Spectrochimica Acta Part A 60 (2004) 1817–1823

Cw-laser thermal lens spectrometry in binary mixtures of water

and organic solvents: composition dependence of the
steady-state and time-resolved signals
Nadine Arnaud, Joseph Georges∗
Laboratoire des Sciences Analytiques, Bât. CPE-308 D, Université Claude Bernard—Lyon 1, 69622 Villeurbanne Cedex, France

Received 2 April 2003; received in revised form 1 September 2003; accepted 1 September 2003

Abstract

The thermal lens effect obtained in binary liquid systems composed of water and ethanol, propanol and acetonitrile has been investigated.
The dependence of dn/dT upon the solvent volume fraction follows polynomials up to sixth order and cannot be precisely predicted using the
additive rule. The sensitivity of the thermal lens method upon the addition of organic solvent in water varies as the temperature-dependent
refractive index gradient to thermal conductivity ratio of the mixture provided that the signal is sampled correctly. Otherwise, especially when
steady-state experiments are carried out, the thermally induced concentration gradient, known as the Soret effect, can change the thermo-optical
properties of the solution locally in the irradiated area and produce an additional signal. This effect depends on the solvent and is maximum
at low solvent composition. At the critical solvent volume fraction of 0.1–0.15, the Soret component may represent up to 25% of the pure
thermal lens signal and has a time constant which is 200–400 times greater than the characteristic time constant of the thermal lens.
© 2003 Elsevier B.V. All rights reserved.

Keywords: Thermal lens spectrometry; Binary liquid systems; Refractive index; Temperature gradient; Concentration gradient

1. Introduction erties due to higher dn/dT and lower k values. Therefore, it

Thermal lens spectrometry (TLS) results from the tem-
perature rise which is subsequent to absorption of optical ra-
diation and nonradiative relaxation of the excited molecules.
Due to the Gaussian profile of the excitation laser, the
temperature gradient produces a refractive index gradient
which maximizes at the beam center and behaves like a
converging or diverging lens depending on whether dn/dT is
positive or negative, respectively [1–4]. The signal is mea-
sured as a change in the beam center intensity of a probe
beam passing through the thermal lens. Besides the laser
power and the solute concentration, the signal is sensitive
to the thermo-optical properties of the medium, namely the
temperature-dependent refractive index and the thermal con-
ductivity or heat capacity. Owing to its low dn/dT and high k
values, water is a poor solvent for thermal lens spectrometry.
On the contrary, alcohols have better thermo-optical prop-
∗ Corresponding author. Tel.: +33-4-724-32627;
fax: +33-472-431-078.
E-mail address: j.georges@cpe.fr (J. Georges).
may be advantageous to use water–solvent mixtures to im-
prove the sensitivity of the TLS method [5,6]. The effect of
added methanol, acetone and acetonitrile on the sensitivity
and detection limits of chromium species using ion chro-
matography and TLS detection has been investigated [7].
For liquid mixtures, the composition dependence of the
photothermal signal is difficult to predict and one needs in-
formation on the thermo-optical properties of the mixtures.
Most often, the relationship between sensitivity and solvent
composition is estimated assuming that dn/dT and k are both
linear functions of the volume fraction of each solvent [6,7].
However, the composition dependence of the refractive in-
dex gradient is generally a more complicated function [8]
and reliable values must come from experiments. Such de-
termination of dn/dT, thermal conductivities or heat capaci-
ties have been made using the thermal lens method [9–14],
but precise determination of dn/dT requires knowledge of
the thermal conductivity, the sample absorbance and a ref-
erence medium to calibrate the experimental set-up if the
optical and geometrical parameters of this set-up are not
known [15]. In addition, it is necessary that the refractive

