8-3 Viscosity vs.

temperature for different

compositions; viscosity of granitic liquids.
Now from the diagrams that you've just seen, you might be beginning to develop the
hope that the complexity of these things will stop with a simple model, where the more
viscous melts on the left will be more or less linear in Arrhenian space.
And then, as we add other depolymerizing agents to them they'll become more curved.
But unfortunately, we have to go to yet another level of complexity, which we observe
in these melts.
And you see it in the following diagram, where you have a variety of silicate melt
compositions.
And you see that the viscosity temperature relationships, on occasion, also cross each
other in these diagrams.
And this is really, in essence, the last level of graphical complexity, which needs to be
resolved in order to describe the compositional and temperature dependence in a
variation of these viscosities.
But you can see one famous, famous example of this crossover which is the system
albite-anorthite.
If you look carefully, you'll see the viscosity temperature relationship for albite and that
for anorthite.
The two end-members of the plagioclase series of feldspar mineralogy.
And you can see that at high temperatures, the albite has a far higher viscosity than the
anorthite.
If you follow both of the curves down, in quotation marks, to the upper-right part of the
diagram, where they are at lowest temperatures, you can see that the viscosity of the
anorthite melt has now risen to be more than that of the albite melt.
This is a fact.
In fact, the very measurements, which were done on these systems almost a century ago
now showed this fundamental discontinuity or discrepancy between high and low
temperature, relative viscosity determinations of anorthite and albite.
Now this was a classic case of a paper which was originating from the logic that one
would take minerals or mineral series, and melt them and look at their respective
viscosities.
After all, plagioclase is one crystallographic series, a solid solution series.
And, therefore, one would anticipate that they might be related in their liquid structure.
But if you look at the third component which is next to them here, diopside, you'll see
that actually the temperature dependence of the diopside viscosity looks very similar to
that of anorthite.
Looks like they're related in some way as a family of behavior, whereas the albite looks
quite different.
And I need only remind you that the anorthite and the diopside both have 50% of SiO
2
,
whereas the albite has 75%, to suggest to you that perhaps the more valid comparison in
terms of structure is between the two liquids with 50%, who indeed have very, very
similar degrees of curvature on their viscosity-temperature relationships.
So, there's a lesson in there.
We have to careful about what lessons we take from solid solutions in applying it to our
anticipation of what liquid structures are going to be related one to the other.
There are significant differences.
Now what do we do about these crossovers.
Well, there's one thing that we can do in a very, very simple manner and it was done
quite some time ago now which is, we simply normalize to a value of Tg.
Tg as you will recall, we will have to define in terms of a certain classical time scale.
But if we do that, then we have a temperature.
And if we normalize to Tg as a constant value of viscosity then we can get a curve
which essentially hangs the data, if you will, on a constant viscosity point.
And in this diagram, which is plotting that normalization to the glass transition
temperature, you can see that in the upper right part of the diagram, all of the curves
converge on a single value by definition, because they're all plotted, normalized to the
same value of glass transition temperature, which is defined as a constant value of
viscosity.
So they are forced to go through the upper right-hand corner.
They all have a series of viscosity measurements associated with them, which of course,
goes certain distance down to the lower left corner.
But cannot go all the way, because all the way means infinite temperature at the lower
left-hand corner.
And you see at the top the word strong and at the bottom the word fragile.
And this was a proposal that was made in the 1960s and 1970s.
There was a lot of discussion of it. And the idea was that these liquids, which are
defined as strong, show very little variation of the slope of their viscosity versus
temperature, as a function of absolute temperature, whereas the fragile ones show great
curvature, great variation in their slope, as they change the absolute temperature of the
system.
And one inference which one can try to make from such a variety of behavior is that the
structures, which exist in the liquids, which have very little variation in slope, and SiO
2

