This article was downloaded by: [University of Stellenbosch]

On: 14 March 2013, At: 20:16

Publisher: Taylor & Francis
Informa Ltd Registered in England and Wales Registered Number: 1072954
Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH,
UK

Mineral Processing and

Extractive Metallurgy Review:
An International Journal
Publication details, including instructions for
authors and subscription information:
http://www.tandfonline.com/loi/gmpr20

CALCULATION OF MINERAL
COMPOSITION FROM CHEMICAL
ASSAYS
a
Bill Whiten
a
Julius Kruttschnitt Mineral Research Centre, The
University of Queensland, Brisbane, Australia
Version of record first published: 27 Dec 2007.

To cite this article: Bill Whiten (2007): CALCULATION OF MINERAL COMPOSITION

FROM CHEMICAL ASSAYS, Mineral Processing and Extractive Metallurgy Review: An
International Journal, 29:2, 83-97

To link to this article: http://dx.doi.org/10.1080/08827500701257860

PLEASE SCROLL DOWN FOR ARTICLE

Full terms and conditions of use: http://www.tandfonline.com/page/terms-

and-conditions

This article may be used for research, teaching, and private study purposes.
Any substantial or systematic reproduction, redistribution, reselling, loan,
sub-licensing, systematic supply, or distribution in any form to anyone is
expressly forbidden.

The publisher does not give any warranty express or implied or make any
representation that the contents will be complete or accurate or up to
date. The accuracy of any instructions, formulae, and drug doses should be
independently verified with primary sources. The publisher shall not be liable
for any loss, actions, claims, proceedings, demand, or costs or damages
 whatsoever or howsoever caused arising directly or indirectly in connection
with or arising out of the use of this material.
Downloaded by [University of Stellenbosch] at 20:16 14 March 2013
 Mineral Processing & Extractive Metall. Rev., 29: 83–97, 2008
Copyright Q Taylor & Francis Group, LLC
ISSN: 0882-7508 print=1547-7401 online
DOI: 10.1080/08827500701257860

CALCULATION OF MINERAL COMPOSITION

FROM CHEMICAL ASSAYS
Downloaded by [University of Stellenbosch] at 20:16 14 March 2013

BILL WHITEN
Julius Kruttschnitt Mineral Research Centre,
The University of Queensland, Brisbane, Australia

The operation of a mineral-processing plant is determined by the

minerals present in the ore. However, it is normal to measure the
chemical assays of samples rather than the mineral composition
and then attempt to convert from assays to mineral composition.
Unfortunately, this conversion is not always possible as there may
be different proportions of some or all of the minerals that give
the same assays. The singular value decomposition provides a con-
venient method to determine the minerals for which the conversion
is possible and to calculate the matrix needed for the calculation of
mineral composition.

Keywords: chemical assays, conversion, mineral compositions,

singular value decomposition

1. INTRODUCTION
A mineral-processing plant performs separations based on the properties
of the minerals present in the ore. To understand and improve the per-
formance of a mineral-processing plant, it is necessary to examine the
behavior of minerals within the plant. While it is possible to measure
the mineral composition of process samples (MLA 2006; QemScan
2006), these measurements tend to be costly and slow compared
with chemical assays. A conversion of chemical assays to mineral com-
position is often used as the first step in the evaluation of plant

Address correspondence to Bill Whiten, Julius Kruttschnitt Mineral Research Centre,

The University of Queensland, Brisbane, Qld 4072, Australia. E-mail: W.Whiten@uq.edu.au
 84 B. WHITEN

performance. While it is noted that the chemical assays can sometimes

be reported for compounds, for clarity in this article I only refer to
elements being reported from chemical and similar assays.
The conversion from mineral composition to chemical assays is done
readily using the chemical composition of the minerals. However, this
conversion is in the opposite direction to that needed for the more usual
requirement to convert from chemical assays to minerals. Unfortunately,
this reverse conversion is not always possible, as different compositions of
certain minerals can result in the same elemental assay values. It may be
Downloaded by [University of Stellenbosch] at 20:16 14 March 2013

necessary to use additional assumptions to make the conversion possible.

