Time dependence in large amplitude oscillatory shear: a rheo-ultrasonic study of fatigue dynamics in a colloidal gel
Christophe Perge, Nicolas Taberlet, Thomas Gibaud, Sebastien Manneville
aa r X i v : . [ c ond - m a t . s o f t ] J a n Time dependence in large amplitude oscillatory shear: arheo-ultrasonic study of fatigue dynamics in a colloidal gel
Christophe Perge, Nicolas Taberlet,
1, 2
Thomas Gibaud, and S´ebastien Manneville Laboratoire de Physique, Universit´e de Lyon - ´Ecole Normale Sup´erieurede Lyon - CNRS UMR 5672, 46 All´ee d’Italie, 69364 Lyon cedex 07,France UFR de Physique, Universit´e Claude Bernard Lyon 1,Universit´e de Lyon, 43 boulevard du 11 novembre, 69100 Villeurbanne,France. (Dated: 8 October 2018)
We report on the response of a yield stress material, namely a colloidal gelmade of attractive carbon black particles, submitted to large amplitude oscil-latory shear stress (LAOStress). At a constant stress amplitude well below itsapparent yield stress, the gel displays fatigue and progressively turns from anelastic solid to a viscous fluid. The time-resolved analysis of the strain response,of the Fourier spectrum, and of Lissajous plots allows one to define two differ-ent timescales τ w < τ f associated with the yielding and fluidization of the gel.Coupling rheology to ultrasonic imaging further leads to a local picture of theLAOStress response in which the gel first fails at the walls at τ w and then under-goes a slow heterogeneous fluidization involving solid–fluid coexistence until thewhole sample is fluid at τ f . Spatial heterogeneities are observed in both the gra-dient and vorticity directions and suggest a fragmentation of the initially solidlikegel into macroscopic domains eroded by the surrounding fluidized suspension. I. INTRODUCTION
Among the huge variety of soft materials, dispersions of colloidal particles are in-volved in a broad range of applications such as paints, food stuff, printing, etc. [Russel(1987); Mezzenga et al. (2005); Gibaud et al. (2012)]. On a more fundamental side,colloidal systems also appear as model systems to study both glass and sol–gel tran-sitions [Sciortino and Tartaglia (2005); Cardinaux et al. (2007); Lu et al. (2008)]. Inparticular, concentrated colloidal dispersions can be categorized as repulsive or attractiveglasses depending on the interaction potential between the particles [Pham et al. (2002);Tanaka et al. (2004)]. For sufficiently attractive particles, low volume fractions may beenough for clusters to form and percolate through the sample, leading to a “gel” struc-ture [Lu et al. (2008)]. Due to the jammed microstructure in concentrated colloids andto the space-spanning microstructure in colloidal gels, these systems all display elasticproperties at rest: the storage modulus G ′ is always much larger than the loss modulus G ′′ when measured close to equilibrium, e.g. through oscillatory shear of small ampli-tude (typically a strain amplitude of less than 1% or a stress amplitude of less than 1 Pa)[Barnes et al. (1989); Larson (1999)].Besides the equilibrium phase behavior of colloidal systems and their (linear) me-chanical properties at rest, their (nonlinear) response to large deformations is also ofoutstanding interest [Buscall et al. (1993); Larson (1999); Pham et al. (2006, 2008);Brader et al. (2010); Koumakis et al. (2013)]. Indeed, under large deformations, col-loidal particles may rearrange by changing neighbors, clusters may break into smalleraggregates, so that the system is able to flow like a liquid, i.e. G ′′ becomes much largerthan G ′ . Although it is virtually involved in any practical application of colloidal sys-tems, this yielding transition (or “unjamming” transition) still raises many theoreticaland experimental challenges [Moorcroft and Fielding (2013)]. At the heart of these dif-ficulties is the issue of the time dependence of the microstructure. Obviously, the timeneeded to “unjam” and fluidize a colloidal system lying in a kinetically arrested state(and more generally any yield stress material) can be anticipated to become very largedepending of the strain amplitude and on the protocol details. On the other hand, whenstrain or stress is decreased on the fluidized colloidal dispersion, it always takes sometime for the microstructure to reform. This leads to so-called “thixotropic” or “rheopec-tic” features [Mewis and Wagner (2009); Møller et al. (2009)] and can be captured bymeasuring the timescales involved in step-up or step-down stress or strain relaxationexperiments [Dullaert and Mewis (2005, 2006)] or through hysteretic loops in up-downflow curves [Chen et al. (1992); Divoux et al. (2013); Ovarlez et al. (2013)]. Even moregenerally, increasingly long timescales are expected for flow dynamics upon approachingthe yield point, which can be thought of as the critical point in a dynamical phase tran-sition [Bocquet et al. (2009); Divoux et al. (2012)]. Such time dependence has recentlybeen illustrated through the existence of diverging timescales under steady shear closeto yielding, not only in colloidal gels [Gopalakrishnan and Zukoski (2007); Gibaud et al.(2010); Sprakel et al. (2011); Lindstr¨om et al. (2012); Grenard et al. (2013)] but also instandard yield stress fluids such as microgels [Uhlherr et al. (2005); Divoux et al. (2010,2011, 2012)] and soft thermo-reversible protein gels [Brenner et al. (2013)].A widespread way to probe yielding in soft materials is to follow the evolution of G ′ and G ′′ as the oscillation amplitude of the stress or strain is progressively increased abovelinear response. This generally allows one to estimate the degree of elastic vs viscousresponse and to infer yield criteria, e.g. from the crossover G ′ = G ′′ in yield stress ma-terials [Petekidis et al. (2002); Pham et al. (2006); Smith et al. (2007); Renou et al.(2010); Datta et al. (2011); Koumakis et al. (2012); Shao et al. (2013); Dimitriou et al.(2013)]. For instance, this approach has been used to unveil a major difference betweenattractive and repulsive colloidal glasses: while the elastic modulus decreases monoton-ically in dense hard sphere-like systems, a two-step yielding is observed in attractiveglasses, which has been related to the destruction of the “cages” made by neighboringattractive particles around a given colloid [Pham et al. (2006); Laurati et al. (2011)].In the last two decades, rheological measurements based on oscillatory shear within thenonlinear regime have made tremendous progress thanks to the development of new toolsin Fourier transform (FT) rheology [Wilhelm (2002)] and to other refined mathematicaltools [Cho (2005); Klein et al. (2007); Ewoldt et al. (2008)]. FT rheology allows oneto go beyond the measurement of the material response at the fundamental oscillationfrequency f (characterized by the storage and loss moduli G ′ and G ′′ extended into thenonlinear regime and usually noted G ′ and G ′′ ) by analyzing higher harmonics in thelarge amplitude oscillatory stress or strain signals [Wilhelm (2002)]. Yet the harmoniccontent lacks a clear physical interpretation. To overcome this drawback of FT rheol-ogy, various approaches based on different decompositions of the nonlinear signal havebeen proposed which provide valuable physical information on the material behavior dur-ing one oscillation cycle [Cho (2005); Klein et al. (2007); Ewoldt et al. (2008)]. Amongthem, the orthogonal decomposition in Chebyshev polynomials proposed by Ewoldt et al.(2008) has the advantages of being unique and directly connected to Fourier coefficients,and of providing measures with a simple physical interpretation in terms of strain hard-ening, strain softening, etc. For more details, a recent enlightening review on such a largeamplitude oscillatory shear (LAOS) rheology can be found in Hyun et al. (2011).In view of the importance of LAOS rheology both for practical applications and forfundamental problems, it is not surprising that this technique, also recently coined“LAOStrain” or “LAOStress” in order to distinguish between strain-controlled or stress-controlled oscillations [La¨uger and Stettin (2010); Dimitriou et al. (2013)], has generatedan ever-increasing amount of literature since the early 1990s. Much more unexpectedis the fact that only a handful of papers are concerned with the possible effect of timedependence on LAOS measurements. For instance, the academic search engine Webof Knowledge returns about 290 and 200 results since 1990 for searches based on thekeywords “large amplitude oscillatory shear” and “Fourier-transform rheology” respec-tively, but less than five results are found when either of these keywords is associatedwith “time dependence” or “thixotropy.” With the notable exceptions of Li et al. (2009)in the context of polymers, of Rogers et al. (2011) in the context of aging glassy starpolymers, and of Dimitriou (2013) in the context of waxy crude oils, it appears that avast majority of works focusing on LAOS experiments take for granted that oscillationsare performed in a “steady-state”, in the sense that the material properties do not sig-nificantly change from one oscillation cycle to another and somehow adapt “instantly”to any change in the oscillations. At the very least, this requires that the timescalescharacteristic of microstructural changes remain much shorter than 1 /f and than anyother timescale involved in the LAOS protocol, such as the inverse of the sweep ratewhen the oscillation amplitude is varied through a continuous ramp or the time intervalbetween steps of constant amplitude, which may indeed become the fastest timescale atlow oscillation frequencies. Unfortunately, as recalled above, yielding often goes alongwith slow dynamics. If the material microstructure evolves on timescales longer thanthose of the LAOS protocol, these dynamics may somewhat complicate the analysis ofthe LAOS results and even question the validity of their interpretation.The purpose of the present paper is twofold. It first aims at attracting the attention ofthe rheology community on the potential influence of time-dependent microstructure onLAOS experiments performed in attractive colloidal systems and more generally in yieldstress materials. Second, we show that crucial complementary information on yieldingunder oscillatory shear can be gained by combining LAOS with ultrasonic imaging. Wemainly focus on time-resolved LAOS experiments under a fixed oscillation amplitude ina colloidal gel known to present “delayed yielding”, namely a carbon black gel. This cor-responds to the limiting case where the material evolves very slowly compared to 1 /f during most of the experiment. The paper is organized as follows. Section II describesour sample and experimental setup. We also briefly recall the LAOStress protocol andanalysis as well as the potential problems inherent to LAOStress. Our results are gath-ered in Sect. III. We first use the LAOStress measures as described in La¨uger and Stettin(2010) and Dimitriou et al. (2013) to characterize the temporal evolution of the oscilla-tory response of our colloidal gel. We then report on local ultrasonic measurementswhich allow us to distinguish between two successive steps in the yielding process of ourattractive colloidal system and to interpret these steps in terms of yielding at the wallsand heterogeneous bulk fluidization. Finally, Sect. IV provides a short discussion of theseresults as well as a few perspectives for future work. II. MATERIAL AND METHODSA. Carbon black gels
