Practical Aspects

PART V I I
of Chemical Analysis

Chapter 35
The Analysis of Real Samples

Chapter 36
Preparing Samples for Analysis

Chapter 37
Decomposing and Dissolving the Sample

Chapter 38
Selected Methods of Analysis
 The Analysis of Real
CHAPTER 35 Samples

The analysis of real samples, such as the soil and rock samples brought back to the earth from
the moon by the Apollo astronauts, is usually quite complex compared to the analysis of materials
studied in laboratory courses. As discussed in this chapter, the choice of analytical method for
real materials is not simple, often requiring consultation with the literature, modification of exist-
ing methods, and extensive testing to determine method validity.
Shown in the photograph is one of the Apollo astronauts taking a core sample of the lunar soil.
Such samples were valuable in determining the geological history of the moon and its relationship
to the history of the earth.

V ery early in this text (Section 1C), we pointed out that a quantitative analysis consists
of a sequence of steps: (1) selecting a method, (2) sampling, (3) preparing a laboratory
sample, (4) defining replicate samples by mass or volume measurements, (5) preparing solu-
© Corbis
tions of the samples, (6) eliminating interferences, (7) completing the analysis by performing
measurements that are related in a known way to analyte concentration, and (8) computing
the results and estimating their reliability. In Chapters 1 through 34, we have focused largely
on steps 6, 7, and 8 and to a lesser extent on steps 2 and 4. We have chosen this emphasis
not because the earlier steps are unimportant or easy. In fact, the preliminary steps may be
more difficult and time consuming than the two final steps of an analysis and may be greater
sources of error.
The reasons for postponing a discussion of the preliminary steps to this point are ped-
agogical. Experience has shown that it is easier to introduce students to analytical tech-
niques by having them first perform measurements on simple materials for which no method
­selection is required and for which problems with sampling, sample preparation, and sample
dissolution are either nonexistent or easily solved. Thus, we have been largely concerned so
far with measuring the concentration of analytes in simple aqueous solutions that have few
interfering species.

35A Real Samples

Determining an analyte in a simple solution is often easier than in a complex solution
because the number of variables that must be controlled is small and the tools available
are numerous and easy to use. With simple systems, our knowledge of the chemical
and measurement principles allows us to anticipate problems and to correct for them.
 35A Real Samples  961