1386-1425/$ – see front matter © 2003 Elsevier B.V. All rights reserved.
doi:10.1016/j.saa.2003.09.019
1818 N. Arnaud, J. Georges / Spectrochimica Acta Part A 60 (2004) 1817–1823

index of the solution is not influenced by other factors than Table 1

the temperature. Especially, in solutions with more than one Thermo-optical parameters and physical constants for water, ethanol,
propanol and acetonitrile at 20 ◦ Ca
constituent, the temperature gradient induces thermal dif-
fusion and migration of molecules in the thermal gradient. Water EtOH PrOH ACN
This tendency of species to diffuse under the influence of a k (W m−1 K−1 ) 0.595 0.167 0.155 0.179
temperature gradient, known as the Soret effect, may result (dn/dT) × 104 (K−1 ) −0.9 −3.87 −3.8 −4.55
in the formation of a concentration gradient leading to a lo- Cp (J g−1 K−1 ) 4.18 2.41 2.31 2.14
cal change in the thermo-optical properties of the solution. ρ (g cm−3 ) 1 0.79 0.8 0.78
tc (ms) 2.95 4.8 5.0 3.9
The Soret effect has been observed in binary liquid mixtures
a Average of literature values.
near a consolute critical point [16–18], in micellar systems
[14,19,20] and in polymer solutions [21]. Depending on the
time scale of data recording, the concentration gradient can refractive index of water, ethanol, propanol, acetonitrile and
affect the thermal lens signal, which may lead to erroneous their mixtures was measured using a thermostated Abbé re-
interpretation of experimental data. fractometer over 1 ◦ C intervals between 14 and 26 ◦ C. The
The aim of this work was to investigate the composition dependence of the refractive index on temperature, dn/dT,
dependence of the thermal lens signal in binary systems and composition at room temperature, dn/dF, was deduced
composed of water and ethanol (EtOH), n-propanol (PrOH) by fitting the data to polynomial functions and then tak-
or acetonitrile (ACN). Time-resolved and steady-state ex- ing the derivative with respect to temperature and compo-
periments were carried out in order to separate the effects sition, respectively. Thermal conductivities of pure liquids
originating from a change in the temperature-dependent re- were taken from the literature (Table 1). For the binary sys-
fractive index due to a change in the macroscopic properties tems, the thermal conductivities have been estimated using
of the solution from the effects produced by the concentra- the Filippov equation which was shown to be better than the
tion gradient created locally in the irradiated region of the additive mass fraction model [22]. Then, the sensitivity of
solution. thermal lens spectrometry in the binary systems relative to
that in pure water was expected to vary as the dn/dT to k
ratio.
2. Experimental
2.3. Signal processing
2.1. Instrumentation
Under cw-laser excitation, the strength of the thermal lens
The dual-beam thermal lens spectrometer was the same as is defined by the inverse focal length:
that described and calibrated in a previous work [15]. It con- 1 P(1 − 10−A ) dn
sisted of an air-cooled argon-ion laser (λ = 488 nm) as the = (1)
f πkωoe
2 dT
excitation laser and a helium-neon laser as the probe beam.
Both beams were focused independently by two lenses and where P and ωoe are the power and the radius of the excita-
the set-up was arranged in a mode-mismatched configura- tion beam into the sample cell, A is the absorbance and k is
tion. The excitation beam was allowed to irradiate the sam- the thermal conductivity of the solution.
ple or blocked using a mechanical shutter. The probe beam The relation for the probe beam intensity change, derived
intensity change was measured through a 1 mm pinhole with from the theoretical model developed by Snook et al. [15,23],
a PIN photodiode and processed with a digital oscilloscope may be written as:
(Tektronix TDS 2002) connected to a PC computer. For each
sample, the signal was derived from the average of several I(t) = I(o)(1 − b(t)θ)2 (2)
recordings. b(t) is the time-dependent term which describes the forma-
tion of the thermal lens signal:
2.2. Reagents and procedures


