SSAC-pv2007.QE522.CC2.

How Do We Estimate Magma Viscosity?

How does the viscosity of a silicate magma

vary with temperature, water content, and
crystal content?

Using a “best-fit” Core Quantitative Issue

Units
regression model to
estimate viscosity of Supporting Quantitative Issues
Models
a hydrous, Graphs
leucogranitic melt.

SSAC - Physical Volcanology Collection

Chuck Connor – University of South Florida, Tampa

© Chuck Connor. All rights reserved. 2007 Edited by Judy Harden 10/24/07 1
Preview

This module presents a calculation of the viscosity of a magma with

varying temperature, water content, and crystal content.

Slides 3-10 give some background on estimates of viscosity in

silicate melts and magmas.

Slide 11 states the problem. What is the density of a hydrous rhyolite

magma?

Slides 12-14 analyze the problem and prompt you to design a plan to
solve it. The problem breaks down into parts: estimating the viscosity
of a rhyolite melt using a statistical model, and estimating the change
in viscosity as crystals are added to the magma.

Slides 15-16 illustrate a spreadsheet that calculates an answer.

Slide 17 discusses the point of the module and provides a broader

volcanological context.

Slide 18 consists of some questions that constitute your homework

assignment.

Slides 19-21 are endnotes for elaboration and reference. 2

Background

What is viscosity?

Viscosity is a measure of a fluid’s

resistance to flow. Think of viscosity as a
coefficient that relates the stress applied to
a fluid and the fluids response. For
example:
du1
th
dx2
du1
where t is the stress (Pa), dx is the
2
Photo by L. Connor
resulting velocity gradient in the fluid, and
h is the viscosity (sometimes referred to as This basaltic lava flow emanating
from a very small vent on Mt. Etna is
shear viscosity in this context). about as viscous as a thick salsa.

3
Background

What are the units of viscosity?

Consider a fluid trapped between

two plates. When a stress is
applied to the upper plate while the
lower plate is held still, a vertical
velocity gradient is created in the
fluid. This velocity gradient is
equivalent to a strain rate. Since

du1
th Figure from Mader, 2006, Volcanic processes as a source
dx2 of statistical data, In: Mader et al., (eds) Statistics in
Volcanology, Geological Society of London, 1-14.

the units of viscosity are Pascal Please check for yourself that the
seconds (Pa s). units of viscosity must be Pa s, given
the equation and diagram.

4
Background

Newtonian and non-Newtonian viscosity

high slope, high viscosity
Any fluid, including silicate
melts, is considered Newtonian
if there is a linear relationship
between stress and strain. That
low slope, low viscosity
is, if:
du1
th
dx2
In a non-Newtonian fluid, the above linear relationship does not hold true. For
example, in a Bingham fluid:
du1
t  to  h
dx2
where to is a yield stress (also called yield strength) required to “get the fluid
moving”. Although still linear, a new term is added to the linear equation.

5
Make sure you can sketch the stress-strain relationship for a Bingham fluid on the graph.
Background

Photo by B. Hill

Why estimate the

viscosity of magma?

Viscosity controls the rate of magma

flow in response to a given pressure
change and the rate at which gas
bubbles and solids (rock fragments
and crystals) move through a magma.
Low viscosity magmas, such as some
basalts, can form very long lava flows
very rapidly. In contrast, rhyolite
magmas are viscous enough to limit
bubble movement. Rather than
escaping by buoyantly rising through
the magma, gas bubbles are trapped
in rhyolite magmas, expand, and
eventually create very explosive The photo shows a mildly explosive eruption of
eruptions. basaltic magma at Cerro Negro volcano,
Nicaragua, in December, 1995. Increasing
viscosity, due to crystal formation and cooling,
contributes to explosive activity. 6
Background

What are the viscosities of magmas?

Photo C. Connor

Magma viscosity varies, with temperature, A platinum sphere (density ~21 g cm-3,
water content, and composition, between melting point about 1768 °C) sinking in
NaAlSi3O8 melt at high pressure and high
about 102 Pa s and about 1011 Pa s. That is
temperature (4.2 GPa and 1700°C). From:
nine orders of magnitude! Magma viscosities K. Funakoshi, A. Suzuki and H. Terasaki,
are often determined experimentally, or Journal of Physics :Condensed Matter 14,
measured in the field. 11343-11347 (2002). 7
Background

Temperature and Viscosity

Viscosity is highly dependent on

temperature. Think of a pool of molten
glass. As the temperature drops, it will
take more and more applied stress to
make the glass flow.

