Trends in Analytical Chemistry 149 (2022) 116547

Contents lists available at ScienceDirect

Trends in Analytical Chemistry

journal homepage: www.elsevier.com/locate/trac

Unified principles of univariate analytical calibration

Paweł Ko
scielniak
Department of Analytical Chemistry, Faculty of Chemistry, Jagiellonian University, 2 Gronostajowa St., 30-387, Krakow, Poland

a r t i c l e i n f o a b s t r a c t

Article history: An attempt has been made to build a revised and logical concept of the univariate analytical calibration
Available online 1 February 2022 in qualitative and quantitative analysis. It focuses on the terms of real (virtual) and model (practical)
calibration functions and their similarity as a condition for correct calibration leading to a reliable
Keywords: analytical result. The similarity of both functions depends on the extent to which the random and sys-
Univariate analytical calibration tematic measurement changes, so-called uncontrollable analytical effects, are eliminated and/or
Calibration principles
compensated. It is assumed that the model function is able to be formulated exclusively with the use of
Calibration function
calibration standards, which can be not only chemical but also mathematical, biological or physical in
Measurement errors
Absolute methods
nature. On the basis of the developed concept it has been proved that every method of qualitative and
quantitative analysis requires analytical calibration. It was justified using the example of analytical
methods commonly regarded as absolute or standardless ones.
© 2022 Elsevier B.V. All rights reserved.

1. Introduction metrological [8] and statistical [9e11] points of view or in relation

to individual analytical methods, including mainly chromatography
Analytical calibration is the basic issue combining analytical [12,13] and inductively coupled plasma mass spectrometry (ICP-
chemistry as a science with chemical analysis understood as a MS) [14]. In the textbooks on analytical chemistry, little is usually
practical examination of the qualitative and quantitative chemical said about calibration, being mostly limited to basic, customary
composition of matter. The generally accepted division of analytical terms, approaches and applications. In vain do they look for an
calibration includes univariate and multivariate calibration. The explicit reference to the mentioned definition of analytical cali-
latter is based on chemometric methods and its goal in analytical bration and its coherent connection with calibration methods used
chemistry is to develop multivariate models that can be used to in chemical analysis.
predict the composition of matter from the measured properties of This state of affairs obviously has negative consequences. Above
the chemical system. In analytical practice, however, univariate all, it is not conducive to didactic goals, since it is difficult to reliably
calibration is much more often used, mainly due to the fact that it is convey specific analytical knowledge on the subject without a
simpler and more accessible in terms of mathematics and inter- unified, universally accepted approach. The lack of a common
pretation [1]. Without this process, which accompanies millions of ground for communication in this area can also become a source of
analyzes performed every day around the world, it is impossible to misunderstandings and ambiguities leading to erroneous and
obtain correct information about the type and amount of compo- incorrect analytical procedures. And yet, no one but a chemist-
nents that make up matter. analyst should be particularly sensitive to “order” and “cleanli-
Thus, it is somewhat of a paradox that while the theory of ness” and set an example of this in his or her actions.
multivariate calibration is the subject of many articles and books, The aim of this work is to at least partially fill the gap in this
including those developed in recent years [2e5], the place of uni- topic in the form of a proposal for a revised theoretical approach to
variate calibration in analytical chemistry, i.e. its theoretical basis univariate analytical calibration. Based on the formulated funda-
and general concept, has not yet received a clear and coherent mentals of calibration, the definition of this process will be
description. The last definition of univariate analytical calibration expanded along with concepts such as calibration standards of
recommended by the International Union of Pure and Applied different nature and calibration functions. Then, the calibration
Chemistry (IUPAC) was published over twenty years ago [6]. In later procedure will be discussed in the context of measurement errors.
articles, this topic was discussed rather from methodological [1,7] Special attention will be paid to the elimination and compensation
of so-called uncontrolled effects and to the consequence of both
these actions from the point of view of the accuracy of analytical
E-mail address: [email protected].

https://doi.org/10.1016/j.trac.2022.116547
0165-9936/© 2022 Elsevier B.V. All rights reserved.
P. Ko
scielniak Trends in Analytical Chemistry 149 (2022) 116547

determinations. Finally, the problem of the indispensability of the function and true analyte value are essentially unrealistic and
analytical calibration process in chemical analysis and the validity impossible to verify experimentally or mathematically.
of the related concepts of “absolute”, “calibration-free” and “stan- When the sample is prepared for analysis, the type or amount of
dardless” methods will be addressed. analyte in the sample to be analyzed takes on a real value, x0. The
relationship between the analytical signal and the type or amount
2. Principle of analytical calibration of analyte is described at this point by the real function, Y ¼ F(x),
which takes the value Y0 for the value x0:
Analytical calibration is defined by IUPAC as “… the operation
that determines the functional relationship between measured Y0 ¼ F(x0) (1)
values (signal intensities at certain signal positions) and analyt-
ical quantities characterizing types of analytes and their amount Although the value of Y0 is measurable, the exact form of the
(content, concentration)” [6]. The theory based on this definition, real function is unknown because it depends on a number of effects
developed in book [15], presents calibration as a mathematical act: and processes that led to the current state of this relationship
“The result of calibration is a model that may have the form of a during the preparation of the sample for measurement. Conse-
conversion factor, a mathematical equation or a graph”. Such a purely quently, the determination of the real result x0 by means of the real
theoretical approach is too general, even abstract, and unrelated to function is impossible.
analytical practice. In particular, it does not provide guidance on how This situation forces the formulation of an additional auxiliary
the functional relationship (calibration model) should be formulated model function, Y ¼ G(x). The role of this function is to replace the
in different analytical situations and how it relates to the different real function in the search for the real value, x0. It should therefore
types of methods used in qualitative and quantitative analysis. meet two basic conditions: to be known and well-defined and to be
To complement the above definition, it is proposed to introduce the most accurate approximation of the real function (G(x) 4 F(x)).
three functions (instead of a single one) that relate the signal to the To fulfill these conditions, a calibration standard (one or more)
analytical result: the true, real, and model functions. This approach, should be used, which should be similar to the sample and properly
inspired by George Box's famous quote “all models are wrong, but prepared for measurement.
some are useful” [16], is illustrated in Fig. 1. Assuming that the approximation of the real function by the
If a sample that an analyst takes for qualitative or quantitative model function, Y ¼ G(x), is accurate, then the inverse form of the
analysis contains a component (analyte) of interest, then before any model function, x ¼ G1(Y), has to be created that is called the
action is taken with the sample, the type of analyte and its quantity evaluation function [6]. Theoretically, it allows the value of Y0 to be
in the sample can be referred to as the true value (type or quantity), transformed into the real result, x0:
xtrue, of the analyte. If it were possible to measure the analytical
signal for that analyte at that moment, then the relationship be- x0 ¼ G1 (Y0) (2)
tween the resulting signal and its true type or quantity,
Ytrue ¼ T(xtrue) could be called the true function. In practice, the approximation of the real function by the model
However, the determination of the true function and the true function is never accurate because the real function is essentially
value of the analyte is not possible in practice because it requires unknown. Therefore, transformation (2) leads to a certain value xx:
the analyst's intervention in the form of preparing the sample for
measurement and performing the measurement. The initiation of xx ¼ G1 (Y0) (3)
even the simplest and shortest analytical steps results in a change
of the true analyte concentration in the sample that continues until which is an approximate measure of the real result, x0. This result
the analytical signal is measured. Thus, the concepts of true can also be considered as the final analytical result.