In fact, academic and industrial scientists are usually interested in materials

that are not as a rule simple. To the contrary, most analytical samples are com-
plex mixtures of components containing in some cases hundreds of species. Such
­materials are frequently far from ideal in matters of solubility, volatility, stability,
and homogeneity, and many steps must precede the final measurement step. In-
deed, the final measurement may be easier and less time consuming than any of
the preceding steps.
For example, we showed in earlier chapters that the calcium ion concentration of
an aqueous solution may be easily determined by titration with a standard EDTA
solution or by potentiometry with a specific-ion electrode. Alternatively, the calcium
content of a solution can be determined from atomic absorption or atomic emission
measurements or by the precipitation of calcium oxalate followed by weighing or by
titrating with a standard solution of potassium permanganate.
All of these methods can be used to determine the calcium content of a sim-
ple salt, such as the carbonate. However, chemists are seldom interested in the
­calcium content of calcium carbonate. More likely, what is needed is the percent-
age of this ­element in a sample of blood, animal tissue, a silicate rock, or a piece of
glass. The analysis thus acquires a new level of complexity. None of these materials
is soluble in water or in dilute aqueous reagents. Before calcium can be determined,
therefore, the sample must be decomposed by high-temperature treatment with con-
centrated reagents. Unless care is taken, we could lose some calcium during this step,
or equally bad, we could introduce some calcium as a contaminant because of the
relatively large quantities of reagent usually needed for sample decomposition.
Even after the sample has been decomposed to give a solution containing calcium
ions, the procedures mentioned in the two previous paragraphs cannot ordinarily be
applied immediately to complete the analysis because the reactions or properties used
are not specific to calcium. Thus, a sample of animal tissue, silicate rock, or glass
almost surely contains one or more components that also react with EDTA, act as
a chemical interference in an atomic absorption measurement, or form a precipitate
with oxalate ion. In addition, the high ionic strength resulting from the reagents used
for sample decomposition would complicate a direct potentiometric measurement.
Because of these difficulties, several additional operations are required to eliminate
interferences before the final measurement is made.
We have chosen the term real samples to describe materials such as those in Real samples are far more complex
the preceding example. In this context, most of the samples encountered in an than most of those that you encounter
­elementary quantitative analysis laboratory course definitely are not real but rather are in the instructional laboratory.
­homogeneous, stable, soluble, and chemically simple. Also, there are well-­established
and thoroughly tested methods for their analysis. There is much value in introducing
analytical techniques with such materials because they permit you to focus on the
mechanical aspects of an analysis. Even experienced analysts use such samples when
learning new techniques, calibrating instruments, or standardizing solutions.
Determining the compositions of real samples frequently demands intellectual skill
and chemical intuition in addition to mechanical aptitude. Frequently, a compromise
must be struck between the available time and the accuracy required. We are often
happy to settle for an accuracy of one or two parts per hundred instead of one or two
parts per thousand, knowing that ppt accuracy may require several hours or even days
of additional effort. In fact, with complex materials, even parts-per-­hundred accuracy
may be unrealistic.
The difficulties encountered in the analysis of real samples stem from their com-
plexity. As a result, the literature may not contain a well-tested analytical route for the
kind of sample under consideration. For such cases, an existing procedure must be
962  CHAPTER 35 The Analysis of Real Samples
The objectives of an analysis
must be clearly defined before
the work begins.
modified to take into account compositional differences between the current ­sample
and samples in the original publication. Alternatively, an entirely new ­analytical
method might need to be developed. In either case, the number of variables that
must be taken into account usually increases exponentially with the number of spe-
cies contained in the sample.
As an example, contrast the problems associated with the inductively coupled
plasma atomic emission analysis of calcium carbonate with those for a real calcium-
containing sample. In the calcium carbonate case, the number of components is
small, and only a few variables are likely to affect the results. The most important
variables are the physical loss of analyte when carbon dioxide evolves as the sample is
dissolved in acid, the effect of the anion of the acid and of the radio-frequency power
on the intensity of the calcium emission line, the position of the plasma with respect
to the spectrometer entrance slit, and the quality of the standard calcium solutions
used for calibration.
Determining calcium in a real sample, such as a bone or a silicate rock, is far
more complex since the sample is insoluble in ordinary solvents and contains a
dozen or more components. The silicate rock sample, for example, can be dis-
solved by fusing it at a high temperature with a large excess of a reagent such as
sodium carbonate. Physical loss of the analyte is possible during this treatment
unless suitable precautions are taken. Furthermore, we must be concerned that
­c alcium could be introduced from the excess sodium carbonate or the fusion
­vessel. ­Following fusion, the sample and reagent are dissolved in acid. With this
step, all the variables affecting the calcium carbonate sample are operating, but
in addition, many new variables are introduced because of the dozens of com-
ponents in the sample matrix. Now, we must take care to minimize instrumental
and chemical interferences brought about by the presence of various anions and
cations in the solution being introduced into the plasma. Thus, the analysis of a
real substance is often a challenging problem and developing a procedure for such
materials is a demanding task.
 35B Choice of Analytical Method
The choice of a method for the analysis of a complex substance requires good judg-
ment based on sound knowledge of the advantages and limitations of the available
analytical tools. In addition, a familiarity with the literature of analytical chemistry is
essential. We cannot be very explicit concerning how an analytical method is selected
because there is no single best way that applies under all circumstances. We can,
however, suggest a systematic approach to the problem and present some generalities
that can aid in making good decisions.
35B-1 Definition of the Problem
❯ The first step, which must precede any choice of method, is to clearly define the ana-
lytical problem. The method selection will be mostly governed by the answers to the
following questions:
■ ■ What is the concentration range of the analyte to be determined?
■ ■ What degree of accuracy is desired?
■ ■ What other components are present in the sample?
■ ■ What are the physical and chemical properties of the gross sample?
■ ■ How many samples will be analyzed?
 35B Choice of Analytical Method  963

