The attached
copy is furnished to the author for internal non-commercial research
and education use, including for instruction at the authors institution
and sharing with colleagues.
Other uses, including reproduction and distribution, or selling or
licensing copies, or posting to personal, institutional or third party
websites are prohibited.
In most cases authors are permitted to post their version of the
article (e.g. in Word or Tex form) to their personal website or
institutional repository. Authors requiring further information
regarding Elseviers archiving and manuscript policies are
encouraged to visit:
http://www.elsevier.com/copyright
Optical Materials
journal homepage: www.elsevier.com/locate/optmat
a r t i c l e
i n f o
Article history:
Received 31 October 2012
Received in revised form 31 December 2012
Accepted 2 January 2013
Available online 8 February 2013
Keywords:
Thermal mirror
Thermo-optical properties
Optical materials
a b s t r a c t
The Thermal mirror technique relies on measuring laser-induced nanoscale surface deformation of a solid
sample. The amplitude of the effect is directly dependent on the optical absorption and linear thermal
expansion coefcients, and the time evolution depends on the heat diffusion properties of the sample.
Measurement of transient signals provide direct access to thermal, optical and mechanical properties
of the material. The theoretical models describing this effect can be formulated for very low optical
absorbing and for absorbing materials. In addition, the theories describing the effect apply for semi-innite and nite samples. In this work, we apply the Thermal mirror technique to measure physical properties of optical glasses. The semi-innite and nite models are used to investigate very low optical
absorbing glasses. The thickness limit for which the semi-innite model retrieves the correct values of
the thermal diffusivity and amplitude of the transient is obtained using the nite description. This procedure is also employed on absorbing glasses, and the semi-innite BeerLambert law model is used to
analyze the experimental data. The experimental data show the need to use the nite model for samples
with very low bulk absorption coefcients and thicknesses L < 1.5 mm. This analysis helped to establish
limit values of thickness for which the semi-innite model for absorbing materials could be used,
L > 1.0 mm in this case. In addition, the physical properties of the samples were calculated and absolute
values derived.
2013 Elsevier B.V. All rights reserved.
1. Introduction
Laser induced surface deformation has been explored in many
photothermal (PT) techniques [14]. These techniques are characterized as fast, non-contacting and highly sensitive spectroscopic
tools for material characterization. The reason is that there are
many different experimental congurations that are important in
many research aspects, such as nondestructive evaluation of solids
and surfaces, from multilayered and thin lms to biological systems [513]. The amplitude of the deformation is directly related
to the optical absorption and linear thermal expansion coefcients,
and the time evolution of the deformation depends on the heat diffusion properties. Thus, measuring the transient effect implies direct quantitate access to thermal, optical and mechanical
properties of the material [1016].
Surface deformation can be generated by modulated [511] or
pulsed [1214] excitation lasers. Briey, the principle of detection
of these PT methods consists of measuring the surface displacement of the sample by pumpprobe methods. For instance, this
deformation can be detected by analyzing the focusing/defocusing
Corresponding author.
E-mail address: astrathngc@pq.cnpq.br (N.G.C. Astrath).
0925-3467/$ - see front matter 2013 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.optmat.2013.01.003
of the probe beam reected off of the sample surface in the far eld
region. In this conguration, namely Thermal Mirror (TM) spectrometry, the method has been recently introduced under cw
Gaussian [1518] and Top-hat [19,20] laser excitations for measurement of thermal diffusivity and thermo-optical properties of
semi-transparent and opaque solids. TM is attractive in that it
can also be used concurrently with the Thermal Lens method to
determine physical properties of uorescent materials [19].
Recent theoretical and experimental works for Gaussian excitation have been developed assuming the sample either as a semiinnite [17] or as a nite [18] media. Under the semi-innite
approximation, the sample is treated as an innite medium along
the z axis with its surface placed at z = 0. For the nite approach,
the sample surfaces are at z = 0,L. The thermoelastic equations
describing the laser induced surface displacement were solved
according to these assumptions and experimental investigations
of very low bulk absorbing materials were compared to the theories. It was shown that for samples with thicknesses under certain
values the nite model should be considered. However, the nite
description allows only solutions for materials with low optical
absorption coefcient (Ae), where the attenuation of light intensity
is considerably negligible along the sample thickness. For absorbing samples, in which the light intensity attenuation follows the
1130
BeerLambert law as eAe z , only semi-innite solutions for the thermoelastic equations are semi-analytically available. Here, we further investigate the nite and semi-innite approaches by
performing experiments in very transparent and absorbing optical
glasses.
In this work, we apply the Thermal mirror technique to measure
physical properties of optical glasses. These glasses have optical
absorption coefcients varying from approximately 61150 m1.
The semi-innite and nite models are used to investigate the
low optical absorbing glasses (Ae < 200 m1). The thickness limit
for which the semi-innite model retrieves the correct values of
the thermal diffusivity and TM amplitude is obtained using the nite description. This procedure is also employed on absorbing
glasses, and the semi-innite BeerLambert law model is used to
analyze the experimental data for Ae > 400 m1.
