Spectrochimica Acta Part B 58 (2003) 1997–2014

Views and criticisms

A discussion about the significance of Absorbance and sample
optical thickness in conventional absorption spectrometry and
wavelength-modulated laser absorption spectrometry
Ove Axnera,1, Jorgen
¨ Gustafssonb, Florian M. Schmidtb, Nicolo` Omenettoa,*,
James D. Winefordnera
a
Department of Chemistry, University of Florida, Gainesville, FL 32611-7200, USA
b
Department of Physics, Umeå University, SE-901 87 Umeå, Sweden

Received 16 December 2002; accepted 21 August 2003

Abstract

One of the most frequently used concepts in atomic absorption spectrometry (AAS) is that of Absorbance, an
easily measurable quantity that is linear with the concentration of the analyte over a large range of concentrations
whenever Beer’s law is valid. Laser-based modulation techniques, however, in particular wavelength modulated diode
laser absorption spectrometry (WMAS), do not measure exactly the same physical quantities as conventional AAS
techniques do, since they do not rely upon separate measurements of the intensities in the presence and in the absence
of the sample, but rather (a certain fraction of) the difference between the signals ‘on’ and ‘off’ resonance. Hence,
Absorbance is not as natural for the modulated laser-based absorption techniques as it is for the conventional AAS
techniques. The entity called sample optical thickness (SOT) appears to be a natural entity as well as a natural unit
(denoted SOT units) for the quantification of signals in WMAS. The present paper discusses the concept of
measurable quantities and their units in WMAS in some detail and compares them (theoretical and practical
considerations) with those of conventional AAS. In particular, it makes a distinction between the ‘observed’ sample
optical thickness and the ‘true’ sample optical thickness and shows how these two entities are related to the
Absorbance entity. Finally, this work also introduces a dimensionless sensitivity of the WMAS technique, and shows
this quantity to be equal to the nth Fourier coefficient of the wavelength modulated, peak-normalized line shape
function.
䊚 2003 Elsevier B.V. All rights reserved.

Keywords: Absorbance; Sample optical thickness; Wavelength modulated absorption; Diode laser absorption spectroscopy

*Corresponding author. Fax: q1-352-392-5362.

E-mail addresses: omenetto@chem.ufl.edu (N. Omenetto), ove.axner@physics.umu.se (O. Axner).
1
On leave from Department of Physics, Umeå University, SE-901 87 Umeå, Sweden.

0584-8547/03/$ - see front matter 䊚 2003 Elsevier B.V. All rights reserved.
doi:10.1016/j.sab.2003.08.006
1998 O. Axner et al. / Spectrochimica Acta Part B 58 (2003) 1997–2014

