3.2 Shear stresses of olloidal dispersions at the glass transition in equilibrium
3.2.3 Linear and non linear rheology
This setion presents the basis of the rheology of olloidal suspensions. In the rest of
the hapter the mode oupling theory is introdued to give a quantitative desription
of the visoelastiity of the system in the linear regime and of the ow behaviour in
the stationary regime. By linear visoelastiity we onsider a system experiening an
osillatory strain. The dependene of the stress to the amplitude
γ
and frequeny ofthe soliitationis used to estimate the elasti and loss ontributionusually represent by
the moduli
G ′
andG ′′
. If we x the frequeny, the linearity is onsidered per denitionfor strains
γ
where these two moduli are onstant. At this point the struture of thesystem is not disturbed and the system is in equilibrium. If the strain exeeds a ertain
value
γ o
this ondition is not respeted anymore and the struture is disturbed leading10 -1 10 0 10 1 10 2
10 -5 10 -3 10 -1 10 1 10 3
dg /dt [s -1 ]
s [P a .s ]
10 -1 10 0 10 1
10 -1 10 0 10 1 10 2
0.01 Hz 0.1 Hz 1 Hz 10 Hz
g [%]
G ´, G ´´ [P a ]
B)
C) A)
D)
Figure3.8: A) step ow experiment performed at dierent shear rate with dierent time
inre-ments: 50
ms
(rosses), 100ms
(squares), 200ms
(irles). The lines present thets obtained from equation 3.10. B) Comparison of the owurves measured
experi-mentally with inreasing and dereasing shear rates (full and dashed lines) with the
stresses measured for dierent waiting times from the step ow experiments: 50
ms
(down triangles), 100
ms
(heked boxes), 1s
(up triangles), 10s
(hollow squares),100
s
(hollow irles) and stationary regime (full irles). C) Strain sweepatdier-entfrequenies. D)Shear stressmeasured asfuntionof thestrain fromthe stepow
experiments at dierent shear rates (hollow symbols) and frequenies (full symbols).
The samesymbolsare used forthe samevalues ofthe frequeny andshear rates. The
dashed lines present the ts fromequation 3.10.
to the relaxation and ow of the system. The simplest representation of suh a system
is the model of Maxwell. In the rest of the experiments the theory was applied to the
frequeny dependent measurementof
G ′
andG ′′
inthe linearregime. For this reasonthemodulihave been measured as funtion of the strain for dierent frequenies in order to
ensure that the deformations used in the frequeny dependent experiment belong for all
the frequenies to the linear regime. An example is given by the gure 3.8 C). A dense
solutionwithan eetivevolume
φ ef f = 0.622
issubjet deformationsweep experiment is presented for adense13.01wt%
solution at14o C
fordierent frequeny. Themoduliareonstant untila strain around4
%
. This solutionrepresents the highest eetive volumefration investigated in this work, for this reason a strain of 1
%
full the ondition oflinearityin allthe frequeny sweep experimentspresented inthis hapter.
The ow urves have been measured in the stationary regime. This means pratially
that the solution is sheared at a onstant shear rate
γ ˙
long enough, to ensure a laminarow resulting in a onstant stress that we will dene as
σ( ˙ γ)
, whih is the prerequisite of the theory. The stationarity has been heked by step ow experiments also for thehighest eetive volume fration (13.01
wt%
solution at 14o C
,φ ef f = 0.622
). Figure3.8 A) presents the experiments performed for dierent shear rates, with dierent time
inrements (50, 100, 200
ms
). No signiant variations ould be observed between thedierenttimes. Forthe lowest shear rates (
γ ˙ = 10 −4 − 10 −1 s −1
). The stressσ( ˙ γ, t)
rstdereases in the very small time, and then inreases again to reah the onstant value
σ( ˙ γ )
. The experimental values ofσ( ˙ γ)
(solidirles) have been reported inthe gure 3.8B) and omparedto the owurves measuredwith inreasing and dereasingshear rates
(full and dashed urves) in the experimental onditions desribed above and with the
measurement for dierent time. No signiant dierene an be observed whih proves
that the ow urves are measured in the stationary regime. On the ontrary a marked
disrepany an be observed if the time of measurement is not long enough. This result
an be diretly ompared to the work of Heymann et al. [219, 220℄ where the authors
investigated the inuene of the measurement time on the ow urves of onentrated
suspensionsof spherial partiles.