1 −1 2mV
b(t) = tan
Thermal lens experiments were carried out with cobalt 2 [(1 + 2m)2 + V 2 ](tc /2t) + 1 + 2m + V 2
nitrate using a 2 mm quartz cuvette. Solutions of appropri- (3)
ate absorbances were prepared by successive dilutions from
stock solutions and absorbances were measured with a con- where m = (ωpc /ωoe )2 and V = Z1 /Zc are geometrical pa-
ventional spectrophotometer. The water–solvent mixtures rameters of the mode-mismatched dual-beam configuration,
were prepared by weighing and the composition was finally and tc = ωoe 2 ρC /4k is the characteristic time constant of
p
expressed as volume fraction of the organic solvent, F. De- the thermal lens effect; ωpc the radius of the probe beam in
pending on the mixture composition, the sample absorbance the cell, Z1 the distance from the waist of the probe beam
was adjusted between 0.004 and 0.06 in order that the ther- to the cell, Zc the confocal distance of the probe beam, and
mal lens signal was remaining approximately constant. The ρ and Cp are the density and specific heat of the medium,
 N. Arnaud, J. Georges / Spectrochimica Acta Part A 60 (2004) 1817–1823 1819

Table 2 where kT = DT /Dm is the thermal diffusion ratio, i.e. the

Experimental parameters used for the theoretical fits coefficient of thermal diffusion compared to that of mass dif-
Excitation laser spot size at cell, ωec 41 ␮m fusion. Following Eqs. (1)–(2) and assuming that dn/dF and
Probe laser spot size at cell, ωpc 100 ␮m dn/dT have the same effect on the propagation of the probe
Excitation laser power at cell 4 mW beam, the probe beam intensity change can be written as:
m 6.3
Zc 7.5 mm I(t) = I(o)(1 − b(t)θ − b (t)θ  )2 (6)
Z1 14 mm
where θ  is related to the amplitude of the Soret signal and
b (t) is the time-dependent term which describes the forma-
respectively. θ is a dimensionless parameter which indicates tion of the thermally induced concentration gradient. This
the strength of the thermal lens: term has been expressed as [25,26]:
(1 − 10−A )P dn  ∞

θ=− (4)  4 (2n − 1)π
λp k 
dT b (t) = 1 − sin
(2n − 1)π 2
n=1
where λp is the wavelength of the probe beam. θ and b(t) can

be calculated using the sample absorbance, the laser power in t
× exp −(2n − 1)2 (7)
the cell, the thermo-optical properties of the solution and the tD
geometrical parameters of the experimental set-up (Table 2)
which has been calibrated in a previous work [15]. where tD is the mass diffusion time.
In mixtures, the temperature gradient produced by re-
laxation of the excited species induces mutual migration
of both components which results in the formation of a 3. Results and discussion
composition-dependent refractive index gradient, dn/dF.
This gives rise to a concentration lens which adds to the 3.1. Changes in dn/dT and dn/dF
thermal lens. The inverse focal length of the new lens has
been expressed as [17,24]: The experimental values obtained for dn/dT are reported
  in Fig. 1. The data show that dn/dT as a function of the
1 P(1 − 10−A ) dn kT dn mixture composition does not follow a simple additive
= − (5)
f πkωoe2 dT T dF rule, but is best represented by fourth, fifth- and sixth-order

Fig. 1. Plots of the temperature-dependent refractive index gradient dn/dT × 104 (K−1 ) as a function of the solvent volume fraction for the water–solvent
binary systems. The dashed lines are polynomial fits to experimental data.
1820 N. Arnaud, J. Georges / Spectrochimica Acta Part A 60 (2004) 1817–1823

Fig. 2. Plots of the composition-dependent refractive index gradient dn/dF at room temperature.

polynomials for ethanol, acetonitrile and propanol, respec-
tively. The refractive index as a function of the mixture
composition at room temperature followed second-, third-
and fourth-order polynomials for propanol, ethanol and
acetonitrile, respectively. The derivative of these functions
allowed the calculation of the concentration-dependent re-
fractive index gradient, dn/dF. As shown in Fig. 2, the three
mixtures exhibit a different behavior: for water–propanol,
dn/dF decreases linearly as the solvent volume fraction
increases while for water–ethanol and water–acetonitrile
systems, the variation is more complicated and exhibits a
sign change at a solvent volume fraction close to 0.7–0.8.