1. Introduction Typical limits of detection (LODs) for the mod-

ulated laser-based absorption techniques are in the
One of the most frequently used concepts in
10y5 –10y6 range, although a few demonstrations
atomic absorption spectrometry (AAS) is that of
of detection in the 10y7 range also have appeared
Absorbance, A w1,2x. The reason for the usefulness w16–27x. These numbers refer to the minimum
and the popularity of Absorbance is that it express-
fractional absorption (also called ‘absorption units’
es an easily measurable quantity that is linear with w24x) obtained for a given atomic or molecular
the concentration of the analyte over a large range
absorption transition (see below) and can be trans-
of concentrations whenever Beer’s law is valid. It
lated into more familiar LODs, expressed either as
is also fairly straightforward to work with since A
absolute amounts (e.g. in g), relative concentra-
consists of the (decadic, i.e. base 10) logarithm of
tions (e.g. pgyml) or number densities of atoms
the ratio of two quantities that are easily measur-
and molecules in the gas phase (e.g. 1ycm3) by
able in conventional AAS: the intensities in the
the simple relation, which holds at low concentra-
absence and in the presence of the analyte, I0 and
tions, between the fractional absorption and the
I, respectively.
product of the absorption cross-section, the number
Wavelength-modulation techniques and the
density and the absorption length (Eqs. (2)–(5)).
associated name ‘derivative spectrometry’ are well
For example, the LOD for ammonia corresponding
known in spectrochemical applications and have
to a fractional absorption of 10y6 was reported as
been extensively used in the past in emission and
2.4=108 moleculesycm3, corresponding to a con-
absorption spectrometry, mainly for improving the
centration of 0.007 mgym3 or 9 pptv w22x. Atomic
selectivity of the techniques in the presence of
number densities down to approximately 1000
spectral interferences (see, for example, the dis-
atomsycm3 can correspond to fractional absorption
cussion given in the book by Alkemade et al. w2x
of 10y7 w20x. In electrothermal AAS, a value of
and an old instrumental paper by O’Haver w3x).
6.5=10y6 for rubidium translates into a character-
The advent of diode lasers and the possibility of
istic mass of 6.5 pg and an absolute LOD of
high frequency modulation of the light source have
approximately 10 fg w18x. It is important to note,
revitalized the field of modulation spectroscopy in
however, that fractional absorption is expressed
the recent years by also improving the signal-to-
as the minimum detectable relative intensity dif-
noise ratios of the spectrometric techniques in
ference, DIyI0, where DI is given by the difference
general. The introduction of a rapid modulation of
between the signals ‘on’ and ‘off’ resonance, i.e.
the wavelength or the frequency of the light is
I0 –I, and is therefore not the Absorbance. The
motivated by the fact that the detection can be
reason stems from the fact that the laser-based
shifted to higher frequencies where the flicker
modulation techniques in fact do not measure
noise (or the so-called 1y f-noise) is substantially
exactly the same physical quantities as convention-
reduced w4–9x. As a consequence, the new laser-
al AAS techniques do. They do not rely upon
based wavelength- andyor frequency-modulated
separate measurements of the intensities in the
absorption techniques, commonly referred to as
presence and in the absence of the sample, I0 and
wavelength-modulation diode laser absorption
I (which are needed for determination of Absorb-
spectrometry (WMAS)2 w6–12x and frequency-
ance), but rather (a certain fraction of) the differ-
modulated absorption spectrometry (FMS) w4–
ence between the signals ‘on’ and ‘off’ resonance,
9,13–15x, are capable of detecting significantly
i.e. DI, as clearly has been pointed out in Refs.
lower concentrations of atoms and molecules than w7,8,11,12,28–30x, often referred to as the absorp-
conventional (unmodulated) AAS techniques.3
tion of the sample. It should be clear that it is
2
Often also abbreviated as WMS, WM-DLAS or WM- neither straightforward nor extremely convenient
DLAAS, or sometimes also referred to as tunable diode laser to compare a technique that measures the differ-
absorption spectrometry (TDLAS).
3
Although many of the general conclusions in the introduc- ence between the signals ‘on’ and ‘off’ resonance
tion are valid also for the FMS technique, we will in this to an entity that is defined in terms of the loga-
paper focus on the WMAS technique because of its wider use. rithm of the ratio of the two intensities, as is the
 O. Axner et al. / Spectrochimica Acta Part B 58 (2003) 1997–2014
case for Absorbance.4 Hence, Absorbance is not
as natural for the modulated laser-based absorption
techniques as it is for the conventional AAS
techniques, wherefore it loses some of its attrac-
tiveness for these techniques.
Although the measurable entities for modulated
laser-based absorption techniques in principle
could be expressed in terms of absorbance units,
they have so far often been given in terms of
fractional absorption. In the limit of validity of
Beer’s law (i.e. in absence of optical saturation
and optical pumping effects), the absorption equa-
tion (see below) contains, in the exponent, the
product of three parameters, either the absorption
coefficient, the absorption path length and the
concentration of analyte or alternatively, the num-
ber density of absorbers, the absorption cross-
section and the absorption path length,5 which
define another well-known entity, called the ‘opti-
cal thickness’ or ‘optical depth’ w2,31,32x. This
entity, which recently has been referred to as the
dimensionless sample optical thickness (sometimes
abbreviated SOT and denoted a6) in the WMAS
literature w28,29x,7 has found use as a natural entity
as well as a natural unit (denoted SOT units) for
quantification of signals and LODs in WMAS
w28,29x. Since sample optical thickness is directly
proportional to the number density of absorbers in
the interaction region, the output of the WMAS
technique, being directly proportional to the sam-
ple optical thickness entity when optically thin
samples are analyzed (which is the range of
4
As an example, it is not even trivial to create an entity for
WMAS that is proportional to the concentration of the analyte
over the same concentration range as is the AAS technique
w29x. On the other hand, it is possible to find alternative
physical measurable entities that are linear with concentration
of the analyte over quite a large range of concentrations, this
range being predominantly lower than that of conventional
AAS techniques.
5
Or as two parameters, the absorptivity and the absorption
path length.
6
This is not to be confused with the a that often is used
for the fraction of radiation absorbed in conventional AAS.
7
As will be further discussed below, the sample optical
thickness is a dimensionless entity, given by the product of
the physical quantities in the exponent of Beer’s law. Although
the correct expression for this would be the dimensionless
sample optical thickness, we will in this work for simplistic
reasons only refer to it as sample optical thickness.
1999
applications where laser-based techniques most
often are used), is then directly proportional to the
number density of absorbers in the interaction
region for small concentrations w29x.
The present paper discusses the concept of
measurable quantities and their units in WMAS in
some detail and compares them (theoretical and
practical considerations) with those of convention-
al AAS. In particular, it makes a distinction
between the ‘observed’ sample optical thickness
and the ‘true’ sample optical thickness and shows
how these two entities are related to the Absorb-
ance entity. The paper also introduces the defini-
tion of the dimensionless sensitivity of the WMAS
technique as the ratio of the nth harmonic DC-
signal-normalized WMAS output, denoted as S̄n,
and the sample optical thickness (with the nth
harmonic DC-signal-normalized WMAS output,
S̄n, being defined as Sn yS0 where Sn is the nth
harmonic output of the WMAS instrumentation
and S0 the signal from an unmodulated light source
in the absence of analytes). For the case investi-
gated here, i.e. in the absence of any associated
intensity modulation, it is shown that the dimen-
sionless sensitivity of the WMAS technique for
nth harmonic detection is given by a very funda-
mental entity in WMAS, namely, the nth Fourier
coefficient of the wavelength-modulated lineshape
function, x̄n.
2. Conventional atomic absorption spectrometry
and the concept of Absorbance
As is well known, Absorbance is defined as the
(decadic) logarithm of the ratio of the intensities
in the absence and in the presence of absorbers,
I0 and I, respectively, i.e. as
I0
A'log . (1)
I
Although Absorbance is dimensionless, scien-
tists often referred to its units as ‘Absorbance
units’. Hence, a sample for which IsI0 y10 is
2000 O. Axner et al. / Spectrochimica Acta Part B 58 (2003) 1997–2014
referred to as having an Absorbance of 1 absorb-
ance unit.8
Under conditions when Beer’s law is valid,
Absorbance has an exceptionally simple form.
Some common ways to write Beer’s law (in its
exponential form) are as
IsI0eykbsI0eyk9bC, (2)
where k is generally referred to as the linear
absorption coefficient or the absorptivity, b the
length of the interaction region, k9 the absorption
coefficient and C the concentration of the analyte
(all expressed in suitable units that make the
exponent dimensionless) w1x, or as
I(n)sI0eyns(n)L, (3)
where n is the number density of the analyte (in
units of 1ym3), s(n) the absorption cross-section
(in units of m2), which is strongly frequency
dependent around an absorbing feature, and L the
length of the interaction region (in units of m)
w29,33x. Note that use is made here (and in the
literature) of different symbols ‘b’ and ‘L’ to
denote absorption path length.
