THERMOCALC Course 2006

1
THERMOCALC Course 2006

• Chemical systems, phase diagrams, tips tricks
• Richard White
• School of Earth Sciences
• University of Melbourne
• rwwhite_at_unimelb.edu.au

2
Outline

• What chemical system to use
• differences between systems
• Choosing a bulk rock composition
• Getting started
• The shape of lines fields
• Starting guesses
• Problem solving
• Using diagrams to interpret rocks
• What diagrams to draw

3
What chemical system to use

• Before you embark on calculating diagrams, you
  need to work out what chemical system to use.
• It must be able to allow you to achieve your aims
• Must be as close an approximation to nature as
  possible
• Using a single system throughout a study provides
  a level of consistency
• If you are modelling both pelites and greywackes
  you could use KFMASHTO for pelites and NCKFMASH
  for greywackes BUT NCKFMASHTO for both is better
• With very different rock types (eg mafic
  pelite) you may have to use different systems

4
What chemical system to use

• The system you choose also depends on what you
  are trying to do
• Forward modelling theoretical scenarios and
  processes in general
• Simpler systems may be used to illustrate these
  more clearly
• Inverse modelling of rocks for P-T info
• Larger systems should be used to get equilibria
  in the right place

5
What chemical system to use

• The rocks minerals tell you what system you
  need to use
• What elements are present in your minerals
• Eg Grt in metapelite at Greenschist has Mn
• MnKFMASH better than KFMASH
• Grt at high -P in metapelite may have significant
  Ca
• NCKFMASH better than KFMASH
• Spinel bearing rocks-need to consider Ti Fe3
• KFMASHTO better than KFMASH
• Getting this right at the beginning saves later
  problems
• It may be tempting to try and use simple systems
  (less calculations)
• If in doubt, the larger system is safer

6
What chemical system to use

• When adding components, we need to consider what
  minerals these components will go in
• THERMOCALC has to be able to write reactions
  between endmembers.
• Must have this component in more than 1 endmember
  and in reality as many as we can
• May involve us adding new phases to the modelling
  that may or may not actually be in our rock.
  mineral stability is relative to other minerals.
• THERMOCALC is simply a tool. It can only give us
  information within the parameters we decide.

7
What chemical system to use

• An example
• The effect of Fe3 on spinel stability.
• Can model spinel in KFMASH, but this doesnt
  consider Fe3
• Could model in KFMASHO, but is this
  satisfactory?- NO
• Why? Must consider other minerals that take up
  Fe3, eg the oxides.
• When modelling the oxides, we should also
  consider Ti (e.g. ilmenite, magnetite, haematite)
• So a better system is KFMASHTO

8
What chemical system to use

• Why is the right system so important
• If we are trying to model rocks, our model system
  must approach that of the rock as closely as
  possible.
• Minor components can have a big influence on some
  minerals hence some equilibria.
• Minor minerals in a rock will change the
  reactions and their positions on a petrogenetic
  grid
• Ignoring a component can artificially alter the
  bulk comp
• Eg a High-T granulite metapelite
• FMAS will show relationships between many
  minerals but they wont be in the right P-T space
  or possibly the right topology.
• The rock will not see any of the FMAS univariant
  equilibria

9
What chemical system to use

• Eg a High-T granulite metapelite cont.
• These rocks will contain melt at peak,
  substantial K, some Ca, Na, H2O (in melt crd)
  and Ti Fe3 in biotite spinel if appropriate.
• KFMASH doesnt do a bad job (backbone of the main
  equilibria) but will make modeling melt oxides
  problematic and ignores plag.
• So to do it properly we need to model our rocks
  in NCKFMASHTO.
• Modeling in these larger systems does have major
  benefits for getting appropriate model bulk rock
  compositions from real rocks
• Thus size is important!

