deforma- ture by comparing particle positions at a minimum of tion alone [4, 28]. A plastic rearrangement is caused by global strain γ(tmin) with those at the following maxi- a local buildup of (elastic) stress as the whole material mum γ(tmax). Irreversible change is measured strobo- is sheared; an opposing buildup is required to reverse −1 scopically, by sampling at times (tmax +tmin ±2πω )/2, it. Such events appear hysteretic, turning “on” during so that we compare the beginning and end of a full pe- forward shear at a global strain γon, and “off” during re- riod of driving that straddles tmin and tmax. Wherever verse shear at γoff, with γon − γoff > 0 as a proxy for the there is no irreversible change to microstructure, any to- activating stress. tal change in that same cycle is by definition reversible. Figure 4a shows hysteresis in a single cycle, using Figure 2 shows changes to microstructure in single cy- γ(t) = γmin as the undeformed state. Using the thresh- 2 cles of deformation, for the entire system and for a single old D0 = 0.015, we obtain a γon at the last video frame 2 2 region. Panels (a,b,c) detect rearrangements with the for which a particle’s Dmin < D0, and γoff at the last 2 2 2 quantity Dmin, computed between 2 instants by measur- frame with Dmin ≥ D0. We require “on” and “off” to ing how much each particle and its 2 nearest “shells” be in the first and second halves of the cycle respec- 2 2 of neighbors move unlike a continuous elastic solid; it is tively, and Dmin ≥ D0 for at least 50% of the intervening the mean squared residual displacement after subtract- frames (for most events this approaches 100%). At the 2 ing the best affine transformation [9]. Dmin is normal- extreme, some rearrangements activate at ∼ γmax but re- 3

a) b) 0.4 0.1 Reversible Irreversible

0.0 0.0 200 µm

c) d) e) f) g)

1 4

3 2

h)

20 µm

FIG. 2: (color online) Local deformation in a plastic event. (a,b) Total (i.e. peak-to-peak, blue) and irreversible (i.e. stro- 2 boscopic, overlaid in red) Dmin for cycles (a) 8 and (b) 20 of shear at γ0 = 0.020, showing clusters of non-affine deformation. One reversible cluster in (b) is boxed, and shown in (c–f). The magnetic needle is at the top of the image; the fixed wall is 2 at the bottom. (c) Detail of a reversible cluster, showing the Dmin of individual particles. Color scale is the same as in (a,b). (d) Local relative displacement of particles in (c) at the minimum (red) and subsequent maximum (blue open circles) of γ, subtracting motion of neighbors within 10a. (e) Streamlines computed from displacements (subtracting motion of neighbors within 40a); square outline is region of (c,d,f). The hyperbolic character of the displacements is evident in the far-field. (f) Micrograph with particle centers (small dots) and the centroid of the 4 particles in each T1 rearrangement (large dots) marked. (g,h) Sequence illustrating a T1 rearrangement. Particles 3 and 4 begin as nearest neighbors but are separated in (h).

as also seen in the looped trajectories of Fig. 4c, and it locally makes strain a multiple-valued function of stress. These behaviors are inconsistent with purely elastic de- formation and consistent with plasticity as described by STZ theory [8, 9, 11].

5 We can now connect our simultaneous observations of |) 6 rheology and microscopic behavior in the steady state. ψ 2 2 P(| Numbers of rearranging particles (Dmin ≥ D0) averaged 0 over the final 3 cycles of each movie at various γ0 are 0 0.5 1.0 plotted in Fig. 4d; behavior changes little over at least 4 |ψ | 100 µm 6 cycles. To measure rheology in Fig. 4e, we model stress as the real part of (G0 + iG00)γ, with γ = eiωt, where FIG. 3: (color online) Rearrangements in portion of material ω is the angular frequency of driving; this gives a stor- at γ0 = 0.02 during transient (cycle 15; chosen to obtain 0 more events than in steady state). Each particle is shown as age modulus G , measuring elastic character, and loss 00 solid dot with size representing the extent of local crystalline modulus G , measuring viscous or plastic character. As ordering |ψ6|; minimum and maximum size signify |ψ6| ' 0.1 discussed above, γon − γoff is a proxy for the local stress and 1. Color is solely to show differences in lattice director. σpl causing the rearrangement, and for its contribution to The centroids of total T1 events (see Fig. 2f) are shown as dissipation (i.e. to G00). Using the relation for dissipation large open circles. Inset: Histograms of |ψ6| in reversible 2 00 per unit area per cycle, wcyc = πγ0 G , and the data in steady state (cycle 20). Curve: all particles. Shaded bars: Fig. 4e, we can estimate the plastic contribution to G00 in particles involved in T1 events (555 out of 5.6 × 104). Dot a reversible or mostly-reversible steady state (γ0 ≤ 0.04), positions and ψ6 are for γ ' hγi. 2 X G00 = G0a2(γi − γi )2 (1) pl πγ2A on off 0 i verse at ∼ γmin. Figure 4b shows that hysteretic plastic- ity grows dramatically in abundance and strength as γ0 where A is the area of observations, 2 refers to each par- is increased. Hysteresis breaks time-reversal symmetry, ticle switching twice per cycle, and the sum estimates 4