In the first case, which most often is used when
conventional atomic absorption is used for analyt-
ical purposes, Absorbance takes the form
I0
Aslog slog(e)k9bCs´bC, (4)
I be referred to as the ‘true’ Absorbance, Atrue, i.e.
where in the last step ´, which is commonly
referred to as the molar absorptivity (or molar
extinction coefficient) when the concentration C
is given in units of mol per liter, equals
log(e)k9f0.4343k9.
In the other case (which is, for example, most
commonly used when absorption spectrometry is
used with laser-based techniques for environmental
monitoring w23,27x), Absorbance takes the form
8
It is here important to note that the word Absorbance is
used for both a physical entity, log(I0 yI) according to Eq. (1),
as well as a part of the units of the entity, i.e. ‘absorbance
units’. The situation is slightly different for the entities that
are used for the WMAS techniques (see further discussion
below).
I0
A(n)slog slog(e)ns(n)L
I(n)
s0.434 ns(n)L. (5)
These expressions clearly show that for the cases
where Beer’s law is applicable, A is linear with
concentration of the analyte, C, or the number
density of absorbers, n, from zero to infinity. The
only practical limit for its validity is when the
detected light becomes so small that stray light
starts to affect the detection.
It is important to note that, with the definition
of Absorbance as given by Eq. (1), any measure-
ment that provides IsI0 y10 is referred to as having
an Absorbance of 1 absorbance unit, even though
the conditions for Beer’s law are not fulfilled. As
a consequence, from such a measurement, one
cannot identify the concentration of the analyte
from Eq. (4), C, to be (´b)y1. This would be
correct only if all requirements for the validity of
Eq. (4) are fulfilled.
In order to be able to handle the most general
types of conditions, we will in this work refer to
the entity log(I0 yI) as the ‘observed’ Absorbance,
Aobs, i.e.
BI E
Aobs'logC F,
0
(6)
D IG
whereas the two entities ´bC and log(e)nsL will
Atrue'´bC'log(e)nsL. (7)
It should be clear that for conventional AAS,
the relation
AtruesAobs (8)
holds as long as Beer’s law is valid.
3. Wavelength-modulated laser absorption spec-
trometry and the concept of sample optical
thickness
In WMAS, the laser wavelength is rapidly mod-
ulated (i.e. scanned) across the line profile at a
 O. Axner et al. / Spectrochimica Acta Part B 58 (2003) 1997–2014
given frequency, f m. The presence of an analyte,
with its non-linear absorption cross-section around
an absorbing feature, results in a detector signal
that has a number of harmonic components of that
particular frequency, i.e. at frequencies of nfm,
where n is an integer, i.e. 1, 2, 3, 4, etc. w6–
8,12,21–23,27–29x. A lock-in amplifier is used to
extract a given harmonic component of the detector
signal, in general the nth harmonic component for
nf-detection.
Since only the part of the detector signal that
originates from a non-linear absorption feature in
the system will create higher harmonic output
components (i.e. for n)1) from a wavelength
modulation, and the absorption feature of the
analyte is strongly non-linear, the laser-based mod-
ulated absorption techniques will only be affected
by the analytical part of the detector signal.9 This
implies that the outputs of the WMAS technique
will in fact be a certain fraction (given by the
instrumental settings, see below) of the difference
between the two entities, i.e. I0yIsDI. This is in
clear contrast to the conventional AAS techniques
that measure two large entities, I0 and I, and make
use of the (logarithm of the) ratio of these.
As indicated by Eqs. (4) and (5), whenever
Beer’s law is valid, Absorbance in conventional
AAS is linear with the concentration of absorber
over a large range of concentrations. As will be
shown below, however, because of its mode of
measurement, the output of the WMAS technique
will not be linear with the concentration or the
number density of absorbers for high number
densities or concentrations of analyte. On the other
hand, due to its ability to reduce noise, WMAS
has the capability to measure far lower concentra-
tions than the conventional AAS techniques.
Hence, WMAS has a larger linear dynamic range
than the conventional AAS techniques (signifi-
cantly extended into the low concentration range).
In addition, although WMAS is not linear with the
number density of absorbers for high sample opti-
cal thickness, it has a larger dynamic range than
conventional AAS even for high sample optical
9
This is valid under the condition that we can neglect the
influence of the associated intensity modulation and the fre-
quency dependence of the transmission, which both can give
rise to background signals in WMSA w28,29x.
2001
thickness. This originates from the fact that the
WMAS technique not only probes the absorption
profile on resonance but also on the wings of the
absorption profiles, which makes it less susceptible
to stray light under high optical thickness condi-
tions w34x. Although these properties expand the
range of applications of the modulated laser-based
absorption techniques, they also put some restric-
tions on the use of them for quantitative assess-
ments as well as demand for a useful alternative
to the concept of Absorbance.
3.1. Absorption and detection nomenclature in
WMAS — the sample optical thickness entity
3.1.1. Detector signals
In WMAS, as well as in many other types of
absorption technique, the detector signal can be
written in terms of the intensity of the light after
the interaction region, I(n), as
S(n)shTI(n), (9)
where T is the transmission of the optical system
in the absence of analyte and h is an instrumental
factor that relates the detector output, S(n), to the
intensity that impinges upon the detector, TI(n)
w28,29x.
3.1.2. Sample optical thickness
As was shown above, under the conditions
where Beer’s law is valid, the intensity of the light
after the interaction region can, for the WMAS
technique, be written as
I(n)sI0eyns(n)L, (10)
where we have explicitly indicated the entities that
are frequency dependent.10
10
We have here assumed that both the transmission of the
optical system in the absence of analytes, T, and the intensity
of the laser, I, are frequency independent. This is obviously an
oversimplification when it comes to the description of back-
ground signals, since the transmission is in general affected
by multiple reflections in optical components, so-called etalon
effects, and the associated intensity modulation. Nevertheless,
it is sufficient for the present purpose of describing the
analytical output signal of the WMAS instrumentation w29x.
2002 O. Axner et al. / Spectrochimica Acta Part B 58 (2003) 1997–2014
As was alluded to above, it has previously been
found practical in WMAS to use the (dimension-
less) sample optical thickness, denoted as a, for
the dimensionless product of the three entities in
the exponent in Eq. (20), i.e. the number density
of analyte, the absorption cross-section and the
length of the interaction region w28,29x. In this
case, Beer’s law can be written as
ya(n)
I(n)sI0e , (11)
where a(n), the sample optical thickness expe- 11
rienced by light with a frequency of n, is identified
with12
a(n)'ns(n)L. (12)
Furthermore, it is convenient to rewrite the
absorption cross-section, s(n), as a cross-section
on resonance, s0, times a peak-normalized line-
shape function, x̄(n), which thus carries the entire
frequency dependence of the absorption cross-
section w28,29x. This implies, among other things,
that also the sample optical thickness, a(n), can
be written in terms of a product of the sample
optical thickness on resonance, a0, and the same
peak-normalized lineshape function, x̄(n), i.e. as
¯
a(n)sa0x(n), (13)
where a0, the sample optical thickness on reso-
nance, is given by
a0'ns0L. (14)
It should be emphasized that the sample optical
thickness on resonance, a0, is the maximum sam-
ple optical thickness that can be achieved for a
given transition and that it is an entity that is
independent of the frequency of the light used.
11
Strictly speaking, ‘sample’ optical thickness should be
called ‘analyte’ optical thickness since n and s in Eq. (12)
refer to the amount only and not to the matrix, which, together
with the analyte, constitute the ‘sample’.
12
a also corresponds to the product k9bC given in Eqs. (2)
and (4).
3.1.3. Background detector signal, analytical
detector signal and fraction of radiation absorbed
As shown below, it is also convenient to rewrite
the detector signal, given by Eq. (9), in terms of
a background detector signal, here denoted as
S BG, and an analytical detector signal, referred to
as S AS, as
S(n)sSBGqSAS(n). (15)
Whenever the number density of absorbers is 0,
also the sample optical thickness, as well as S AS,
becomes 0. Eq. (9) then shows that the background
detector signal, i.e. the detector signal in absence
of analytes, S BG (in this work also referred to as
the unmodulated detector signal in the absence of
analytes, S0), can be written as
SBGsS0sS(as0)shTI0. (16)
Since the background detector signal does not
depend on the analyte, it is independent of the
frequency of the light, provided that the associated
intensity modulation is 0 and the transmission is
frequency independent.
The analytical detector signal, S AS, can be writ-
ten as the difference between the total detector
signal and the background detector signal by the
use of Eqs. (9) and (16), i.e. as
SAS(n)sS(n)ySBG
shTI(n)yhTI0 (17)
syw1yeya0x¯ (n)xhTI0.
Furthermore, as can be seen from Eq. (16),
S AS(n) can also be conveniently written in terms
of the unmodulated detector signal in the absence
of analytes, S0, as
SAS(n)syw1yeya0x¯ (n)xS0. (18)
Here, the expression within brackets describes
the entire dependence of the analytical detector
signal on sample optical thickness as well as on
the frequency of the light, and is commonly
referred to as the fraction of radiation absorbed
 O. Axner et al. / Spectrochimica Acta Part B 58 (2003) 1997–2014 2003