Fig. 1. Concept of analytical calibration based on the terms of true, Y ¼ T(x), real, Y ¼ F(x), and model, Y ¼ G(x), functions (virtual analytical steps and terms are denoted by dotted
lines; for details see text).

2
P. Ko
scielniak Trends in Analytical Chemistry 149 (2022) 116547

The processes of creating a model function and its approxima- nonlinear, usually increasing form. Moreover, in some calibration
tion and transformation are fundamental, integral, and necessary methods it is transformed to a decreasing function (e.g. in indirect
elements of analytical calibration. Thus, it can be said that method) or increasing and decreasing function over different an-
analytical calibration consists of approximating the real rela- alyte concentration ranges (e.g. in titration) [1].
tionship between the signal, Y, and the type, b, or amount, c, of
an analyte in a sample by means of a model function, and then
2.2. Calibration types
applying this function to transform the signal obtained for the
analyte in the sample to the analytical result.
Depending on the type of univariate model function used,
analytical calibration can be broadly divided into empirical cali-
2.1. Calibration in qualitative and quantitative analysis bration and theoretical calibration. In the first case, the model
function is formulated on the basis of the performed experiment,
In qualitative analysis, the measure of the type of analyte sought sensory perception or observation, and in the second e on the basis
in the sample is the value of the quantity characteristic for the of a mathematical description of physicochemical phenomena and
applied measurement method (e.g. wavelength, migration time, processes occurring in the sample during the analysis. In many
current voltage, etc.), i.e. illustratively speaking e the position of cases the calibration is also of a complex nature to varying degrees
the measurement signal on the axis of this type of quantity. In the (empirical-theoretical or theoretical-empirical) when, in order to
example shown in Fig. 2 this quantity is the migration time. When better represent the real function, empirical information is sup-
the sample measurement image is mapped sufficiently accurately ported by theoretical information or vice versa.
with the image of standard, any two signals at the standard and Calibration standards are a necessary part of any calibration
sample images of the same position are theoretically evidence of process. They can be of different nature: chemical, biological,
the presence of a particular component in both the standard and physical or mathematical. The common feature of calibration
sample. So, if one of these signals corresponds to the analyte in the standards is that they directly or indirectly allow the assignment of
standard, the other one is an evidence of this analyte in the sample. a measurement signal to a known, well-defined type or amount of
In some cases a single analyte can be identified by two (as shown in analyte. The standards are thus used to formulate a model function.
Fig. 2) or more signals. A widely recognized and used method of analytical calibration is
In qualitative analysis, the real and model functions describe the the empirical calibration performed with a chemical standard. This
relationship between the position of the signal and the type of the is a synthetic or (less commonly) natural material, single or multi-
analyte. From the mathematical point of view, they are then component, containing an analyte of known type or amount. In
discrete functions, as shown in Fig. 3A. special cases, a chemical standard contains a known type or
In quantitative analysis, the measure of analyte quantity is the amount of a substance that reacts with the analyte, or a known type
intensity of the measurement signal. Measurements generally or amount of an isotope of the element being determined.
consider the single signal with the highest intensity. The real and The similarity of a chemical standard to the sample to be
model functions, i.e. the relationships between the signal intensity analyzed is ensured by making its properties (physical, biological,
and the analyte amount, are continuous functions (see Fig. 3B) structural) and chemical composition as similar as possible (e.g.
because the signal intensity is a continuous quantity. The model using matrix matching technique [17]) or by selecting a material
function is usually formulated using two or more calibration with such characteristics. The model function, Y ¼ F(x), is formu-
standards. Over a range of analyte amounts, it can take a linear or lated from measurements made with the standard under

Fig. 2. Example of quantitative analysis performed using capillary electrophoresis: when the sample and standard images are matched to each other, the position of two signals on
the migration time axis (at 6.9 and 8.2 min) corresponding to the analyte (magenta dye, M) in the standard indicate the presence of this analyte in the sample.

3
P. Ko
scielniak Trends in Analytical Chemistry 149 (2022) 116547

Fig. 3. Analytical calibration in qualitative (A) and quantitative (B) analysis: analyte of type b0 or of quantity c0 in a sample is identified or determined as of type bx or of quantity cx
on the basis of model function, Y ¼ G (x), matched to real function, Y ¼ F(x), and of signal value Y0 measured for a sample.