Carbon black (CB) particles are colloidal carbonated particles with typical sizes rang-ing from 80 to 500 nm that result from the partial combustion of hydrocarbon oils[van der Waarden (1950); Samson et al. (1987)]. Those particles are widely used in theindustry, for instance to reinforce mechanical properties or insure the conductivity of plas-tic and rubber materials [Donnet et al. (1993)]. When dispersed in oil, these carbon blackparticles are weakly attractive, with interactions of typical strength U ∼ k B T , and ag-gregate into a space-spanning fractal network [Trappe and Weitz (2000); Trappe et al.(2007)]. These colloidal dispersions therefore form gels at low concentrations as describedabove in the introduction.Here, we focus on carbon black particles (Cabot Vulcan XC72R, density 1.8) dispersedat 6% wt. in a mineral oil (density 0.838, viscosity 20 mPa.s, Sigma Aldrich). Inorder to use ultrasonic imaging in this optically opaque material, the gel is seeded with1% wt. hollow glass microspheres (Potters Sphericel, mean diameter 6 µ m, mean density1.1 g.cm − ) as already described in Gibaud et al. (2010) and in Grenard et al. (2013).Hollow glass microspheres act as ultrasonic contrast agents in an otherwise acousticallytransparent dispersion, without altering significantly its rheological properties. Thissystem also has the great practical advantage of being chemically stable and insensitiveto evaporation: the same gel can be studied for several weeks provided an adequate pre-shear protocol is used to resuspend the seeding microspheres and rejuvenate the gel priorto each experiment.When a constant stress is applied from rest in a standard creep experiment, CB gelswere shown to display “delayed yielding”: these gels [Gibaud et al. (2010); Grenard et al.(2013)] as well as other colloidal gels [Gopalakrishnan and Zukoski (2007); Sprakel et al.(2011); Lindstr¨om et al. (2012)] first seem to remain solid before undergoing a rathersudden fluidization after a well-defined time τ f . This strongly time-dependent behaviorwas modeled in the framework of activated dynamics, predicting quasi-exponential decayof the fluidization time τ f with the applied stress σ [Gopalakrishnan and Zukoski (2007);Sprakel et al. (2011); Lindstr¨om et al. (2012)].Besides time dependence, Gibaud et al. (2010) have inferred from one-dimensionalultrasonic imaging that the yielding process of CB gels involves wall slip and spatiallyheterogeneous fluidization. Very recent investigations have also shown that this delayedyielding is affected by boundary conditions. Indeed, using rough boundaries instead ofsmooth walls introduces an additional timescale associated with an initial power-law creepregime that ends with the gel failure at the moving wall and whose duration diverges as apower-law as the applied shear stress approaches the yield stress [Grenard et al. (2013)].In the present paper, we shall use LAOS data analysis together with a newly developedtwo-dimensional ultrasonic imaging technique in order to complement the LAOStressexperiments already briefly reported in Gibaud et al. (2010). B. LAOStress protocol
Our rheological measurements are performed in a smooth Couette cell of height 50 mmwith a rotating inner bob of radius 48 mm made out of PEEK, and a fixed outer cupof radius 50 mm made out of PMMA. The bob is a hollow cylinder in order to reduceits inertia. This Couette cell is mounted on a stress-imposed rheometer (ARG2, TAInstruments). To ensure an initial reproducible gel state, each measurement is carriedout by first pre-shearing the suspension at 500 s − then at −
500 s − for 20 s each inorder to break up any large aggregate. Second, we let the gel rest so that it can reformby applying a zero shear rate for 2 s and a zero shear stress for 10 s. Third, we monitorthe linear viscoelastic properties of the gel by measuring the elastic modulus G ′ and lossmodulus G ′′ through oscillatory shear at a low stress amplitude of 0.5 Pa and frequency f = 1 Hz for 60 s. A steady state is reached within a few seconds and we find typicalelastic and loss moduli G ′ ≃ . Pa and G ′′ ≃
150 Pa (see Fig. 2). The sample is thenleft to rest for 10 s by applying a zero shear stress. Finally, we start the large amplitudeoscillatory shear stress experiment at time t = 0 by imposing oscillations of constantfrequency f over time. The temperature is controlled by a water circulation around theCouette cell and fixed to 25 ± . ◦ C for all experiments. All the data presented here wereobtained with the same oscillation frequency f = 1 Hz. The amplitude of the stress-controlled oscillations is noted σ . In Sect. III A below, we shall only briefly report on anexperiment where σ is swept from small to large values in order to highlight how timedependence arises in a standard ramp protocol. All other experiments will be conductedover time at a fixed σ and we shall mainly focus on a single specific experiment at σ = 11 Pa. C. Potential problems with LAOStress
It is important to keep in mind that in a time-dependent material, a number of ex-perimental problems may arise when performing LAOStress experiments even with asingle-head stress-controlled rheometer [La¨uger and Stettin (2010)]. Indeed, the rheome-ter uses a feedback loop to account for the instrument and bob inertia in order to ensurethat the stress applied on the sample is sinusoidal with the prescribed amplitude. Whenthe material structure evolves significantly over a small number of successive oscillations,as can be the case for stress-induced fluidization, the feedback may not be efficient enoughfor the amplitude to match the commanded value and for stress oscillations to remainsinusoidal, especially for large amplitudes and/or low frequencies and if the feedback in-volves several full oscillation cycles. The most common problem is that when the samplefluidizes, the stress due to inertia (referred to as “inertia” stress) increases sharply whilethe rheometer keeps the same total stress (also referred to as “raw” stress), so that the ac-tual stress felt by the sample (referred to as “sample” stress) falls below the commandedvalue. However, most recent rheometers (including the one used in the present study)use a “direct strain oscillation” (DSO) procedure to correct for inertia in real-time withineach oscillation cycle, which usually minimizes deviations from the required sinusoidalstress [La¨uger and Stettin (2010)]. Another issue arises from the fact that our specificrheometer uses a small additional position control superimposed to the stress control toavoid any drift in the mean zero position. This corresponds to a “soft-lock” torque whichreduces the achieved torque. This torque reduction is usually negligible for high samplestiffnesses but may significantly affect the stress signal as the stiffness of the sample goesdown due to progressive fluidization.More generally, even for non-thixotropic, time-independent materials in a true steadystate, inertia may seriously complicate the analysis of LAOStress data [Dimitriou et al.(2013)]. In particular, in its standard configuration, our rheometer software providesaccess only to the total stress waveform. In other words, in the analysis of the time-resolved strain and stress data presented below, the stress signal σ ( t ) = σ total ( t ) containsboth the sample stress σ sample ( t ) and the contribution from the geometry and instrumentinertia σ inertia ( t ). Assuming that the strain response γ ( t ) is sinusoidal with amplitude γ , the amplitude of the oscillatory inertia stress for our small-gap Couette geometry is: σ inertia , = e πHR ω Iγ , (1)where e = 2 mm is the gap width, R = 48 mm is the bob radius, H = 50 mm is theheight of the cell, ω = 2 πf , and I is the total effective moment of inertia includingthat of the instrument and of the geometry [see also the discussion in Appendix B ofDimitriou et al. (2013) for the case of a cone-plate geometry]. Therefore, for a com-manded stress amplitude σ , the contribution of inertia can be neglected as long as: γ ≪ πHR eω I σ . In the case of the present experiments, the total inertia is measured tobe I = 301 µ Nms − , where the main contribution comes from our large custom-madebob ( I bob = 282 µ Nms − ). For f = 1 Hz and σ = 11 Pa, the above condition reads γ ≪ γ ⋆ ≃
16. Although this value should only be taken as a rough estimate (sincethe strain is no longer sinusoidal much below γ ⋆ ), we can confidently neglect inertia andidentify the total stress with the sample stress as long as γ .