or the relative absorption in Refs. w1,2x since it sionally broadened system, the time-dependent
corresponds to the entity (I0yI)yI0. ¯ ¯ d,n¯ a,fm,t),
peak-normalized lineshape function, x(n
can then be written as
3.2. Wavelength modulation
1
3.2.1. The modulation procedure x¯ Žn¯ d,n¯ a,fm,t.s w , (22)
n¯
y d ¯
yn
x

a cosŽ 2pfmt.z~2q1 |

In WMAS, the frequency of the light is being

sinusoidally modulated according to
n(t)sncqna cosŽ2pfmt., (19)
where nc is the laser center frequency, na the laser
frequency-modulation amplitude and f m the mod-
ulation frequency w28,29x. Correspondingly, the
detector output, S(n), will be time dependent, i.e.
S(n(t)), according to
S(n(t))sSŽnc,na,fm,t.shTI0eya0x¯ (nc,na,fm,t), (20)
where n(t) is described by Eq. (19) and where nc
and na are instrumental (user-chosen and user-
controllable) parameters. In fact, the detector sig-
nal will not only be time dependent but will also
be periodic with a period of 1y f m through the
where n¯ d and n¯ a are referred to as the width-
normalized detuning and width-normalized fre-
quency-modulation amplitude, formally given by
(n0ync)yDn and na yDn, respectively w28,29x.
3.3. Output from WMAS
3.3.1. General expressions
As also has been shown previously in Refs.
w28,29x, the nth harmonic output from the lock-in
amplifier in a WMAS instrumentation, Sn(n¯ d,n¯ a),
where n is the harmonic component detected,13
can be written as
2 t
SnŽn¯ d,n¯ a.s
t | SŽn¯ ,n¯ ,f ,t. cosŽ2pnf t.dt.
0
d a m m (23)

time-dependent peak-normalized lineshape func-

tion x̄(nc,na,fm,t). It has been found convenient to rewrite the
detector signal, S(n¯ d,n¯ a,fm,t), in terms of the back-
3.2.2. An example — a Lorentzian lineshape ground and analytical detector signals given by
function Eq. (15). This implies that the nth harmonic
As an example, the time-dependent peak-nor- background and analytical outputs from the lock-
in amplifier from a WMAS instrumentation,
malized lineshape function, x̄(nc,na,fm,t), can, for
a collisionally broadened system, i.e. for a system SBG
n (n¯ d,n¯ a) and SAS
n (n
¯ d,n¯ a), can be written as
described by a Lorentzian shaped lineshape func-
2 t
tion, be written as SBG ¯ d,n¯ a.s
n Žn
t |S 0
BG
cosŽ2pnfmt.dt (24)