is the upper most curve that you see in this diagram, are structures that simply do not
change much with temperature.
Such that the structure of the liquid that you're looking at 1600 Celsius is the same as
1800 Celsius is the same as 1200 Celsius.
And, therefore, the resistance to flow and the slope of viscosity versus temperature is
more or less a constant.
Whereas, the fragile liquid structures that are the description of those on the lower right,
imply that there is a lot of configurational or indeed, in general, structural change as
these liquids are heated up.
And that, therefore, the liquids, which are being compared to produce such a curve of
viscosity versus temperature, one liquid, one chemistry, and one crucible, is
nevertheless involving a comparison of very different structures of the liquid as we go
through temperature.
If you will, an ordered liquid at low temperature and a very disordered liquid at high
temperature.
And if there is a large amount of order-disorder possible in the temperature range of the
viscosity measurements, then in principle we could infer that this could cause a strong
variation in the slope of the viscosity-temperature relationship, as we see in fragile
liquids.
So what are we to do?
We have curved viscosity-temperature relationships.
They appear to break all the rules of simple parameterization that we might try to use,
based on stoichiometry, based upon relative silica contents, based upon the individual
identity of cations once we step into multi-component liquids.
Well, one proposal has been made already in the 1960s as a basis for viscous flow at all
in liquid systems.
And it's called the Adam-Gibbs model after the two authors of the study.
And it's based on the simple principle or notion that configurational entropy in these
systems is inversely proportional to viscosity.
The idea being that if you have more entropy then you have more opportunities for
reorganization or relaxation of the system.
And, therefore, it's easy to shear and to accommodate shear.
And, therefore, less resistance to shear is produced and, therefore, a lower viscosity.
Now if this is true and if we had sufficient thermodynamic data on configurational
entropy, then it would seem that the wise thing to do is to normalize the viscosity
temperature relationship to configurational entropy.
And that's what's done in this following diagram.
If you look at the top part, you can see the strongly curved temperature dependence of
the viscosity of anorthite of CaAl
2
Si
2
O
8
.
And if you look at the bottom, you can see the same data plotted normalized to the
configurational entropy in the denominator of the one on T axis.
And you can see that it linearizes the plot.
So it's a strong statement to suggest that it is the configurational entropy which is
expressing the ease with which viscous systems can flow.
If I reduce configurational entropy, viscosity will increase.
If I increase configurational entropy, viscosity would decrease.
Now there are several contributors to configurational entropy.
Temperature itself, of course, contributes to the absolute value of configurational
entropy, and of entropy in general in these systems.
But chemistry, of course, also directly contributes to configurational entropy.
Because a part of what we're describing is the configuration, is the mixing or the
unmixing, the order or the disorder of individual cations on particular sites within the
liquid structure.
And so you can clearly see that there will be entropy of mixing as well, which we
should be able to observe if the entropy theory for the origin of viscosity is correct.
Now in the next diagram, I'd like to show you what water does to the viscosity of a
haplogranitic liquid.
A calc-alkaline rhyolite, one which has approximately a tectosilicate structure, is about
70 to 75% silica.
And the rest is albite and orthoclase-like components in the chemical sense.
These are the most viscous natural liquids that we encounter.
These are rhyolitic eruptions and calc-alkaline arcs across the Earth.
And in the dry state, these have the highest viscosity of any natural volcanic liquid that
we know.
But we also know they have a healthy appetite for water at depth.
And they dissolve several, several weight percent of water at modest depths of a few
kilometers from which individual eruptive systems actually start on the day of the
eruption.
And what you can see here is a plot of the logarithm of viscosity versus the added water
content to a calc-alkaline rhyolite.
And you can see clearly a strong drop in the viscosity as you add the first half, the first
one weight percent of water to these liquids.
And then you can see a relative leveling off or flattening or no further dependence of the
viscosity upon the water content.
And so this strong, water dependence of viscosity, especially in these high silica
systems, these rhyolitic systems, which are the most explosive ones that we know of on
Earth, is probably a direct link between the two things.
Between the physical properties on one side and between the explosive behavior on the
other side.
And we'll return to this a little later.
If you see the parametrization, the very first non-Arrhenian parametrization which was
possible, which was back in 1996.
Then you can see here the log of the viscosity plotted versus water content on the left
hand side.
And you can see this strongly nonlinear decrease of the viscosity with added water
content.
And you can see on the other side, the temperature dependence of the viscosity, of
individual liquids which have different varying amounts of water.
And it runs all the way from 0.01 weight percent of water to 8 weight percent of water.
You can see as you go from left to right on the right-hand part of the diagram, that
you're going for more or less an Arrhenian, a linear expression in this plot for the dry
melt or the nominally 100 of a weight percent water melt, but as you go towards the
right-hand side, you see that the, the curves become more curved.
They deviate more strongly from Arrhenian behavior.
But interestingly enough, they do it in a slightly different way than peralkaline melts do.
If you look to the upper right part of the diagram, you'll see, despite the fact that these
things are becoming less Arrhenian that the slopes of the viscosity-temperature
relationship, particularly in the upper right-hand part of the diagram, are still remaining
relatively low.
It's as though they're falling away with their value of the slope, as opposed to the
peralkaline melts that we saw an example of earlier, where there's more of a steepening
of the slope as you go to these very high water contents.