An Arrhenian model of viscosity is one in

which viscosity is exponentially
dependent on temperature: This glass is an alkaline silicate melt
with low viscosity at 1000 °C and very
 E  “workable” into shapes at about 800 to
h(T )  ho exp 
 RT 
900 °C. At about 700 °C, the viscosity
is high enough for the shape to not
deform under its own weight.
where ho is viscosity under standard Learn More about the
temperature conditions, E is the Arrhenian model
activation energy, R is the universal gas
constant, and T is temperature. Learn More about E/RT 8
Background

For most silicate melts, even Arrhenian models do not work!

Over the last 25 years, magma Water “de-polymerizes” melts,

physicists have discovered silicate changing their viscosities.
melts are often non-Arrhenian,
largely because of the additional
impact of water on viscosity. Water
breaks chains of silica polymers in
melts, and the shorter polymers
result in a lower viscosity. One
non-Arrhenian model of viscosity is
in the form of the Vogel-Fulcher-
Tammann (VFT) equation:
b( H2 O)
log h(T , H2 O)  a( H2 O) 
T  c( H2 O)
where a, b, and c are constants that
must be determined for specific
melt compositions using lab
experiments. Images created by Bill Rose 9
Background

For silicate magmas, even non-Arrhenian models do not work!

Even when coefficients a, b, and c are estimated for the VFT

equation for a specific melt composition, other factors can
strongly influence viscosity. Picture the sudden onset of
crystallization, a virtual snowstorm of microlites (small
feldspar crystals) in an ascending magma. These crystals
increase the bulk viscosity of the melt-crystal mixture.

Viscosity of the melt-crystal magma

mixture can be estimated with:
 52
  
hmagma  hmelt  1   
o
where  is the volume fraction of crystals
and o is known as the maximum packing
fraction of a solid. Microlites in a now frozen magma
10
Problem

Calculate the viscosity of a rhyolite melt with 2 weight percent

water at 900 °C and with 0.03 volume fraction crystals.

Use a form of the VFT model to

solve the problem.

More generally, it is useful to

use the VFT model to explore
the nature of rhyolite magma
viscosity. Namely, how does
viscosity vary as a function of:

• Weight percent water?

The volcanic Isla San Luis off the coast of Baja, California,
• Temperature? has a Holocene rhyolite dome at its center. The dome is the
• Volume fraction of crystals in mass of dark rocks forming a steep mound, indicative of the
high viscosity of this magma when it erupted. (Photo from
the magma? the Smithsonian Global Volcanism Network webpage).

11
Designing a Plan, Part 1

Given the water content, Give answer in Pascal

temperature, and volume seconds (Pa s).
fraction of crystals in the
magma, calculate the
viscosity. Notes:
(1) Coefficients used in the VFT model remain
to be determined at this stage.
You will need to: (2) Different models for calculating magma
viscosity from melt viscosity exist; a
• implement the VFT comparatively simple model based on
model to solve for melt spherical “crystals” will be implemented
viscosity. here.
(3) This VFT model is actually compositionally
• calculate the magma specific, it is only valid for leucogranitic
viscosity from melt melts. A leucogranitic melt is a rhyolitic
viscosity, given the magma enriched in aluminum, often
volume fraction of characterized by biotite phenocrysts and a
crystals suspended in pale colored groundmass.
the magma.
12
Designing a Plan, Part 2

Calculate the melt viscosity.

We need to put the VFT model into practice.

A form of the VFT model proposed by Hess and Dingwell (1996, Viscosities of hydrous leucogranitic melts:
A non-Arrhenian model, American Mineralogist, volume 81, 1297-1300) is used to solve the problem.

Hess and Digwell (1996) found a best-fit regression model for

non-Arrhenian viscosity: The values of these

b1  b2 ln( w)

coefficients are:

logh  a1  a2 ln( w) 

a1 = -3.54

T  c1  c2 ln( w)

a2 = 0.83
b1 = 9601
b2 = -2366
where h is viscosity, w is weight percent water, T is temperature (K), c1 = 196
and a1, a2, b1, b2, c1,and c2 are coefficients of their regression c2 = 32
model.

Hess and Dingwell estimated the values of the coefficients in

the model by regression. That is, they found the best-fit
values of these six coefficients using data from 111 viscosity Learn more
experiments conducted at various temperatures on rhyolites about regression
with various water contents between 0.2 and 12.3 wt % and
temperatures between about 750 K and 1473 K. Their model
is not valid for other conditions! 13
Designing a Plan, Part 3

Estimate how the viscosity changes with Recall that melt is by

the addition of crystals to the magma. definition a liquid,
whereas magma contains
The change in viscosity of a melt, with the addition of solid melt + solids and gas.
particles like crystals or xenoliths, is a complex problem. A
simplified answer is provided by the following equation.