The concentration range of the analyte may well limit the number of feasible
methods. For example, if we wish to determine an element present at the parts-per-
billion or parts-per-million level, gravimetric or volumetric methods can be ruled
out, and spectrometric, potentiometric, and other more sensitive methods become
likely candidates. For components in the parts-per-billion and parts-per-million
range, even small losses resulting from coprecipitation or volatility and contamina-
tion from reagents and apparatus are major concerns. In contrast, if the analyte is a
major component of the sample, these considerations are less important, and a classic
analytical method may well be adequate for the task.
The accuracy required is vitally important in choosing a method and in the way it
is performed because the time required to complete an analysis increases dramatically
with demands for higher accuracy. Therefore, to improve the reliability of analytical
results from, say, 2 to 0.2% relative may require an increase in the analysis time by a
factor of 100 or more. Consequently, we should always carefully consider the degree
❮ The time required to carry out
an analysis increases, often in
an exponential manner, with the
of accuracy that is required before undertaking an analysis. desired level of accuracy.
The demands for accuracy frequently dictate the procedure chosen for an analy-
sis. For example, if the allowable error in the determination of aluminum is only a
few parts per thousand, a gravimetric procedure is probably required. On the other
hand, if an error of, say, 50 ppt can be tolerated, a spectroscopic or electroanalytical
­approach may be preferable.
The way in which an analysis is done is also affected by accuracy requirements.
If precipitation with ammonia is chosen for the analysis of a sample containing
20% aluminum, the presence of 0.2% iron is of serious concern if accuracy in the
parts-per-thousand range is demanded and if a preliminary separation of the two
elements is necessary. If an error of 50 ppt is tolerable, however, the separation of
iron is not necessary. This tolerance can also govern other aspects of the method.
For example, 1-g samples can be weighed to perhaps 10 mg and certainly no closer
than 1 mg. In addition, less care is needed in transferring and washing the precipi-
tate and in other time-consuming operations of the gravimetric method. The wise
use of shortcuts is not a sign of carelessness but a recognition of the importance of
efficiency. The question of accuracy, then, must be clearly resolved before begin-
❮ Often, you can save considerable
time by the use of permissible
shortcuts in an analytical
ning an analysis. procedure.
To choose a method for the determination of one or more species in a sample,
you frequently need to know what other elements or compounds are present. If such
information is lacking, a qualitative analysis must be undertaken in order to iden-
tify components that are likely to interfere in the various methods under consider-
ation. As we have noted frequently, most analytical methods are based on reactions
and physical properties that are shared by more than a single element or compound.
Thus, measuring the concentration of a given element by a method that is simple and
straightforward in the presence of one group of elements or compounds may require
many tedious and time-consuming separations in the presence of others. A ­solvent
suitable for one combination of compounds may be totally unsatisfactory when used
for another. It is very important to know the approximate chemical composition of
the sample before selecting a method for the quantitative determination of one or
❮ Itidentify
is frequently necessary to
the components of a
sample before undertaking a
more of its components. quantitative analysis.
We must also consider the physical state of the sample in order to determine
whether it must be homogenized, whether volatility losses are likely, and whether
its composition may change under laboratory conditions due to the absorption or
the loss of water. We must also determine how to decompose or dissolve the sample
without loss of analyte. Preliminary tests of one sort or another may be needed to
provide this type of information.
964  CHAPTER 35 The Analysis of Real Samples
A little extra time spent in the
library can save a tremendous
amount of time and effort in the
laboratory.
The technology for computer-
based scientific information
retrieval provides an efficient
means of surveying the
analytical literature. For
example, complete archives of
all American Chemical Society
journals are available online.
Finally, the number of samples to be analyzed is an important criterion in select-
ing a method. If there are many samples, considerable time can be spent in calibrating
­instruments, preparing reagents, assembling equipment, and investigating shortcuts since
the cost of these operations can be spread over the large number of samples. If, however,
only few samples will be analyzed, a longer and more tedious procedure with few of these
preparatory operations may be the wiser choice from an economic standpoint.
Once we have answered the preliminary questions, we can then consider possible
approaches to the problem. Sometimes, experience provides a straightforward route
for the analysis. In other instances, we must speculate on problems that are likely to
be encountered in the analysis and how they can be solved. By the time these prob-
❯ lems have been considered, some methods probably will have been eliminated from
consideration and others put on the doubtful list. Usually, however, we first turn to
the analytical literature to profit from the experience of others.
35B-2 Investigating the Literature
A list of reference books and journals covering various aspects of analytical chemistry
appears in Appendix 1. This list is not exhaustive but rather one that is adequate for
most work. It is divided into several categories. In many instances, the division is
­arbitrary since some works could be logically placed in more than one category.
We usually begin a search of the literature by referring to one or more of the
treatises on analytical chemistry or to those devoted to the analysis of specific types
of materials. In addition, it is often helpful to consult a general reference work
­relating to the compound or element of interest. From this survey, a clearer picture
of the problem at hand may develop, including the steps that are likely to be diffi-
cult, the separations that must be made, and the pitfalls to be avoided. Occasionally,
all the answers needed or even a set of specific instructions for the analysis may be
❯ found. Alternatively, journal references that lead directly to this information may
be discovered. Sometimes, we find only a general notion of how to proceed. Several
possible methods may appear suitable; others may be eliminated. At this point, it is
often helpful to consider reference works concerned with specific substances or spe-
cific techniques. If these works do not provide the desired information, the various
analytical journals may be consulted. Monographs on methods for completing the
analysis are often valuable in deciding among several possible techniques.
The most difficult aspect of using analytical journals is locating articles pertinent
to the problem at hand. The various reference books can be useful since most con-
tain many references to the original journals. The key to a thorough search of the
literature, however, is Chemical Abstracts, which was published in print form from
1907 through 2009. As personal computers became available, manual literature
searches were replaced by computer-based searches. Two major electronic databases
available through Chemical Abstracts are CAplus and CAS REGISTRY. In 2011,
the CAplus database contained some 34 million references to the scientific litera-
ture and to patents. The CAS REGISTRY is the authoritative source for chemical
names and ­structures and uniquely identifies chemical substances through a CAS
Registry ­Number. These databases and several others can be accessed by computer
with ­SciFinder and STN software. SciFinder is a web-based program with a graphical
interface that permits scientists to search the large bibliographic, structure, and reac-
tion databases. The STN software is intended primarily for information profession-
als and features a command-language interface. Computer-aided literature searches
through such programs have reduced dramatically the time required for a careful
literature search—from months or days to hours or even minutes.
 35B Choice of Analytical Method  965