Fig. 3. Change in probe beam intensity upon step excitation for cobalt nitrate in water–propanol mixture (F = 0.1). The thin line is the fit of Eq. (2) to the
experimental curve within the 0–400 ms time range after the beginning of excitation. The thicker line is the fit of Eq. (6) to the overall experimental curve.
 N. Arnaud, J. Georges / Spectrochimica Acta Part A 60 (2004) 1817–1823 1821

3.2. Time-resolved signal
A typical curve for the probe beam intensity change un-
der step excitation is shown in Fig. 3. The initial fast change
corresponds to the formation of the temperature-dependent
refractive index gradient (thermal lens signal) while the sec-
ond part corresponds to the slower build-up of the ther-
mally induced concentration gradient (Soret signal). Both
processes can be easily separated because the Soret compo-
nent builds up with a time constant which is much greater
than the characteristic time constant of the thermal lens.
3.2.1. Relative sensitivity of the TLS method
In pure solvents, where the Soret effect is not efficient,
the fit of Eq. (2) using the calculated values of b(t) and θ
was valid until steady-state is reached. On the contrary, in

Fig. 4. Plots of the relative sensitivity of the thermal lens method as a function of the solvent volume fraction for the (a) water–ethanol, (b) water–propanol
and (c) water–acetonitrile binary systems. The dotted lines represent the relative sensitivities expected from the dn/dT to k ratios.
1822 N. Arnaud, J. Georges / Spectrochimica Acta Part A 60 (2004) 1817–1823

Fig. 4. (Continued ).

most binary mixtures, the fit was limited to the short period,
typically 400 ms, within which mass diffusion is inoperative
and the probe beam intensity change is only determined by
the temperature-dependent refractive index gradient. The fit
to the initial fast intensity change allowed the determination
of the transient steady-state intensity from which the ther-
mal lens signal was obtained. The data, expressed as the
response obtained in the binary systems relatively to that in
pure water under the same experimental conditions, show
that the sensitivity adequately follows the dn/dT to k ratio