In some cases it will be possible to determine some minerals uniquely,
while other minerals do not have a unique proportion as determined by
the given elemental assays. A matrix analysis using the singular value
(Press et al. 1992) can be used to determine when the conversion from
elemental assays to the mineral composition is possible. Often, it is poss-
ible to determine some but not all of the proportions of the minerals
present. This article shows how the singular value decomposition can
identify which mineral proportions can be determined and give the coeffi-
cients needed to calculate these mineral proportions.
Previous literature on this conversion has typically avoided the
problems of multiple solutions by restricting the application to particular
geological rock types (Cohen and Ward 1991; Rosen et al. 2004), selecting
one of the possible solutions using optimization methods such as linear
programming (Braun 1986; Podkovyrov et al. 2003), or simply by restrict-
ing the conversion to cases where there are sufficient assays (Paktunc
1998). It has been found that the proposed general methods can give differ-
ing answers (Cohen and Ward 1991; Rosen et al. 2004). This article sepa-
rates the cases where unique compositions can be obtained from those
where multiple mineral compositions will satisfy the chemical assays. For
the latter case, the range of possible compositions can be calculated, but
additional information is needed to give a unique composition.

2. CONVERSION OF MINERAL COMPOSITION

TO ELEMENTAL ASSAYS
The transform from minerals to elements is the reverse of the required
transform from elements to minerals. However, the transform from miner-
als to elements provides the information needed for calculating the
required transform. Each mineral is assumed to have a known elemental
 CALCULATION OF MINERAL COMPOSITION FROM CHEMICAL ASSAYS 85

composition. In some cases a mineral can have a variable composition

and an assumed composition or a range of compositions is needed.
The proportions of each mineral in a sample can be expressed in a
vector where the proportion by mass of the jth mineral is given in the
jth entry mj, in the vector m. Similarly, the proportion of the ith element
will be in the ith entry ai, in the vector a. The ith elemental assay by mass
of the jth mineral is given by ci,j, which is the value in the ith row and jth
column of the matrix C.
Then, for the conversion from mineral composition m to the elemen-
Downloaded by [University of Stellenbosch] at 20:16 14 March 2013

tal assays a, we have the assay for the ith element as the sum of the pro-
ducts of the fraction of that element in the jth mineral times the fraction
of that mineral in the sample
X
ai ¼ ci;j mj ð2:1Þ
j

or in matrix notation
a ¼ C m: ð2:2Þ
Thus, this conversion uses the elemental composition of minerals that
are always available and is a simple summation of the amounts of each
element contributed by each mineral.

3. CONVERSION OF ELEMENTAL ASSAYS

TO MINERAL COMPOSITION
The conversion of interest is the reverse of that in the previous section,
namely given that the assays a determine the mineral composition m. This
requires solving the matrix Equation 2.2 for m given the values of the
assays a and the conversion matrix C. When C is square and nonsingular,
this is straightforward. As C is generally sparse, the equation solution
often is easily obtained by carefully selecting the order in which the ele-
ments of a are determined, in a manner that corresponds to reorganizing
the rows and columns of C so that it is lower triangular. Rather than solve
the equations for m, it generally is more convenient to obtain an explicit
formula for the mineral composition m in terms of the the assays a. This
is when the inverse matrix R with elements ri,j, of the mineral-to-assay
conversion matrix C is used:
X
mi ¼ ri;j aj ð3:1Þ
j
 86 B. WHITEN

or in matrix notation
m ¼ R a: ð3:2Þ

Each column of R gives the amount that the corresponding assay contri-
butes to the amounts of the minerals present. Unlike the C matrix, which
contains only positive elements, the R matrix usually will have negative
terms. These negative terms indicate the amount for which the corre-
sponding element must be compensated by an increased amount of
another element.
Downloaded by [University of Stellenbosch] at 20:16 14 March 2013