conditions characteristic of the analyte, usually the same conditions 3. Calibration in the context of measurement errors
under which the Y0 signal for the analyte contained in the sample is
measured. The role of analytical calibration is not only to make it possible
For the construction of the model function and identification of to identify or determine an analyte in a sample, but also to do so
the analyte in qualitative analysis, a single standard is usually with as much accuracy and precision as possible. The measure of
sufficient, for which one or more signal position values are usually accuracy is the statistically significant difference between the
established (see Fig. 3A). In some cases, e.g. when the analyte to be analytical result obtained, xx, and the true type or amount of
identified is an organic compound or a whole sample, a series of analyte, xtrue, in the sample before it was subjected to any
standards whose characteristic measurement signals are collected analytical process. The measure of precision is the random dif-
in a computer database are used for analysis. In special situations, ference in analytical results obtained in so-called parallel analyses,
a sample is identified based on measurement information that is, performed in the same way and under the same experi-
collected for another sample (reference sample) of unknown mental conditions. The accuracy and precision of an analytical
composition. result are thus determined by any systematic and random changes
In quantitative analysis, a model function is formulated using at in the true function before it becomes, at the time of measure-
least two standards with different concentrations of an analyte ment, the real function, and then by the systematic and random
(element or chemical compound). Usually these are synthetic difference between the real function and its representation, the
standards, prepared by the analyst himself, or commercially avail- model function.
able. As chemical standards, the so-called certified reference ma- Changes in the analytical signal that occur both during sample
terials may be used, i.e. natural materials (more rarely synthetic) of preparation for measurement and during measurement, resulting
known amount of selected components, determined usually by in the transformation of the real function to the model function, can
interlaboratory studies [18]. Based on the measurement signal in- be called analytical effects. These effects can be controllable and
tensity values obtained for the standards, a mathematical, usually uncontrollable. Controlled analytical effects include, for example,
very simple form of the model function is determined. In general, changes caused by a targeted action by the analyst to decrease or
the aim is for the model function to be a linear function, i.e. to be increase the concentration of an analyte in a sample by dilution or
sufficiently well fitted to the real function in its linear form (see concentration, respectively. Effects of this type can usually be
Fig. 3B). calculated and corrected at the stage of analytical result calculation.
In both qualitative and quantitative analysis, other substances in During qualitative and quantitative analysis, however, there are
known, well-defined amounts are sometimes added to the sample also such changes of the analytical signal that are partially or
and to standards containing a known amount of an analyte in order completely out of the analyst's control. These uncontrolled
to enrich the analytical calibration role in various ways. Such a analytical effects can be both random and systematic. Although
substance is the so-called internal standard, which is added to the analyst is usually aware of the risk of their occurrence and
compensate for random measurement fluctuations (internal stan- usually tries to prevent them accordingly, he or she may overlook or
dard method), or a reagent reacting with the analyte, which, when even neglect them while performing the analysis. As a result,
subjected to measurements, indirectly indicates the amount of control over the entire analytical process is lost in a sense. Un-
analyte in the sample (indirect method) [1]. controlled effects manifest themselves by changing the position
In creating an empirical model using chemical standards, the and intensity of the analytical signal, i.e. they are important in both
analyst does not get into the theoretical aspects of the dependence qualitative and quantitative analysis.
of the analytical signal on the type or amount of analyte. Although
in some cases the laws and rules underlying this dependence (e.g. 3.1. Uncontrolled analytical effects
Lambert-Beer law, Nernst formula, etc.) may be helpful, calibration
using chemical standards is a universal procedure, essentially in- Uncontrolled effects can be caused by many factors manifesting
dependent of the chosen measurement technique. themselves at different stages of the analytical process. The clas-
Selected analytical methods based on empirical and theoretical sification of these effects covering all possible factors is, of course, a
calibration using different types of standards will be presented and matter of convention. The division presented below is the author's
discussed later in this article. proposal.
4
P. Ko
scielniak Trends in Analytical Chemistry 149 (2022) 116547

Uncontrolled effects are primarily caused by the analyst himself there is sometimes the problem of baseline, which varies more or
(the so-called human factor) as a result of incorrect or careless less randomly between spectra. The detection effect can also be
behavior at various stages of the analytical process. The magnitude related to natural phenomena underlying the measurement
of these changes depends primarily on the analyst's knowledge and method (a typical example is the phenomenon of self-absorption of
skills, that is, on his or her professional abilities and qualifications. radiation emitted in the emission spectrometry method, causing a
Personal factors such as tiredness, nervousness, hurry, etc. play a change in analytical signal intensity out of proportion to the
large role. The minimization of the human factor is also not favored amount of analyte in the sample).
by a routine, “automatic” approach to individual analytical activ- The signal measured for a specific type or amount of analyte can
ities, resulting, for example, from performing analyses according to also be affected by other components both naturally present in the
a single, unchanging analytical method over a long period of time. sample (native) and introduced during sample preparation for
The basic effects include preparative effect. Under this term we measurement. These components then take on the role of inter-
understand signal changes caused by such sample processing that ferents, and the signal change caused by them is the so-called
results in uncontrolled change (loss, less often gain) of analyte interference effect. If the effect comes exclusively from the
amount in the sample. The analyte can be partially lost e.g. when native components of the sample, then it is called a matrix effect,
changing the aggregate state of the sample (in order to make it while if the interferents are components added to the sample
suitable for the given measurement method) or when separating its during sample processing, then the induced changes are called a
components. The process of changing the amount (or rarely the blank effect. The interference effect can originate at the stage of
type) of sample can also take place outside the purposive, sample preparation for measurement (e.g. due to added reagents),
controlled action of the analyst as a result of e.g. an induced but can also be induced during measurement of the analytical
chemical reaction. A preparative effect is also involved when the signal as a result of phenomena and processes occurring at this
change in signal results directly from physical changes in the stage.
sample or standard (e.g. solution viscosity), or from changes in Finally, a specific effect is the speciation effect. It occurs when
environmental conditions under which the sample and standard an analyte contained in a sample unexpectedly for the analyst
are processed (e.g. temperature, humidity, illumination). changes its chemical form and at the same time changes its mea-
The instrumental effect is caused by the action of various surement sensitivity. As with the interference effect, this change
instrumental components used in the analytical process to process can occur before measurement (e.g. as a result of a chemical reac-
the sample prior to measurement. In this case, the source of tion) or at the time of measurement, when it involves a change in
random changes in the analytical signal are all natural imperfec- that form which is responsible for causing the analytical signal in
tions in the design and operation of these instruments, including the detection system (e.g. a change from atoms to analyte ions in
the measurement systems that characterize the measurement atomic absorption spectrometry).
method. However, as a result of instrument malfunction, signal Uncontrolled effects are revealed by a change in the analytical
changes can also be systematic. signal either directly or indirectly by changing the type or amount
The instrumental measurement system is the source of separate of analyte, as illustrated in Fig. 4.
specific measurement changes occurring in the detection system.
This phenomenon can therefore be called a detection effect. These 3.2. Elimination and compensation of uncontrolled effects
changes are manifested, for example, by the limited ability of the
system to respond proportionally to the analyte concentration, The natural way to avoid uncontrolled effects revealed during
which is natural for each detector. Another phenomenon is the so- sample handling is to employ various means of eliminating them.
called measurement trend, which consists of a successive increase Effectiveness of these actions largely depends on proper identifi-
or decrease in signal intensity over time. In spectrometric methods cation of the type of these effects and their sources, which are

Fig. 4. Pathways of the various uncontrolled effects.