10
differences between systems

• Will concentrate on going from smaller to larger
  systems.
• New phases to add
• New endmembers to existing phases
• Start with petrogenetic grids in particular
  invariant points.
• Need to consider the phase rule
• Relationships are different for adding different
  numbers of phases components
• V C - P 2
• V, Variance C, Number of components P, Number
  of phases
• And Schreinermakers rules

11
Some examples I
12
Some examples II
13
Building up to bigger systems I

• Building up from KFMASH for example to KFMASHTO,
  NCKFMASH or NCKFMASHTO requires several
  intermediate steps.
• The grid can only be built up one component at a
  time
• Each of the new sub-system topologies has to be
  determined
• To go from KFMASH to KFMASHTO we have to make the
  datafiles and calculate the grids for the
  sub-systems KFMASHO KFMASHT before we make the
  KFMASHTO datafiles and grid.

14
Building up to bigger system II
15
Building up to bigger systems III
16
Building up to bigger systems IV
17
Building up to bigger systems V
18
Building up to bigger systems VI
19
Building up to bigger systems VII

• On P-T grids we can get either more or less
  invariants.
• KFMASH to KFMASHTO More
• KFMASH to NCKFMASH Less
• Overall more possibilities for more fields in
  pseudosections
• The controlling subsystem reactions are still
  present but
• may involve additional phases, or
• be present as higher variance relations
• Will shift in P-T space

20
Bulk compositions

• Pseudosections require that a bulk rock
  composition in the model system is chosen.
• For diagrams that are directly related to
  specific rocks this bulk rock info should be
  derived from the rocks themselves
• But must reduce the measured bulk to the model
  system-must be done with care
• Thus, choosing a bulk rock composition will
  depend on your interpretation of a volume of
  equilibrium
• May be different for different rocks
• May vary over the metamorphic history

21
Bulk compositions

• Ways of estimating bulk rock composition
• XRF- good if you have large volumes of
  equilibration.
• Quantitative X-Ray maps-good for analysing
  smaller compositional domains. Clarke et al.,
  2001, JMG, 19, 635-644
• Modes and compositions-Less reliable,but can work
  on simple rocks.
• Wt bulks have to be converted to mole to use
  in THERMOCALC
• Mol wt / mw
• The amount of H2O has to generally be guessed if
  not in excess.
• Fe3 may also require guess work, or measured
  another way

22
Bulk compositions III

• The bulk rocks we use in THERMOCALC are
  approximations of the real composition as many
  minor elements are ignored
• The further our model system is from our real
  system the harder it is to accurately reproduce
  the mineral development of rocks.
• Eg. Using KFMASH to model a specific metapelite
  raises problems with ignoring Na, Ca, Ti, Fe3.
• Location and variance of equilibria, modifying
  our bulk rock so it is in KFMASH.

23
Bulk compositions IV

• Scales of equilibration we are trying to model.
• Commonly we interpret the scale of equilibration
  to be smaller than a typical XRF sample size
• Our prograde and peak scale of equilibration may
  have been large but if we are trying to model
  retrograde processes this scale may be small
• Our rocks may contain distinct compositional
  domains, driven by a slow diffuser eg. Al
• High-Mn garnet cores may be chemically isolated
  from the rest of the rock
• We need to adjust our bulk composition to
  accommodate these features

24
Bulk compositions V

• How do we adjust our bulk
• Use a smaller scale method for estimating bulk
  such as X-ray maps
• Useful only on quite small scales
• Can directly relate measured compositions to
  textures and hence effective bulk compositions
• Modify the bulk composition using the modes
  compositions given by THERMOCALC
• Can model progressive partitioning by doing this
  in steps
• Cheap simple, but still need to do the
  petrography mineral analyses to establish the
  nature of the element distribution

25
Bulk compositions VI

• Two examples involving removing the cores of
  garnets from our bulk rock
• E.g, 1. Using X-ray maps to remove garnet cores
  in prograde-zoned garnets.
• Based on a paper by Marmo et al 2002, JMG
• In this paper different amounts of core garnet
  are removed to model the prograde mineral
  assemblage development in the matrix.
• E.g. 2. Using THERMOCALC to remove the cores of
  large garnets so that the retrograde evolution of
  a rock can be assessed.
• Will show how this is done

26
Example 1
27
Example 1
28
Example 1
29
E.g. 2
30
E.g. 2
31
E.g. 2 removing the garnet cores

• Calculate the full bulk equilibria at the
  desired P T.
• There is a new facility to change min props
  called rbi
• We can use rbi to set our bulk comp via info on
  the modes compositions of minerals
• rbi info can be output in the log file