Fig. 5. Plots of the Soret signal, expressed as percent of the pure thermal lens signal, as a function of the solvent volume fraction.
 N. Arnaud, J. Georges / Spectrochimica Acta Part A 60 (2004) 1817–1823
of the binary system (Fig. 4). On the contrary, if the sig-
nal was taken at the final steady-state, i.e. within a few sec-
onds, the true thermal lens signal would be affected by the
concentration gradient. It is clear from Fig. 3 that the ef-
fect of the concentration gradient depends on the excitation
duration and the time at which the signal is sampled af-
ter the beginning of excitation. It is expected that working
with short irradiation time and fast data collection would
prevent the effect of mass diffusion. However, it has been
shown in a previous work [27] that even upon chopped exci-
tation and lock-in detection, a concentration gradient builds
up slowly to a steady-state and changes the probe beam
intensity.
3.2.2. Amplitude and kinetics of the Soret component
The Soret signal was taken as the difference between the
transient probe beam intensity, as deduced from Eq. (2),
and the final steady-state intensity. As shown in Fig. 5, the
effect of mass diffusion is more important with ethanol
and propanol than with acetonitrile. With both alcohols, the
error produced by the concentration gradient may represent
up to 25% of the pure thermal lens signal when the solvent
volume fraction is within the range 0.05–0.25. This result
may be explained by two factors: (i) dn/dF is much smaller
with acetonitrile than with ethanol and propanol; (ii) as the
solvent composition increases, dn/dT increases while dn/dF
decreases, which contributes to minimize the effect of mass
diffusion with respect to the thermal lens signal. At inter-
mediate solvent compositions, the concentration gradient
had a reverse effect and the final steady-state intensity was
significantly greater than the transient steady-state intensity,
which is reported by negative values in Fig. 5. The negative
contribution of the Soret effect is unexpected within this
concentration range given that the measured dn/dF values
are positive. It is likely that the concentration gradient in-
duces local changes of the solution properties including
absorbance, the thermal conductivity and the refractive in-
dex. Although it is difficult to account for each of them
individually, these changes could result in a slight decrease
of the thermal lens signal.
As shown in Fig. 3, the slow component is perfectly fitted
by Eq. (6), retaining terms up to n = 4 in the summation
of Eq. (7). The fit allowed the determination of the mass
diffusion time tD , which, by analogy to the thermal time
constant tc , has been defined as [21]:
ω2
tD = ec
4Dm
1823
where Dm is the coefficient of mass diffusion. Diffusion
coefficients of 3.5 × 10−6 cm2 s−1 have been derived from
experimental tD values of 1.2 s for both water–ethanol and
water–propanol systems at solvent volume fractions of 0.15
and 0.1, respectively. Such a data treatment has previously
been validated for micelles and polymers in aqueous so-
lutions [21]. However, the values obtained in the present
work are about four times smaller than the mutual diffusion
coefficients measured for the same mixtures at the same
composition using the Taylor dispersion technique [28,29].
Further work is required to understand the reason of this
disagreement.
References
[1] N.J. Dovichi, CRC Crit. Rev. Anal. Chem. 17 (1987) 357.
[2] J. Georges, Talanta 41 (1994) 2015.
[3] R.D. Snook, R.D. Lowe, Analyst 120 (1995) 2051.
[4] S. Bialkowski, Photothermal spectroscopy methods for chemical anal-
ysis, in: J.D. Winefordner (Ed.), Chemical Analysis, vol. 134, Wiley,
New York, 1996.
[5] N.J. Dovichi, J.M. Harris, Anal. Chem. 51 (1979) 728.
[6] W.A. Weimer, N.J. Dovochi, J. Appl. Phys. 59 (1986) 225.
[7] M. Sikovec, M. Franko, M. Novic, M. Veber, J. Chromatogr. A 920
(2001) 119.
[8] W.B. Li, P.N. Segre, R.W. Gammon, J.V. Sengers, J. Chem. Phys.
101 (1994) 5058.
[9] M.C. Gupta, S.-D. Hong, A. Gupta, J. Moacanin, Appl. Phys. Lett.
37 (1980) 505.
[10] R.D. Snook, R.D. Lowe, M.L. Baesso, Analyst 123 (1998) 587.
[11] K. Seidman, A. Payne, J. Chem. Educ. 75 (1998) 897.
[12] S.M. Colcombe, R.D. Snook, Anal. Chim. Acta 390 (1999) 155.
[13] C.V. Bindhu, S.S. Harilal, V.P.N. Nampoori, C.P.G. Vallabhan, Opt.
Eng. 37 (1998) 2791.
[14] M. Franko, C.D. Tran, J. Phys. Chem. 95 (1991) 6688.
[15] M. Fischer, J. Georges, Anal. Chim. Acta 322 (1996) 117.
[16] F.D. Hardcastle, J.M. Harris, Appl. Spectrosc. 40 (1986) 606.
[17] M. Giglio, A. Vendramini, Appl. Phys. Lett. 25 (1974) 555.
[18] M. Giglio, A. Vendramini, Phys. Rev. Lett. 34 (1975) 561.
[19] J. Georges, T. Paris, Anal. Chim. Acta 386 (1999) 287.
[20] N. Arnaud, J. Georges, Spectrochim. Acta Part A 57 (2001) 1085.
[21] J. Georges, Spectrochim. Acta Part A 59 (2003) 519.
[22] G. Cai, H. Zong, Q. Yu, R. Lin, J. Chem. Eng. Data 38 (1993) 332.
[23] J. Shen, R.D. Lowe, R.D. Snook, Chem. Phys. 165 (1992) 385.
[24] F.D. Hardcastle, J.M. Harris, Appl. Spectrosc. 40 (1986) 606.
[25] K.J. Zhang, M.E. Briggs, R.W. Gammon, J.V. Sengers, J. Chem.
Phys. 104 (1996) 6881.
[26] K.J. Zhang, M.E. Briggs, R.W. Gammon, J.V. Sengers, J.F. Douglas,
J. Chem. Phys. 111 (1999) 2270.
[27] N. Arnaud, J. Georges, Anal. Chim. Acta 445 (2001) 239.
[28] K.R. Harris, T. Goscinska, H.N. Lam, J. Chem. Soc., Faraday Trans.
(8) 89 (1993) 1969.
[29] L. Hao, D.G. Leaist, J. Chem. Eng. Data 41 (1996) 210.