5
P. Ko
scielniak Trends in Analytical Chemistry 149 (2022) 116547

differently situated on the analytical procedure plan. This is shown that effect compensation differs from the process of elimination in
in Fig. 4. that it does not eliminate the uncontrolled effects, but merely in-
The prerequisite for reducing the influence of the human factor volves equalizing them in the sample and standard.
is that analyses should be performed only by qualified staff, with a In an empirical calibration performed using a chemical stan-
high level of knowledge and skills, maintaining care and caution dard, it is easiest to compensate for instrumental effects because it
during the work. Instrumental and detection effects may not be a is sufficient to maintain instrumental conditions at the same opti-
major problem if the instruments used are of high quality, proven mum level during sample and standard preparation for measure-
reliability and low maintenance. In special cases where there are, ment. The detection effect is compensated for just as easily by using
for example, strong time trends or baseline shifts, special correction the same instrument for both sample and standard measurements
procedures are used [19]. and keeping the conditions of the measurements the same.
In contrast to instrumental and detection effects, speciation Compensating for preparative effects is more difficult, although
effects can be difficult to eliminate if the analytical procedure is it can be achieved to some extent by subjecting the standard to the
relatively complex and involves the use of different types of same preparative treatments to which the sample was subjected.
chemical reactions. The preparative effect can also be difficult to However, it must be taken into account that the analyte in the
eliminate effectively. This is because no sample processing is in standard may be subject to this effect to a different degree than the
practice free from partial loss of analyte. The degree of this phe- native analyte due to the different chemical environment and
nomenon should in each case be well recognized by preliminary potentially different chemical form. For speciation effects, it is very
experiments and then reduced as much as possible. The amount of important that the chemical form of the analyte remains the same
analyte lost can also be quantified (e.g. by the recovery method) in the sample and in the standard during the calibration procedure.
and the final analytical result can be corrected on this basis. If the analyte is present in several chemical forms in the sample, the
The interference effect can be eliminated in basically two ways. analyte in the standard need not take all of these forms, but should
The universal way is to remove the interferents from the sample or be in the form in which the analyte is to be determined in the
to isolate the analyte from the sample matrix by appropriately sample.
selected laboratory techniques (e.g. by extraction, crystallization, The most difficult effect to compensate for effectively is the
gaseous diffusion, etc.). Another approach is to add an appropri- interference effect. If the type and number of interferents causing
ately selected reagent to the sample to eliminate interferents by this effect in the sample are known, it is possible to compensate for
chemical means. their effect by adding them in sufficient quantity to the standard. It
Progressive elimination of uncontrolled effects causes the two is very difficult or even impossible in practice to reproduce the
analyte values, true, xtrue, and real, x0, to become increasingly sample composition in the standard in the case of unknown native
similar, as can be seen in Fig. 5. When the effects are completely interferents. It is best to use appropriately selected reference ma-
eliminated, the real analyte value becomes a measure of the true terials or the sample matrices without analyte for this purpose, if of
analyte value in the sample, i.e. x0 ≈ xtrue. course such materials are available.
When proceeding with an analytical calibration, the analyst is One unconventional calibration method, namely the standard
forced to use a standard. Importantly, however, this constraint addition method [1], offers a particular opportunity to compensate
simultaneously provides an opportunity to make the standard for interference effects. The method involves adding standard so-
similar to the sample. If the sample and standard are similar, then lutions with increasing analyte concentrations to the sample in
all uncontrolled effects occurring during the analytical procedure such a way that the concentrations of the other sample components
should, in theory, manifest themselves in the same way and with (including potential interferents) are kept constant. Interferences
appropriate strength with respect to both the sample and the are then compensated for, even if the type and amount of inter-
standard. As a result, compensation for these effects occurs. Note ferents present in the sample are not known.

Fig. 5. Impact of uncontrolled effects on an analyte in the sample during its preparation and measurement; due to elimination of effects the real analyte value approaches the true
value (x0 z xtrue).

6
P. Ko
scielniak Trends in Analytical Chemistry 149 (2022) 116547

The compensation of effects is offered by the calibration process relationship can be completely described on the basis of physical
and is therefore closely related to the representation of the real constants and universal quantities”, and further: “gravimetric,
function by the model function. The more accurate the approxi- volumetric and coulometric methods … may be considered abso-
mation of the two functions is, the more complete the compensa- lute methods provided that the experimental conditions are chosen
tion process is. Progressive compensation of effects promotes a so that their performance is theoretically predictable (preferably
progressive approximation of the analytical result, xx, to the real 100% performance)”. From a metrological point of view, methods
result, x0, as well as the real result, x0, to the true result, xtrue.Thus, such as gravimetry, volumetry and coulometry are considered
after complete compensation, the analytical result becomes an “primary methods”, i.e. “…whose operation is completely
estimate of the true value of the analyte in the sample, i.e. described and understood … and whose results are, therefore,
xx ≈ xtrue. This is illustrated schematically in Fig. 6. accepted without reference to a standard of the quantity being
Analytical calibration thus leads to an accurate analytical result measured” [21].
either by complete elimination of uncontrolled effects or by their What is characteristic of the definitions mentioned above is
complete compensation. Elimination of an effect thus does not their idealistic approach to analytical methods. After all, it is diffi-
require its compensation (e.g. once an interference effect has been cult to find in analytical chemistry a method whose “operation is
eliminated with a special reagent, there is no need to reconstruct completely described and understood” or “the functional rela-
the composition of interferents in the standard), although if it is tionship can be completely described on the basis of physical
known that the elimination of an effect may be incomplete, it constants and universal quantities”. Nor can it be required that the
should be compensated. Similarly, compensation for effects method be characterized by 100% efficiency in practice. These
(e.g. instrumental effects) does not require their elimination, rigorous conditions in themselves call into question the actual ex-
although any small reduction increases the chance of their com- istence of “absolute” methods, including even those that are
plete compensation. The processes of elimination and compensa- commonly regarded as such. It is also difficult to agree that
tion of uncontrolled effects are thus complementary activities in methods considered as “absolute methods” or “primary methods”
the sense that, taken together, they provide the best chance of need not require a “standard of the quantity being measured”.
achieving an accurate assessment of the true value of the analyte in The above comments and caveats support the thesis that every
the sample from the analytical result obtained. analytical method requires analytical calibration. The calibration
nature of a method is demonstrated by the following features:
4. “Absolute” analysis
 the type or amount of analyte cannot be determined directly
from the analytical signal, i.e. from a real function with a fixed
As a consequence of the previous discussion, the fundamental
form that does not rely on uncontrolled effects,
question should be asked: can chemical analysis be performed
 the method requires the use of a standard (not necessarily in
without analytical calibration? In general, based on scientific
chemical form) necessary for the formulation of an empirical or
sources, the answer to these questions is positive. The terms “ab-
theoretical model function that approximates the real function.
solute”, “calibration-free”, “standard-free” or “standardless” anal-
ysis (methods), often met in the literature as synonyms, generally
The analysis of above characteristics will provide a basis for
refer to analyses that are performed e according to the authors e
verifying the calibration nature of some analytical methods (shown
without the need for calibration.
in Table 1) for which such a nature is overlooked or questioned.
The above terms are to some extent clarified and distinguished
by IUPAC. According to the standpoint expressed in Ref. [6], the
terms “standards-free” or “standardless” methods should be 4.1. Empirical calibration
distinguished from “calibration-free” or “absolute” methods and
can be used only for those methods that require calibration, but the 4.1.1. Sensory method
calibration functions “… are transferable over time and between The most direct analytical method (in particular, of qualitative
laboratories under standardized operating conditions” (e.g. x-ray analysis) is the sensory method, which involves determining the
fluorescence spectrometry). In another IUPAC work [20], an type of analyte in a sample using the human senses. This takes
analytical method can be considered absolute if “… the functional advantage of the natural ability of humans to detect, distinguish,

Fig. 6. Impact of uncontrolled effects on an analyte in both the sample and standard during its preparation and measurement; due to compensation for effects the analytical result
value approaches the true value (xx z xtrue).