32
Adjusting bulk from calculated modes

• Bulks can be set/adjusted using the mineral
  modes(mole prop.) and the mineral compositions
• Uses the rbi code (rbi read bulk info)
• You can make thermocalc output the rbi info into
  the log file using the command printbulkinfo
  yes

33
Adjusting bulk from calculated modes

• The bulk rock can be read from rbi code in
  thetcd file instead of the usual mole oxide s
  

34
Bulk compositions

• We can use the method shown in e.g. 2 for any
  phase or groups of phases
• This is how we make melt depleted compositions
  for example.
• We can divide a bulk rock into model
  compositional domains
• Again, what we do here is determined by our
  petrography interpretation of what processes
  may go on

35
Getting started

• In most of the pracs you will be largely
  finishing partly completed diagrams
• In reality, you will need to start from scratch
• Knowing where to start is not always
  straight-forward
• It is easy to accidentally calculate a metastable
  higher variance assemblage rather than the stable
  lower variance one
• Some rocks are dominated by high variance
  assemblages in big systems (eg greywackes,
  metabasics)
• If your system has lots of univariant lines you
  can look at them

36
Getting started

• In large systems, there are few if any univariant
  reactions that will be seen
• Need to look for higher variance equilibria
• There are some smaller system equilibria that
  form the backbone for larger systems
• The classic KFMASH univariant equilibria occur as
  narrow fields in bigger systems in pelites
• NCFMASH univariant equilibria in metabasics may
  still be there in some form in bigger systems

37
Getting started

• In most cases the broad topology of a
  pseudosection will be well enough understood that
  you will know what some of the equilibria will
  be.
• Follow logic most metapelites see the reaction
  bi sill g cd in some form
• Look at diagrams in the same system and with
  similar bulks to your samples
• Sometimes you may be trying to calculate a
  diagram in an unusual bulk or one that hasnt be
  calculated by anyone
• Diagrams that are dominated by high variance
  equilibria may be hard to start.
• What is the right equilibria to look for

38
Getting started

• There are two ways to approach this problem
• Calculate part of a T-X or P-X diagram from a
  known bulk to your unknown bulk
• Work your way across the diagram, find an
  equilibria that occurs in your new bulk and build
  up your P-T pseudosection from there
• Use the dogmin code in THERMOCALC to try and
  find the most stable assemblage at P-T
• This is a Gibbs energy minimisation method
• May not be able to calculate the most stable
  assemblage and your answer could be a red
  herring.
• Method 1 is far more reliable, and if possible
  should be used in preference to method 2

39
e.g.
40
Drawing up your diagram

• It is always wise to sketch the diagram as you go
• No need to make this sketch an in-proportion and
  precise rendering of the phase diagram-thats
  what drawpd is for
• The sketch is there to help you draw the diagram
  and for labelling
• Very small fields have to be drawn bigger than
  they really are

41
(No Transcript)
42
Shapes of fields lines

• Most assemblage field boundaries on a
  pseudosection are close to linear
• Strongly curved boundaries do occur and can be
  difficult to calculate in one run
• Very steep very shallow boundaries reactions
  can also present problems
• For shallow boundaries calculate P at a given T

calctatp ask You are prompted at each
calculation calctatp yes You input P to get
T calctatp no You input P to get T
43
Curved boundaries I
44
Curved boundaries II

• In T-X P-X sections, X is always a variable so
  near vertical lines require very small X-steps to
  find them.
• Curved lines with two X solutions have to be
  done over small T or P ranges
• Overall changing the P, T or X range will help as
  will changing the variable being calculated
• Changing from calc T at P to calc P at T.