1. As will be shown belowin Sect. III B, this corresponds to the main part of the experiment, where fluidizationremains spatially limited. We have also checked that the stress signal remains sinusoidaland that its amplitude matches the commanded stress to within better than 5% over thesame period of time (see Sect. III B 2 below).Finally, knowing the strain waveform γ ( t ), the sample stress can in principle be recov-ered from the total stress by differentiating γ ( t ) twice and by using: σ sample ( t ) = σ total ( t ) − σ inertia ( t ) = σ total ( t ) − e πHR I ¨ γ ( t ) . (2)However, in the present work, we chose to stick with the rheometer measurement σ total ( t )for the following reasons: (i) taking the second derivative of γ ( t ) strongly enhances ex-perimental noise, (ii) LAOStress measures are computed from γ ( t ) (see Sect. II D below)and are insensitive to the correction in Eq. (2), and (iii) even more importantly, we shallshow through local measurements that the strain field becomes spatially heterogeneousas fluidization proceeds and the contribution of inertia accordingly increases. In theevent of spatial heterogeneity, the LAOStress framework gets clearly questionable andthe corresponding measures become hardly interpretable. D. LAOStress data analysis1. Fourier transform rheology
Besides the usual measurements at the fundamental oscillation frequency that aredirectly provided by the rheometer ( G ′ = G ′ , G ′′ = G ′′ , and the phase shift δ = δ between the input stress and the strain response at ω ), we use standard tools fromFourier transform rheology to compute the full complex Fourier spectrum J ⋆ ( ω ) of thestrain response γ ( t ) and the discrete Fourier components J ⋆n = J ⋆ ( nω ), whose real andimaginary parts will be noted respectively J ′ n and J ′′ n , with n ∈ N .As we shall check below in Fig. 4, odd harmonics strongly dominate in the strainresponse. Therefore, following Dimitriou et al. (2013) and using the adequate normal-ization for the Fourier coefficients, γ ( t ) can be written as: γ ( t ) = σ X n odd { J ′ n cos( nω t ) + J ′′ n sin( nω t ) } , (3)where σ ( t ) = σ cos( ω t ) is the input commanded stress.As a crude way to quantify nonlinearity, we define the total harmonic distortion (THD)of the strain signal as: THD( γ ) = s X n odd ≥ | J ⋆n | | J ⋆ | . (4)In order to test for the presence of harmonics in the input stress signal, we will similarlydefine a total harmonic distortion THD( σ ) based on the Fourier coefficients of σ ( t ) (seeFig. 5).Since we are interested in time dependence, the Fourier transform is performed in atime-resolved fashion by analyzing successive waveforms recorded over one period by therheometer during long experiments where the oscillation amplitude and frequency arekept constant. As expected in a time-dependent material, the Fourier coefficients willthemselves depend slowly on time, so that they should in principle be noted J ′ n ( t ) and J ′′ n ( t ), where t is the slow timescale at which the Fourier transform is performed. Inthe following however, we shall use the same symbol t for denoting the time within oneoscillation cycle and over the whole experiment. The same comment holds for the THD.
2. LAOStress measures
LAOStress data are further analyzed as described in La¨uger and Stettin (2010) andDimitriou et al. (2013). Here, we only briefly recall the definitions of the variousLAOStress measures based on the rheological measurements and their interpretations.First, from the compliance coefficients J ′ n , two different elastic measures are introduced:(i) the zero-stress elastic compliance J ′ M , defined as: J ′ M ≡ d γ d σ (cid:12)(cid:12)(cid:12)(cid:12) σ =0 = X n odd ( − ( n − / nJ ′ n = J ′ − J ′ + 5 J ′ − J ′ + · · · , (5)and (ii) the large-stress elastic compliance J ′ L , defined as: J ′ L ≡ γσ (cid:12)(cid:12)(cid:12) σ = σ = X n odd J ′ n = J ′ + J ′ + J ′ + J ′ + · · · . (6)These two compliances can be geometrically interpreted by considering the classical Lis-sajous plot of the strain response γ ( t ) vs the input shear stress σ ( t ) : J ′ M and J ′ L respec-tively correspond to the slope of the tangent to the Lissajous plot at σ = 0 and to theslope of the line joining both extremities of the cycle at σ = ± σ . This is illustrated inFig. 6(a,c,e).Similarly, one introduces two viscous measures based on the shear rate response ˙ γ ( t ):(i) the zero-stress fluidity φ ′ M , defined as: φ ′ M ≡ d ˙ γ d σ (cid:12)(cid:12)(cid:12)(cid:12) σ =0 = X n odd ( − ( n − / n ω J ′′ n = ω ( J ′′ − J ′ + 25 J ′′ − J ′′ + · · · ) , (7)and (ii) the large-stress fluidity φ ′ L , defined as: φ ′ L ≡ ˙ γσ (cid:12)(cid:12)(cid:12)(cid:12) σ = σ = X n odd nω J ′′ n = ω ( J ′′ + 3 J ′′ + 5 J ′′ + 7 J ′′ + · · · ) . (8)Equations (7) and (8) simply correspond to Eqs (5) and (6) where the shear rate˙ γ is used instead of the strain γ and the fluidity coefficients nω J ′′ n are used insteadof the compliance coefficients J ′ n . Consequently, φ ′ M and φ ′ L have a similar geometricinterpretation but for Lissajous plots of ˙ γ ( t ) vs σ ( t ) as shown in Fig. 6(b,d,f). J ′ M and φ ′ M characterize the nonlinear behavior at the minimum stress σ = 0, while J ′ L and φ ′ L quantify nonlinearity at the maximum stress σ = ± σ . In order to grasp theintra-cycle elastic behavior with a single measure, Dimitriou et al. (2013) proposed toconsider the relative ratio R of the change in compliance within a LAOStress cycle: R ≡ J ′ L − J ′ M J ′ L = 4 J ′ − J ′ + 8 J ′ − J ′ + . . .J ′ + J ′ + J ′ + · · · . (9)A large positive value of R corresponds to pronounced intra-cycle stress softening whilea negative value indicates stress stiffening. Therefore, R is referred to as the “stress-softening index.” Finally, a similar ratio Q , referred to as the “stress-thinning index,” isdefined for the intra-cycle change of fluidity: Q ≡ φ ′ L − φ ′ M φ ′ L ≃ J ′′ − J ′′ + 56 J ′′ − J ′′ + . . .J ′′ + 3 J ′′ + 5 J ′′ + · · · , (10)with Q >
Q <
0) indicating stress-thinning (resp. stress-thickening) behaviorwithin an oscillation cycle. As already noted above for the Fourier coefficients and forthe THD, all the various LAOStress measures introduced in this section are expected toslowly vary with time in our time-dependent colloidal gel.