Step owand strainsweep experimentsare very loseinthe sensethat they desribe the
evolution of the system as funtion of the time and strain and have been the objet of
reentstudies[208,221,222℄. Ifweonsider thestepowexperimentsofthegure3.8A)
theoriginoftherstdeayintheshorttimeisnotfullyunderstoodandmustberelatedto
shorttime relaxationproess. Nevertheless theexperimentalpointsinthe shorttimesfor
the lowshearrates superpose intoamaster urve,whihan be desribed asapowerlaw
withanexponent -0.4. Then thestressinreases againquasilinearlyfortheslowest shear
rate to nally relax and reah a onstant value. The two experimentsan be ompared,
if we onsider the evolution of the stress as a funtionof the strain. Forthis the moduli
of the strain sweep are rst transformed into omplex modulus
G ∗ = ((G ′ ) 2 + (G ′′ ) 2 ) 0.5
and thenonvertedinstress
σ ∗ = G ∗ /γ
. Conerningthe step owexperimentsthe strain isobtained by multiplyingthe timet
withthe shear rateγ ˙
. Tothis high volumefrationthe system is in the glassy state and the omplex modulus is almost onstant between
10
−2
and 10
Hz
and the value at the plateauG ∗ p
is approximately equal to 23P a
. Theomparison is presented in the gure 3.8D).
σ ∗
is inreasing linearly with the strain forthe small strain and relax for the higher strain. This kind of resaling underlines the
orrespondene of the two experiments in the linear regime. For the non linear regime a
diret omparison is obtained if the stress is measured for both experiments are done at
the samestrainand shearrate. Forthis purpose we have todene thestrain-rate dened
as
γ ˙ 0 = γω
. The omparisonisdone onthe gure3.9, wherebothexperimentssuperpose eahother to desribe the seond relaxationof the stress.A simple approah lose to the one proposed by Wyss et al. [208℄ was used to aount
of the linear visoelastiity dened by the omplex modulus
G ∗ ( ˙ γ 0 = ˙ γ)
(in the linearregime) atthe lowerstrains and of the stationarity and ow at higherstrains dened by
the onstant stress
σ( ˙ γ )
.The model onsists on a simple Maxwell element, dened with its general equation of
movement:
dγ dt = 1
G dσ
dt + σ
η
(3.9)We onsider now a steady shear (
dγ/dt = ˙ γ
),G = G ∗ ( ˙ γ 0 = ˙ γ )
the linear visoelastiontribution,
η = σ( ˙ γ)/ γ ˙
withσ( ˙ γ)
the stress inthe stationaryregime. Consideringthat att = 0
,σ(t) = 0
, the resolution of this equation gives the following expression for thestress:
σ( ˙ γ, t) = σ( ˙ γ) (1 − exp( − G ∗ ( ˙ γ 0 = ˙ γ) ˙ γt/σ( ˙ γ)))
(3.10)Thisexpressionhavebeenusedforthestepowandstrainsweep experimentsonsidering
G ∗ ( ˙ γ 0 = ˙ γ) = G ∗ p = 23 P a
(see g. 3.8 A) and D) and g. 3.9). This simple approahdesribes very well the seond relaxation observed experimentally in the glassy regime
for both experiments. Of ourse this kind of experiment is muh more ompliated and
the orrelation between the time, frequeny, strain and shear rate has to be addressed
moreindepth. Nevertheless itshows learlythe onnetionbetweenlinearandnonlinear
rheology whih willbe presented inthe rest of the hapter.