45
Starting Guesses

• THERMOCALC uses the starting guesses in the tcd
  file as a point from which to begin the
  calculation.
• These starting guesses have to
• Be reasonably close to the actual calculated
  results
• Have common exchange variables in the right order
  for the minerals eg. XFe ggtbigtcd
• This may mean having to change the starting
  guesses to calculate different parts of the
  diagram
• When changing starting guesses, it is best to
  create a new tcd file and change the guesses in
  that so your original file remains unchanged.
• This way you will always have all the files
  needed to calculate the whole diagram

46
Changing starting guesses

• A good way to ensure starting guesses are
  appropriate is to use output comps as starting
  guesses.
• These can be written to the log file in the form
  shown on the left
• To do this the following script printguessform
  yes goes into the tcd file
• There are a few tricks to remember when doing
  this, especially with phases with the same coding
  separated by a solvus
• Have to ensure the starting guess is on the right
  side of the solvus

47
Common problems with starting guesses

• THERMOCALC wont calculate all or part of a given
  equilibria
• THERMOCALC gives the same composition for two
  similar minerals that should be separated by a
  solvus
• Eg. Ilm-hem, mt-sp, pl-ksp
• THERMOCALC sometimes gives a different answer to
  one calculated earlier with different starting
  guesses or even with the same starting guesses
• THERMOCALC gives a bomb message regarding chl
  starting guesses.

48
THERMOCALC wont calculate all or part of a given
equilibria

• Four problems can cause this
• Your line is outside your specified P-T range
• Your P-T range is too broad
• Your line is very steep/flat or is curved
• Your starting guesses are too far from a solution
• The solution to problem 4 is to use the
  compositions from the log file on the part of
  the equilibria you can calculate or from a nearby
  equilibria you can calculate.
• If its the first line on a diagram, have a
  guess from another tcd file in the same system
  or use your rock info
• You can also calc part of a T/P-x section from a
  known bulk that works with your starting guesses
• Adjust you starting guesses as you work across
  the diagram

49
liq 8 q(L) 0.1825 fsp(L) 0.2236
na(L) 0.5086 an(L) 0.003065 ol(L)
0.001511 x(L) 0.9256 h2o(L) 0.6519
--------------------------------------------------
------------------ P(kbar) T(C) q(L)
fsp(L) na(L) an(L) ol(L) x(L)
h2o(L) 6.82 820.0 0.1837 0.3422
0.3649 0.01560 0.004747 0.6510 0.4315
mode liq ksp pl cd
g ilm sill q
0.2253 0.1498 0.08311 0
0.1392 0.01302 0.05505 0.3345
50
THERMOCALC gives the same composition for two
similar minerals that should be separated by a
solvus

• Restricted to minerals that have identical coding
  but rely on distinct starting guesses to get each
  of the 2 solutions.
• Particularly problematic close to the solvus top
• Caused by the starting guesses generally being
  too similar or both too close to only one of the
  solutions
• Solution Change starting guesses so they are
  less similar and on opposite sides of the solvus

51
A univariant example in KFMASHTO
P(kbar) T(C) x(he) y(he) z(he)
x(mt) y(mt) z(mt) 2.60
877.9 0.9464 0.8203 0.06514 0.9750
0.1730 0.4069 209sp 167opx 29liq
10ilm 220q 56mt 35cd 129g 11ksp
2.70 544.0 0.9913 0.03152 0.1763
0.9913 0.03152 0.1763 mt sp
2.80 548.1 0.9908 0.03197 0.1751
0.9908 0.03197 0.1751 mt sp
In the last two results both spinel and magnetite
have a magnetite composition
In pseudosections this feature can cause the
calculation to fail or may give perfectly
sensible looking P-T conditions for an
equilibria if it is near the solvus top, but with
the wrong composition
52
THERMOCALC sometimes gives a different answer to
one calculated earlier with different starting
guesses or even with the same starting guesses

• Different starting guesses may give different P-T
  answers
• Especially when you have some very complex phases
  where the G-x surface is bumpy (gets stuck in a
  hole)
• Also a problem when you have a mineral that may
  have a solvus (composition flicks from one side
  of the solvus to the other)
• Solution Go back to well behaved equilibria that
  lead to your trouble area. Follow the
  compositions carefully (tco) the change in P-T
  should be accompanied with a sudden change in
  some of the mineral compositions. Change starting
  guesses to close to the right answers, with
  allowances for solvii.
• If problem persists email roger with the tcd and
  log files

53
(No Transcript)
54
THERMOCALC gives a bomb message regarding chl
starting guesses.