Dn2
x̄Žnc,na,fm,t.s w z2 2
, and
yn0yncyna cosŽ2pfmt.~ qDn
x |

2 t
(21) SAS ¯ d,n¯ a.s
n Žn
t |S 0
AS
Žn¯ d,n¯ a,fm,t. cosŽ2pnfmt.dt,

where Dn is the half-width-half-maximum

(HWHM) of the homogeneously (collisionally) (25)
broadened peak-normalized lineshape function for
the analyte. respectively.
It has been found convenient to express the 13
As also has been shown previously, Sn(n¯ d,n¯ a) is nothing
instrumental entities (i.e. nc and na) in terms of else than the nth Fourier coefficient of the detector signal
width-normalized entities. For the case of a colli- w28,29x.
2004 O. Axner et al. / Spectrochimica Acta Part B 58 (2003) 1997–2014

The first important observation to make is that a0<1). This implies that it is possible to rewrite
the integral over the background detector signals the exponential function in Eq. (26) using its
does not make any contribution to the output from series expansion, i.e. as
the WMAS instrumentation (again under the con-
dition that the transmission of the optical system eyx¯ (n¯ d,n¯ a,fm,t)a0f1yx¯ Žn¯ d,n¯ a,fm,t.a0q« (27)
and the intensity of the laser light are independent
of frequency) since S BG is a constant, and an This implies that the fraction of radiation absorbed,
integral over a trigonometric function is 0 for an defined in Eq. (18), takes a particularly simple
integer (or a sufficiently large) number of periods. form,
This implies in turn that the nth harmonic output
from the WMAS instrumentation, Sn(n¯ d,n¯ a), is 1yeyx¯ (n¯ d,n¯ a,fm,t)a0fx¯ Žn¯ d,n¯ a,fm,t.a0. (28)
given solely by the nth harmonic analytical output
from the lock-in amplifier, i.e. SASn (n
¯ d,n¯ a), which, Whenever this expression is valid, we will refer
by the use of Eqs. (18) and (25), can be written the sample as being ‘optically thin’. This implies,
as in turn, that it is possible to write the nth harmonic
analytical output from the WMAS instrumentation,
2 t as well as the nth harmonic output from the
SnASŽn¯ d,n¯ a.sy
t | w1ye
0
¯ n(n
ya0x ¯ d,n
¯ a,t)
x WMAS instrumentation, as

=cosŽ2pnfmt.dtØS0. (26) SnŽn¯ d,n¯ a.sSnASŽn¯ d,n¯ a.

2 t .
Eq. (26) represents an exact expression for the
fy
t 0 |
x¯ Žn¯ d,n¯ a,fm,t. cosŽ2pnfmt.dtØa0S0
nth harmonic output from a WMAS set-up, valid
for both optically thin and optically thick condi- (29)
tions, under the assumptions that the associated
intensity modulation can be neglected and that the Furthermore, it should be noticed that the inte-
transmission does not show any frequency gral in Eq. (29) is nothing else than the nth
dependence. Fourier coefficient of the wavelength-modulated
To the authors’ knowledge, this expression can- peak-normalized lineshape function w28,29x,
not be solved analytically under general conditions, x¯ n(n¯ d,n¯ a), defined as
and so it has to be solved numerically in the
general case. There are, however, two limiting 2 t

cases for which there exist analytical solutions to x¯ nŽn¯ d,n¯ a.s
t | x¯ Žn¯ ,n¯ ,f ,t. cosŽ2pnf t.dt.
0
d a m m