some particles rearrange plastically during deformation. a) 0.02 This regime is due to a stable population of rearrange- ments, comprising just ∼1% of particles, suggesting that 3 0.01 to reliably observe it, &10 particles must be studied. It is reminiscent of a limit cycle, a closed trajectory ⟩ γ

⟨ in phase space that a nonlinear system may evolve to-

– 0 ward [20], and which describes simulations of cyclically- off γ sheared athermal frictionless jammed [16] and unjammed −0.01 γ0 = 0.01 particles [17]. Limit cycles break time-reversal symme- try, as seen in the looped trajectories of Fig. 4c, and

−0.02 γ0 = 0.02 so are much more general than the linear dynamics of the reversible steady state in dilute non-Brownian sus- pensions [29]. Our finding of limit cycles may depend −0.02 −0.01 0 0.01 0.02 γon – ⟨γ⟩ weakly on the duration of the experiment, in that ther- N b) NA Npl mal or mechanical noise could cause sporadic further re- 104 d) 300 103 laxations [30, 31]. 102 200 10 Considering the results discussed in this work, both in 100 G', G'' e) our experiments and published elsewhere [11, 14–19], we (N/m) 0 see 3 regimes of steady-state cyclic deformation: (1) Far 0 0.05 G' c) γon – γoff 10−7 below yielding (γ0  γy), response is truly elastic and G'' time-reversible, with no rearrangements. Nonetheless, some particle motions may be non-affine due to disor- der [4, 28]. (2) As γ → γ , microscopic plasticity grows 10−8 0 y 0.001 0.01 0.1 rapidly. Rheological response is still dominated by elas- 10 µm 0 00 γ0 ticity (G  G ), and the material is stroboscopically static [14–18], but time-reversibility is broken [16, 17]. 00 FIG. 4: (color online) Hysteresis of rearrangements. (a) Plasticity contributes to G but may not dominate. (3) micro Global strain γ at which individual particles rearrange γ0 = γy marks the appearance of irreversible plastic- 2 (Dmin ≥ 0.015) and reverse, in steady state for γ0 = 0.01 ity in the steady state and is a clearly-defined yielding (+) and γ0 = 0.02 (◦). Shaded regions show limits of γ. The transition [15, 16, 18, 19]. Much of the system may be farther an event falls from the diagonal line, the more hys- nonetheless reversible in a given cycle (see Fig. 4d or teretic it is. (b) Histogram of hysteresis at γ0 = 0.04 (open, Movie SM3) [11, 22, 32]. On the other hand, the rheo- black curve), 0.02 (shaded, red), 0.01 (solid, blue). (c) Parti- 00 cle trajectories break time-reversal symmetry. A rearranging logical yielding transition, wherein G increases and elas- portion of the system is shown for one cycle in the steady ticity declines, is gradual; at the microscopic level it is state (γ0 = 0.02). Colors distinguish the particles. The effect due to both reversible and irreversible plasticity. of microscope vibration is reduced by subtracting average y (here, vertical) motion of entire visible system. (d) Num- Our work shows that in an experimental jammed ma- ber of particles Npl in plastic rearrangements in steady state, 4 terial, plasticity and irreversibility can become decoupled as function of γ0. Points show number (out of 5.6 × 10 ) in the steady-state oscillatory response: the material can with total (+) or irreversible (•) D2 ≥ 0.015. Hysteretic min host many microscopic plastic rearrangements that cou- particles (◦) have γon − γoff exceeding the largest change in ple to the bulk stress and dissipate energy, yet do not γ between video frames. (e) Oscillatory rheology. N: Esti- mated enhancement of G00 above zero-plasticity level (dashed give rise to global irreversibility. This strongly suggests line), based on microstructure (see text). a qualitative difference between microstructural yielding (the transition to irreversibility) and rheological yielding: rearrangements and bulk plasticity are necessary but not the elastic energy built up and then dissipated, for each sufficient for irreversibility. Differences between the re- 2 particle in Fig. 4b. This estimate, made by choosing D0 stricted, self-organized STZ-like rearrangements of the only, is shown in Fig. 4e. It is of the same order as the reversible steady state, and a more general population 00 actual increase in G at γ0 = 0.04. under steady shear, may shed light on models of STZ Using simultaneous bulk rheometry and particle track- populations [8], or other measures of static and dynami- ing under shear, we have studied the nature and mechan- cal structure [2, 7, 33]. ical role of microscopic plastic events in a soft jammed material. This material can evolve to a steady state in We thank John Brady, Andrea Liu, Martin van Hecke, which mechanical response is primarily elastic and mi- and Ye Xu for helpful discussions. This work was sup- crostructure is unchanged by each cycle [15], and yet ported by the Penn NSF MRSEC (DMR-1120901). 5