3.2.4 Theory
Mirosopi approah
The next setions providea full desription of the theory developed as mentioned in the
introdutionbythe professorMatthiasFuhsand hisoworkers. Weonsider
N
spherialpartiles with radius
R H
dispersed in a volumeV
of solvent (visosityη s
) with imposedhomogeneous, and onstant linear shear-ow. The ow veloity points along the
x
-axisand its gradient along the
y
-axis. The motion of the partiles (with positionsr i (t)
fori = 1, . . . , N
)is desribed byN
oupledLangevin equations [206℄ζ dr i
dt − v solv (r i )
= F i + f i .
(3.11)Solventfrition ismeasuredby theStokesfrition oeient
ζ = 6πη s R H
. TheN
vetorsF i = − ∂/∂r i U ( { r j } )
denote the interpartile fore on partilei
deriving from potentialFigure3.9: Shear stress measured as funtion of the strain from the strain sweep (full symbols)
andstepowexperiments (hollowsymbols). The samesymbolsareused forthe same
values of shear rates
γ ˙
and strainratesγ ˙ 0 = ωγ
.interations with allother partiles;
U
is the potential energy whih depends onallpar-tiles' positions. The solvent shear-ow is given by
v solv (r) = ˙ γ y x ˆ
, and the Gaussianwhite noise foresatises (with
α, β
denoting diretions)h f i α (t) f j β (t ′ ) i = 2ζ k B T δ αβ δ ij δ(t − t ′ ) ,
where
k B T
is the thermal energy. Eah partile experienes interpartile fores, solvent frition, and random kiks. Interation and frition fores on eah partile balane onaverage,so thatthe partilesare atrest inthe solventonaverage. The Stokesianfrition
is proportional to the partile's motion relative to the solvent ow at its position; the
latter varies linearly with
y
. The random fore on the level of eah partile satises theutuation dissipation relation.
An important approximation in Eq. (3.11) is the neglet of hydrodynami interations,
whihwould arise fromthe propertreatmentof the solvent owaround movingpartiles
[33,206℄. Inthefollowingwewillarguethatsuheetsanbenegletedathighdensities
where interpartile fores hinder and/or prevent strutural rearrangements, and where
the system is lose to arrest into an amorphous, metastable solid. Another important
approximation in Eq. (3.11) is the assumption of a given, onstant shear rate
γ ˙
, whihdoes not vary throughout the (innite) system. We start with this assumption in the
philosophy that, rst, homogeneous states should be onsidered, before heterogeneities
and onnement eets are taken into aount. All diulties in Eq. (3.11) thus are
onneted to the many-body interations given by the fores
F i
, whih ouple theN
Langevin equations. In the absene of interations,
F i ≡ 0
, Eq. (3.11) leads tosuper-diusive partilemotion termed'Taylordispersion' [206℄.
While formulation of the onsidered mirosopi model handily uses Langevin
equa-tions, theoretial analysis proeeds more easily from the reformulation of Eq. (3.11) as
Smoluhowski equation. It desribes the temporal evolution of the distribution funtion
Ψ( { r i } , t)
of the partilepositions∂ t Ψ( { r i } , t) = Ω Ψ( { r i } , t) ,
(3.12)employingthe Smoluhowski operator [33, 206℄,
Ω =
the system relaxes intoa unique stationarystate atlong times, so that
Ψ(t → ∞ ) = Ψ s
holds. Homogeneous, amorphous systems are studied so that the stationary distribution
funtion
Ψ s
is translationally invariant but anisotropi. Negleting ageing, the formal solutionof the Smoluhowskiequation withinthe ITT approah an be broughtintotheform[41, 44℄
where the adjoint Smoluhowski
Ω †
operator arises from partial integrations. It ats on the quantities tobeaveraged withΨ s
.Ψ e
denotes the equilibrumanonial distribution funtion,Ψ e ∝ e −U/(k B T )
, whihis the time-independentsolution of Eq. (3.12) forγ ˙ = 0
;in Eq. (3.14), it gives the initial distribution at the start of shearing (at
t = 0
). Thepotential part of the stress tensor
σ xy = − P N
i=1 F i x y i
entered viaΩΨ e = ˙ γ σ xy Ψ e
. Thesimple, exat result Eq. (3.14) is entral to the ITT approah as it onnets steady
state propertiesto time integrals formed with the shear-dependent dynamis. The latter
ontains slowintrinsipartile motion.