the integral:
● for small sample optical thickness and (30)
● for small frequency-modulation amplitudes.
The output from the WMAS instrumentation
These two situations will be covered in separate
under optically thin conditions can thereby be
sections below, since they are either of direct
written succinctly as
interest for the user of the WMAS techniques or
they can provide important information about the
SnŽn¯ d,n¯ a.fyx¯ nŽn¯ d,n¯ a.a0S0
general properties of the WMAS technique. (31)
syx¯ nŽn¯ d,n¯ a.a0hTI0,
3.3.2. Small sample optical thickness (optically
thin samples) but arbitrary frequency-modulation where we have in the last step made use of Eq.
amplitudes (16).
In most of the applications of WMAS, the Eq. (31) shows that, for small sample optical
optical thickness of the sample is small (i.e. thickness, the nth harmonic output from the
 O. Axner et al. / Spectrochimica Acta Part B 58 (2003) 1997–2014
WMAS instrumentation, Sn(n¯ d,n¯ a), is indeed pro-
portional to the amount of light being absorbed
(i.e. Aa0I0) with the proportionality constant con-
taining, among other things, the nth Fourier coef-
ficient of the wavelength-modulated
peak-normalized lineshape function, x¯ n(n¯ d,n¯ a).
Furthermore, for small sample optical thickness,
the unmodulated analytical detector signal on res-
onance, S AS (nds0), can, according to Eq. (18),
be written as
SASŽnds0.syhTI0w1yeya0xfya0S0.
This implies that it is also possible to write the
nth harmonic output from the WMAS instrumen-
tation, Sn(n¯ d,n¯ a), in terms of the unmodulated
analytical detector signal on resonance, SAS(nds
0),
SnŽn¯ d,n¯ a.sx¯ nŽn¯ d,n¯ a.SASŽnds0..
This expression clearly shows that, for small sam-
ple optical thickness, the WMAS technique meas-
ures only a certain fraction of the analytical
detector signal, given by the nth harmonic com-
ponent of the peak-normalized lineshape function,
also referred to as the nth Fourier coefficient of
the peak-normalized lineshape function,
¯xn(n¯ d,n¯ a).14
The most suitable way to normalize the WMAS
signal is to create the ratio Sn(n¯ d,n¯ a)yS0, with S0
being the unmodulated detector signal in the
absence of analyte. This ratio, which will be
referred to as the nth harmonic DC-signal-normal-
ized WMAS output, and will be denoted as
S¯ n(n¯ d,n¯ a), can then be expressed (Eq. (31)) as
SnŽn¯ d,n¯ a.
S¯ nŽn¯ d,n¯ a.' syx¯ nŽn¯ d,n¯ a.a0.
S0
This expression shows that the nth harmonic
14
In the case when the associated intensity modulation is
taken into account, the fraction of the analytical detector signal
that is measured is given by a sum of various combinations
of Fourier coefficients of the peak-normalized lineshape func-
tion and the associated intensity modulation of the laser light,
as, for example, the expression within the curly brackets in
Eq. 76 in Ref. w29x.
2005
DC-signal-normalized WMAS output, S¯ n(n¯ d,n¯ a),
is, for small sample optical thickness, given by a
product of two physically fundamental entities: the
nth Fourier coefficient of the wavelength-modulat-
ed lineshape function x¯ n(n¯ d,n¯ a) and the optical
thickness of the sample on resonance, a0.
x¯ n(n¯ d,n¯ a) has previously been characterized for
both Lorentzian and Gaussian lineshapes w35,36x.
It has been shown, for example, that the nth
Fourier coefficient of a wavelength-modulated
lineshape function from a homogeneously broad-
(32) ened absorption profile (i.e. for an analyte with a
Lorentzian absorption profile) will take values up
to 0.34 for 2f-detection, up to 0.18 for 4f-detection
and up to 0.12 for 6f-detection, depending on the
values of n¯ c and n¯ a w29,35x.
From Eq. (34), it can be seen that the absolute
value of the ratio of the nth harmonic DC-signal-
(33) normalized WMAS output, S¯ n(n¯ d,n¯ a), and the nth
Fourier coefficient of the wavelength-modulated
lineshape function, x¯ n(n¯ d,n¯ a), is always given by
the sample optical thickness on resonance, a0. As
will be shown further below, x¯ n(n¯ d,n¯ a) also takes
the role as the (dimensionless) sensitivity of the
WMAS technique. A formally correct notation for
the ratio of S¯ n(n¯ d,n¯ a) and x¯ n(n¯ d,n¯ a) would thereby
be the sensitivity-normalized and DC-signal-nor-
malized nth harmonic WMAS output and be denot-
ed as S¯ (n¯ ,n¯ ). This implies that S¯ (n¯ ,n¯ ) can be
n d a n d a
written as
S¯ nŽn¯ d,n¯ a.
) x¯ Žn¯ ,n¯ . ).
S¯ nŽn¯ d,n¯ a.'
n d a
(35)
Whenever the medium is optically thin, so that
(34) Eq. (34) is valid, this implies, in fact, that the
sensitivity-normalized and DC-signal-normalized
nth harmonic WMAS output becomes identical to
the sample optical thickness on resonance of the
sample, i.e.
S¯ nŽn¯ d,n¯ a.sa0. (36)
This expression thereby shows one of the unique
2006 O. Axner et al. / Spectrochimica Acta Part B 58 (2003) 1997–2014
roles of the sample optical thickness entity for the
WMAS technique.15
As will be shown below, S¯ n(n¯ d,n¯ a) will only be
identical to a0 under optically thin conditions.
Under (partly) optical thick conditions S¯ n(n¯ d,n¯ a)
will depend on a0 in a more complicated manner.
We will therefore refer to S¯ n(n¯ d,n¯ a) as the
‘observed’ sample optical thickness, whereas a0
will be referred to as the ‘true’ sample optical
thickness, in an analogous way to what was done
for conventional AAS above, i.e. Eqs. (6) and (7).
Before investigating how the ‘observed’ sample
optical thickness relates to the ‘true’ sample optical
thickness for other situations, it is worth com-
menting upon the use of units. In an analogous
way to what was done for Absorbance, for which
a sample with an Absorbance value of ‘x’ was
referred to as a sample with ‘x’ absorbance units,
the ‘observed’ sample optical thickness will be
referred to as having ‘x’ SOT units whenever
S¯ n(n¯ d,n¯ a) takes a numerical value of ‘x’. If the
sample optical thickness is sufficiently small so
that Eq. (36) is valid, the ‘observed’ sample optical
thickness will be identical with a ‘true’ sample
optical thickness of ‘x’ SOT units. It is important
to note, however, that every sample for which
S¯ n(n¯ d,n¯ a) takes a numerical value of ‘x’ will be
referred to as having an ‘observed’ sample optical
thickness of ‘x’ SOT units, irrespective of whether
the conditions for the validity of Eq. (36) are
fulfilled or not (i.e. irrespective of whether the
sample optical thickness is small or not). This
implies that if these conditions are not fulfilled,
one cannot automatically make use of Eq. (36)
and from Eq. (14) conclude that the ‘true’ sample
optical thickness is ‘x’ SOT units and that the
number density of absorbers is given by xy(s0L).
This is in full analogy with the discussion carried
out for Absorbance in the last paragraph of Section
2: a given measurement yielding an Absorbance
15
Since the SOT on resonance is directly proportional to
the concentration or the number density of absorbers, as
illustrated in Eq. (14), Eq. (36) shows that also the sensitivity-
normalized and DC-signal-normalized nth harmonic WMAS
output signal is proportional to the concentration or the number
density of absorbers for the cases when the SOT<1.
of ‘x’ absorbance units can be used to attribute to
the analyte concentration the value of xy(´b) only
if the conditions for Beer’s law are valid.
3.3.3. Small frequency-modulation amplitudes but
arbitrary sample optical thickness
For small modulation amplitudes, i.e. for n̄a<
1, it is possible to derive an analytical expression
for the general expression for the nth harmonic
output from a WMAS set-up given above, Eq.
(26), that is valid for an arbitrary sample optical
thickness, by Taylor expanding the expression for
the lineshape function around the laser center
frequency w29x. As is shown by Eq. (4) in Ref.
w34x, the analytical expression for the general
expression for the nth harmonic output from a
WMAS set-up can be written as
SnŽn¯ d,n¯ a.fZnŽn¯ d,n¯ a.a0eya0S0, (37)
where Zn(n¯ d,n¯ a) is a function of the width-nor-
malized frequency-modulation amplitude, n̄a, and
the order of the harmonic detected, n, which,
according to Eq. (30) from Ref. w29x, can be
written as
Žn̄a.n
ZnŽn¯ d,n¯ a.s x¯ (n))n¯ sn¯ d, (38)
2ny1
where x̄(n)±n¯ sn¯ d is the nth derivative of the peak-
normalized lineshape function evaluated for the
normalized laser center frequency that corresponds
¯ ¯ d.16
to nsn
As can be seen from Eq. (37), the nth harmonic
output of the WMAS instrumentation, Sn(n¯ d,n¯ a),
is related to the ‘true’ sample optical thickness
(i.e. a0) by a product of two functions, one linear
and one exponentially decaying. This implies that
the nth harmonic output for small frequency-
modulation amplitudes will increase linearly with
the ‘true’ sample optical thickness for optical thin
samples (for which exp(ya0)f1), in agreement
16
This shows clearly that the nth harmonic output of the
WMAS signal is proportional to the nth derivative of the
lineshape function for small modulation amplitudes. This is
also the reason the WMAS technique is sometimes referred to
as derivative spectroscopy.
 O. Axner et al. / Spectrochimica Acta Part B 58 (2003) 1997–2014
with Eq. (31), i.e. as
SnŽn¯ d,n¯ a.fZnŽn¯ d,n¯ a.a0S0, (39)
before it will reach a maximum, after which it will
start to decrease exponentially. This implies that
Eq. (37) predicts that the ‘observed’ sample optical
thickness is a multi-valued function of the ‘true’
sample optical thickness. It is this multi-valueabil-
ity that is the reason the WMAS technique also
can be used for assessing high sample optical
thickness, as was demonstrated in Ref. w34x.17
One can easily find the ‘true’ sample optical
thickness for which the ‘observed’ sample optical
thickness will take its maximum value by equal-
izing the derivative of Eq. (37) with respect to
a0 to 0. One then finds that the maximum signal
for a single peak transition is obtained for a ‘true’
sample optical thickness of unity (i.e. a0s1).
Inserting this into the expressions for the nth
harmonic output of the WMAS instrumentation,
Eq. (38), shows that the maximum value of the
nth harmonic output from a WMAS instrumenta-
tion with a small frequency-modulation amplitude
will be
SnŽn¯ ds0, n¯ a<1.maxsSnŽn¯ ds0, n¯ a<1, a0s1.
sZnŽn¯ ds0, n¯ a.ey1S0
(40)
This expression thus shows that when small
frequency-modulation amplitudes are used, the nth
harmonic output of the WMAS instrumentation
takes a maximum value of Zn(n¯ a)ey1S0 for a ‘true’
sample optical thickness of unity. The fact that the
maximum value of Sn(n¯ d,n¯ a<1) is obtained for a
‘true’ sample optical thickness of unity (i.e. a0s
17
Although having different physical origins, this behavior
is similar to that of fluorescence self-absorption in the case of
a narrow lines source excitation. In WMAS, the decrease in
the output signal strength with increasing SOT originates from
the fact that the modulated part of the detector signal decreases
for high sample optical thickness, while in fluorescence it
originates from the inverse square root dependence of the
signal upon the number density of absorbers, which sets in at
high concentrations.
2007
1) supports the conclusion that the sample optical
thickness is a basic and useful measure for WMAS.
Fig. 1 shows the behavior of the 2nd, 4th, 6th
and 8th harmonic output strengths as a function of
sample optical thickness for modulation amplitude
of 0.20 for a one-peak transition according to Eqs.
(37) and (38). As can be seen from the figure,
the functional form of Eq. (37) with respect to
a0, i.e. a0 exp(ya0), is the same for all single
peak transitions, in all elements, irrespective of
the prevalent broadening mechanism, probed with
a small frequency-modulation amplitude. Also, this
shows the special role of sample optical thickness
for the WMAS technique.
As also can be seen from Fig. 1, the 4f, 6f and
8f harmonic output strengths have a comparable
magnitude for this particular frequency-modulation
amplitude. Due to their strong dependence on the
modulation amplitude, however, each higher-order
harmonic component will be significantly smaller
than its precursor for frequency-modulation ampli-
tudes smaller than that used in Fig. 1.
3.3.4. Arbitrary frequency-modulation amplitude
and arbitrary sample optical thickness
It is now very appealing to write also the general
expression for the nth harmonic output from the
WMAS instrumentation, Eq. (26), valid for any
‘true’ sample optical thickness and for any fre-
quency-modulation amplitude, in terms of the
unmodulated analytical detector signal, as was
done for small sample optical thickness in Eq.
(33). Although not as straightforward, it is possible
to formally write Eq. (26) by the use of the
general expression for the unmodulated analytical
detector signal on resonance, given by Eq. (18),
as
2 t 1yeya0x¯ n(n¯ d,n¯ a,t)
SnŽn¯ d,n¯ a.s
t | 0 1yeya0
=cosŽ2pnfmt.dtØSAS. (41)
It can easily be concluded that the integral
becomes independent of a0 for small sample opti-
cal thickness (whereby Eq. (41) simplifies to Eq.
(33)). For sample optical thickness that does not
adhere to the requirement of a0<1, or for width-
2008 O. Axner et al. / Spectrochimica Acta Part B 58 (2003) 1997–2014