E. Ultrasonic imaging
Just like standard rheology provides information that can be reliably interpreted pro-vided the sample and the strain/flow field remain homogeneous, the LAOS measuresintroduced above are meant to describe homogeneous situations only. For instance, ithas been shown that wall slip may lead to the presence of even harmonics in LAOS signalsand in some cases to nonperiodic responses [Graham (1995)]. Of course, the interpreta-tion of LAOS measures becomes even more problematic in the case of bulk solid–fluidcoexistence such as that frequently encountered in yield stress materials under steadyshear. In order to test for strain heterogeneity in our CB gel, we use ultrasonic imaging asalready introduced by Gibaud et al. (2010). Here we adapt the previous one-dimensionaltechnique to two-dimensional imaging based on a newly developed multi-channel setup.Our ultrasonic setup is sketched in Fig. 1. It has been described at length inGallot et al. (2013) to which the reader is referred for full details. In brief, our techniquerelies on the use of an array made of 128 piezoelectric transducers that emit and receivelow-intensity ultrasonic pulsed waves with a central frequency of 15 MHz. The totallength of the transducer array is 32 mm. The array is set vertically (i.e. along the vortic-ity direction z ) in a large water tank surrounding the Couette cell at about 30 mm from FIG. 1. Sketch of the experimental arrangement showing the Couette cell coupled to the rheome-ter and the ultrasonic transducer array used for imaging the local displacement of the samplefrom one oscillation cycle to another. the outer wall of the cup. A plane pulse is emitted by firing all 128 transducer elementssimultaneously. This pulse gets scattered by the hollow glass microspheres seeding theCB gel and the backscattered signals are received by each of the 128 transducers andstored on a personal computer through a real-time processing (RTP) board. From the128 backscattered signals, an ultrasonic image is formed and yields a “speckle image” S ( r, z, t ) of the scatterer distribution across the gap of the Couette cell (with r the radialdistance to the inner bob and z the vertical position from the top of the transducer array)at the time t when the pulse was sent (neglecting the ultrasonic travel time from/to thetransducer array which amounts to about 40 µ s.).The spatial resolution in the radial direction is roughly given by the ultrasonic wave-length 100 µ m and the spacing of 250 µ m between two adjacent transducers on the arraygives the vertical resolution. A much finer resolution on the scatterer displacements(down to a few microns) is obtained by cross-correlating successive images S ( r, z, t ) and S ( r, z, t + δt ) as described in Manneville et al. (2005) and Gallot et al. (2013). Due to thelimited spatial resolution along z and since the displacements are expected to be mainlyhorizontal, the correlation is performed only on the horizontal direction r . Under a con-stant shear stress (or strain rate), this leads to images of the displacement field ∆( r, z, t )that are directly proportional to the velocity field v ( r, z, t ) projected along the acousticpropagation axis. In Sect. III C below, we set the time interval δt to be a multiple of theoscillation period in order to focus on the displacement of the sample between differentcycles and emphasize the slow time dependence of our colloidal gel under LAOS. III. RESULTSA. Ramp experiment
Our first experimental results are reported in Fig. 2 and concern a standard oscillationprocedure where the imposed stress amplitude is linearly swept up starting from thelinear regime up to a fully nonlinear state at a fixed frequency f = 1 Hz. Shown withfilled symbols are the viscoelastic moduli G ′ (in blue) and G ′′ (in red) recorded by therheometer as a function of the stress amplitude σ for a “fast” sweep rate of 34 mPa.s − .Initially, the gel displays solidlike behavior with G ′ ≫ G ′′ . It then apparently yields( G ′ ≃ G ′′ ) when σ reaches a characteristic yield value σ y ≃ . G ′′ ≫ G ′ .Surprisingly, if one uses a slower sweep rate of 7 mPa.s − , G ′ and G ′′ (open symbols)
10 12 14 16 18 2010 G ′ , G ′′ ( P a ) σ (Pa) FIG. 2. Standard LAOStress sweep experiment illustrating the yielding and fluidization of a6% wt. carbon black gel at frequency f = 1 Hz. Elastic modulus G ′ (blue) and loss modulus G ′′ (red) measured when sweeping up the oscillatory stress amplitude σ for two different sweeprates: 34 mPa.s − (filled symbols) and 7 mPa.s − (open symbols). show a similar yet much sharper stress dependence and cross at σ y ≃ . G ′ and σ y are expected to increasein a material subject to structural consolidation [Negi and Osuji (2010)]. Here, the sameCB gel rather yields at a lower stress when submitted to similar stresses for longertimes. It can be shown through a more systematic investigation that the apparent yieldstress consistently decreases as the sweep rate is decreased. This behavior hints at a“fatigue” process under oscillatory shear. Consequently, the evolution of the viscoelasticmoduli shown in Fig. 2, in particular the various steps and drops in G ′ and G ′′ , cannotbe interpreted in terms of a steady-state response and these measurements call for atemporally resolved picture of the yielding dynamics under LAOS.As a final note, we emphasize that the data points in Fig. 2 are not equally spacedwith σ although a linear sweep is imposed. For the slower sweep, the stress amplitudeeven decreases for σ ≃
15 Pa. This problem is typical of the difficulties to reach thecommanded stress amplitude in a strongly time-dependent material and results fromboth the increasing inertia and the feedback loop of the rheometer as mentioned abovein Sect. II C.
B. Constant amplitude experiment: LAOS stress rheology
For the rest of this paper, we shall mainly focus on a single LAOStress experimentperformed at a constant stress amplitude σ = 11 Pa and a constant frequency f =1 Hz over more than 4 10 s. This experiment is performed at a sufficiently low stressamplitude, so that the various steps of the yielding dynamics can be easily distinguished.We first discuss the standard rheological measurements performed by the rheometer andthen turn to Fourier transform rheology and LAOStress measures.
1. Standard rheological measurements
Figure 3(a) shows the temporal evolution of the viscoelastic moduli G ′ and G ′′ togetherwith examples of stress and strain waveforms over one oscillation cycle taken at threedifferent stages of the experiment. Figure 3(b) and 3(c) respectively report the phase0 −2 G ′ , G ′′ ( P a ) t = 315 s t = 16087 s σ ( P a ) t = 30000 s t (s) γ δ / π −2 −1 γ −5 −4 −3 −2 d γ / d t( s − ) t (10 s) a) τ w τ f b)c)d) FIG. 3. Time-resolved LAOStress experiment illustrating the yielding and fluidization of a6% wt. carbon black gel under an oscillatory stress of constant amplitude σ = 11 Pa andfrequency f = 1 Hz. (a) Time evolution of the elastic modulus G ′ (blue) and loss modulus G ′′ (red). Initially, the gel displays solidlike behavior ( G ′ ≫ G ′′ ), apparently yields ( G ′ ≃ G ′′ )at t ≃ . s, and seems to flow like a liquid thereafter ( G ′ ≪ G ′′ ). Insets: “instantaneous”strain response γ ( t ) (red solid curves) as a function of the imposed sinusoidal shear stress σ ( t )(black dashed curves) over one oscillation period at three different times indicated by squaresymbols in the main graph. (b) Time evolution of the phase shift δ = tan − ( G ′′ /G ′ ) between σ ( t ) and γ ( t ). (c) Time evolution of the strain amplitude γ . (d) Time derivative d γ / d t ( t ) ofthe strain amplitude response shown in (c). Gray dashed lines indicate the two characteristictimes discussed in the text: the time τ w such that G ′ ( τ w ) = G ′′ ( τ w ), which defines apparentyielding, and the time τ f at which d γ / d t ( t ) reaches a global maximum, which corresponds tofull fluidization as inferred from ultrasonic imaging. shift δ of the strain response relative to the stress input and the strain amplitude γ asrecorded by the rheometer as a function of time. Here, the phase shift δ is computedbased on the sample stress while the stress waveforms in Fig. 3(a) are raw signals. Thesestandard rheological data clearly reveal a strong time dependence of the colloidal gelunder study. Although σ lies below the apparent yield stresses inferred from both theamplitude sweeps of Fig. 2, the CB gel slowly turns from an elastic solid to a viscousliquid if given enough time to yield. Indeed, one has G ′ ≫ G ′′ and δ ≃ t . . swhile G ′ ≪ G ′′ and δ ≃ π/ t & s (see also the sinusoidal waveforms at t = 315and 3 10 s that are respectively in phase and in phase quadrature). Meanwhile, thestrain amplitude γ increases from about 0.01 by more than two orders of magnitude toreach an asymptotic value of about 5.1From the slow evolution of the strain amplitude γ shown in Fig. 3(c), one may computethe time derivative d γ / d t , which should not be confused with the amplitude of theoscillatory shear rate ˙ γ = ω γ . The temporal evolution of d γ / d t over the wholeexperiment is plotted in Fig. 3(d). After a long period of time where it takes very smallvalues, d γ / d t goes through a first local maximum, which nicely corresponds to the pointat which G ′ and G ′′ coincide (and therefore δ = π/ τ w associated with the yielding dynamics of a CB gel under LAOStress and which we willrefer to as “apparent yielding.” In the present case, one has τ w = 16910 ±
10 s and γ ≃ .
04 at τ w . At this point, the strain response has become strongly nonlinear as canbe checked from the waveform at t = 16087 s in the inset of Fig. 3(a).Furthermore, a second characteristic timescale τ f can be inferred from Fig. 3(d), whichcorresponds to the global maximum reached by d γ / d t , here for τ f = 28950 ±
50 s.Between τ w and τ f , the rheological response looks very complex. From the fact that G ′′ > G ′ , the sample appears fluidlike. The general trend is that both viscoelasticmoduli decrease sharply while γ strongly increases, but the details are rather bumpy andfluctuating as can be seen from d γ / d t . When τ f is reached, γ ≃ . G ′ , G ′′ , δ , and γ eventually enter a much smoother final relaxation towardssteady state.The temporal behavior of our gel characterized by two characteristic timescales isreminiscent of the response of attractive glasses in LAOStrain sweep experiments byPham et al. (2006) and Laurati et al. (2011) who have defined two successive yieldingprocesses as a function of strain amplitude: a first one at γ = 0 . γ = 1–3 interpretedas the dislocation of the cages formed by the neighboring colloids. Although the orderof magnitudes of γ ( τ w ) and γ ( τ f ) are strikingly consistent with the two characteristicstrains γ y and γ y seen in LAOStrain, it is however precipitate to draw any conclusionin the absence of more local insights. In order to investigate the LAOStress response inmore details, we turn to nonlinear analysis in the next two paragraphs.