Unfortunately, the inverse of C does not always exist; thus, it is

necessary to test if the inverse exists and, if it does not, to determine
which elements of R can be calculated for a partial conversion of assays
to minerals. It is here that the singular value decomposition is useful.
The singular value decomposition (Golub and Reinsch 1970; Press
et al. 1992), often abbreviated to SVD, factors a matrix C so that it is
equal to the form U D V0 where U and V are orthogonal (i.e., U U0 ¼ I
and V V0 ¼ I where I is the unit matrix and 0 denotes transpose) and D
contains zeros except on the main diagonal. The matrix D is the same
shape as C, which is not necessarily square, and the nonzero diagonal
elements of D are positive and given in decreasing order. The singular
value decomposition is known to be robust in that it is always possible
to perform it in an accurate manner. The SVD factorization is seldom
done by hand, but reliable computer software is available for its calcu-
lation. Functions for the SVD factorization are included in Matlab
(2006) and Press et al. (1992), and are available from Netlib (2006).
The number of zero or nearly zero columns of D tell the extent to
which the matrix C is singular and the corresponding columns of V,
which we will define as Vo (the null space of C), are vectors that when
multiplied by C give a zero result.
Defining E as the matrix obtained from D by replacing any nearly
zero diagonal elements by zero and then replacing the nonzero elements
by their inverse, we can write
R ¼ V E 0 U0 : ð3:3Þ
When C is square and nonsingular, this is simply the inverse of C;
otherwise, R is known as the pseudoinverse or the Moore Penrose
inverse. When there are more assays than minerals, this value of R pro-
vides the least squares solution (i.e., a best-fit solution) of Equation 2.2
for m. In this case, provided that the assays are consistent and C is not
 CALCULATION OF MINERAL COMPOSITION FROM CHEMICAL ASSAYS 87

singular, then R gives the correct mineral composition. Example 1 is a

case of this type.
In the case when there are more minerals than assays, the pseudo-
inverse provides one of the possible solutions to Equation 2.2 for m
(but not necessarily a valid mineral composition). The complete solution
for the vector m, given the assays a and the mineral-to-assay conversion
matrix C, requires the null space Vo of C, as determined above using the
singular value decomposition and is
Downloaded by [University of Stellenbosch] at 20:16 14 March 2013

m ¼ R a þ Vo x ð3:4Þ

where x is an arbitrary vector that provides the degree of possible vari-

ation in the solution for m. When there are no columns in Vo, this equa-
tion reduces to Equation 3.2 and gives the unique values for m. Left
multiplying this equation by C gives the original Equation 2.2, showing
that m is indeed a solution of Equation 2.2. This formula shows that
any mineral that is independent of the value of x is determined as a
unique value from the sample assays. Examples 2 and 5 demonstrate this
behavior.
When Vo has one or more columns, the condition for the proportion
of a mineral in the sample to be calculated from the sample assays, is that
the corresponding row of Vo is all zero. Minerals that have nonzero
values in the corresponding row of Vo can occur in varying proportions
that still give the same assay values.
For mineral composition we are only interested in the solutions that
give positive values for m. The positive solutions for m form a polytope,
the corners of which can be calculated by solving all the equation sets
obtained by setting a given number of rows in the expression for m in
terms of x (Equation 3.4) equal to zero and solving for x, then selecting
only those solutions that give positive values of m. The number of rows
selected is equal to the dimension of x. If the matrix for conversion from
minerals to assays does not require the assays and mineral contents sum
to 1, it is necessary to add an additional condition that the sum of the
proportions of minerals present is less than or equal to 1.
If M is the matrix formed with columns giving the mineral compo-
sition of the corners of the polytope (thus defining the convex hull),
the possible valid solutions for the mineral composition m can also also
be expressed as

m¼Mz
 88 B. WHITEN

where the elements of z are positive and sum to 1. Again, left multiplying
by C gives Equation 2.2 and, thus, all of these solutions are consistent
with the given assay values.

4. ADDITIONAL CONSTRAINTS ON MINERAL COMPOSITION

Unfortunately, it is often the case that there are more minerals than
assays and, thus, the proportions of all the mineral present cannot be
determined uniquely. The previous section shows how to find which
Downloaded by [University of Stellenbosch] at 20:16 14 March 2013