Fig. 1. The 2nd, 4th, 6th and 8th harmonic output strengths on resonance of a Lorentzian-broadened one-peak transition (represented
by the curves a, b, c and d, respectively) as a function of sample optical thickness for a modulation amplitude of 0.20 according
to Eqs. (37) and (38). Note that although every second even-order harmonic signal (i.e. S2, S6, S10, etc.) in reality is negative,
whereas the other even-order harmonics are positive (i.e. S4 , S8, etc.), we have plotted all of them being positive for display
purposes.

normalized frequency-modulation amplitudes that
are not necessarily small, the expression above
shows how large is the fraction of the unmodulated
analytical signal on resonance that appears in a
given nth harmonic output from the WMAS instru-
mentation for a given sample optical thickness. It
can be shown, by numerical integration, that the
integral decreases as a function of sample optical
thickness for large sample optical thickness, in a
manner similar to, but not identical to, that of Eq.
(37) w34x.
3.4. Absorbance vs. sample optical thickness
In the previous section, it was shown that it is
natural and convenient to express the output from
the properly normalized WMAS signal, S¯ n(n¯ d,n¯ a),
in SOT units. It is in most cases trivial to express
it also in absorbance units.
As was concluded by Eq. (36), the ‘observed’
sample optical thickness, S¯ n(n¯ d,n¯ a), can be related
to the ‘true’ sample optical thickness entity as
S¯ nŽn¯ d,n¯ a.fa0 (42)
under the conditions that a0<1.
Furthermore, according to the definitions of
‘true’ Absorbance and ‘true’ sample optical thick-
ness above, i.e. Eqs. (7) and (12), the relation
between the ‘true’ Absorbance and the ‘true’
sample optical thickness entity is
 O. Axner et al. / Spectrochimica Acta Part B 58 (2003) 1997–2014
Atrueslog(e)a0f0.434a0,
which is valid without constrains.
As also was concluded above (Eq. (8)), under
the condition that Beer’s law is valid, the
‘observed’ Absorbance is equal to the ‘true’
Absorbance, i.e.
AobssAtrue.
This implies that the ‘observed’ sample optical
thickness, S¯ n(n¯ d,na), is related to the ‘true’ Absorb-
ance according to
S¯ nŽn¯ d,n¯ a.s2.30Atrue
only when a0<1, and to the ‘observed’ Absorb-
ance according to
S¯ nŽn¯ d,n¯ a.s2.30Aobs
only when both a0<1 and under the conditions
that Beer’s law are valid.
Hence, a sample with an ‘observed’ sample
optical thickness of 0.01 SOT unit has an Absorb-
ance of 0.0043 absorbance units w29x. Conversely,
a sample with an Absorbance of 0.01 absorbance
unit has an ‘observed’ sample optical thickness of
0.023 SOT unit. However, it is important to
remember that the expressions above are only valid
under the conditions given above, of which the
most important is that a0<1 (e.g. (0.05). This
implies, for example, that a sample with an
‘observed’ sample optical thickness of 1 SOT unit
cannot be assigned an Absorbance of 0.434
absorbance units.
When the condition a0<1 is not fulfilled, the
connection between the ‘true’ and ‘observed’ sam-
ple optical thickness can no longer be made by
Eq. (42), but needs instead to be made through
any of the other expressions given above, i.e. Eq.
(37) for the case with small frequency-modulation
amplitudes and Eq. (26) or Eq. (41) in the more
general case.
In conclusion, although it is possible to relate
the absorbance unit formally to the SOT unit by
2009
(43) Eq. (43), a WMAS measurement always provides
a measurement of the ‘observed’ sample optical
thickness, which, however, not always is identical
to the ‘true’ sample optical thickness. The two are
the same only when a0<1. This is fortunately
also the most common range of application of the
technique. As a consequence, in order to make
comparisons with other techniques more easy, and
(44) although it has become commonplace to quote the
LODs for the WMAS in SOT units (or in terms
of fractional absorbance), it should be fairly simple
to quote the LODs also in absorbance units by the
use of Eq. (45) or Eq. (46). In the examples given
in the introduction, in which it was concluded that
(45) the WMAS technique has shown to be able to
measure fractional absorption in the 10y5 –10y6
region (and occasionally down to 10y7) expressed
in SOT units, this implies that the WMAS tech-
nique is capable of measuring absorbance down
well into the 10y6 absorbance unit region and
(46)
sometimes even into the 10y7 absorbance unit
region.
3.5. The sensitivity of WMAS
The expressions above also suggest a convenient
way to define the dimensionless sensitivity of the
WMAS technique. The dimensionless sensitivity
of the WMAS technique can most conveniently be
defined as the ratio of the nth harmonic DC-signal-
normalized WMAS output, S¯ n(n¯ d,n¯ a), and the sam-
ple optical thickness on resonance, a0.18 Eq. (34)
shows that for this simple case considered in this
work, i.e. when the effect of associated intensity
modulation can be neglected, the dimensionless
sensitivity of the WMAS technique is, for the
cases when the sample optical thickness <1,
identical to the nth Fourier coefficient of the
wavelength-modulated peak-normalized lineshape
18
This sensitivity is referred to as the dimensionless sensi-
tivity since it differs from the (conventional) sensitivity used
when calculating, for example, limits of detection (which
necessarily has to be given in the ratio of units of the output
signal and the concentration).
2010 O. Axner et al. / Spectrochimica Acta Part B 58 (2003) 1997–2014

Fig. 2.
 O. Axner et al. / Spectrochimica Acta Part B 58 (2003) 1997–2014 2011

Fig. 2. The Absorbance entity and the nth harmonic DC-signal-normalized WMAS output, S¯ n(n¯ d,n¯ a), are presented in the panels A,
B and C, for 2nd, 4th and 6th harmonic detection, respectively, as a function of the sample optical thickness on resonance, a0. In
all panels, curve a represents conventional AAS, whereas the curves b–g represent various detunings as well as various modulation
amplitudes. The curves b–d represent the situation on resonance, whereas the three curves e–g correspond to the situation with the
laser detuned half of a HWHM of the width of the absorption profile (i.e. n̄ds0.5 ). The three curves in each group represent width-
normalized modulation amplitudes that are half of the optimum, the optimum and twice the optimum for each harmonic, respectively.
The optimum width-normalized modulation amplitudes for the three harmonics considered (2nd, 4th and 6th, respectively) are 2.2,
4.1 and 6.1, respectively w29x. Hence, the curves b–g in panel A represent S¯ 2(n¯ d,n¯ a) with the following combinations of width-
normalized detuning and width-normalized modulation amplitude (n¯ d, n¯ a ): b—(0, 1.1); c—(0, 2.2); d—(0, 4.4); e—(0.5, 1.1); f—
(0.5, 2.2) and g—(0.5, 4.4).