35B-3 Choosing or Devising a Method

After defining the problem and investigating the literature for possible approaches,
we must next decide on the laboratory procedure. If the choice is simple and obvi-
ous, analysis can be undertaken directly. Frequently, however, the decision requires
the exercise of considerable judgment and ingenuity. In addition, experience, an
­understanding of chemical principles, and perhaps intuition all come into play.
If the substance to be analyzed occurs widely, the literature survey usually yields
several alternative methods for the analysis. Economic considerations may dictate a
method that will yield the desired reliability with the least expenditure of time and
effort. As mentioned earlier, the number of samples to be analyzed is often a deter-
mining factor in the choice.
In many cases, existing instrumentation and expertise in particular methods may
play a large role in the choice of methods. There is no reason to consider a mass spec-
trometric approach if a mass spectrometer is not available in the laboratory or in the
vicinity. Likewise, appropriate expertise may not be at hand for a newly purchased
instrument until weeks after installation. Hence, choices may be limited because of
the laboratory environment.
Investigation of the literature does not always reveal a method designed specifi-
cally for the type of sample in question. Ordinarily, however, we will find procedures
for materials that are at least similar in composition to ours. We must then decide
❮ Preliminary laboratory testing
may be needed to evaluate
proposed changes to established
whether the variables introduced by the differences in composition are likely to have methods.
any influence on the results. This judgment can be difficult, and we may still be
uncertain as to the effects. Experiments in the laboratory may be the only way of
making a wise decision.
If we conclude that existing procedures are not applicable, consideration must be
given to modifications that may overcome the problems imposed by the variation in
composition. Again, the complexity of the chemical system may dictate that we can
propose only tentative alterations. Whether these modifications will accomplish their
purpose without introducing new difficulties can be determined only in the laboratory.
After considering existing methods and their modifications, we may decide that
none fits the problem and an entirely new procedure must be developed. For this
approach, all the facts on the chemical and physical properties of the analyte must
be organized and considered. Several possible ways of performing the desired mea-
surement may become evident from this information. Each possibility must then be
examined critically, while considering possible influences of the other components in
the sample as well as to the reagents that must be used for solution or decomposition.
At this point, we must try to anticipate sources of error and possible interferences
due to interactions among sample components and reagents. It may be necessary to
devise strategies to circumvent such problems. The conclusion of such a preliminary
survey generally produces one or more tentative methods worth testing. Usually, the
feasibility of some of the steps in the procedure cannot be determined without pre-
liminary laboratory testing. Certainly, critical evaluation of the entire procedure can
come only from careful laboratory work.