minerals can be determined with unique values. Additional equations

are needed if the remaining minerals are to be determined. Some possi-
bilities for additional equations are discussed in the paragraphs below.
If it is known that Relation 2.2 (or 2.1) contains all the minerals in
the sample but there are some elements present for which assays are
not available, a relation stating that the proportions of the minerals must
add to 1 can be included as part of the matrix C with a corresponding
value of one in the assay vector. Example 3 is a case of this type.
Another approach that is commonly used is to add conditions that cer-
tain minerals occur in known ratios. This may at times be adequate for sam-
ples from an ore body but within a processing plant, where the aim is to alter
the proportions of the different minerals present, it is clearly inadequate.
Similar to the previous paragraph, the ratio condition can be added as an
additional row (or rows) to the conversion matrix C, this time requiring a
0 value as the corresponding entry in the vector a. Example 4 shows how
this type of condition can be used. This method is approximate, depends
on local conditions, and thus should only be used with considerable care.
A more quantitative method is to determine additional ratios of cer-
tain mineral groups in the sample by using selective chemical-dissolution
methods. Different dissolution treatments can often provide sufficient
information to determine the proportions of several minerals that are
not uniquely determined by the elemental assays. The ratios determined
are added as extra rows in the assay to mineral conversion, similar to the
ratio case above and Example 4.
Another way in which additional equations can be obtained is to use
the amount of minor elements present in the minerals to provide extra
assays to increase the number of assays available (Zhang and Whiten
2001). This method requires that the amounts of minor elements in the
minerals present remain constant. An analysis that takes into account
the relative accuracies of the proportions of minor elements is possible
 CALCULATION OF MINERAL COMPOSITION FROM CHEMICAL ASSAYS 89

(Zhang 2000). Certain methods of elemental analysis can make the

assays of more elements available at a minimal additional cost.

5. EXAMPLES
The following examples illustrate different cases that can occur. The first
example is a case with more assays than minerals. This is followed by
examples with one less assay than minerals and the final example has
two less assays than minerals.
Downloaded by [University of Stellenbosch] at 20:16 14 March 2013

Example 1: More Assays Than Minerals

To determine the proportions of Galena, Sphalerite, Chalcopyrite, Pyrrho-
tite, and nonsulfide gangue (NSG) from Pb, Zn, Cu, Fe, S, and NSG assays.
The matrix C to determine the assays from the minerals is

Galena Sphalerite Chalcopyrite Pyrrhotite NSG

Pb 0.866 0 0 0 0
Zn 0 0.565 0 0 0
Cu 0 0 0.345 0 0
Fe 0 0.100 0.305 0.604 0
S 0.134 0.335 0.350 0.396 0
NSG 0 0 0 0 1

The singular value decomposition C ¼ U D V0 then gives

U¼ 0
0.3981 0.9098 0.0217 0.0234 0.1131

0
0.2495
0.1316
0.8937 0.0142 0.3487
0
0.1698
0.0945 0.1709
0.9121 0.3180
0
0.5856
0.3355 0.4061 0.3870 0.4794
0
0.6384
0.1830
0.0822
0.1326
0.7312

1.0000 0 0 0 0 0
D¼ 1.0000 0 0 0 0
0 0.9675 0 0 0
0 0 0.8524 0 0
0 0 0 0.5576 0
0 0 0 0 0.3032
0 0 0 0 0
V¼ 0
0.4447 0.8955 0.0140 0.0082
0
0.4273
0.1985
0.8820 0.0076
0
0.4761
0.2334 0.2762
0.8016
0
0.6269
0.3228 0.3815 0.5978

1.0000 0 0 0 0
 90 B. WHITEN

We see that each column of D contains a nonzero value; hence, it is

possible to determine all the minerals from the available assays. The
matrix contains the inverse of the nonzero elements of D and is
E¼ 1.0000 0 0 0 0
0 1.0336 0 0 0
0 0 1.1731 0 0
0 0 0 1.7933 0
0 0 0 0 3.2982
0 0 0 0 0
Downloaded by [University of Stellenbosch] at 20:16 14 March 2013

and, thus, the value of R ¼ V E0 U0 is

R¼ 1.1400
0.0456
0.0415
0.0626 0.0955 0

0.0698 1.5547
0.1963
0.2959 0.4513 0

0.1043
0.3214 2.6055
0.4419 0.6740 0

0.0256
0.3719
1.5356 1.5472 0.1653 0
0 0 0 0 0 1.0000

This value of R can be checked by multiplying it into columns contain-

ing the assays of an individual mineral, which gives a mineral composition
containing only the single mineral. The columns of C give the assays of the
minerals so this check consists of verifying that R C is the unit matrix.
For example, the following percent assays when multiplied by R give
the corresponding mineral composition:
Pb 4.8 Galena 5.54
Zn 5.7 Sphalerite 10.07
Cu 0.3 Chalcopyrite 0.83
Fe 4.8 Pyrrhotite 5.83
S 6.7 NSG 77.7