2. Fourier transform rheology
Figure 4(a) shows the power spectrum | J ⋆ ( ω ) | of the strain response at time t =16087 s already displayed in the middle inset of Fig. 3(a) and replotted in Fig. 4(b)with a thick black line. This particular time was purposely chosen slightly below τ w so as to correspond to a time when the harmonic content is among the largest. Thepower spectrum is normalized by the contribution of the odd harmonics P n odd | J ⋆n | .The apparent non-zero width of the fundamental peak and of the third harmonic resultsfrom the windowing used for the Fourier transform. We can check in Fig. 4(a) that oddharmonics dominate in the power spectrum up to the 9 th harmonic. Therefore, in thefollowing, summation over odd harmonics will always be restricted to n = 9. Indeed, eventhough the second harmonic | J ⋆ | is two orders of magnitude smaller than the fundamental | J ⋆ | and one order of magnitude smaller than the third harmonic | J ⋆ | , even harmonicsbecome comparable to odd harmonics with n ≥
9. This non-negligible content in evenharmonics may be related to a small amount of wall slip around τ w as will be furtherdiscussed in Sect. III C below. Figure 4(a) also shows that | J ⋆n | roughly decreases as 1 /n which is typical of a triangular signal. Finally, considering only the odd harmonics upto n = 9 allows one to reconstruct accurately the strain response as shown by the greenline in Fig. 4(b).In Fig. 5, we plot the total harmonic distortion of the strain response THD( γ ) definedby Eq. (4) as a function of time over the whole experiment corresponding to Fig. 3.The three different regimes unveiled in the previous section appear even more clearlyon THD( γ ). For t < τ w , the odd harmonics are seen to grow almost linearly with timeup to about 15% (see black squares): as the amplitude of the strain slowly increases,nonlinearity builds up in the initially elastic material which departs from the linearregime right from the start of the LAOStress oscillations. The contribution of even2 −6 −4 −2 | J ⋆ | / X n o dd | J ⋆ n | f (Hz) t (s) γ a)b) FIG. 4. Fourier transform rheology of the LAOStress experiment shown in Fig. 3 at time t = 16087 s. (a) “Instantaneous” normalized power spectrum of the strain response γ ( t ) shownin the middle inset of Fig. 3(a). The dashed line shows the power-law behavior | J ⋆ ( ω ) | ∼ /ω .(b) Reconstruction of the experimental signal (thick black line) at time t = 16087 s based onthe Fourier transform of the instantaneous strain response γ ( t ) with different odd harmoniccontents: fundamental only (red), fundamental and third harmonic (blue), and up to the 9 th harmonic (green). t (10 s) T H D τ w τ f FIG. 5. Time evolution of the total harmonic distortion of the strain response THD( γ ) computedusing Eq. (4) over odd harmonics up to n = 9 ( (cid:4) ) and of the stress signal THD( σ ) (blue • ).Open symbols show the THD computed over even harmonics up to n = 8 for the strain response( (cid:3) ) and for the stress signal (blue ◦ ). Gray dashed lines indicate τ w and τ f . Same experimentas in Fig. 3. harmonics to the spectrum remains limited (below 2%, see empty squares) in this initialphase although it becomes measurable as soon as t & s pointing to possible wallslip. Apparent yielding at τ w coincides with a strong peak in THD( γ ), which falls offvery quickly down to about 5% for t & τ w . As already noted above, the intermediatetimes τ w < t < τ f are characterized by complex fluctuations. We also note that the evenharmonics become comparable to the odd harmonics for τ w < t . s, which hintsto important effects of wall slip as will be confirmed below through ultrasonic imaging.Finally, THD( γ ) goes through a final well-defined maximum that coincides with τ f .Interpreting this maximum is however tricky since at this stage of the yielding process,the total stress signal includes a significant contribution from inertia. Indeed, the strain3amplitude exceeds the rough criterion γ ≃ t & . s.This corresponds to the point where THD( σ ) computed over the odd harmonics (seefilled blue circles in Fig. 5) starts to increase significantly and then passes through amaximum at τ f . Note that the even harmonics of σ ( t ) (empty circles) remain completelynegligible throughout the experiment while those of γ ( t ) (empty squares) only get fullynegligible for t & s. We also checked that in this final regime, the amplitude of thestress signal progressively decays from the commanded value, eventually reaching about9 Pa instead of 11 Pa (data not shown). Together with the fact that asymptoticallyTHD( γ ) ≃ THD( σ ), this clearly shows that the steady state reached for t > τ f is afluidlike state where inertia dominates the rheological recordings.
3. Lissajous plots and LAOStress measures −0.01−0.00500.0050.01 γ −10 −5 0 5 10−0.06−0.0300.030.06 σ (Pa) ˙ γ ( s − ) −0.02−0.0100.010.02 −10 −5 0 5 10−0.2−0.100.10.2 σ (Pa) −4−2024 −10 −5 0 5 10−20−1001020 σ (Pa) a)b) c) J ′ M J ′ L d) φ ′ M φ ′ L e)f) FIG. 6. “Instantaneous” Lissajous plots [ σ ( t ), γ ( t )] (top) and [ σ ( t ), ˙ γ ( t )] (bottom) at (a,b) t =315 s, (c,d) t = 16087 s, and (e,f) t = 3 10 s corresponding to the waveforms shown in theinsets of Fig. 3(a). Red dashed lines show the slopes J ′ M (top) and φ ′ M (bottom). Blue dashedlines indicate J ′ L (top) and φ ′ L (bottom). See also Supplementary Movie 1 for the full data set. The temporal build-up of nonlinearity and the subsequent transition to a fluidizedstate can also be directly visualized through the Lissajous curves of Fig. 6 (see also Sup-plementary Movie 1). There, the three waveforms shown in Fig. 3(a) and representativeof the various yielding steps are graphed as strain γ vs stress σ parametrized by time t (top row in Fig. 6). The bottom row reports the shear rate ˙ γ ( t ) [computed from γ ( t )]as a function of σ ( t ). The initial ellipses [Fig. 6(a,b)] point to quasi-linear viscoelasticresponse and the fact that the semiminor axis is much smaller than the semimajor axisindicates the predominance of elasticity over viscosity (also note the change in verticalscales from one set of Lissajous plots to the other). Close to apparent yielding at τ w , thecycle gets strongly distorted due to nonlinearity [Fig. 6(c,d)]. Eventually, the Lissajousplot recovers an elliptical shape yet now dominated by viscosity [Fig. 6(e,f)]. Note thatthe phase shift (i.e. the orientation of the semimajor axis relative to the horizontal) inFig. 6(e) becomes larger than π/
2. This is because the stress signal corresponds to thetotal stress, which gets dominated by the inertial term in Eq. (2) at the late stages of theexperiment. Since the inertial term involves a second derivative of γ ( t ), it correspondsto a phase shift of π which explains the ellipse orientation and the negative slope for J ′ L in Fig. 6(e) as well as the self-intersecting cycle in Fig. 6(f).As recalled in Sect. II D 2 and as indicated by dashed lines in Fig. 6, Lissajous plots canbe further analyzed in terms of their slopes at the minimum stress and of the slopes of thelines joining their extremities in order to estimate the various LAOStress elastic ( J ′ M and J ′ L ) and viscous ( φ ′ M and φ ′ L ) measures. The same measures can be inferred from theirexpressions in terms of the Fourier coefficients given by Eqs. (5)–(8). It can be checkedin Figs. 7 and 8 that these two different ways of measuring the LAOStress observablesas well as the corresponding ratios R and Q are in excellent quantitative agreement.Estimates based on Lissajous plots are naturally subject to more experimental scatter,4 −0.2−0.10 J ′ L ( P a − ) −3 −0.2−0.10 J ′ M ( P a − ) −3 Φ ′ L ( P a − . s − ) t (10 s) Φ ′ M ( P a − . s − ) a)b)c)d) τ w τ w τ w τ w τ f FIG. 7. Time evolution of the LAOStress measures defined in Eqs. (5)–(8): (a) J ′ L , (b) J ′ M ,(c) φ ′ L , and (d) φ ′ M . Black squares correspond to the estimation based on the Fourier coefficientswhile red circles are estimated from the slopes in the Lissajous plots. Insets are enlargementsfor t < τ f . Gray dashed lines indicate τ w and τ f . Same experiment as in Fig. 3. especially φ ′ M due to the differentiation of the strain signal and to the determinationof the tangent at σ = 0. The most prominent feature of the LAOStress measures thatappears in Fig. 7 is their time dependence for t > τ f : both compliances J ′ M and J ′ L turnnegative and then strongly decrease while the fluidities φ ′ M and φ ′ L sharply increase.However, as already evidenced above, the data for t > τ f are dominated by the geometryinertia and mix information on the sample behavior with the contribution of the inertiastress. Therefore, although the increase of fluidity is eventually expected for a yieldingmaterial, LAOStress measures should only be considered as representative of the sampleresponse for t < τ f .The insets in Fig. 7 shows enlargements of J ′ M , J ′ L , φ ′ M , and φ ′ L for t < τ f . Itcan be seen that the compliances J ′ L and J ′ M both weakly increase for t < τ w , whichcorresponds to the global slow softening of the CB gel already observed in the initialresponse of the elastic modulus G ′ [see Fig. 3(a)]. The corresponding fluidities φ ′ M and φ ′ L remain very close to zero as expected for a solidlike material. Interestingly, the stress-softening index R remains fairly negligible for t < τ w [see Fig. 8(a)]. In other words, theintra-cycle response (or at least the difference in maximum-stress and minimum-stressbehaviors) does not show any pronounced stress-thinning nor stress-thickening althoughnonlinearity progressively builds up for t < τ w and reaches a maximum at t ≃ τ w . Notethat for t < τ w , the stress-thinning index Q cannot be defined since φ ′ M ≃ φ ′ L ≃
