1.4 Definitions for binary mixtures and multicomponent systems
2.1.2 Ideal MCT Glass transition lines
Transition lines from ergodic to non-ergodic for several values of the volume fractionΞ¦ are shown in Figure2.1. Once in the usual dimensionless units of the hard sphere diameterπand once in the units of the mean interparticle distanceπβ1β3, where the potential is given as:
π½ Μπ’(π₯) =Μ ΜπΎπβΜπ Μπ₯
Μ
π₯ =π½ π’(π₯πΜ β1β3) =πΎπ1β3 πβπ Μπ₯πβ1β3
Μ
π₯ , so that Μπ=ππβ1β3 andΜπΎ=πΎπ1β3 (2.1) with the particle density beingπ= 6Ξ¦βπin units ofπ. From the shape of the potential one can already expect that the coupling parameter πΎ needs to grow as fast asππ to be large enough for the transition into the solid phase. Only in the limit of weak screeningπ β 0, where the1βπnature of the potential is dominant the slope of the transition line is smaller. Again as expected, for lower densities/volume fractions the transition line goes to higherπΎ values. Rescaled by the interparticle distance the lines for lowΞ¦ (0.05 and5 Γ 10β4) almost collapse, which indicates that they are in the dilute limit, where the hard sphere part of the potential is negligible. On the contrary, coming closer to the MCT glass transition for hard spheres atΞ¦ = 0.516[19] the hard sphere interaction comes into play and lowers the transition line considerably.
A recent work by Yazdi et al. [28] computed the transition line in the dilute limit without considering hard sphere interactions. They used the HNC closure for the computation of structure factors and found
1Note thatπ(π)is equivalent to the correlation of the density fluctuations on infinite time scalesπ‘ββ
2.1 Critical parameters and features at MCT transition lines
Figure 2.1: MCT ideal glass transition lines for different volume fractionsΞ¦as indicated in the legend of the right panel. In the left panel coupling parameterπΎand screening parameterπare given in units of the hard sphere diameterπ(as in Equ.1.1), in the right panel they are scaled to the mean interparticle distanceπβ1β3given by the particle densityπ= 6Ξ¦βπ. In the rescaled view the curves ofΞ¦ = 0.05and Ξ¦ = 5 Γ 10β4almost collapse, they are indicating the dilute limitΞ¦β0of the transition line.
that the glass transition line in units of the interparticle distance (ΜπΎ, Μπ) closely resembles the equation
ΜπΎπ(Μπ) = ΜπΎOCP πΜπ
1 +Μπ+Μπ2β2, (2.2)
whereπΎOCP is the limit of the so-called one component plasma (π = 0) where no screening microions are present. Yazdi et al. [28] derived Equ. 2.2 on the basis of the Lindemann criterion [54], which states that a crystal melts if the root mean squared displacementββ¨π2β©is bigger than a certain fraction of the mean interparticle distance (usually about10%). This is done in the harmonic approximation
Μ
π’β²β²(Μπ= 1)β¨Μπ2β©β const., withπ’Μβ²β²(Μπ= 1)being the second derivative of the Yukawa potential at the mean interparticle distanceΜπ. With the assumption thatβ¨Μπ2β©is constant along the transition line, Equation2.2 follows from π’Μβ²β²(Μπ = 1) = const. It is clear that this approximation is rather imprecise: Not only is the distance to the nearest neighbours assumed to be exactly Μπ = 1(the mean interparticle distance), interactions with particles further away are also neglected completely.
Taking results from simulations of dusty plasmas (also obeying the HSY potential) Vaulina et al. [55]
found that for Μπ β² 8 Equation2.2 agrees well with thecrystalmelting transition line. They found a value ofΜπΎOCP = 106(or 172 in units of the Wigner-Seitz radius) for the one-component-plasma limit.
Therefore it is an intriguing result that the MCT glass transition line obtained by Yazdi et al. [28] has the same form, just with a3.45times larger value for the OCP limit: ΜπΎOCP = 366. However, the comparison in Figure2.2shows that this simple form does not fit to the results obtained here. Especially for low Μπ, where Yazdi et al. [28] see a good agreement, the shape of Equ.2.2misses the data points. Even more, the OCP limit for this data is rather in the range ofΜπΎOCP= 150 β 160, half the value obtained by [28]. A reason for this discrepancy could be that the HNC closure relation tends to underestimate the principal peak of π(π) (see Figure 1.4), which results in an overestimation of the critical valuesπΎπ. The here obtained factor of about1.5compared to the crystal melting line seems to be more reasonable, especially due to the fact that MCT usually predicts transitions to be at lower densities or lower coupling strengths compared to simulations and experiments. For instance, the experimental hard sphere glass transition is found at the volume fractionΞ¦ = 0.575[18] which is lower than the MCT predictionΞ¦ = 0.516.
0 1 2 3 4 5 6 7 8
k
102 103
104 =5Γ10
4
=0.48
equ.2.2,=0.5 OCP=162 equ.2.3, OCP=151, b=0.9,c=0.5
10 1 100 101 102
k
100 101 102
ex p( k )
Figure 2.2: Trying to fit Equations2.2and2.3to the MCT glass transition lines shown above in Fig-ure2.1. In the right panel all curves are multiplied byexp(βΜπ)to get rid of the exponential growth of the curves. Obviously the very simple form given in Equation2.2(black line) does not fit to the results, not even for small Μπ. Curves for higher packing fractionsΞ¦run almost parallel to the low density line and show the same behaviour as the curve for the low density limit.