7
P. Ko
scielniak Trends in Analytical Chemistry 149 (2022) 116547

and evaluate the intensity of sensory impressions, particularly The uniqueness of gravimetry, revealed by the specific percep-
smell, color, and taste. Sensory ability is also used to compare tion of this method compared to other analytical methods, is due to
whether the analyte under study (sample component, sample) several reasons. One of them is that the response of the measuring
produces an impression of greater or lesser intensity than another instrument (measurement signal) is in this case measured in mass
analyte (reference). In this way, the analysis performed is semi- units, i.e. in the same units in which the analytical result is ob-
quantitative. Is this natural character of sensory analysis evidence tained. This gives the misleading impression that the result is
that it does not require any calibration process? determined directly from the weight reading, without the use of
Not, it isn't, because every human being has a well-developed any standard or model function, i.e. without analytical calibration.
sensory memory, which means that once a sensory impression In addition, unlike in other analytical methods, the result is ob-
(color, taste, smell) induced by a specific substance (analyte) is tained immediately, without the need to perform calculations or
assigned to a specific, appropriate receptor. Sensory memory, use any graph. Moreover, the analytical result is generally the same
therefore, acts as a kind of biological standard, which we acquire as the weight reading, because the weight automatically matches
in life through the acquisition of empirical information, and the its mass to the mass of the sample being analyzed (as schematically
memorized relationship between the impression (analytical signal) indicated in Fig. 7).
and the type of analyte is an empirical model function adapted to The gravimetric method is one of the most precise and accurate
the real situation (real function). analytical methods. This is due, on the one hand, to the naturally
As in any other chemical analysis, errors of both random and high accuracy with which the mass of the analytical weight can be
systematic nature are made in sensory analysis, resulting from determined, and, on the other, to the simple, very well-developed
various types of uncontrolled effects. Errors result, for example, and reliable construction of the analytical balance. Nevertheless,
from disturbances of perceived sensory impressions by compo- the indications of the balance are always subject to some random
nents other than the analyte, accidently present in the sample or in fluctuations, which affects the precision of the results. Further-
its environment. The color, odor or taste characteristic of the ana- more, both elements e the mass of the weight and the construction
lyte may then be altered by these substances. If these changes are of the balance e are subject to changes over time, which then
caused by substances introduced into the sample during the anal- become a source of uncontrolled effects and, consequently, a source
ysis, we are dealing with a typical interference effect, whereas if of analytical errors. Gravimetric results can also be systematically
they are caused by components in the sample environment, we can erroneous as a result of temperature and environmental conditions.
rather speak of a preparative effect. Other sources of preparative It is a well-known rule that the temperature of the substance to be
effect are factors that determine the environmental conditions such weighed should be brought to ambient temperature (to compen-
as temperature, humidity, color or illumination. sate for its effect on the analytical signal), and that disturbance of
The results of the sensory analysis are particularly exposed to the position of the analytical balance may affect the weight reading
the human factor. Sensory impressions perceived by a human are of the substance to be weighed.
naturally subjective and the accuracy of the analysis depends In addition to the undeniable advantages of the gravimetric
largely on both the individual minimum sensory sensitivity of the method, the disadvantage is that the standard (weight) is not able
analyst (the so-called sensory minimum) and his ability to recog- to be treated in the same way as the sample at the stage prior to
nize and remember sensations (so-called sensory memory). The measurement. Uncontrolled effects occurring at this stage in rela-
phenomenon of the so-called sensory fatigue, i.e. a gradual tion to the sample cannot therefore be compensated for, but can
decrease in the ability to perceive this feeling over time, is also only be eliminated and the accuracy of the analysis is determined
known. Memory and sensory model may even be completely lost or by the degree of this elimination. Care must therefore be taken in
indicate a completely erroneous relationship between our sensory particular not to lose any part of the sample or to introduce other
perception and the type or amount of analyte. As the ongoing substances in the form of impurities into the sample during its
pandemic period teaches us, such cases may be caused by, for preparation for weighing. An error due to this preparative effect
example, coronavirus disease (COVID-19).

4.1.2. Gravimetric method

Methods that do not require calibration in quantitative analysis
primarily include the gravimetric method. In this case, the analyst
uses a balance, which is the measuring instrument. The calibration
standard in gravimetry is a physical standard in the form of an
analytical weight of a well-defined and known mass. The rela-
tionship between the response of the balance and the mass of the
weight is an empirical model function, shown in Fig. 7 as a graph of
a linear function. Based on this model and the measurement result
obtained for the sample, the analytical result, i.e. the mass of the
sample, is determined.

Table 1
Empirical and theoretical analytical calibration based on different standard and
models used in different analytical methods.

Standard type Model type Calibration type Analytical method

Biological Empirical Empirical Sensory

Physical Gravimetry
Chemical Volumetry
Fig. 7. Analytical calibration in gravimetry: weight of sample (analytical result), mx, is
Mathematical Mathematical, fixed Theoretical Coulometry
obtained based on weight of weight (physical standard), mw, and on the corresponding
Mathematical, flexible LIBS
responses of a balance (measurement instrument), Y0 and Ym.