5
hmagma  hmelt 1    
2

 o

Here the crystals are taken to be spheres (not a particularly good

assumption about the shape of many crystals, but a good leading Approaching the maximum
order approximation nevertheless). The viscosity then changes packing fraction in a melt
according to a power law. As the crystal spheres become more populated with “ideally
abundant, they pack closer together until they occupy much of spherical” crystals. Although
the total volume, with melt only left in the open spaces between the “crystals” are densely
spheres. This is the maximum packing fraction of the solid, o. packed, there is still room for
Assume o = 0.6 melt in the gaps.
14
Carrying Out the Plan: Spreadsheet to Calculate the Viscosity

B C D E F
2 Calculate the Viscosity of a Rhyolite Magma using the VFT Model
3
4 Coefficients of the VFT Model A cell containing a
5 a1 -3.54
number that is given
6 a2 0.83 information
7 b1 9601
8 b2 -2366
9 c1 196 A cell containing a number
10 c2 32 that is a constant for this
11 problem
12 Max. packing
13 fraction (Ø o ) 0.6
14
15 Given Conditions
A cell containing a
16 Temperature (K) 1173
17 Water Content (wt %) 2
formula
18 Vol. fraction of crystals 0.03
19
20 Calculated results
21 log (melt viscosity)(Pa s) 5.37303
22 melt viscosity (Pa s) 236064.1
23 magma viscosity (Pa s) 268361.9

At this point, be sure to implement this spreadsheet and check

that your formulas duplicate the values shown. You will need
this spreadsheet to complete the end-of-module assignment.
15
Carrying Out the Plan: Spreadsheet to Calculate the Viscosity
B C D E F
2 Calculate the Viscosity of a Rhyolite Magma using the VFT Model While the calculation of a single viscosity value
3
4 Coefficients of the VFT Model is useful, it does not allow you to see trends.
5 a1 -3.54 The spreadsheet is easily modified to show
6 a2 0.83
7 b1 9601 results of numerous calculations graphically.
8 b2 -2366
9 c1 196
10 c2 32 Add cells to your spreadsheet so you
11
12 Max. packing
can graph results and observe trends.
13 fraction (Ø o ) 0.6
14
15 Given Conditions
A cell containing a
16 Temperature (K) 1173 number that is
17 Water Content (wt %) 2
18 Vol. fraction of crystals 0.03
given information
19
20 Calculated results
21 log (melt viscosity)(Pa s) 5.37303
A cell containing a
22 melt viscosity (Pa s) 236064.1 formula
23 magma viscosity (Pa s) 268361.9
24
10
25 Log (Melt Viscosity) (Pa s)
9

Log (viscosity) (Pa s)

26 Temperature (K)
27 water content (wt %) 1000 1200 1400
8
28 0.5 9.490653 6.838885 5.052167
7
6 1000 K
29 1 8.401542 6.022749 4.434252
30 1.5 7.721182 5.516466 4.052193 5 1200 K
31 2 7.217990 5.143743 3.771548 4 1400 K
32 2.5 6.815539 4.846682 3.548257 3
33 3 6.478611 4.598688 3.362109 2
34 3.5 6.187923 4.385241 3.202084 1
35 4 5.931723 4.197508 3.061484 0
36 4.5 5.702294 4.029700 2.935923 0 5 10
37 5 5.494286 3.877809 2.822367
38 5.5 5.303829 3.738944 2.718627 Wate r Conte nt (wt %)
39 6 5.128033 3.610943 2.623071
40 6.5 4.964677 3.492150 2.534447
41 7 4.812019 3.381267 2.451773
16
42 7.5 4.668662 3.277256 2.374266
43 8 4.533474 3.179272 2.301290
What you have done
You have calculated the viscosity of a hydrous rhyolite magma using the non-Arrhenian
VFT model and a simplified model to account for volume fraction of crystals.

The viscosity of silicate melts is fundamental to the nature of magma transport. Viscosity controls the form
of lava flows and strongly contributes to the nature of volcanic eruptions. In fact, viscosity is a powerful
constraint on the nature of many geological phenomena, such as convective flow in the mantle, the
deformation of the ductile lower crust, and particle transport in the atmosphere and in rivers.

You have discovered that it is possible to estimate the viscosity of magmas using highly nonlinear models
that account for the affects of temperature, water content, and volume fraction of crystals.

Although based in physical principles, these models are statistical in nature. That is, the coefficients of the
VFT model were estimated from experimental data using a sophisticated regression technique. It is likely,
therefore, that these models will improve with additional experiments. In addition, it is clear that it is
inappropriate to use such models outside of the range of experimental data from which they are derived.

Useful papers that discuss magma viscosity in detail:

Shaw, H.R., 1972, Viscosities of magmatic silicate liquids: An empirical method of prediction. American
Journal of Science, 272, 870-889. (Shaw was one of the first people to discuss the relationships between
thermodynamic properties of silicates and physical properties of magmas).