Another shape for the melting line in Yukawa systems was proposed in a recent work by Tolias et al. [56]:
ΜπΎπ(Μπ) = ΜπΎOCP πΜπ
1 +πΜπ+ (πβ2)Μπ2 (2.3)
They found values ofπ = 1.2 andπ = β0.43for the transition from crystal to liquid. For Μπ β² 8the data points obtained here are well represented by Equation 2.3with the choice π = 0.9andπ = 0.5. Interestingly, this is quite close to the intrinsic dependency ofπΎonπ, conveyed by the definition of the potential (cf. Equ.1.2), which isπΎ = πBβππef f2 ππβ(1 +πβ2)2. This similarity is a hint that the effective chargeπef f(π)at the glass transition line only has weak dependence on the screening parameterπ. We will see this also in the later discussions (cf. Section 2.3). Similar to the situation in crystal melting [55,56], also here the fit is only good for low screeningΜπ β²8. For largerΜπthe right panel of Figure2.2 reveals a different behaviour which is clearly not explicable by a simple argument like the harmonic approximation. An explanation could be that in the high screening limit Μπ > 10 the HSY potential resembles more and more the hard sphere potential, so that eventually the growth ofΜπΎπ(Μπ)is a bit slower than exponential.
An important question is to what extent the hard sphere part of the potential plays a role. This question is addressed in the following. In Figure2.3we see the dependence ofπΎπon the volume fractionΞ¦. For the left panel the screening parameter π is held constant in units of the hard sphere diameter, which results in an initially very fast decay of the MCT transition line with increasing density. This is due to the dependency ofπΎon the particle density. For low screeningπ= 3and in the rangeΞ¦ = 0.3 β 0.45the curve is almost flat, in this range the particle density does not change very quickly. Finally the transition line drops drastically again, when it comes closer to the MCT hard sphere transition atΞ¦ = 0.516. In the right panel of Figure 2.3the fixed parameter is Μπ, which means that the screening length is always the same fraction of the interparticle distance. Therefore the particle density does not play a role. Here, the increase of the volume fraction can be interpreted as an increase of the hard sphere diameter of the particles while the density is kept constant. As expected the transition line stays almost flat until a volume fraction ofΞ¦ = 0.45, where the drop occurs which is caused by the hard sphere interactions.
For a closer look at the shape of the transition lines close to the hard sphere transition limit all curves are presented again in Figure 2.4but now rescaled with a division byπΎπ(Ξ¦ = 0.515), the critical coupling parameter for the highest computed volume fraction Ξ¦ = 0.515. In rescaled units (right panel), this
2.1 Critical parameters and features at MCT transition lines
Figure 2.3: MCT ideal glass transition lines for different values of the screening parameter. In the left panel lines are shown for a constant parameterπin units of the hard sphere diameter, in the right panelΜπΎ(Ξ¦)is shown for constantΜπin units of the mean interparticle distanceπβ1β3. In rescaled units the dependency on the volume fraction only sets in when the hard sphere part of the potential is not negligible any more at aboutΞ¦ = 0.45
Figure 2.4: The same transition lines as in Figure2.3above, but now the curves are rescaled with a division by the value for the highest volume fractionπΎπ(Ξ¦ = 0.515). Again, lines for constantπ(left panel), where increasingΞ¦ means a density increase, are compared to the lines for constant Μπ(right panel), where increasingΞ¦means a diameter increase. Interestingly, there is a region where the critical coupling parameterΜπΎπgrows with increasing volume fractionΞ¦.
reveals a growing of the critical coupling parameter ΜπΎπ with increasing volume fraction fromΞ¦ = 0.3 to a maximum atΞ¦ = 0.45. This maximum is higher for for stronger screening. More precisely, for
Μπ= 8.87it is7 %higher than the base line, while for Μπ= 2.42it is only a3 %effect. It seems that the hard sphere interactions at a volume fraction of aboutπ = 0.45have a softening effect which is more pronounced if the repulsive interactions are rather short ranged (i.e. for largeΜπ). Only aboveΞ¦ = 0.47 the hard cores start to contribute to the stiffness of the system. In principle this means that there should be a reentrant glass-liquid-glass transition from liquid to glass and back to liquid induced by an increase of the volume fractionΞ¦. However, increasingΞ¦at constant ΜπΎand constantΜπis equivalent to increasing the radius of the particles without changing the Yukawa part of the potential; experimentally this cannot be achieved. Moreover, it is not clear whether the effect is not a feature that comes in by the rescaled mean spherical approximation (RMSA) used for the computation of the structure factors. Still, it seems interesting to verify if this effect can be seen in simulations. Lai et al. [27,57] observed a similar reentrant transition but they used a differently modified RMSA version, their results are not comparable to the data
presented here. As the left panel of Figure2.4shows, an experimentally more realistic increase of the particle density, while keeping the screening lengthπand the coupling parameterπΎ constant, does not yield such a reentrant transition.