In ITT, the evolution towards the stationary distribution at innite times is
approxi-mated by following the slowstrutural rearrangements,enoded in the transient density
orrelator
Φ q (t)
. It isdened by [41, 44℄Φ q (t) = 1
NS q h δ̺ ∗ q e Ω † t δ̺ q(t) i ( ˙ γ=0) .
(3.15)It desribesthe fate ofan equilibriumdensity utuationwith wavevetor
q
, where̺ q = P N
j=1 e iq·r j
, under the ombined eet of internal fores, Brownian motionand shearing.Note that beause of the appearane of
Ψ e
in Eq. (3.14), the average in Eq. (3.15) anbe evaluated with the equilibrium anonial distribution funtion, while the dynamial
evolution ontains Brownianmotion and shear advetion. The normalizationis given by
S q
theequilibriumstruturefator[33,206℄forwavevetormodulusq = | q |
. Theadvetedwavevetor enters inEq. (3.15)
q(t) = q − γt q ˙ x y ˆ ,
(3.16)q (0)
t>0 t=0
q (t)
Figure3.10: Shear advetion of a utuation with initial wavevetor in
x
-diretion,q(t = 0) = q (1, 0, 0) T
, and adveted wavevetor at later timeq(t>0) = q (1, − γt, ˙ 0) T
. At alltimes,
q(t)
is perpendiular to the planes of onstant utuation amplitude. Note that the magnitudeq(t) = q p
1 + ( ˙ γt) 2
inreases with time. Brownian motion,negleted in this sketh,would smear out the utuation.
where unit-vetor
y ˆ
points iny
-diretion) The time-dependene inq(t)
results from theane partile motion with the shear ow of the solvent. Translational invariane
un-der shear ditates that at a time
t
later, the equilibrium density utuationδ̺ ∗ q
has anonvanishing overlap only with the adveted utuation
δ̺ q(t)
(see g. 3.10), where anon-deorrelatingutuation is skethed under shear. In the ase of vanishing Brownian
motion, viz.
D 0 = 0
in Eq. (3.13), we ndΦ q (t) ≡ 1
, beause the adveted wavevetortakes aount of simple ane partile motion [223℄. The relaxation of
Φ q (t)
thusher-alds deay of strutural orrelations. Within ITT, the time integralover suh strutural
deorrelationsprovides anapproximation to the stationarystate:
Ψ s ≈ Ψ e + γ ˙ 2k B T
∞
Z
0
dt X
k
k x k y S k ′
k NS k(t) 2 Φ 2 k (t) Ψ e ̺ ∗ k(t) ̺ k(t)
,
(3.17)with
S k ′ = ∂S k /∂k
[224℄. Thelastterminbrakets inEq. (3.17)expresses, thattheexpe-tation value of a general utuation
A
in ITT-approximation ontains the (equilibrium) overlap with the loal struture,h ̺ ∗ k ̺ k A i ( ˙ γ=0)
. The dierene between the equilibrium andstationarydistributionfuntionsthen followsfromintegratingovertimethespatiallyresolved (viz. wavevetor dependent) density variations.
Thegeneralresultsfor
Ψ s
,the exat oneof Eq. (3.14)and theapproximationEq. (3.17), an beapplied toompute stationaryexpetationvalues like for example thethermody-nami transverse stress,
σ( ˙ γ) = h σ xy i /V
. Equation (3.14) leads to an exat non-linearGreen-Kuborelation:
σ( ˙ γ) = ˙ γ
∞
Z
0
dt g(t, γ) ˙ ,
(3.18)where the generalizedshear modulus
g(t, γ) ˙
depends on shear rate viathe Smoluhowski operator fromEq. (3.13)g(t, γ ˙ ) = 1
k B T V h σ xy e Ω † t σ xy i ( ˙ γ=0) .