• This is a minor, specific problem that commonly
  pops up with highly ordered phases chl and some
  of the carbonates
• THERMOCALC cant handle exact solutions (ie.
  output results from a log file) as starting
  guesses in chlorite.
• Simply nudge the numbers slightly and it should
  work

55
Other common problems

• There are a range of things that can go wrong
  with calculating mineral equilibria and drawing
  phase diagrams
• These have an equally broad range of sources
  ranging from user errors to bugs in the code
• Remember there are uncertainties in every
  calculation
• The standard deviation on each calculation can be
  provide by thermocalc using calcsdnle yes in
  the tcd file
• These are 1? errors given so they should be
  doubled to give 2? uncertainties- based on
  uncertainty of enthalpy only

56
Other problems

• Here I calculated T at P so we only have an
  uncertainty on T
• 2? uncertainty is 18
• Notice we also have uncertainties on mineral
  composition and mineral modes
• Can be considered when contouring diagrams

57
Other problems

• Thermocalc does not reproduce my assemblages
• How different are they (one phase extra or
  missing)
• Is it a minor or major phase (look at the rocks)
• Eg in modelling some metagranites I found that
  thermocalc calculated a small amount of
  sillimanite (0.2-0.6) that wasnt in the rock,
  same problem with plag in some pelites
• This is not the end of the world but the diagram
  looks a bit wrong
• Look at the uncertainties on the modes, are they
  bigger than the mode itself

58
Other problems

• In this metapelite, the presence or absence of
  minor plag is not constrained
• Similar problems can occur with any mineral

59
Other problems

• What causes a discrepancy between observed and
  modelled assemblages
• The modelling is not in the right system
• There is a component and phase we cant model
  that is in the rock
• Our method for estimating bulk has problems (look
  at analytical uncertainties)
• The thermo and or a-x relationships are incorrect
• The eqm assemblage in the rocks has been
  misidentified
• Always go back and look at the rocks again, have
  a good look for that mineral, there may only be a
  few grains of it

60
Other problems

• In the case of minor sill in a metagranite, I
  found that the measured biotite was a little more
  aluminous than the calculated biotite
• A rock made up of bi-pl-ksp-q-ilm plotted in the
  bi-pl-ksp-ilm-sill field
• A very minor adjustment to the bulk rock
  composition gets rid of sill
• Remember there are analytical uncertainties in
  measuring bulks

61
Other problems

• Crashes!!!
• These still occasionally occur
• Look at the error output, is the cause obvious
  from this and can you fix it
• If not, contact Roger, with an explanation of
  what happened, your tcd file, the log file, and
  information of what version of thermocalc you
  were using and on what platform
• Thermocalc cant find a solution
• Just returns a series of numbers
• Commonly this is a starting guess issue, or
  choice of P-T window

62
Other problems

• I get a solution but it is in the wrong P-T area
• Generally this reflects 2 solutions, one is
  metastable
• Common on curved equilibira
• Can generally be avoided by either changing the
  P-T window or by changing from calc T at P to
  calc P at T or vice versa
• Can also occur if you have accidentally changed
  some of the a-x relations
• Always keep spare original tcd files

63
Other problems

• You just cant calculate the equilibria you know
  is there, or cant calculate all of it
• Barring starting guess or slope of line problems,
  sometime thermocalc just may struggle with a
  particular calc
• Look at the part you can calculate
• There is info in the output that can help
• Try changing the P-T window and P/T increments
• Can sometimes set a mode or composition parameter

64
Other problems
65
What diagrams to draw I

• It is not always obvious what diagrams to draw
  to show a particular feature of our rocks or to
  highlight a given process
• Our basic pseudosections are
• P-T pseudosections
• T-X/P-X pseudosections
• Compatibility diagrams
• More complex diagrams include
• X-X pseudosections (constructed by hand)
• M-X pseudosections (constructed by hand)
• T-V pseudosections
• T-a P-a pseudosections

66
What diagrams to draw II

• P-T pseudosections
• A series of these diagrams can show the textural
  development in different rocks/domains
• The compositions of the different bulks can be
  shown on a compatibility diagram e.g. AFM
• Open system processes and mineral fractionation
  can be shown on a series of P-T pseudosections
• P-T pseudosections are the mainstay diagram for
  analysing rocks
• But some other diagrams can show much