8
P. Ko
scielniak Trends in Analytical Chemistry 149 (2022) 116547

will not be correctable at the weighing stage and will adversely are becoming technically more and more perfect and reliable, so
affect the accuracy of the analytical result irrespective of the that measuring portions of very small volumes of liquid and their
physical perfection of the standard and the measuring instrument. addition and mixing with another liquid, and even the gradual
Finally, it should be noted that the gravimetric method is often reduction of the volume of the titrant portion as the end point is
used to determine an element or group of elements that are part of approached, can be performed with great precision and accuracy.
a weighed chemical compound. Also, many times the analyte is However, titration is not free from uncontrolled effects. Typical
subjected to a chemical reaction, and its amount is indirectly errors resulting from preparative and measurement effects include:
indicated by the weighed reaction product. In both these cases, the “droplet error” e consisting of exceeding the titration endpoint as a
amount of analyte is calculated on the basis of information of a result of adding the smallest unit volume of titrant to the sample,
theoretical nature e knowledge of the formula of the chemical “run-off error” caused by the phenomenon of solution adhesion on
compound or stoichiometry of the reaction used. The gravimetric the vessel walls, or “reading error” consisting of incorrect deter-
method can therefore be regarded as a method based on empirical mination of the solution volume due to the parallax effect. A critical
calibration with elements of theoretical calibration. step in the titration procedure is the accurate determination of the
end point.
4.1.3. Volumetric method A major source of systematic errors is effects related to the
The volumetric method (titration) is an empirical method in chemical reactions taking place. They are mainly of the nature of
which the titrant acts as a chemical standard [1]. The titrant is interference effects and most often result from the fact that the
gradually added to the sample (or vice versa) inducing a chemical selected reaction does not occur only with the selected analyte, but
reaction with a well-defined stoichiometry. As the reaction pro- also with other components of the sample. The influence of sub-
gresses, the equivalence point (stoichiometric point) is reached, i.e. stances accompanying the analyte in the sample can also be caused
the state where the amount of standard added is chemically by such effects as a shift in the equilibrium of the reaction (e.g. due
equivalent to the amount of analyte. The change in the amount of to the common ion effect), a change in the concentration product of
analyte as the reaction proceeds can be followed visually or with the solubility of precipitates (e.g. due to the salt effect). A speciation
the chosen measuring instrument to form an empirical model effect occurs when the analyte may be present in the sample at
function. The graphical form of this function (calibration graph) is least partially in a different chemical form than the predicted form
termed a titration curve. The image of the equivalent point on this and in this different form does not participate in the selected
curve is the so-called the titration end point. It can be the mini- chemical reaction.
mum, maximum or inflection point of the titration curve. Because of the necessary knowledge of the stoichiometry of the
Titration, commonly regarded as a “classical” analytical method, reaction and its exact course, titration, like the gravimetric method,
is not only subject to calibration, but is in fact an empirical cali- is a partially theoretical calibration method.
bration method with its own specificity expressed in the following
features:
4.2. Theoretical calibration

 the standard is mixed with the sample (and not prepared

4.2.1. Coulometric method
separately from the sample) and contains a compound of a well-
Coulometry is an electrochemical method based on two Fara-
defined concentration that reacts with the analyte (and not an
day's laws of electrolysis, which state that (I) the amount of
analyte of known concentration),
chemical change produced by current at an electrode-electrolyte
 the analytical result is calculated from the stoichiometry of the
boundary is proportional to the quantity of electricity used and
reaction and the volume of sample (or titrant) needed to reach
(II) the amounts of chemical changes produced by the same
the end point of the titration (and not from the analytical signal
quantity of electricity in different substances are proportional to
obtained for the analyte in the sample),
their equivalent weights. Both laws expressed mathematically lead
 the calibration graph (model function) is a representation of the
to a formula describing the fixed relationship between the electric
theoretical course of the reaction (real function) before and after
charge, Q, required for an electrochemical reaction involving a
the equivalent point is reached, so it has a different form before
substance of mass m:
and after the titration endpoint.
n,F
Thus, titration escapes to some extent the general rules of Q¼ ,m (4)
M
analytical calibration. However, like in any calibration method, the
determination of the analytical result requires a representation of where n is the number of electrons involved in the reaction, M is
the real function by a model function. In addition, like other cali- the molecular weight of the substance under test and F is Faraday's
bration methods, it is a way to obtain an analytical result using a constant. On the basis of this relation, it is possible to determine the
standard. unknown quantity (mass) of a substance subjected to a specific
Titration is one of the calibration methods providing the highest electrochemical reaction by measuring the corresponding value of
precision of analytical results. This is mainly due to its simplicity. electric charge as an analytical signal.
Although the chemical reactions leading to the determination of an Equation (4) meets the requirements for the absolute method
analyte are often multi-step and require from the analyst consid- posed in the work [20], since the functional relationship is
erable chemical knowledge and laboratory skills, these processes e completely described on the basis of physical constants and uni-
after careful preparation of the reagents and equipment e do not versal quantities. Nevertheless, it should be taken into account that
cause significant errors due to their very good reproducibility. The Faraday's laws of electrolysis are nevertheless valid only when
titration process usually does not require additional instrumental well-defined experimental conditions are met, above all: the stoi-
sample processing steps, thus sources of accidental measurement chiometry of electron reaction is known, and the electrode reaction
errors are few. proceeds with 100% efficiency.
The possibility of full mechanization and automation of the In practice, these conditions are not possible to fulfill. Side re-
entire titration process is also of great importance for achieving actions may occur with e.g. the solvent, the electrode material, the
precise results. The instruments for titration, the so-called titrators, components of supporting electrode or the products of the
9
P. Ko
scielniak Trends in Analytical Chemistry 149 (2022) 116547