(3.19)In ITT, the slow stress utuations in
g (t, γ) ˙
are approximated by following the slow strutural rearrangements, enoded in the transient density orrelators. The generalizedmodulus beomes, using the approximation Eq. (3.17), or, equivalently, performing a
mode oupling approximation [44, 45,216℄:
g(t, γ) = ˙ k B T
Summation over wavevetors has been turned into integration in Eq. (3.20) onsidering
aninnite system.
The familiar shear modulus of linear response theory desribes thermodynami stress
utuations in equilibrium, and is obtained from Eqs. (3.19,3.20) by setting
γ ˙ = 0
[33,158,225℄. WhileEq. (3.19)then givesthe exatGreen-Kuborelation,the approximation
Eq. (3.20) turns into the well-studied MCT formula. For nite shear rates, Eq. (3.20)
desribes how ane partile motion auses stress utuations to explore shorter and
shorterlengthsales. Therethe eetivefores, asmeasuredby thegradientofthediret
orrelationfuntion,
S k ′ /S k 2 = nc ′ k = n∂c k /∂k
,beomesmaller,andvanishasymptotially,c ′ k→∞ → 0
; the diret orrelation funtionc k
is onneted tothe struture fator via theOrnstein-Zernike equation
S k = 1/(1 − n c k )
, wheren = N/V
is the partile density.Note, that the equilibrium struture sues to quantify the eetive interations, while
shear just pushesthe utuationsaround onthe 'equilibriumenergy landsape'.
Strutural rearrangements of the dispersion aeted by Brownian motion is enoded in
the transient density orrelator. Shear indued ane motion, viz. the ase
D 0 = 0
,is not suient to ause
Φ k (t)
to deay. Brownian motion of the quiesent orrelatorΦ ( ˙ k γ=0) (t)
leads at high densities to a slow strutural proess whih arrests at long timesin (metastable) glass states. Thus the ombination of strutural relaxation and shear
is interesting. The interplay between intrinsi strutural motion and shearing in
Φ k (t)
is aptured by
(i)
rst a formally exat Zwanzig-Moritype equation of motion, and(ii)
seondamodeouplingfatorisationinthememoryfuntionbuiltwithlongitudinalstress
utuations [41, 44℄. The equation of motionfor the transient density orrelatorsis
∂ t Φ q (t) + Γ q (t)
wherethe initialdeay rate
Γ q (t) = D 0 q 2 (t)/S q(t)
generalizesthe familiarresultfrom lin-ear response theory to adveted wavevetors; it ontains Taylor dispersion. The memoryequationontainsutuatingstressesand similarlylike
g(t, γ) ˙
inEq. (3.17),isalulatedin mode ouplingapproximation
where we abbreviated
p = q − k
. The vertex generalizes the expression in the quiesent ase [44℄:V qkp (t, t ′ ) = S q(t) S k(t ′ ) S p(t ′ )
q 2 (t) q 2 (t ′ ) V qkp (t) V qkp (t ′ ) , V qkp (t) = q(t) · k(t) nc k(t) + p(t) nc p(t)
.
(3.23)Withshear, wavevetors inEq. (3.23) are adveted aording to Eq. (3.16).
Equations (3.17,3.21,3.22), with the spei example of the generalized shear modulus
Eq. (3.20),forma losedset of equationsdetermining rheologialproperties ofa sheared
dispersion from equilibrium strutural input [41, 44℄. Only the stati struture fator
S q
is required to predit(i)
the time dependent shear modulus within linear response,g lr (t) = g (t, γ ˙ = 0)
, and(ii)
the stationary stressσ( ˙ γ)
from Eq. (3.18). The loss andstorage moduli of small amplitude osillatory shear measurements [33, 158℄ follow from
Eq. (3.19) inthe linear response ase
(i) G ′ (ω) + i G ′′ (ω) = iω
∞
Z
0
dt e −i ω t g(t, γ ˙ = 0) .
(3.24)While,inthelinearresponseregime,modulusanddensityorrelatoraremeasurable
quan-tities, outsidethe linearregime, both quantities serve as toolsin the ITT approahonly.