35B-4 Testing the Procedure

Once a procedure for an analysis has been selected, we must decide whether it can
be applied directly to the problem at hand or it must be tested. The answer to this
question is not simple and depends on a number of considerations. If the method
chosen is the subject of a single, or at most a few, literature references, there may
966  CHAPTER 35 The Analysis of Real Samples
The National Institute of
Standards and Technology is an
important source for standard
reference materials. For literature
describing standard reference
materials and other standards,
see http://www.nist.gov.
be real value in a preliminary laboratory evaluation. With experience, we become
more and more cautious about accepting claims regarding the accuracy and applica-
bility of a new method. All too often, statements found in the literature tend to be
overly optimistic. A few hours spent in testing the procedure in the laboratory may
be enlightening.
Whenever a major modification of a standard procedure is undertaken or an
­attempt is made to apply it to a type of sample different from that for which it was
designed, a preliminary laboratory test is again recommended. The effects of such
changes simply cannot be predicted with certainty.
Finally, a newly devised procedure must be extensively tested before it is adapted
for general use. We now consider the means by which a new method or a modifica-
tion of an existing method can be tested for reliability.
The Analysis of Standard Samples
The best way to evaluate an analytical method is to analyze one or more standard
samples whose analyte composition is reliably known. For this technique to be
­effective, however, it is essential that the standards closely resemble the samples to
be analyzed with respect to both the analyte concentration range and the overall
composition.
Occasionally, standards suitable for method testing can be synthesized by thor-
oughly homogenizing weighed quantities of pure compounds. Such a procedure is
generally inapplicable, however, when the samples to be analyzed are complex, such
as biological materials, soil samples, and many forensic samples.
❯ Section 8E-3 discusses the general methods for validation of analytical results.
The National Institute of Standards and Technology sells a variety of standard refer-
ence materials (SRM) that have been specifically prepared for validation purposes. 1
Most standard reference materials are substances commonly encountered in industry
or in governmental and academic studies. Over 1200 standard reference materials are
available, in categories such as engineering, food and agriculture, health and clinical,
­environmental, high-purity, and industrial materials. Specific substances include fer-
rous and nonferrous metals; ores, ceramics, and cements; environmental gases, liq-
uids, and solids; primary and secondary chemicals; clinical, biological, and botanical
­samples; fertilizers; and glasses. The concentration of one or more components in these
SRMs is certified by the Institute based on measurements using (1) a previously vali-
dated reference method; (2) two or more independent, reliable measurement methods;
or (3) results from a network of cooperating laboratories, technically competent and
thoroughly familiar with the material being tested. Several industrial concerns also offer
various kinds of standard materials designed for validating analytical procedures.
When standard reference materials are not available, the best we can do is to
­prepare a solution of known concentration whose composition approximates that
of the sample after it has been decomposed and dissolved. Such a standard gives no
information at all concerning the fate of the substance being determined during the
important decomposition and solution steps.
Using Other Methods
The results of an analytical method can sometimes be evaluated by comparing them
with data obtained from an entirely different method, particularly if we know the
1
See U.S. Department of Commerce, NIST Standard Reference Materials Catalog, 2011 ed., NIST
Special Publication 260, Washington, D.C.: U.S. Government Printing Office, 2011. The catalog
can be downloaded from the NIST web site at http://www.nist.gov.
 35C Accuracy in the Analysis of Complex Materials   967

reliability of the reference method. The second method should be based on ­chemical
or instrumental principles that differ as much as possible from the method being
considered. Because it is unlikely that the same errors influence both methods, if we
obtain results comparable to the reference method, we can usually conclude that our
new method is satisfactory. Such a conclusion does not apply to those aspects of the
two methods that are similar.