electrolysis undergoing secondary electrode reaction. Other com- determined from the measured spectra by identifying the observed
ponents of the sample may also react, producing the typical inter- spectral lines.
ference effect. A separate problem in coulometry is the accuracy of LIBS has many advantages over other elemental analysis tech-
determining the amount of electricity required for an electro- niques, such as high potential for in situ and real-time analysis, the
chemical reaction. In addition, when the electrode potential is held ability to analyze samples in any phase (solid, liquid, or gas) at the
constant, accurate and reproducible results can be obtained by microscale and without destroying them, broad elemental coverage
minimizing the residual current, which is virtually impossible to (including lighter elements), and extremely fast measurement
eliminate completely. times (typically a few seconds for single point analysis). On the
Due to the fact that the function (4) is therefore subject to other hand, the LIBS signal depends not only on the analyte con-
numerous uncontrolled effects, it cannot be treated as a real centration, but also on the composition and aggregate state of the
function and the coulometric method as an absolute method. matrix. This is because the nature of the plasma and the behavior of
Function (4) should therefore be considered as a mathematical the atoms depend on many variables related to the origin of the
model function. It is not an exact representation of the actual state ablated material: composition, crystallinity, optical reflectivity,
(the real function), but only an approximation of the real function optical transmittance, and surface morphology. For this reason, in
to varying degrees depending on the extent to which the experi- many cases it is very difficult to obtain or purchase standards with
mental conditions deviate from those required theoretically. appropriate properties and composition, and the empirical cali-
The question may be asked: what in this case is a calibration bration process is often the most tedious and lengthy step in the
standard? As assumed earlier, a standard should directly or indi- analysis.
rectly allow the practical assignment of a measurement signal to a In this situation, theoretical modeling in the LIBS method is
known, well defined type or amount of analyte, i.e. the formulation particularly justified. The first basic model was developed by Ciucci
of a model function. In coulometry, the model function is estab- et al. [23]. The construction of this model is based on several as-
lished by Faraday's laws, so they act as mathematical standards (in sumptions: (a) the plasma composition is representative of the
place of chemical standards usually used in electrochemical actual material composition prior to the ablation, (b) in the actual
methods). temporal and spatial observation window, the plasma is in local
It is worth noting that a mathematical standard derived from thermal equilibrium (LTE) conditions, (c) the radiation source is
such an unquestionable source as Faraday's law cannot be subject optically thin, (d) self-absorption effect doesn't occur. Assuming
to any modifications that take into account the current state of the such simplified conditions, the dependence of the analytical signal,
sample and any uncontrolled effects that the analyte in the sample l (integral line intensity measured at wavelength l) on the con-
Iki
undergoes. It therefore provides no way to compensate for these centration of the emitting atomic species (i.e. element with a
effects. The only way to increase the reliability of determinations is determined electric charge), c, are defined by formula:
therefore to eliminate these effects.
In spite of the reservations against coulometry as a calibration-
free method, it must be clearly stated that this method finds a gk ,eðEK =KB TÞ
unique place among analytical methods. This is due to the Iki
l ¼ F , Aki , ,c (5)
Us ðTÞ
extremely strict definition, not found in other methods, of the
dependence of the analytical signal on the quantity of a substance. where Ns is the number density of the emitting atoms for each
Owing to this, under specific, well-recognized conditions and with species, Aki e the probability of transition between the upper (k)
the use of modern measuring instruments, it is possible to obtain and lower (i) energy states, gk and Ek e the degeneracy and energy
results of exceptional precision and accuracy. of upper level k, respectively, KB e the Boltzmann constant, T e the
plasma temperature; and Us(T) e the partition function for the
emitting species at the given plasma temperature. By experimen-
4.2.2. LIBS method tally measuring the intensity of the plasma emission spectrum and
For many years, attempts have been made to create theoretical querying the parameters Aki, gk, and Ek in the spectral database, one
models based on a mathematical description of physicochemical can convert the spectral line to a point in a two-dimensional
phenomena and processes occurring during the analysis performed Boltzmann plane [23].
with a particular measurement method. Such models are intended A great advantage of the model (5) is the possibility to obtain
to replace conventional empirical calibration and free the analyst results taking into account (at least to some extent) the matrix
from the process of preparing chemical standards and related effect. A good picture of the present state of affairs is given by the
problems. However, these studies are extremely difficult and usu- recently obtained results of the multicomponent analysis of Ant-
ally do not lead to satisfactory results (e.g. in case of analysis by arctic soils, which are partly summarized in Table 2 [24]. A signif-
flame atomic absorption spectrometry (FAAS) [22]). Most often it icant improvement in this respect is provided by the additional use
turns out that the processes to be taken into account are too little of internal standards, but then the calibration loses its theoretical
recognized to be given a sufficiently precise mathematical form, value.
and making approximations to them in various ways is usually
unreliable. The difficulties are enhanced by the fact that models
often have to be formulated individually for particular analytes or Table 2
Results of multicomponent analysis of a soil sample obtained by LIBS using theo-
even types of samples analyzed. Separate problems arise when it is
retical model and by inductively coupled plasma optical emission spectrometry
necessary to include in the theoretical model uncontrolled effects (ICP-OES) in the conventional calibration way [23].
(e.g. interference) to which the sample is subjected.
Analyte Contents (%) Analyte Contents (%)
The most spectacular theoretical modeling results have been
achieved for the laser-induced breakdown spectroscopy (LIBS) Calculated Experimental Calculated Experimental
method. In this method a powerful laser pulse is focused on the SiO2 34.7 ± 2.7 52.9 ± 0.5 MgO 3.2 ± 0.8 4.7 ± 0.1
sample surface, resulting in the ejection of its material, which leads Al2O3 15.5 ± 1.6 16.9 ± 0.2 Na2O 1.9 ± 0.2 4.4 ± 0.0
to the formation of a plasma plume. The emitted radiation from the Fe2O3 11.9 ± 1.7 0.3 ± 0.2 TiO2 3.2 ± 0.3 1.7 ± 0.0
CaO 9.7 ± 1.5 8.8 ± 0.1 K2O 0.7 ± 0.1 0.5 ± 0.2
plasma plume is recorded, and the composition of the sample is
10
P. Ko
scielniak Trends in Analytical Chemistry 149 (2022) 116547

It is obvious, however, that despite all efforts to make such analytical result can approximate the true result even if these ef-
mathematical models as model (5) as improved as possible, they will fects have not been eliminated beforehand. This aspect of an
always be only an approximation of the actual dependence of the important aspect of analytical calibration is not generally recog-
analytical signal on the amount of analyte in the sample, i.e. a real nized and appreciated.
function. Therefore, they should be treated as mathematical model It follows from the above considerations that one should accept
functions formulated on the basis of mathematical standards, the necessity of analytical calibration in qualitative and quanti-
i.e. mathematical forms (approximate as a rule) given to particular tative analysis, and not treat any calibration approaches that do
phenomena, processes and properties characteristic for a given not require the use of a chemical standard as calibration-free
analytical system. In the process of creating these models, informa- procedures. Instead, more attention should be focused on the
tion obtained experimentally is often used, which gives them a rules for proper calibration and, in particular, on the proper and
theoretical-empirical character. Due to the fact that they can be efficient use of standards. In this context, any activities facilitating
formulated in a rather arbitrary way, they can be called “flexible” e.g. the availability of certified materials and analyte-free stan-
models in contrast to “fixed” models represented by equation (4) dard samples, as well as research on the development of new
resulting from Faraday's laws. methods for empirical calibration (i.e. new proposals for sample
and standard handling) should be considered particularly valu-
5. Conclusions able. Theoretical calibration efforts are also very important, but
they should not overshadow the role and importance of empirical
The vision of the basis of analytical calibration presented in this calibration in those cases where current knowledge does not
paper is based on the observation of the essence of analytical allow the creation of sufficiently reliable standards and mathe-
research. The aim of the analyst's work is to look deep into the matical models.
matter and determine the type or quantity of components con-
tained in it. In other words, every chemical analysis is, in a sense, a Declaration of competing interest
quest for the objective truth that the surrounding reality conceals.
This goal can never be fully achieved, but it can be approached as The author declares that he has no known competing financial
closely as possible with the tools created by the analyst. Chemical interests or personal relationships that could have appeared to
analysis is thus a confrontation between what is unattainable (true, influence the work reported in this paper.
theoretical) and what can be created and achieved in practice.
Analytical calibration as part of chemical analysis is subject to References
this logic. Hence, such terms as “standard” (in the extended sense),
“model function”, and “analytical result” are the practical equiva- [1] P. Ko
scielniak, M. Wieczorek, Univariate analytical calibration methods and
lents of the terms “sample”, “real function”, “true type or amount of procedures. A review, Anal. Chim. Acta 944 (2016) 14e28. https://doi.org/
10.1016/j.aca.2016.09.024.
analyte.” The latter are largely hypothetical, unknown to the ana- [2] W.K. Ha €rdle, L. Simar, Applied Multivariate Statistical Analysis, Springer-
lyst, while the former are known and are approximations of the Verlag, 2015.
[3] A.C. Olivieri, G. Escandar, H.C. Goicoechea, A.M. de la Pen ~ a, Fundamentals and
latter. Just as the composition and properties of a sample can never
Analytical Applications of Multiway Calibration, Elsevier, 2015.
be faithfully reproduced in a model, the form of the true function [4] A.C. Olivieri, Introduction to Multivariate Calibration. A Practical Approach,
cannot be accurately approximated by a model function, and the Springer Nature, 2018.
actual type or amount of analyte in the sample at the time the [5] K. Backhous, B. Erichson, S. Gensler, R. Weiber, T. Weiber, Multivariate Anal-
ysis: An Application-Oriented Introduction, Springer Gabler, 2021.
analytical signal is measured can only be approximated by the [6] K. Danzer, L.A. Curie, Guidelines for calibration in analytical chemistry. Part 1.
analytical result obtained. This approach is, in the author's opinion, Fundamentals and single component calibration, Pure Appl. Chem. 70 (1998).
entirely natural. https://doi.org/10.1351/pac199870040993, 993-114.
[7] L. Cuadros-Rodriguez, L. G
amiz-Gracia, E. Almansa-Lopez, J.M. Bosque-Sendra,
The same logic underlies the thesis that calibrations must be
Calibration in chemical measurement processes. II. A methodological
performed regardless of analytical circumstances. One cannot approach, Trends Anal. Chem. 20 (2001) 620e636. https://doi.org/10.1016/
expect that the “truth” in the form of the true kind or quantity of S0165-9936(01)00111-X.
[8] L. Cuadros-Rodriguez, L. Ga
miz-Gracia, E. Almansa-Lo
pez, J. Laso-Sa
nchez,
substance in the matter under investigation can be revealed
Calibration in chemical measurement processes. I. A metrological approach,
directly, without any auxiliary action. That action is analytical Trends Anal. Chem. 20 (2001) 195e206. https://doi.org/10.1016/S0165-
calibration, which makes it possible to approach and determine the 9936(00)00093-5.
analytical objective, but always by an indirect route by means of a [9] C. Osborne, Statistical calibration e a review, Int. Stat. Rev. 59 (1991)
309e336. https://doi.org/10.2307/1403690.
standard and a model function. The type of standard is not relevant [10] J.N. Miller, Basic statistical methods for analytical chemistry. Part 2. Calibra-
here, because, as has been shown, a mathematical or biological tion and regression methods. A review, Analyst 116 (1991) 3e14. https://
standard can successfully perform the same task as a chemical doi.org/10.1039/AN9911600003.
[11] Ł. Komsta, Chemometric and statistical evaluation of calibration curves in
standard (commonly regarded as only possible and correct). pharmaceutical analysis e a short review on trends and recommendations,
The degree to which the analytical objective is approached de- J. AOAC Int. 95 (2012) 669e672. https://doi.org/10.5740/jaoacint.SGE_Komsta.
pends on the magnitude of the errors made, which are derived from [12] L.E. Vanatta, D.E. Coleman, Calibration, uncertainty, and recovery in the
chromatographic sciences, J. Chromatogr. A 1158 (2007) 47e60. https://
changes in the measurement signal called analytical effects in this doi.org/10.1016/j.chroma.2007.02.040.
paper. Calibration is an analytical step characterized by its own [13] L. Cuadros-Rodriguez, M. Bagur-Gonza
lez, M. Sa
nchez-Vin ~ as, A. Gonza