The transient orrelator and shear modulus provide a route to the stationary averages,
beause they desribe the deay of equilibrium utuations under external shear, and
their time integral provides an approximation for the stationary distribution funtion,
see Eq. (3.17). Determination of the frequeny dependent moduliunder large amplitude
osillatoryshearhas beomepossiblereentlyonly[226℄,and requiresanextensionofthe
present approah totime dependent shear rates inEq. (3.13) [221℄.
Universal aspets
The summarized mirosopi ITT equations ontain a bifuration in the long-time
be-havior of
Φ q (t)
, whih orresponds to a non-equilibrium transition between a uid and a shear-molten glassy state. Close to the transition, (rather) universal preditions anbe made about the non-linear dispersion rheology and the steady state properties. The
entral preditions are introdued in this setion and summarized in the overview given
by g. 3.11. It is obtained from the shemati model whih is also used to analyse the
data, and whihis introdued insetion 3.2.4.
A dimensionless separation parameter
ε
measures the distane to the transition whihis situated at
ε = 0
. A uid state (ε < 0
) possesses a (Newtonian) visosity,η 0 (ε <
0) = lim γ→0 ˙ σ( ˙ γ)/ γ ˙
, and shows shear-thinning upon inreasingγ ˙
. Via the relationη 0 = lim ω→0 G ′′ (ω)/ω
, the Newtonian visosity an also be taken from the loss modulus atlow frequenies, where
G ′′ (ω)
dominates over the storage modulus. The latter varies likeG ′ (ω → 0) ∼ ω 2
. Aglass(ε ≥ 0
),intheabsene ofow, possessesanelastionstantG ∞
,whihanbemeasuredinthe elastishear modulus
G ′ (ω)
inthelimitoflowfrequenies,-2 0 2 4 6 8
Figure3.11: Overviewofthepropertiesofthe F
( ˙ γ)
12
-modelharateristiforthe transitionbetween uid andyielding glass. The upper panel shows numerially obtained transientor-relators
Φ(t)
forε = 0.01
(blak urves),ε = 0
(red),ε = − 0.005
(green), andε = − 0.01
(blue). The shear rates are| γ/Γ ˙ | = 0
(thik solid lines),| γ/Γ ˙ | = 10 −6
(dotted lines), and
| γ/Γ ˙ | = 10 −2
(dashed lines). For the glass state atε = 0.01
(blak),
| γ/Γ ˙ | = 10 −8
(dashed-dotted-line) is also inluded. All urves were alu-lated withγ c = 0.1
andη ∞ = 0
. The thin solid lines give the fatorization result Eq. (3.25) with saling funtionsG
for| γ/Γ ˙ | = 10 −6
; labela
marks the ritiallaw (3.27), and label
b
marks the von Shweidler-law (3.28). The ritial glass form fatorf c
is indiated. The inset shows the ow urves for the same valuesfor
ε
. The thin blak bar shows the yield stressσ + c
forε = 0
. The lower panelshowsthe visoelasti storage (solidline) andloss(broken line)moduli forthe same
values of
ε
. The thin green lines are the Fourier-transformed fatorization result Eq. (3.25) with saling funtionG
taken from the upper panel forε = − 0.005
.The dashed-dotted lines show the t formula Eq. (3.39) for the spetrum in the
minimum-region with
G min /v σ = 0.0262
,ω min /Γ = 0.000457
atε = − 0.005
(green)and
G min /v σ = 0.0370
,ω min /Γ = 0.00105
atε = − 0.01
(blue). The elastionstantat the transition
G c ∞
ismarked also.G ′ (ω → 0, ε ≥ 0) → G ∞ (ε)
. Here thestoragemodulusdominatesoverthelossone,whihdrops like
G ′′ (ω → 0) ∼ ω
. Enforing steady shear ow melts the glass. The stationarystress of the shear-molten glass always exeeds a (dynami) yield stress. For dereasing
shear rate, the visosity inreases like
1/ γ ˙
, and the stress levels o onto the yield-stress plateau,σ( ˙ γ → 0, ε ≥ 0) → σ + (ε)
.Closetothe transition,the zero-shearvisosity
η 0
,the elastionstantG ∞