0. The5 R t (10 s) Q τ w τ f a)b) FIG. 8. Time evolution of (a) the stress-softening index R = ( J ′ L − J ′ M ) /J ′ L and (b) the stress-thinnning index Q = ( φ ′ L − φ ′ M ) /φ ′ L inferred from the data of Fig. 7. Black squares correspondto the estimation based on the Fourier coefficients while red circles are estimated from the slopesin the Lissajous plots. Gray vertical dashed lines indicate τ w and τ f . time dependence of all LAOStress measures changes markedly at t ≃ τ w after which theysharply increase. This reflects the progressive elasticity loss and fluidity build-up in thematerial. As seen in Fig. 8, this change of behavior results in values of both indices R and Q of the order of unity for τ w . t . τ f . This corresponds to significant apparentstress softening and stress thinning of the CB gel within one oscillation cycle. Here again,for t > τ f , the predominance of inertia prevents one from drawing any conclusion fromthe final relaxation of R and Q . C. Constant amplitude experiment: ultrasonic imaging
The LAOStress analysis presented above provides interesting insights into the delayedyielding of our CB gel. In particular, it allows one to clearly define two different timescales τ w and τ f characteristic of the yielding dynamics. However, it assumes that the sampleremains homogeneous over time. In other words, the above data can only be consideredas the result of a spatial average over the whole sample, including boundary layers thatmay have a crucial influence in the occurrence of apparent wall slip. We now turn toultrasonic imaging in order to check for the sample homogeneity.Figure 9 shows two spatiotemporal diagrams of the ultrasonic speckle images S ( r, z, t )recorded during the same experiment as that analyzed in Figs. 3–8 above with a samplingfrequency f US = 0 . S at a fixed vertical position z in the middle of the Couette cell [Fig. 9(a-b)] and at a fixed position r in the middle of the gap [Fig. 9(c-d)]. The distance r ismeasured from the inner bob while z is measured from the top of the transducer array.In both cases, the horizontal axis corresponds to the time t elapsed since the applicationof oscillatory stress and white vertical dashed lines indicate the two times τ w and τ f inferred previously from the rheological data.Three different regimes can be distinguished in Fig. 9. (i) For t < τ w , the horizontallines in the spatiotemporal diagrams show that the sample comes back to the exact sameposition from one cycle to another. This indicates a reversible motion of the seedingparticles consistent with an elastic response of the gel. (ii) For τ w < t . τ f , solidlikeresponse is lost and a complex, heterogeneous speckle pattern is observed. (iii) Fulldecorrelation of the speckle pattern in both r and z directions is achieved for t & τ f .More precisely, for t & τ w , the speckle traces in Fig. 9(a) take up a constant slopethroughout the gap. This is indicative of total slippage of the sample at the walls fromone cycle to another. Therefore, apparent yielding at t = τ w can be attributed to wall6 r ( mm ) t (10 s) z ( mm ) t (10 s) τ w τ f a)c) b)d) FIG. 9. Ultrasonic imaging in a 6% wt. carbon black gel under a constant oscillatory stress ofamplitude σ = 11 Pa at frequency f = 1 Hz. The figure displays spatiotemporal diagramsof the speckle images S ( r, z, t ) coded in linear color levels and recorded simultaneously to therheological data shown in Fig. 3. See also Supplementary Movie 2 for an animation of thespeckle images. (a-b) S ( r, z , t ) at depth z = 15 mm from the top of the transducer array(corresponding roughly to the middle height of the Couette cell). (c-d) S ( r , z, t ) at horizontalposition r = 1 mm from the inner rotating cylinder (corresponding to the middle of the gap).(b) and (d) are enlargements of (a) and (c) within the fluidization regime τ w < t < τ f . Theultrasonic sampling frequency is equal to f US = 0 . f = 1 Hz. White dashed lines in (a) and (b) indicate τ w and τ f . Red lines in (c) and(d) indicate the boundary between solidlike and fluidlike regions. slip as a result of failure at the inner bob. This most probably explains the presence ofeven harmonics in γ ( t ) for t ≃ τ w as seen above in Fig. 5. Then, for t & . s, thespeckle pattern loses its temporal correlation for r . . µ m) within 5 oscillation cycles. We take this decorrelation as an evidencefor irreversible motion of the scatterers and for local fluidlike behavior of the sample. Asalready discussed by Gibaud et al. (2010), irreversibility most likely corresponds to thesedimentation of the seeding particles within the fluidized material. For r & . t ≃ . s [see Fig. 9(b)].This means that for about one hour, the sample remains locally solidlike in most of thegap and that this solid coexists with a small fluidized region close to the inner movingwall. As seen in Fig. 5, even harmonics in the strain response become negligible whentotal wall slip gives way to solid–fluid coexistence at t ≃ . s. For t & . s,fluidization also starts at the outer cup. The solidlike material in the bulk then getsprogressively eroded from both ends by the surrounding fluidized suspension until fullfluidization (i.e. full decorrelation of the speckle pattern) is achieved at t ≃ . s atleast for the specific value of z considered in the spatiotemporal diagram of Fig. 9(a,b).This sequence of events is similar to that reported by Gibaud et al. (2010) based onone-dimensional ultrasonic velocimetry at middle height of a Couette cell for a largeroscillation amplitude of σ = 15 Pa. However, here for σ = 11 Pa, wall slip is ob-served to occur after a long delay τ w while it is observed right from the application ofoscillatory stress at the larger amplitude. Moreover, the present two-dimensional ex-periments provide an additional view over the vertical direction: Fig. 9(c-d) shows thatsolid–fluid coexistence also occurs along the vorticity direction although it starts later(around t ≃ . s) than in the radial direction. We recall that Fig. 9(c-d) correspondsto r = 1 mm where solidlike behavior persists for longer times so that it is not surpris-ing that the speckle remains vertically homogeneous for longer times. We also note that7 t = 315 s z ( mm ) t = 23250 s t = 24000 s t = 25000 s r (mm) t = 25750 s t = 27250 s t = 30000 s D i s p l a ce m e n t( µ m ) τ w τ f t (s) FIG. 10. Images of the displacement field ∆( r, z, t ) between two successive ultrasonic pulsessent every δt = 1 /f US = 5 s at various times during the yielding process. The displacements arecoded in linear color levels. Same experiment as in Figs. 3–9. See also Supplementary Movie 2. fluidization starts from the bottom of the Couette cell then at the top before propagatingtowards the middle of the cell.Full fluidization in both r and z directions is observed for t ≃ . s after which thespeckle remains uncorrelated i.e. the whole sample has become fluidlike. This fluidizationtime is about 10% shorter than τ f ≃ . s inferred above from rheological data.However, the ultrasonic images analyzed in Fig. 9 only correspond to specific values of z and r and to a given vertical slice of our CB gel. In view of the radial and verticalheterogeneities, it is most probable that the sample is also heterogeneous in the azimuthaldirection so that t ≃ . s naturally underestimates the actual fluidization time of thewhole three-dimensional sample. Moreover, the red solid–fluid boundaries in Fig. 9(b,d)are only indicative and such boundaries would most likely change if we used a finerultrasonic sampling frequency and a truly quantitative criterion for speckle decorrelation.We conclude that the timescale τ f inferred from rheology alone corresponds to full three-dimensional fluidization of the CB gel.In order to directly confirm the fluidization scenario deduced from speckle images,Fig. 10 displays a few displacement maps computed by cross-correlating successive speckleimages at different times during yielding (see Supplementary Movie 2 for the full dataset). Figure 10 confirms that no significant motion is observed as long as t < τ w while for τ w < t < τ f , strongly heterogeneous displacement maps are reported, first through theexistence of a thin fluidized band close to the inner bob and later through the growth offluidized pockets from both cylindrical walls and from the bottom then from the top ofthe Couette cell. Finally, for t & τ f , the displacement map appears as spatially erraticover the whole sampled volume. This simply translates the fact that speckle images inthe fluidized state are decorrelated and that irreversible motions from one oscillationcycle to another are too large to be reliably captured by our cross-correlation algorithm. IV. DISCUSSION AND CONCLUSIONSA. Two timescales in the fatigue and fluidization of CB gels under LAOStress