Standard Addition to the Sample

When standard reference materials and different analytical methods are not appli-
cable, the standard addition method may prove useful. In addition to being used to
analyze the sample, the proposed procedure is tested against portions of the sample
to which known amounts of the analyte have been added. The effectiveness of the The standard addition method
method can then be established by evaluating the extent of recovery of the added is ­described in Section 8D-3.
quantity. The standard addition method may reveal errors arising from the way the ­Applications of standard additions
methods are presented in Chapters 21,
sample was treated or from the presence of the other elements and/or compounds in 26, and 28.
the matrix.

  Accuracy in the Analysis

35C of Complex Materials
To provide a clear idea of the accuracy that can be expected for the analysis of a com-
plex material, data on the determination of four elements in a variety of materials are
presented in Tables 35-1 to 35-4. These data were taken from a much larger set of
results collected by W. F. Hillebrand and G. E. F. Lundell of the National Bureau of
Standards (now NIST) and published in the first edition of their classical book on
inorganic analysis.2
The materials analyzed were naturally occurring substances and items of com-
merce. They were specially prepared to give uniform and homogeneous samples and
were distributed among chemists who were, for the most part, actively engaged in
the analysis of similar materials. The analysts were allowed to use the methods they
considered most reliable and best suited for the analysis. In most instances, special

Table 35-1
Determination of Iron in Various Materials*
Number of Average Average Relative
Materials Iron, % Analysts Absolute Error Error, %
Soda-lime glass 0.064 (Fe2O3) 13 0.01 15.6
Cast bronze 0.12 14 0.02 16.7
Chromel 0.45 6 0.03 6.7
Refractory 0.90 (Fe2O3) 7 0.07 7.8
Manganese bronze 1.13 12 0.02 1.8
Refractory 2.38 (Fe2O3) 7 0.07 2.9
Bauxite 5.66 5 0.06 1.1
Chromel 22.8 5 0.17 0.75
Iron, ore 68.57 19 0.05 0.07

W. F. Hellebrand and G.E.F. Lundell, Applied Inorganic Analysis, (New York: Wiley, 1929), p. 878. Reprinted by
permission of the Lundell Estate.

2
W. F. Hillebrand and G. E. F. Lundell, Applied Inorganic Analysis, New York: Wiley, 1929,
pp. 874–87.

Unless otherwise noted, all content on this page is © Cengage Learning.

968  CHAPTER 35 The Analysis of Real Samples

Table 35-2
Determination of Manganese in Various Materials*
Number of Average Average Relative
Material Manganese, % Analysts Absolute Error Error, %
Ferro-chromium 0.225 4 0.013 5.8
Cast iron 0.478 8 0.006 1.3
0.897 10 0.005 0.56
Manganese bronze 1.59 12 0.02 1.3
Ferro-vanadium 3.57 12 0.06 1.7
Spiegeleisen 19.93 11 0.06 0.30
Manganese ore 58.35 3 0.06 0.10
Ferro-manganese 80.67 11 0.11 0.14

W. F. Hellebrand and G.E.F. Lundell, Applied Inorganic Analysis, (New York: Wiley, 1929), p. 880. Reprinted by
permission of the Lundell Estate.