Example 2: Insufficient Assays

If we take the previous example and remove the assays for Cu & Fe
giving the value of C as

Galena Sphalerite Chalcopyrite Pyrrhotite NSG

Pb 0.866 0 0 0 0
Zn 0 0.565 0 0 0
S 0.134 0.335 0.350 0.396 0
NSG 0 0 0 0 1
 CALCULATION OF MINERAL COMPOSITION FROM CHEMICAL ASSAYS 91

and the singular value decomposition of C is

U¼ 0 0.9269 0.3537
0.1255

0 0.1391
0.6342
0.7605
0 0.3486
0.6875 0.6371
1.0000 0 0 0
D¼ 1.0000 0 0 0 0
0 0.8908 0 0 0
0 0 0.7242 0 0
Downloaded by [University of Stellenbosch] at 20:16 14 March 2013

0 0 0 0.4009 0
V¼ 0 0.9535 0.2958
0.0581 0
0 0.2193
0.8129
0.5396 0
0 0.1370
0.3323 0.5562 0.7493
0 0.1549
0.3760 0.6293
0.6622
1.0000 0 0 0 0

This time we see that D has a zero column (column 5), showing
that no unique solution is possible. Column 5 of V has nonzero ele-
ments in rows 3 and 4, corresponding to minerals Chalcopyrite and
Pyrrhotite, which thus do not have unique values determined by the
given assays.
Next, the value of R is calculated as before as

R¼ 1.1547 0 0 0
0 1.7699 0 0

0.1939
0.7430 1.2531 0

0.2194
0.8406 1.4177 0
0 0 0 1.0000

This can be used to determine the proportions of Galena, Sphalerite,

and NSG (rows 1, 2, and 4), but not Chalcopyrite and Pyrrhotite.
Given some particular assay values

Pb 0.0346
Zn 0.0565
S 0.0810
NSG 0.7500

For these assays, the range of possible mineral compositions is

determined by the two limiting compositions
 92 B. WHITEN

Mineral M1 M2

Galena 0.0400 0.0400

Sphalerite 0.1000 0.1000
Chalcopyrite 0 0.0300
Pyrrhotite 0.1065 0.0800
NSG 0.7500 0.7500

It is seen that the minerals that can be uniquely determined have the
Downloaded by [University of Stellenbosch] at 20:16 14 March 2013

same values in both columns, whereas the compositions for Chalcopyrite

and Pyrrhotite vary. Also, only the composition M2 sums to one, so if
this condition applies a unique composition that is obtained. Both of
these compositions give the original assay values when multiplied by
the conversion matix C.

Example 3: Addition of the Condition That All Minerals are Known

For this case, the condition that all the minerals in the sample are
included is added to Example 2. An additional row of 1s is appended
to the matrix C of the previous example and the assay vector is extended
by appending a 1. The matrix C for converting minerals to assays
becomes

Galena Sphalerite Chalcopyrite Pyrrhotite NSG

Pb 0.866 0 0 0 0
Zn 0 0.565 0 0 0
S 0.134 0.335 0.350 0.396 0
NSG 0 0 0 0 1
1 1 1 1 1 1

For this case, the matrix D from the singular value decomposition is

D¼ 2.3915 0 0 0 0
0 0.9709 0 0 0
0 0 0.7887 0 0
0 0 0 0.4406 0
0 0 0 0 0.0279

Now there are no zero columns in D, but now D contains a rather

small value in the last element, indicating a possible sensitivity to assay
errors. The matrix R ¼ V E0 U0 for conversion from assays is
 CALCULATION OF MINERAL COMPOSITION FROM CHEMICAL ASSAYS 93

R¼ 1.1547 0 0 0 0
0 1.7699 0 0 0

6.5770
2.3471
21.7391
8.6087 8.6087
5.4222 0.5771 21.7391 7.6087
7.6087
0 0 0 1.0000 0

This matrix will give unique values for the mineral composition.
However, to use this matrix a 1 needs to be appended to the assay vector,
to apply the additional condition and match the dimension of this matrix.
Downloaded by [University of Stellenbosch] at 20:16 14 March 2013

Again, we can test this matrix for conversion of the columns of C to the
corresponding mineral compositions.