,and the yieldstress
σ +
show universal anomalies as funtions of the distane to the transition: the visositydivergesinapower-lawη 0 (ε → 0 − ) ∼ ( − ε) −γ
withmaterialdependentexponentγ
around2 − 3
,the elastionstantinreaseslikeasquare-rootG ∞ (ε → 0+) − G c ∞ ∼ √
ε
,andthe dynamiyieldstress
σ + (ε → 0+)
alsoinreases withinniteslopeaboveitsvalueσ c +
at the bifuration. The quantitiesG c ∞
andσ c +
denote the respetive values at thetransition point
ε = 0
, and measure the jump in the elasti onstant and in the yieldstress atthe glass transition; inthe uid state,
G ∞ (ε < 0) = 0
andσ + (ε < 0) = 0
hold.The desribed results follow from the stability analysis of Eqs. (3.21,3.22) around an
arrested, glassy struture
f q
of the transient orrelator [44, 45℄. Considering the time window whereΦ q (t)
is metastable and lose to arrest atf q
, and taking all ontrolpa-rameters like density, temperature, et. to be lose to the values at the transition, the
stabilityanalysis yields the 'fatorization' between spatial and temporaldependenies
Φ q (t) = f q c + h q G ( t/t 0 , ε, γt ˙ 0 ) + . . . ,
(3.25)where the (isotropi) glass form fator
f q c
and ritial amplitudeh q
desribe the spatialpropertiesof the metastableglassystate. The ritialglassform fator
f q c
givesthelong-lived omponent of density utuations, and
h q
aptures loal partile rearrangements.Both an betaken as onstants independent on shear rate and density, asthey are
eval-uated from the verties in Eq. (3.23) at the transition point. All time-dependene and
(sensitive)dependene onthe externalontrolparametersis ontainedinthe funtion
G
,whihoften isalled '
β
-orrelator' and obeys the non-linear stability equationε − c ( ˙ γ) ( ˙ γt) 2 + λ G 2 (t) = d
The two parameters
λ
andc ( ˙ γ)
in Eq. (3.26) are determined by the stati struturefator at the transitionpoint,and take values around
λ ≈ 0.73
andc ( ˙ γ) ≈ 3
forS q
takenfrom Perus-Yevik approximation [33℄ for hard sphere interations [44, 45, 227℄. The
transition point then lies at paking fration
φ c ≈ 0.52
(indexc
for ritial), and theseparation parameter measures the relative distane,
ε = C (φ − φ c )/φ c
withC ≈ 1.3
.The'ritial'exponent
a
isgivenbytheexponentparameterλ
viaλ = Γ(1 − a) 2 /Γ(1 − 2a)
[162, 164℄.
The time sale
t 0
in Eq. (3.27) provides the means to math the funtionG (t)
to themirosopi, short-time dynamis. The Eqs. (3.21,3.22) ontain a simplieddesription
of the short time dynamisin olloidaldispersions via the initialdeay rate
Γ q (t)
. Fromthis model for the short-time dynamis, the time sale
t 0 ≈ 1.6 10 −2 R 2 H /D 0
is obtained.Solvent mediated eets on the short time dynamis are well known and are negleted
in
Γ q (t)
in Eq. (3.21). If hydrodynami interations were inluded in Eq. (3.21), all of the mentioned universal preditions of the ITT approah would remain true. Onlythe value of
t 0
will be shifted and depend on the short time hydrodynami interations.This statementremainsvalid,as long asthe hydrodynami interations donot aet the
mode oupling vertex in Eq. (3.23). In this sense, hydrodynami interations an be
inorporated into the theory of the glass transition, and amount to a resaling of the
mathingtime
t 0
, only.Obviously,the mathing time
t 0
alsoprovidesanupperut-o forthetime windowofthestrutural relaxation. At times shorterthan
t 0
the spei short-time dynamis matters.The ondition
γt ˙ 0 ≪ 1
follows and translates intoarestrition for the aessible rangeofshear rates,
γ ˙ ≪ γ ˙ ∗
, where the upper-ut o shear rateγ ˙ ∗
is onneted to the mathingtime.
The parameters