67
(No Transcript)
68
What diagrams to draw III

• T-X P-X pseudosections
• A series of these diagrams can show the effects
  of a progressive process e.g melt loss
• The compositions of the different bulks can be
  shown on a compatibility diagram e.g. AFM
• The X-axis can be simple e.g. XFe or complex
  e.g. Xmelt-loss, between two bulk rock
  compositions
• Open system processes and mineralogical
  fractionation can be shown on a T-X or P-X
  pseudosections
• If the P-T path can be simplified to vertical and
  horizontal segments then the P-T path can be
  shown for a range of rocks on a single diagram
• T-X P-X pseudosections are a very flexible and
  adaptable diagram

69
(No Transcript)
70
Lack of retrogression

• Lets look at how much melt must be lost from
  granulites to allow the preservation of
  dominantly anhydrous assemblages
• For most rocks gt70 of the melt produced has to
  be lost
• Look at simple 1melt loss event scenario

71
(No Transcript)
72
(No Transcript)
73
What diagrams to draw IV

• Compatibility diagrams
• The compositions of the different bulks can be
  shown on a compatibility diagram e.g. AFM
• Use is limited by having enough phases to
  project from
• A series of diagrams can illustrate the
  assemblage development on a wide range of rocks
• The diagrams can use complex axes
• Good summary diagrams

74
(No Transcript)
75
What diagrams to draw V

• More complex diagrams
• These diagrams are relatively uncommon and many
  are constructed by hand using THERMOCALC output
• Some of these, e.g. X-X pseudosections, will
  become more common when their construction is
  automated in THERMOCALC

76
(No Transcript)
77
(No Transcript)
78
(No Transcript)
79
Contours

• Phase diagrams can contoured for mineral modes
  and mineral compositions
• These are very useful for illustrating more
  information about changes that occur in rocks
• Remember there are uncertainties on these
  calculations, so avoid taking the numbers too
  literally
• Mode contours are mole or mole proportion-Not
  Volume
• The mineral modes are calculated on a one oxide
  total basis to normalise the effects of molecular
  oxide sums
• this normalisation serves to make them
  approximate to volume

80
Contours

• Composition contours use the composition
  variables in the a-x relationships
• To compare with analysed minerals you may have to
  rework your analysis into thermocalc style
• Some are proportions eg XFe (opx) some are site
  fractions eg yAl (opx)
• The number of oxygens in some endmembers may
  differ from that commonly reported in analyses
  tables
• Eg micas in thermocalc are calculated on 11ox,
  analyses commonly given as 22ox- this affects
  mole fraction numbers

81
Contours

• Contouring can be enabled using the
  scripts setiso yes or setiso x(bi), for
  composition or setmodeiso yes
  zeromodeiso no setmodeiso bi
  zeromodeiso no
• You will then be prompted for some values either
  as a list of numbers or start end
  interval

82
(No Transcript)
83
E. g. 1
84
Using diagrams to interpret rocks I

• We can use phase diagrams to interpret rocks in
  many ways
• Constraining P-T conditions, P-T paths
• Interpreting reaction textures
• Modeling open closed system processes
• Fluid/ melt generation
• But just because you can explain your rocks
  using a particular diagram doesnt mean that
  explanation is the right one.
• We can explain many reaction textures in
  metapelites using only a P-T grid, but this does
  not mean a rock actually experienced any of the
  univariant equilibria!

85
Using diagrams to interpret rocks II

• The best way to avoid a specious interpretation
  of your rocks is to use as much rock-based
  information as possible
• Pseudosections based on real compositions
• Contouring diagrams for modal proportions
• Using a realistic chemical system
• Detailed petrography
• There are a number of useful ways to more closely
  model rocks

86
Interpreting rocks e.g. 1

• Interpretation of some reaction textures in some
  Fe-rich metapelites.
• The rocks developed distinct compositional
  domains
• Each domain preserves a slightly different
  metamorphic history
• We can use the information from different domains
  to better constrain our history

87
E. g. 1
88
E. g. 1
89
E. g. 1
90
E. g. 1
91
E.g 2

• Take an anticlockwise P-T path
• Convert to linear segments
• Can see effects on a range of bulk rock comps
• Allows us to infer more of the P-T path and
  reconfirm a path derived from one bulk with
  evidence from another

92
E.g. 2