O’Hern (2013), arXiv:1301.7492v1. [18] D. Fiocco, G. Foffi, and S. Sastry, Phys. Rev. E 88, 020301 (2013). ∗ Electronic address: [email protected] [19] G. Petekidis, A. Moussa¨ıd,and P. N. Pusey, Phys. Rev. † Electronic address: [email protected] E 66, 051402 (2002). [1] R. G. Larson, The Structure and Rheology of Complex [20] S. H. Strogatz, Nonlinear dynamics and Chaos: with ap- Fluids (Oxford, 1998). plications to physics, biology, chemistry, and engineering [2] M. Chen, Annu. Rev. Mater. Res. 38, 445 (2008). (Westview, 1994). [3] D. T. N. Chen, Q. Wen, P. A. Janmey, J. C. Crocker, [21] K. Masschaele, B. J. Park, E. M. Furst, J. Fransaer, and and A. G. Yodh, Annu. Rev. Condens. Matter Phys. 1, J. Vermant, Phys. Rev. Lett. 105, 048303 (2010). 301 (2010). [22] See Supplemental Material at ??? for movies, details of [4] M. van Hecke, J. Phys: Cond. Matter 22, 3101 (2010). methods and additional material characterization. [5] A. D. Gopal and D. J. Durian, Phys. Rev. Lett. 75, 2610 [23] C. F. Brooks, G. G. Fuller, C. W. Frank, and C. R. (1995). Robertson, Langmuir 15, 2450 (1999). [6] P. Schall, D. A. Weitz, and F. Spaepen, Science 318, [24] S. Reynaert, C. F. Brooks, P. Moldenaers, J. Vermant, 1895 (2007). and G. G. Fuller, J. Rheol. 52, 261 (2008). [7] M. L. Manning and A. J. Liu, Phys. Rev. Lett. 107, [25] A. Kabla and G. Debr´egeas,Phys. Rev. Lett. 90, 258303 108302 (2011). (2003). [8] M. L. Falk and J. S. Langer, Annu. Rev. Condens. Matter [26] G. Picard, A. Ajdari, F. Lequeux, and L. Bocquet, Eur. Phys. 2, 353 (2011). Phys. J. E 15, 371 (2004). [9] M. L. Falk and J. S. Langer, Phys. Rev. E 57, 7192 [27] D. Weaire and N. Rivier, Contemp. Phys. 25, 59 (1984). (1998). [28] A. Zaccone and E. Scossa-Romano, Phys. Rev. B 83, [10] A. S. Argon, Acta Metallurgica 27, 47 (1979). 184205 (2011). [11] M. Lundberg, K. Krishan, N. Xu, C. S. O’Hern, and [29] L. Cort´e,P. M. Chaikin, J. P. Gollub, and D. J. Pine, M. Dennin, Phys. Rev. E 77, 041505 (2008). Nat. Phys. 4, 420 (2008). [12] S. Slotterback, M. Mailman, K. Ronaszegi, M. van Hecke, [30] P. Richard, M. Nicodemi, R. Delannay, P. Ribi`ere,and M. Girvan, and W. Losert, Phys. Rev. E 85, 021309 D. Bideau, Nat. Mater. 4, 121 (2005). (2012). [31] V. B. Nguyen, T. Darnige, A. Bruand, and E. Clement, [13] J. Ren, J. A. Dijksman, and R. P. Behringer, Phys. Rev. Phys. Rev. Lett. 107, 138303 (2011). Lett. 110, 018302 (2013). [32] P. H´ebraud,F. Lequeux, J.-P. Munch, and D. J. Pine, [14] N. V. Priezjev, Phys. Rev. E 87, 052302 (2013). Phys. Rev. Lett. 78, 4657 (1997). [15] N. C. Keim and P. E. Arratia, Soft Matter 9, 6222 (2013). [33] K. Chen, M. L. Manning, P. J. Yunker, W. G. Ellenbroek, [16] I. Regev, T. Lookman, and C. Reichhardt (2013), Z. Zhang, A. J. Liu, and A. G. Yodh, Phys. Rev. Lett. arXiv:1301.7479. 107, 108301 (2011). [17] C. F. Schreck, R. S. Hoy, M. D. Shattuck, and C. S.