lez-
analytical errors. This is due to the contribution of the standard, Casado, A.M. Go
mez-Sa
ez, Principles of analytical calibration/quantification
for the separation sciences, J. Chromatogr. A 1158 (2007) 33e46. https://
always inherently imperfect, introducing laboratory and measure- doi.org/10.1016/j.chroma.2007.03.030.
ment errors (in empirical calibration) or mathematical approxi- [14] N. Miliszkiewicz, S. Walas, A. Tobiasz, Current approaches to calibration of LA-
mations and corrections (in theoretical calibration). From the ICP-MS analysis, J. Anal. Atom. Spectrom. 30 (2015) 327e338. https://doi.org/
10.1039/C4JA00325J.
imperfection of the calibration standard arises the imperfection of
[15] K. Danzer, Analytical Chemistry. Theoretical and Metrological Fundamentals,
the model function, which may only have some resemblance to the Springer, 2007.
real function, and this in turn results in reduced precision and ac- [16] G.E.P. Box, Science and statistics, J. Am. Stat. Assoc. 71 (1976) 791e799.
curacy of the analytical results. On the other hand, it is only the https://doi.org/10.1080/01621459.1976.10480949.
[17] K. Kleigrewe, A.C. So € hnel, H.U. Humpf, A new high-performance liquid
contribution of the model to the analytical calibration that allows chromatography-tandem mass spectrometry method based on dispersive
uncontrolled analytical effects to be compensated for, so that the solid phase extraction for the determination of the mycotoxin fusarin C in

11
P. Ko
scielniak Trends in Analytical Chemistry 149 (2022) 116547

corn ears and processed corn samples, J. Agric. Food Chem. 59 (2011) [22] B.V. L'vov, D.A. Katskov, L.P. Kruglikova, L.K. Polzik, Absolute analysis by flame
10470e10476. https://doi.org/10.1021/jf2026814. atomic absorption spectroscopy: present status and some problems, Spectrochim.
[18] M. Thompson, S.L.R. Ellison, R. Wood, Harmonized guidelines for single- Acta B 31 (1976) 49e80. https://doi.org/10.1016/0584-8547(76)80005-5.
laboratory validation of methods of analysis, Pure Appl. Chem. 74 (2002) [23] A. Ciucci, M. Corsi, V. Palleschi, S. Rastelli, A. Salvetti, E. Tognoni, New pro-
835e855. https://doi.org/10.1351/pac200274050835. cedure for quantitative elemental analysis by laser-induced plasma spec-
[19] K.H. Liland, T. Almøy, B.H. Mevik, Optimal choice of baseline correction for troscopy, Appl. Spectrosc. 53 (1999) 960e964. https://doi.org/10.1366/
multivariate calibration of spectra, Appl. Spectrosc. 64 (2010) 1007e1016. 0003702991947612.
https://doi.org/10.1366/000370210792434350. [24] J.M. Anzano, A. Cruz-Conesa, R.J. Lasheras, C. Marina-Montes, L.V. Pe
rez-
[20] A. Hulanicki, IUPAC commission on general aspects of analytical chemistry, Arribas, J.O. C

aceres, A.I. Vela
squez, V. Palleschi, Multielemental analysis of
absolute methods in analytical chemistry, Pure Appl. Chem. 67 (1995) 1905. Antarctic soils using calibration free laser-induced breakdown spectroscopy,
https://doi.org/10.1351/pac199567111905. Spectrochim. Acta B 180 (2021) 106191. https://doi.org/10.1016/j.sab.2021.
[21] W. Richter, Primary methods of measurement in chemical analysis, Accredit. 106191.
Qual. Assur. 2 (1997) 354e359. https://doi.org/10.1007/s007690050165.

12