The most striking result of the present paper is the existence of two well-definedtimescales τ w and τ f that characterize the time-dependent yielding of CB gels underLAOStress. The experiment investigated in details in Sections III B and III C shows thatthese timescales can reach several 10 s. This explains why a standard amplitude sweepcannot be used to estimate reliably the LAOStress response of CB gels. In particular,8these colloidal gels yield even for stress amplitudes well below the apparent yield stressesinferred from amplitude sweeps. When the stress amplitude is kept constant, rheologicalmeasurements coupled to ultrasonic imaging show that (i) for t < τ w , the gel behavesas an elastic solid subject to a slow fatigue process, (ii) it yields locally at the walls for t ≃ τ w and subsequently shows apparent wall slip for t & τ w , (iii) for τ w < t < τ f , thesample is constituted of macroscopic fluidized regions that coexist with solidlike domainswhich get progressively eroded, and (iv) for t > τ f , the sample is eventually fully fluidizedand rheological data are dominated by inertia.Interestingly, recent creep experiments by Grenard et al. (2013) on the same CB gels ina roughened Couette geometry have also revealed the existence of two different timescales,noted τ c and τ f , associated with a similar scenario. Under a steady shear stress σ , thematerial was shown to first slowly creep until it fails at the inner bob at t = τ c and thento get progressively and heterogeneously fluidized until τ f . Such a striking similaritybetween creep and LAOStress suggests that the same delayed yielding mechanism is atplay under both constant stress and oscillatory stress. We also expect this scenario tobe relevant to other soft materials showing delayed yielding under creep such as thermo-reversible gels [Gopalakrishnan and Zukoski (2007); Brenner et al. (2013)], silica gels ordepletion gels [Sprakel et al. (2011)]. The general picture involves failure at the walls andsubsequent erosion of the solid material into macroscopic pieces surrounded by larger andlarger fluidized regions. This picture is reminiscent of that found in Laponite suspensionsunder steady shear yet under smooth boundary conditions only Gibaud et al. (2009).Clearly, the influence of surface roughness on the yielding processes under LAOS shouldalso be addressed in future work. B. Local vs global characterization of time-dependent yielding under LAOStress
In the course of this paper, we have highlighted the interest of both Fourier transformrheology and LAOStress measures to characterize a time-dependent material under anoscillatory stress. Our results also reveal important limitations of such approaches basedon global rheological data alone. Indeed, ultrasonic imaging has allowed us to evidencestrong spatial heterogeneities for τ w < t < τ f in both gradient and vorticity directions.Such heterogeneity questions the interpretation (if not the validity) of the spatial averagesprovided by the rheometer. Obviously, in the presence of slippage at the walls, the globalstrain γ only gives an indication on the oscillation amplitude and is not representativeof the actual strain within the material.So far, we have only focused on a single experiment at σ = 11 Pa. In view of thecomplexity of the fatigue and yielding scenario, a systematic investigation as a functionof σ is out of the scope of the present paper. Still, it is interesting to show some globalrheological data for increasing values of σ is the form of the Pipkin-like diagram drawnin Fig. 11. Lissajous plots are gathered as a function of time t (instead of frequency ω inthe classical Pipkin representation) and of the stress amplitude σ . The data investigatedso far are shown in the bottom line. The temporal evolution of the Lissajous plots isseen to change sharply upon increasing σ . This indicates that the ratio τ w /τ f is nota constant and strongly depends on σ . In particular, the elastic behavior is no longerobserved for stresses larger than 11 Pa and for t/τ f = 0 .
01. This is due to the factthat at larger stresses wall slip occurs as soon as the stress is applied in agreement withGibaud et al. (2010). Moreover, the fact that Lissajous plots are not closed loops for thehighest stress amplitudes and for the shorter times clearly points to a time dependenceof the material on the timescale of one cycle (here 1 s). A detailed study of the localtwo-dimensional yielding scenario at stresses larger than 11 Pa and of the dependence ofthe two timescales τ w and τ f on σ is part of ongoing work.9 σ (Pa) t/τ f FIG. 11. Pipkin-like diagram of the time-dependent response of a CB gel to LAOStress: Lis-sajous plots ( σ , γ ) at different times t/τ f in the yielding process and for different amplitudes σ of the imposed oscillatory stress. The oscillation frequency is fixed to 1 Hz in all cases. C. Open questions and perspectives
To the best of our knowledge, the present experiments provide the first direct evidencefor successive failure at the walls and heterogeneous bulk fluidization of a colloidal gel un-der LAOStress. Our macroscopic interpretation of the two timescales τ w and τ f based onultrasonic imaging obviously differs from the microscopic picture based on bond stretch-ing and cage breaking proposed for the two characteristic yield strains found in attractiveglasses under both LAOStrain and constant strain [Pham et al. (2006); Laurati et al.(2011)]. Whether or not it is possible to reconcile these different pictures stands outas an open issue. In particular, time-resolved LAOS experiments (instead of amplitudesweeps) close to yielding coupled to local measurements at scales larger than the indi-vidual colloids would be very interesting to confirm the steady-state interpretation forattractive glasses. Similar local measurements are also in line to show possible spatialcoexistence of the soft and solid states recently evidenced in the oscillatory yielding ofglassy star polymers [Rogers et al. (2011)].More generally, a theoretical framework for LAOS is currently lacking for time-dependent materials. Such a framework is however required to understand at least theglobal data shown in Fig. 11. While including the effects of boundaries and slippage atthe walls still seems out of reach, models that account for some degree of thixotropy andheterogeneity under LAOS are underway. For instance, the recent isotropic-kinematichardening model devised by Dimitriou (2013) should be very fruitful in interpretingexperimental data on time-dependent materials, although its application to delayedyielding in colloidal gels remains questionable.Finally, intra-cycle information based on ultrasonic imaging is also technically avail-able since the temporal resolution of the setup described in Sect. II E can reach 10,000frames per second. Therefore, future work will not only focus on investigating more sys-tematically the influence of the stress amplitude but also on a spatially and temporallyresolved study of the evolution of the strain field within one cycle.0 ACKNOWLEDGMENTS
The authors wish to thank J¨org L¨auger at Anton Paar as well as Aloyse Franck andRaoul Smith at TA Instruments for technical advice on data acquisition and analysis. Wealso thank Vincent Grenard for preliminary experiments and Thibaut Divoux, GarethMcKinley, and Stefan Lindstr¨om for fruitful discussions. This work was funded by theInstitut Universitaire de France and by the European Research Council under the Euro-pean Union’s Seventh Framework Programme (FP7/2007-2013) / ERC grant agreementNo. 258803. TG acknowledges funding from the french Agence Nationale de la Recherche(grant No. ANR-11-PDOC-027).
H. A. Barnes, J. F. Hutton, and K. Walters.
An Introduction to Rheology . Elsevier (Amsterdam), 1989.L. Bocquet, A. Colin, and A. Ajdari. A kinetic theory of plastic flow in soft glassy materials.
Phys. Rev.Lett. , 103:036001, 2009.J. M. Brader, M. Siebenb¨urger, M. Ballauff, K. Reinheimer, M. Wilhelm, S. J. Frey, F. Weysser, andM. Fuchs. Nonlinear response of dense colloidal suspensions under oscillatory shear: Mode-couplingtheory and fourier transform rheology experiments.
Phys. Rev. E , 82:061401, 2010.T. Brenner, S. Matsukawa, K. Nishinari, and R. Johannsson. Failure in a soft gel: Delayed failure andthe dynamic yield stress.
J. Non-Newtonian Fluid Mech. , 196:1–7, 2013.R. Buscall, J. I. McGowan, and A. J. Morton-Jones. The rheology of concentrated dispersions of weaklyattracting colloidal particles with and without wall slip.
J. Rheol. , 37:621–641, 1993.F. Cardinaux, T. Gibaud, A. Stradner, and P. Schurtenberger. Interplay between spinodal decompositionand glass formation in proteins exhibiting short-range attractions.
Phys. Rev. Lett. , 99:118301, 2007.L. B. Chen, C. F. Zukoski, B. J. Ackerson, H. J. M. Hanley, G. C. Straty, J. Barker, and C. J. Glinka.Structural changes and orientational order in a sheared colloidal suspension.
Phys. Rev. Lett. , 69:688–691, 1992.K. S. Cho. A geometrical interpretation of large amplitude oscillatory shear response.
J. Rheol. , 49:747–758, 2005.S. S. Datta, D. D Gerrard, T. S. Rhodes, T. G. Mason, and D. A. Weitz. Rheology of attractiveemulsions.
Phys. Rev. E , 84:041404, 2011.C. J. Dimitriou.
The rheological complexity of waxy crude oils: Yielding, thixotropy and shear hetero-geneities . PhD thesis, Massachusetts Institute of Technology, 2013.C. J. Dimitriou, R. H. Ewoldt, and G. H. McKinley. Describing and prescribing the constitutive responseof yield stress fluids using large amplitude oscillatory shear stress (laostress).
J. Rheol. , 57:27–70, 2013.T. Divoux, D. Tamarii, C. Barentin, and S. Manneville. Transient shear banding in a simple yield stressfluid.
Phys. Rev. Lett. , 104:208301, 2010.T. Divoux, C. Barentin, and S. Manneville. From stress-induced fluidization processes to herschel-bulkleybehaviour in simple yield stress fluids.
Soft Matter , 7:8409–8418, 2011.T. Divoux, D. Tamarii, C. Barentin, S. Teitel, and S. Manneville. Yielding dynamics of a herschel-bulkleyfluid: a critical-like fluidization behaviour.
Soft Matter , 8:4151–4164, 2012.T. Divoux, V. Grenard, and S. Manneville. Rheological hysteresis in soft glassy materials.