Example 4: Addition of a Known Ratio for Some Minerals

In this example, the case in Example 2 is used, with the additional
assumed knowledge that the proportions of Chalcopyrite and Pyrrhotite
are equal. This adds an additional row to the mineral to assay matrix that
gives the condition for this ratio to be applied. Corresponding to this
additional row, a zero needs to be appended to the elemental assay vec-
tor. The new C matrix is

Galena Sphalerite Chalcopyrite Pyrrhotite NSG

Pb 0.866 0 0 0 0
Zn 0 0.565 0 0 0
S 0.134 0.335 0.350 0.396 0
NSG 0 0 0 0 1
0 0 0 1
1 0

The singular value decomposition of this matrix has no zero columns

in D and, hence, these assays with the condition that Chalcopyrite and
Pyrrhotite occur is equal proportions, is sufficient to give the propor-
tions of all the minerals present.
The matrix for conversion from assays to minerals is

R¼ 1.1547 0 0 0 0
0 1.7699 0 0 0

0.2074
0.7948 1.3405 0 0.5308

0.2074
0.7948 1.3405 0
0.4692
0 0 0 1.0000 0

The last column of R is always multiplied by the zero added to the

assay vector corresponding to the ratio row, and so can be removed.
 94 B. WHITEN

Example 5: A Case with Less Assays

If we take example 1 and remove the assays for Pb, Zn, and NSG, for C we get:

Galena Sphalerite Chalcopyrite Pyrrhotite NSG

Cu 0 0 0.345 0 0
Fe 0 0.100 0.305 0.604 0
S 0.134 0.335 0.350 0.396 0

For this case, the singular value decomposition gives two zero
Downloaded by [University of Stellenbosch] at 20:16 14 March 2013

columns in D and the corresponding columns of V (i.e., Vo) are

0.8930 0

0.4441 0
0 0
0.0735 0
0 1.0000

and only the third row of Vo is zero, showing that only the proportion of
chalcopyrite can be determined using these assays. The pseudoinverse in
this case is
R¼
0.5615
1.1001 1.5122

1.3348
1.8862 2.9594
2.8986 0 0

1.2427 1.9679
0.4900
0 0 0

Using the assays available the proportions of Galena, Sphalerite,

Pyrrhotite, and NSG cannot be determined uniquely using this matrix.
For the following sample assay values:

Cu 0.0103
Fe 0.0675
S 0.0810

the four mineral compositions

Mineral M1 M2 M3 M4

Galena 0 0 0.2411 0.2411

Sphalerite 0.1199 0.1199 0 0
Chalcopyrite 0.0300 0.0300 0.0300 0.0300
Pyrrhotite 0.0767 0.0767 0.0966 0.0966
NSG 0 0.7734 0 0.6324
 CALCULATION OF MINERAL COMPOSITION FROM CHEMICAL ASSAYS 95

define the limits (the convex hull) of the range of compositions obtained
from the assays given above. The proportions given for Chalcopyrite are
the same in each column, showing that a unique value for the proportion
of Chalcopyrite has been determined. As in Example 2, these composi-
tions can be multiplied by the matrix C to verify that they all give the
assay values from which they were derived.
In the case where there are no additional minerals in the sample, the
range of possible compositions is given by columns M2 and M4, which
both sum to 1. However, in this case no additional unique mineral com-
Downloaded by [University of Stellenbosch] at 20:16 14 March 2013

positions are available.

6. CONCLUSIONS
Mineral-processing plants perform separations of valuable minerals from
waste minerals based on the properties of the minerals; hence, the com-
position of ores and of processing samples in terms of the mineral con-
tent is of considerable use for the evaluation of detailed plant
performance. Elemental assays of samples are readily available and rela-
tively cheap compared with determination of the proportion of minerals
present. Hence, it is of interest to determine when mineral composition
can be determined reliably from elemental assays.
The conversion of the proportion of minerals present in an ore sam-
ple to the elemental assays is easily done using a matrix (C) of the assays
of the individual minerals. This matrix also provides the necessary infor-
mation for the reverse calculation that determines mineral composition
from elemental assays. Simple cases can often be done manually, as
the matrix can sometimes be reorganized into a triangular form. The gen-
eral case can become more complex with possibly some minerals not
determined uniquely by the assays. The singular value decomposition
provides the general method for determining which mineral composi-
tions can be found from a given set of assays and determining the coeffi-
cients needed to calculate the composition of these minerals from the
assays. All variations of the general case can be conveniently handled
using the singular value decomposition algorithm.
Nonzero values in the vectors defining the null space (calculated using
the singular value decomposition) of the matrix C identify the minerals that
do not have a unique proportion determined by the assays. In some cases
additional conditions on the mineral composition can be used to allow
the determination of the proportions of more minerals from the assays.
 96 B. WHITEN

A Matlab (Matlab 2006) function for the calculation of the assay to

mineral matrix is given in the Appendix.