function, x¯ n(n¯ d,n¯ a).19 This shows clearly that the
dimensionless sensitivity depends both on the
width-normalized laser center frequency, n̄d, and
the width-normalized frequency-modulation ampli-
tude, n̄a.
A corresponding definition of a dimensionless
19
In the case when the associated intensity modulation is
taken into account, the sensitivity of the technique is given by
a sum of various combinations of Fourier coefficients of the
peak-normalized lineshape function and the associated inten-
sity modulation of the laser light, as, for example, the expres-
sion within the curly brackets in Eq. 76 in Ref. w29x.
sensitivity of the conventional AAS technique
would be as the ratio of Absorbance and the
sample optical thickness. According to the discus-
sion above, this relation is for all conditions given
by Eq. (46), i.e. it should be a constant equal to
log(e)s0.434.
A comparison between the dimensionless sen-
sitivities of conventional AAS and the WMAS
technique is given in Fig. 2 for a variety of
harmonics, detuning and frequency-modulation
amplitudes, under the condition that the sample
optical thickness <1. The three panels show
2012 O. Axner et al. / Spectrochimica Acta Part B 58 (2003) 1997–2014
Absorbance and the nth harmonic DC-signal-nor-
malized WMAS output, S¯ n(n¯ d,n¯ a), for 2nd, 4th
and 6th harmonic detection, respectively, as a
function of the sample optical thickness on reso-
nance, a0. In all panels, curve a represents con-
ventional AAS, whereas the curves b–g represent
a variety of situations for the WMAS technique
(various detunings as well as various modulation
amplitudes) for the three particular harmonics
considered. The curves b–d represent the situation
on resonance, whereas the three curves e–g cor-
respond to the situation with the laser detuned half
of a HWHM of the width of the absorption profile
(i.e. n̄ds0.5). The three curves in each group
represent width-normalized frequency-modulation
amplitudes that are half of the optimum, the
optimum and twice the optimum for each harmon-
ic, respectively. The optimum width-normalized
modulation amplitudes for the three harmonics
considered (2nd, 4th and 6th, respectively) are
2.2, 4.1 and 6.1, respectively w29x.
The figure shows clearly that the dimensionless
sensitivity of the WMAS technique (i.e. the slope
of the curves) depends on the order of the har-
monic detected, the detuning, as well as the fre-
quency-modulation amplitudes used, whereas the
sensitivity for conventional AAS is constant (since
it is always measured at a constant center frequen-
cy and does not make use of any modulation).
Fig. 2 gives further support to the notion that
the nth Fourier coefficient of the wavelength-
modulated lineshape function, x̄n(nc,na), which
can be obtained either by calculation or by cali-
bration, as well as the sample optical thickness (or
more correct the sample optical thickness on res-
onance, a0) are basic entities of importance in
WMAS. An alternative way of saying this is that
the sample optical thickness on resonance of the
sample is simply given by the ratio of the nth
harmonic normalized WMAS output, S¯ n(n¯ c,n¯ a),
and the sensitivity, i.e. the nth Fourier coefficient
of the wavelength-modulated lineshape function,
x̄n(nc,na), for all laser center frequencies and for
all frequency-modulated amplitudes.
4. Summary and conclusions
This work has stressed the importance of the
well-known dimensionless product of the three
entities in the exponent in Beer’s law, i.e. the
number density of analyte, the absorption cross-
section and the lengths of the interaction region in
WMAS. This critical absorption measure has been
termed the (dimensionless) sample optical thick-
ness, and denoted by a. By introducing a peak-
normalized lineshape function, x(n ¯ ¯ d,n¯ a), the
sample optical thickness could, in turn, be decom-
posed into the sample optical thickness on reso-
nance, a0, and the peak-normalized lineshape
function, x(n ¯ ¯ d,n¯ a).
This work has also identified the DC-signal-
normalized nth harmonic WMAS output and the
sensitivity-normalized and DC-signal-normalized
nth harmonic WMAS output, S¯ n(n¯ c,n¯ a) and
S¯ n(n¯ c,n¯ a), defined by Eqs. (34) and (35), respec-
tively, as two useful output entities in WMAS.
The latter is of particular importance since it
represents the ‘observed’ sample optical thickness.
The treatment above has shown, by Eq. (34),
that, under certain assumptions, the nth harmonic
DC-signal-normalized WMAS output, S¯ n(n¯ c,n¯ a), in
the case of a small sample optical thickness on
resonance, can be expressed in terms of the sample
optical thickness on resonance, a0, with the nth
Fourier coefficient of the wavelength-modulated
peak-normalized lineshape function, x̄n(nc,na), as
the proportionality factor. It was also shown (by
Eq. (36)) that the ‘observed’ sample optical thick-
ness, S¯ n(n¯ c,n¯ a), takes the value of the ‘true’ sample
optical thickness, i.e. a0, for optically thin samples
(for which a0<1). It was thereby concluded that
it is possible to relate the Absorbance and the
‘observed’ sample optical thickness in WMAS to
each other (through the ‘true’ sample optical thick-
ness) either as
Af0.434S¯ nŽn¯ c,n¯ a. (47)
or as
S¯ nŽn¯ c,n¯ a.f2.30A (48)
only under the conditions that both Beer’s law and
that a0<1 are valid.
When the condition a0<1 is not fulfilled,
 O. Axner et al. / Spectrochimica Acta Part B 58 (2003) 1997–2014
however, the connection between the ‘observed’
and the ‘true’ sample optical thickness (and there-
by between the ‘observed’ sample optical thickness
and Absorbance) needs to be made through any
of the other expressions given above, i.e. Eq. (37)
for the case with small frequency-modulation
amplitudes and Eq. (26) or Eq. (41) in the more
general case. This implies, for example, that a
sample for which the ‘observed’ sample optical
thickness is 0.01 SOT units has an Absorbance of
0.0043 absorbance units, respectively, whereas a
sample for which the ‘observed’ sample optical
thickness WMAS output is 1 SOT unit does not
have an Absorbance of 0.43 absorbance units.
Despite the simple relation between Absorbance
and the nth harmonic DC-signal-normalized
WMAS output that is valid for small sample
optical thickness and the frequent use of the
Absorbance entity in the literature, we have found
it most convenient to consistently express the
WMAS signal in terms of sample optical thickness
on resonance. The two main reasons for this are
the following:
(i) First, the sample optical thickness comes out
as a fundamental and the most natural entity that
describes the concentration or the number density
of analytes in the description of the WMAS
technique.
(ii) Secondly, by keeping the sample optical
thickness as a fundamental entity as well as unit,
the dimensionless sensitivity of the WMAS tech-
nique becomes very simple: for example, in the
case when the associated intensity modulation can
be neglected, the sensitivity is simply given by the
nth Fourier coefficient of the wavelength-modulat-
ed peak-normalized lineshape function. Also, the
functional form of the WMAS outputs as a func-
tion of the number density of absorbers (or con-
centration) becomes very simple for the cases
when small frequency-modulation amplitudes are
being used.
The three following conclusions were therefore
emphasized:
● the ‘observed’ Absorbance can be related to the
‘true’ Absorbance whenever Beer’s law is valid;
● the ‘true’ Absorbance can always be related to
2013
the ‘true’ sample optical thickness by either of
Eq. (47) or Eq. (48) and
● the ‘true’ sample optical thickness is directly
given by the ‘observed’ sample optical thickness
whenever the sample is optically thin (i.e.
a0<1).
Since the latter condition is met for the most
common range of application of the WMAS tech-
nique, it is possible to relate the ‘observed’ sample
optical thickness to the ‘observed’ as well as ‘true’
Absorbance by Eq. (47) or Eq. (48) for most
types of measurements using WMAS, and conse-
quently, in the end, to quote the LODs for the
WMAS techniques also in absorbance units.
Finally, this work has also introduced a dimen-
sionless sensitivity of the WMAS technique. When
the sample optical thickness <1, the dimension-
less sensitivity of the WMAS technique is equal
to the nth Fourier coefficient of the wavelength-
modulated peak-normalized lineshape function,
x̄n(nc,na). This is very convenient since expres-
sions for the nth Fourier coefficient of the wave-
length-modulated peak-normalized lineshape for
both Lorentzian and Doppler broadened functions
have been explicitly given in Refs. w35,36x.
References
w1x J.D. Ingle, S.R. Crouch, in: K.M. Lafferty (Ed.), Spec-
trochemical Analysis, Prentice-Hall, Englewood Cliffs,
New Jersey, 1988, p. 1988.
w2x C.T.J. Alkemade, T. Hollander, W. Snelleman, P.J.T.
Zeegers, Metal Vapours in Flames, Pergamon Press,
Oxford, 1982.
w3x T.C. O’Haver, Derivative and wavelength modulation
spectrometry, Anal. Chem. 51 (1979) 91A–100A.
w4x G.C. Bjorklund, Frequency-modulation spectroscopy: a
new method for measuring weak absorption and disper-
sion lineshapes, Opt. Lett. 5 (1980) 15–17.
w5x G.C. Bjorklund, M.D. Levenson, W. Lenth, C. Ortiz,
Frequency modulation (FM) spectroscopy—theory of
lineshapes and signal-to-noise analysis, Appl. Phys. B
32 (1983) 145–152.
w6x D.S. Bomse, A.C. Stanton, J.A. Silver, Frequency mod-
ulation and wavelength modulation spectroscopies: com-
parison of experimental methods using a lead-salt diode
laser, Appl. Opt. 31 (1992) 718–731.
w7x J.A. Silver, Frequency-modulation spectroscopy for
trace species detection: theory and comparison among
experimental methods, Appl. Opt. 31 (1992) 707–717.
2014 O. Axner et al. / Spectrochimica Acta Part B 58 (2003) 1997–2014
w8x J.M. Supplee, E.A. Whittaker, W. Lenth, Theoretical
description of frequency-modulation and wavelength
modulation spectroscopy, Appl. Opt. 33 (1994)
6294–6302.
w9x F.S. Pavone, M. Inguscio, Frequency-modulation and
wavelength-modulation spectroscopies—comparison of
experimental methods using an AlGaAs diode-laser,
Appl. Phys. B 56 (1993) 118–122.
w10x J. Reid, D. Labrie, Second-harmonic detection with
tunable diode lasers—comparison of experiment and
theory, Appl. Phys. B 26 (1981) 203–210.
w11x D.T. Cassidy, J. Reid, High-sensitivity detection of trace
gases using sweep integration and tunable diode lasers,
Appl. Opt. 21 (1982) 2527–2530.
w12x D.T. Cassidy, J. Reid, Atmospheric pressure monitoring
of trace gases using tunable diode lasers, Appl. Opt. 21
(1982) 1185–1190.
w13x D.T. Cassidy, J. Reid, Harmonic detection with tunable
diode lasers—two-tone modulation, Appl. Phys. B 29
(1982) 279–285.
w14x G.R. Janik, C.B. Carlisle, T.F. Gallagher, Two-tone
frequency-modulation spectroscopy, J. Opt. Soc. Am. B
3 (1986) 1070–1074.
w15x D.E. Cooper, R.E. Warren, Frequency modulation spec-
troscopy with lead-salt diode lasers: a comparison of
single-tone and two-tone techniques, Appl. Opt. 26
(1987) 3726–3732.
w16x H. Groll, K. Niemax, Multielement diode laser atomic
absorption spectrometry in graphite tube furnace and
analytical flames, Spectrochim. Acta Part B 48 (1993)
633–641.
w17x A. Zybin, C. Schnurer-Patschan,
¨ K. Niemax, Measure-
ments of C2F4Cl2, CCl4, CHF3 and O2 by wavelength
modulated laser absorption spectroscopy of excited Cl,
F and O in a DC discharge applying semiconductor
diode lasers, Spectrochim. Acta Part B 48 (1993)
1713–1718.
w18x C. Schnurer-Patschan,
¨ A. Zybin, H. Groll, K. Niemax,
Improvement in detection limit in graphite furnace diode
laser atomic absorption spectrometry by wavelength
modulation technique, J. Anal. At. Spectrom. 8 (1993)
1103–1107.
w19x H. Groll, C. Schnurer-Patschan,
¨ Y. Kuritsyn, K. Niemax,
Wavelength modulation diode laser atomic absorption
spectrometry in analytical flames, Spectrochim. Acta
Part B 49 (1994) 1463–1472.
w20x J. Koch, A. Zybin, K. Niemax, Narrow and broad band
diode laser absorption spectrometry—concepts, limita-
tions and applications, Spectrochim. Acta Part B 53
(2002) 1547–1561.
w21x A.W. Mantz, A review of spectroscopic applications of
tunable semiconductor lasers, Spectrochim. Acta Part A
51 (1995) 2211–2236.
w22x P. Werle, Spectroscopic trace gas analysis using semi-
conductor diode lasers, Spectrochim. Acta Part A 52
(1996) 805–822.
w23x P. Werle, Tunable diode laser absorption spectroscopy:
recent findings and novel approaches, Infrared Phys.
Technol. 37 (1996) 59–66.
w24x V. Liger, A. Zybin, Y. Kuritsyn, K. Niemax, Diode-laser
atomic-absorption spectrometry by the double-beam-
double-modulation technique, Spectrochim. Acta Part B
52 (1997) 1125–1138.
w25x P. Ljung, O. Axner, Measurements of rubidium in
standard reference samples by wavelength-modulation
diode laser spectrometry in graphite furnace, Spectro-
chim. Acta Part B 52 (1997) 305–319.
w26x A. Zybin, C. Schnurer-Patschan,
¨ M.A. Bolshov, K.
Niemax, Elemental analysis by diode laser spectroscopy,
Trends Anal. Chem. 17 (1998) 513–520.
w27x P. Werle, A review of recent advances in semiconductor
laser based gas monitors, Spectrochim. Acta Part A—
Mol. Biomol. Spectrosc. 54 (1998) 197–236.
w28x P. Kluczynski, O. Axner, Theoretical description based
upon Fourier analysis of wavelength-modulation spec-
trometry in terms of analytical and background signals,
Appl. Opt. 38 (1999) 5803–5815.
˚
w29x P. Kluczynski, J. Gustafsson, A.M. Lindberg, O. Axner,
Wavelength modulated absorption spectrometry—an
extensive scrutiny of the generation of signals, Spectro-
chim. Acta Part B 56 (2001) 1277–1354.
˚
w30x P. Kluczynski, A.M. Lindberg, O. Axner, Characteriza-
tion of background signals in wavelength modulation
spectrometry in terms of a Fourier-based theoretical
formalism, Appl. Opt. 40 (2001) 770–782.
w31x B.W. Shore, The Theory of Coherent Atomic Excitation,
Wiley, NY, 1990.
w32x C. Vadla, M. Movre, R. Beuc, J. Franzke, H.D. Wize-
mann, K. Niemax, Optimization of lead metastable
production in a low pressure argon discharge, Spectro-
chim. Acta Part B 55 (2000) 1759–1769.
w33x P.W. Milonni, J.H. Eberly, Lasers, Wiley, New York,
NY, 1988.
w34x J. Gustafsson, N. Chekalin, D. Rojas, O. Axner, Exten-
sion of the dynamic range of the 2f-wavelength modu-
lated diode laser absorption spectrometry technique—
detection of atoms under optically thick conditions,
Spectrochim. Acta Part B 55 (2000) 237–262.
w35x O. Axner, P. Kluczynski, A.M.˚ Lindberg, A general
non-complex analytical expression for the nth Fourier
component of a wavelength-modulated Lorentzian line-
shape function, J. Quant. Spectrosc. Radiat. Trans. 68
(2001) 299–317.
˚
w36x P. Kluczynski, A.M. Lindberg, O. Axner, Wavelength
modulation diode laser absorption signals from doppler
broadened absorption profiles, J. Quant. Spectrosc.
Radiat. Trans., in press.