Phys. Rev.Lett. , 110:018304, 2013.J.-B. Donnet, R. C. Bansal, and M.-J. Wang.
Carbon black: Science and technology . Marcel DekkerInc. (New York), 1993.K. Dullaert and J. Mewis. Thixotropy: Build-up and breakdown curves during flow.
J. Rheol. , 49:1213–1230, 2005.K. Dullaert and J. Mewis. A structural kinetics model for thixotropy.
J. Non-Newtonian Fluid Mech. ,139:21–30, 2006.R. H. Ewoldt, A. E. Hosoi, and G. H. McKinley. New measures for characterizing nonlinear viscoelasticityin large amplitude oscillatory shear.
J. Rheol. , 52:1427–1458, 2008.T. Gallot, C. Perge, V. Grenard, M.-A. Fardin, N. Taberlet, and S. Manneville. Ultrafast ultrasonicimaging coupled to rheometry: Principle and illustration.
Rev. Sci. Instrum. , 84:045107, 2013.T. Gibaud, C. Barentin, N. Taberlet, and S. Manneville. Shear-induced fragmentation of laponitesuspensions.
Soft Matter , 5:3026–3037, 2009.T. Gibaud, D. Frelat, and S. Manneville. Heterogeneous yielding dynamics in a colloidal gel.
Soft Matter ,6:3482–3488, 2010.T. Gibaud, N. Mahmoudi, J. Oberdisse, P. Lindner, J. S. Pedersen, C. L. P. Oliveira, A. Stradner, andF. Cardinaux. New routes to food gels and glasses.
Faraday Discuss. , 158:267–284, 2012.V. Gopalakrishnan and C. F. Zukoski. Delayed flow in thermo-reversible colloidal gels.
J. Rheol. , 51:623–644, 2007.M. D. Graham. Wall slip and the nonlinear dynamics of large amplitude oscillatory shear flows.
J.Rheol. , 39:697–712, 1995.V. Grenard, T. Divoux, N. Taberlet, and S. Manneville. Timescales in creep and yielding of attractivegels. 2013. To appear in Soft Matter, e-print cond-mat/1310.0385, DOI: 10.1039/C3SM52548A.K. Hyun, M. Wilhelm, C. O. Klein, K. S. Cho, J. G. Nam, K. H. Ahn, S. J. Lee, R. H. Ewoldt, and G. H.McKinley. A review of nonlinear oscillatory shear tests: Analysis and application of large amplitudeoscillatory shear (laos).
Prog. Polym. Sci. , 36:1697–1753, 2011.C. O. Klein, H. W. Spiess, A. Calin, C. Balan, and M. Wilhelm. Separation of the nonlinear oscillatory response into a superposition of linear, strain hardening, strain softening, and wall slip response. Macromolecules , 40:4250–4259, 2007.N. Koumakis, A. Pamvouxoglou, A. S. Poulos, and G. Petekidis. Direct comparison of the rheology ofmodel hard and soft particle glasses.
Soft Matter , 8:4271–4284, 2012.N. Koumakis, J. F. Brady, and G. Petekidis. Complex oscillatory yielding of model hard-sphere glasses.
Phys. Rev. Lett. , 110:178301, 2013.R. G. Larson.
The Structure and Rheology of Complex Fluids . Oxford University Press, 1999.J. La¨uger and H. Stettin. Differences between stress and strain control in the non-linear behavior ofcomplex fluids.
Rheol. Acta , 49:909–930, 2010.M. Laurati, S. U. Egelhaaf, and G. Petekidis. Nonlinear rheology of colloidal gels with intermediatevolume fraction.
J. Rheol. , 55:673–706, 2011.X. Li, S.-Q. Wang, and X. Wang. Nonlinearity in large amplitude oscillatory shear (laos) of differentviscoelastic materials.
J. Rheol. , 53:1255–1274, 2009.S. B. Lindstr¨om, T. E. Kodger, J. Sprakel, and D. A. Weitz. Structures, stresses, and fluctuations in thedelayed failure of colloidal gels.
Soft Matter , 8:3657–3664, 2012.P. Lu, E. Zaccarelli, F. Ciulla, A. B. Schofield, F. Sciortino, and D. A. Weitz. Gelation of particles withshort-range attraction.
Nature , 453:499–503, 2008.S. Manneville, L. B´ecu, P. Grondin, and A. Colin. High-frequency ultrasonic imaging: a spatio-temporalapproach of rheology.
Colloids Surfaces A , 270–271:195–204, 2005.J. Mewis and N.J. Wagner. Thixotropy.
Adv. Colloid Interface Sci. , 147–148:214–227, 2009.R. Mezzenga, P. Schurtenberger, A. Burbidge, and M. Michel. Understanding foods as soft materials.
Nature Mater. , 4:729–740, 2005.P. C. F. Møller, A. Fall, V. Chikkadi, D. Derks, and D. Bonn. An attempt to categorize yield stress fluidbehaviour.
Phil. Trans. R. Soc. Lond. A , 367:5139–5155, 2009.R. L. Moorcroft and S. M. Fielding. Criteria for shear banding in time-dependent flows of complex fluids.
Phys. Rev. Lett. , 110:086001, 2013.A. Negi and C. Osuji. Time-resolved viscoelastic properties during structural arrest and aging of acolloidal glass.
Phys. Rev. E , 82:031404, 2010.G. Ovarlez, L. Tocquer, F. Bertrand, and P. Coussot. Rheopexy and tunable yield stress of carbon blacksuspensions.
Soft Matter , 9:5540–5549, 2013.G. Petekidis, A. Moussa¨ıd, and P. N. Pusey. Rearrangements in hard-sphere glasses under oscillatoryshear strain.
Phys. Rev. E , 66:051402, 2002.K. N. Pham, A. M. Puertas, J. Bergenholtz, S. U. Egelhaaf, A. Moussa¨ıd, P. N. Pusey, A. B. Schofield,M. E. Cates, M. Fuchs, and W. C. K. Poon. Multiple glassy states in a simple model system.
Science ,296:104–106, 2002.K. N. Pham, G. Petekidis, D. Vlassopoulos, S. U. Egelhaaf, P. N. Pusey, and W. C. K. Poon. Yieldingof colloidal glasses.
Europhys. Lett. , 66:624–630, 2006.K. N. Pham, G. Petekidis, D. Vlassopoulos, S. U. Egelhaaf, W. C. K. Poon, and P. N. Pusey. Yieldingbehavior of repulsion- and attraction-dominated colloidal glasses.
J. Rheol. , 52:649–676, 2008.F. Renou, G. Stellbrink, and G. Petekidis. Yielding processes in a colloidal glass of soft star-like micellesunder large amplitude oscillatory shear (laos).
J. Rheol. , 54:1219–1242, 2010.S. A. Rogers, B. M. Erwin, D. Vlassopoulos, and M. Cloitre. Oscillatory yielding of a colloidal star glass.
J. Rheol. , 55:733–752, 2011.W. B. Russel.
The Dynamics of Colloidal Systems . The University of Wisconsin Press (London), 1987.R.J. Samson, G. W. Mulholland, and J. W. Gentry. Structural analysis of soot agglomerates.
Langmuir ,3:272–281, 1987.F. Sciortino and P. Tartaglia. Glassy colloidal systems.
Adv. Phys. , 54:471–524, 2005.Z. Shao, A. S. Negi, and C. O. Osuji. Role of interparticle attraction in the yielding response of microgelsuspensions.
Soft Matter , 9:5492–5500, 2013.P. A. Smith, G. Petekidis, S. U. Egelhaaf, and W. C. K. Poon. Yielding and crystallization of colloidalgels under oscillatory shear.
Phys. Rev. E , 76:041402, 2007.J. Sprakel, S. B. Lindstr¨om, T. E. Kodger, and D. A. Weitz. Stress enhancement in the delayed yieldingof colloidal gels.
Phys. Rev. Lett. , 106:248303, 2011.H. Tanaka, J. Meunier, and D. Bonn. Nonergodic states of charged colloidal suspensions: Repulsive andattractive glasses and gels.
Phys. Rev. E , 69:031404, 2004.V. Trappe and D. A. Weitz. Scaling of the viscoelasticity of weakly attractive particles.
Phys. Rev. Lett. ,85:449–452, 2000.V. Trappe, E. Pitard, L. Ramos, A. Robert, H. Bissig, and L. Cipelletti. Investigation of q-dependentdynamical heterogeneity in a colloidal gel by x-ray photon correlation spectroscopy.
Phys. Rev. E , 76:051404, 2007.P.H.T. Uhlherr, J. Guo, C. Tiu, X.-M. Zhang, J.Z.-Q. Zhou, and T.-N. Fang. The shear-induced solidliq-uid transition in yield stress materials with chemically different structures.
J. Non-Newtonian FluidMech. , 125:101–119, 2005.M. van der Waarden. Stabilization of carbon black dispersion in hydrocarbons.
J. Colloid Sci. , 5:317–325, 1950.M. Wilhelm. Fourier-transform rheology.