Table 35-3
Determination of Phosphorus in Various Materials*
Number of Average Average Relative
Material Phosphorus, % Analysts Absolute Error Error, %
Ferro-tungsten 0.015 9 0.003 20
Iron ore 0.014 31 0.001 2.5
Refractory 0.069 (P2O5) 5 0.011 16
Ferro-vanadium 0.243 11 0.013 5.4
Refractory 0.45 4 0.10 22
Cast iron 0.88 7 0.01 1.1
Phosphate rock 43.77 (P2O5) 11 0.5 1.1
Synthetic mixtures 52.18 (P2O5) 11 0.14 0.27
Phosphate rock 77.56 (Ca3(PO4)2) 30 0.85 1.1

W. F. Hellebrand and G.E.F. Lundell, Applied Inorganic Analysis, (New York: Wiley, 1929), p. 882. Reprinted by
permission of the Lundell Estate.

Table 35-4
Determination of Potassium in Various Materials*
Potassium Number of Average Average Relative
Material Oxide, % Analysts Absolute Error Error, %
Soda-lime glass 0.04 8 0.02 50
Limestone 1.15 15 0.11 9.6
Refractory 1.37 6 0.09 6.6
2.11 6 0.04 1.9
2.83 6 0.10 3.5
Lead-barium glass 8.38 6 0.16 1.9

W. F. Hellebrand and G.E.F. Lundell, Applied Inorganic Analysis, (New York: Wiley, 1929), p. 883. Reprinted by
permission of the Lundell Estate.

precautions were taken, and the results are consequently better than can be expected
from the average routine analysis.
The numbers in the second column of Tables 35-1 through 35-4 are best values,
obtained by the most painstaking determination of the measured quantity. Each is
considered to be the true value for calculating the absolute and relative errors shown
in the fourth and fifth columns. The fourth column was computed by discarding
extremely divergent results, determining the deviation of the remaining individual

Unless otherwise noted, all content on this page is © Cengage Learning.

 35C Accuracy in the Analysis of Complex Materials   969

data from the best value (second column), and averaging these deviations. The fifth
column was obtained by dividing the data in the fourth column by the best value
(second column) and multiplying by 100%.
The results shown in these tables are typical of the data for 26 elements reported
in the original publication. We conclude that analyses reliable to a few tenths of a
percent relative are the exception rather than the rule in the analysis of complex mix-
tures by ordinary methods and that, unless we are willing to invest a huge amount of
time in the analysis, errors on the order of 1 or 2% must be accepted. If the sample
contains less than 1% of the analyte, we must expect even larger relative errors.
These data show that the accuracy obtainable in the determination of an element
depends strongly on the nature and complexity of the substrate. Thus, the relative
­error in the determination of phosphorus in two phosphate rocks was 1.1%. In a
synthetic mixture, the error was only 0.27%. The relative error in an iron determina-
tion in a refractory was 7.8%. In a manganese bronze having about the same iron
content, the relative error was only 1.8%. In this example, the limiting factor in the
accuracy is not in the completion step but rather in the dissolution of the samples
and the elimination of interferences.
The data in these four tables are more than 80 years old, and it is tempting to
think that analyses carried out with more modern tools and additional experience
❮ Fundamental sources of
systematic and random error
that plagued us 80 years ago
are likely to have significantly better accuracy and precision. A study by S. Abbey still haunt us today.
suggests that is simply not the case however.3 For example, the data, which were
taken from Abbey’s paper, reveal no significant improvement in silicate analyses of
standard reference glass and rock samples in the 43-year period from 1931 to 1974.
Indeed, the standard deviation among participating laboratories appears to be larger
in later years.
The data in Tables 35-1 through 35-5 show that we must be skeptical of the
­a ccuracy of analytical results on real samples, even if we perform the analysis
ourselves.

Table 35-5
Standard Deviation of Silica Results*
Standard Deviation
Year Reported Sample Type Number of Results (% absolute)
1931 Glass 5 0.28†
1951 Granite 34 0.37
1963 Tonalite 14 0.26
1970 Feldspar 9 0.10
1972 Granite 30 0.18
1972 Syenite 36 1.06
1974 Granodiorite 35 0.46

Reprinted (adapted) with permission from S. Abbey, Anal. Chem., 1981, 53, 529A. Copyright 1981
American Chemical Society.

3
S. Abbey, Anal. Chem., 1981, 53, 529A, DOI: 10.1021/ac00227a718.

Unless otherwise noted, all content on this page is © Cengage Learning.