ACKNOWLEDGEMENTS
Pertti Lamberg provided information on the possibility of using selective
dissolution methods to add information on the proportions of certain
minerals groups in the sample.
Downloaded by [University of Stellenbosch] at 20:16 14 March 2013

REFERENCES
Braun, G. E., 1986, ‘‘Quantitative analysis of mineral mixtures using linear
programming.’’ Clays and Clay Minerals, 34(3), pp. 330–337.
Cohen, D. and Ward, C. R., 1991, ‘‘SEDNORM—A program to calculate a
normative mineralogy for sedimentary rocks based on chemical analyses.’’
Computers & Geosciences, 12(9), pp. 1235–1253.
Golub, G. H. and Reinsch, C., 1970, ‘‘Singular value decomposition and least
squares solutions,’’ Numerische Mathematik, 14, pp. 403–420.
Matlab, 2006, Available at http://www.mathworks.com/
MLA, 2006, Available at http://www.jkmla.com/
Netlib, 2006, Available at http://www.netlib.org and http://www.netlib.org/
lapack/
Paktunc, A. D., 1998, ‘‘MODAN: An interactive computer program for estimat-
ing mineral quantities based on bulk composition.’’ Computers & Geos-
ciences, 24(5), pp. 425–431.
Podkovyrov, V. N., Graunov, O. V., and Cullers, R. L., 2003, ‘‘A linear program-
ming approach to determining the normative composition of sedimentary
rocks.’’ Mathematical Geology, 35(4), pp. 459–475.
Press, W. H., Teukolsky, S. A., Vetterling, W. T., and Flannery, B. P., 1992,
Numerical Recipes in FORTRAN: The Art of Scientific Computing, Cambridge,
UK: Cambridge University Press.
QemScan, 2006, Available at http://www.intellection.com.au/
Rosen, O. M., Abbyasov, A. A., and Tipper, J. C., 2004, ‘‘MINLITH—An experi-
ence based algorithm for estimating the likely mineralogical composition of
sedimentary rocks from bulk chemical analyses’’ Computers & Geosciences,
30(6), pp. 647–661.
Zhang, N., 2000, Determination of Mineral Composition from Elemental Assays,
M.Eng.Sci. Thesis, The University of Queensland, Queensland, Australia
(unpublished).
Zhang, N. and Whiten, W. J., 2001, ‘‘Determining mineral composition from
assays.’’ International Heavy Minerals Conference, Fremantle, WA: AusIMM,
pp. 81–85.
 CALCULATION OF MINERAL COMPOSITION FROM CHEMICAL ASSAYS 97

APPENDIX: MATLAB PROGRAM FOR CALCULATION OF ASSAY

TO MINERAL CONVERSION MATRIX
function [R,V0] ¼ assay2mineral(C,delta)
% assay2mineral Calculate matrix for assay to mineral con-
version
% 6=9=06 Matlab7 W.Whiten
%
% [R,V0] ¼ assay2mineral(C,delta)
Downloaded by [University of Stellenbosch] at 20:16 14 March 2013

% C Mineral to assay conversion matrix

% delta Small number used to test for essentially zero
(optional)
%
% R Assay to mineral conversion matrix
% V0 Null space of C

% value for delta if not given

if(nargin < 2)
delta ¼ 1e-5;
end

% singular value decomposition

[U,D,V] ¼ svd(C);

% null space of C
V0 ¼ V(:,sum(D,1) < delta);

% rows of null space that are not zero

idx ¼ find(sum(abs(V0),2) > delta);

% pseudo inverse matrix

E ¼ D;
E(E < delta) ¼ 0;
E(E > 0) ¼ 1.=E(E > 0);
R ¼ V E0 U0;

% remove rows for minerals not given unique values

R(idx,:) ¼ NaN;

return
end