Hess, K-U., and D.B. Dingwell, 1996, Viscosities of hydrous leucogranitic melts: A non-Arrhenian model,
American Mineralogist, 81, 1297-1300. (the VFT method used in this module)

Spera, F., 2000, Physical properties of magma, In: Sigurdsson et al., eds., Encyclopedia of Volcanoes,
Academic Press, 171-190. (an accessible discussion)
17
End of Module Assignments

1. Make sure you turn in a spreadsheet showing the worked examples.

2. Plot a graph that shows the dependence of rhyolite magma viscosity on crystal content for
hydrous melts H20 = 2 (wt %), 6 (wt %), and 12 (wt %) at 900 °C. Now consider a magma
rising through a volcano conduit. How does water content, crystal content, and temperature
affect the rate of ascent, all other things being equal? Which factor would you say is most
important and why?
3. Each of the six coefficients (a1-c2) used in the VFT model is estimated. Suppose the error in
these estimates is about 20%. That is, the coefficients could be 20% lower or 20% higher
than the values reported here. Given this uncertainty, what is the range of viscosities that you
estimate for a rhyolite melt that is 4 weight percent water at 900 °C.
4. The viscosity of some brands of peanut butter is about 8000 Pa s. Use your spreadsheet to
investigate the range of conditions (water content, temperature, crystal content) under which
rhyolite magmas would have approximately this viscosity.
5. What happens to the viscosity of peanut butter when you add whole peanuts to the mixture?
Why?
6. The ‘obsidian problem’ refers to the fact that some rhyolite magmas reach the surface very
quickly, but do not erupt explosively. Rather, they form effusive eruptions that make lava
domes of very glassy (crystal poor) lava. Use your spreadsheet to investigate this problem.
Under what circumstances might rhyolite magmas form such domes rather than erupt in a
spectacular explosion?

18
More about the Arrhenian Model

The Arrhenian model is named for Svante Arrhenius, who

developed a method of predicting the increased speed of a
chemical reaction with increased temperature. His equation has
the form:
 E 
k  A exp  
 RT  More about E,R,T

where k is the rate coefficient that describes how much faster the reaction will
proceed, A is the Arrhenius coefficient, which varies with the specific chemical
reaction, E is the activation energy, R is the gas constant, and T is temperature.

Note that this is the exact form of the Arrhenian viscosity model, with A replaced
by the viscosity of the fluid at some standard temperature condition.

Arrhenian models are also common in statistics and used to predict the higher
rate of failure of just about anything at higher temperatures. For more about
the statistical application of the model, see:
http://www.itl.nist.gov/div898/handbook/apr/section1/apr151.htm
19
Return to Slide 8
ERT
E stands for activation energy, literally the energy “hump” or barrier, which must be
overcome for a reaction to proceed. In viscosity, the energy that must be applied to
get it all flowing. The unit of activation energy is the joule mol-1, that is, the energy
required to get one mole of substance to react, or flow.

R stands for the universal gas constant and is equal to Boltzmann’s constant x
Avagadro’s number. R relates energy to temperature, T. The units of Boltzmann’s
constant, found experimentally are joule per K, k = 1.380 x 10−23 J K-1, so the
universal gas constant is R=8.314 J mol-1 K-1.
Prove to yourself that E/RT is dimensionless. What happens to E/RT as
T increases but the activation energy remains constant? What happens
to the value of exp[-E/RT] as T increases?

So in the Arrhenian model, the energy required to make the fluid react (in this
case react by flowing) under standard conditions is constant. The viscosity
decreases with temperature because the ratio –E/RT decreases with
increasing temperature. Remember, this is only a model, because in reality
what is happening is that the number and length of silica polymer chains is
decreasing with increasing temperature. The model is a symbolic, in this
case mathematical, representation of reality.
20
Return to Slide 8
The Nature of Regression

In statistics, regression is used to determine the relationship between two or more

variables. This relationship is usually characterized by the magnitude of
coefficients. For example, in linear regression, y= mx +b, the relationship between
variables y and x is determined by the magnitude of the coefficients m and b.

Uncertainty exists in any regression

model, because the relationship
between variables is not ideal. In the
plot shown, the relationship between
cumulative volume and time looks
linear, but there is “certainly uncertainty”
(i.e., measured points plot off the line).

The best-fit line is a statistical model,

that is, a model that includes some
A linear regression showing the relationship between
component of random (or unexplained) two variables (time, and cumulative eruption volume)
variation. When we use a statistical for eruptions of Cerro Negro volcano, Nicaragua.
Actual values are plotted as solid circles, the line is
model to estimate viscosity, we accept the best-fit regression. Data from Hill et al., 1998,
the fact that some random variation will GSA Bulletin.

not be accounted for by the model!

21
Return to Slide 13