2. Theory
The TM experiment was performed in the dual beam modemismatched conguration, as shown in Fig. 1. In this conguration,
the sample is excited and probed by single-mode TEM00 Gaussian
cw laser beams with radii x0e and x1p (on the sample surface at
z = 0), respectively [17]. The excitation laser beam induces a nonuniform temperature gradient deforming the sample surface
accordingly. The probe beam is reected off of the sample surface
and senses the deformation along the z-axis. The amplitude of the
TM effect is directly related to the absorption coefcient and surface bulging, and the time evolution of the deformation depends
on the heat diffusion properties. The deformation evolves following the NavierStokes thermoelastic equation [21]. Due to the axially symmetric radial nature of the Gaussian heat source, the
problem is treated in cylindrical coordinates and, in the quasi-static approximation, the solution is proposed by introducing the scalar displacement potential W and the Love function was described
in details in Ref. [21]. Using the integral transform method it is
possible to write the surface displacement, uz(r,z = 0,t), as [18]
uz r; 0; t 2aT 1 m
r Z
2
a2 Ta; k; t a2
where aT is the linear thermal expansion coefcient, m is the Poissons ratio and Jn(x) represents the Bessel function of the rst kind.
T(a,k,t) is the temperature in the HankelFourier space and h(a,k),
for a nite sample of thickness L, is given by [18]
ha; k
1
2
1 2L a cosh2La
2
f2La sinh2La
For a semi-innite sample, h(a,k) = 1 [15]. Eq. (1) does not consider
sample-uid heat coupling this assumption has been investigated
in Ref. [22] and was found to be a good approximation when air is
used as uid. Assuming that the radial dimensions of the sample are
much larger than the radii of probe and excitation beams on the
sample; the solution of the heat diffusion equation can be calculated for a material whose absorption of light follows the BeerLambert law (BLM) yielding [16]
Ta; k; t
x2 a2
P e Ae /
0e
Qke 8
2pqc
2 k2
eDa
ds:
p
Q k Ae 2=p= A2e k2 is the FourierCosine transform of the
azimuthal source term Sz eAe z . S(z) accounts for the attenuation
of light along the sample thickness. In the limit of very low bulk
absorption coefcient, Low
Model (LAM), Ae ? 0 and
pAbsorption
S(z) = 1, yielding Q k 2pdk. d(k) represents the Dirac delta
function. Pe = P0(1 R), P0 is the excitation power, R is the sample
surface reectivity, / measures the fraction of the absorbed energy
converted to heat, ke is the excitation beam wavelength. D, q and c
are the thermal diffusivity, mass density and the specic heat of the
sample, respectively.
The general BLM representation of the displacement can be
semi-analytically calculated only for the semi-innite approximation. On the other hand, the approximate LAM presents solutions
for both nite and semi-innite descriptions.
The temperature rise distribution induces a surface displacement that acts as an optical element, causing a phase shift of the
probe beam reected off of the sample surface as U(r,t) = (2p/
kp)2[uz(r,0,t) uz(0, 0, t)] [17]. kp is the probe beam wavelength.
By introducing the characteristic TM thermal time t c x20e =4D,
and using Eq. (1), the phase shift for the semi-innite BLM can
be written as [16]
USIBLM g; t
hTM
tc
1 2
in which
2 2
p
p
a x t
x0e
2 tc t Ae x0e exp 4tc0e
t Ae x20e Erfc t2ap
tc
f a; t
p 4
pa a2 A2e
a3 aA2e
p
2tc
tax0e
p
A3e 3a2 Ae Erf
2
2 tc
a3 a2 A2e
2 2
3
2
39
p
=
Ae a2 x20e
t
A
x
e
0e
5 Erfc
5 :
p
2 a3 41 exp 4
;
4t c =t
2 tc
x20e
coshLa 1
La sinhLa
Z t x2
p
2s
2
0e
J0 ax0e gm 1
e 8 1 tc a ds da;
UFLAM g; t hTM
tc
1131
USILAM g; t hTM
x20e
tc
x2
e
0e
8
12tcsa2 ds
p
J0 ax0e gm 1da:
Here, the parameter hTM measures the amplitude of the TM effect and is given by
hTM
P e Ae aT 1 m/
;
kp k
2
R 1
iVg i Ug; t dg
0 exp1
R1
It I0
:
exp1 iVg dg
0
D
hTM/Pe
aT
Ae
qc
Ref. [25].
107 m2s1
m1 W1
107 K1
m1
106JK1 m3
QX
Q98
LSCAS-Ce
SLMn
LSCAS-Cr
3.6 0.3
114 3
83a
6.03 0.05
1.97 0.05
2.8 0.2
746 43
99 5
48.5 0.7
2.48a
5.4 0.4
85 3
76 4
7.5 0.9
2.42 0.08
7.3 0.3
(3.4 0.1) 103
90 4
460 18
1.98 0.05
7.1 0.3
(7.8 0.4) 103
89 4
1158 30
2.40 0.05
1132
(a)
(b)
Fig. 2. Normalized Thermal mirror signal I(t)/I(0) for the QX samples of different
thicknesses under excitation of P0 = 0.82 W. Circle: experimental data; dashed line:
best curve tting using the semi-innite LAM model; solid line: best curve tting
using the nite LAM model. The experimental parameters are Z1 = 8.8 cm, ZC =
1.31 cm, x0e = 71 lm, x1p = 360 lm, m = 25.7, and V = 6.71.
Fig. 4. Experimental results for the phosphate Q98 and LSCAS-Ce glasses: (a) shows
the thermal diffusivities and (b) the parameter hTM/P e obtained from tting to the
nite, Eq. (6), and semi-innite, Eq. (7), models for different thicknesses. The
experimental parameters are Z1 = 8.3 cm, ZC = 1.47 cm, x0e = 104 lm, x1p = 316 lm,
m = 9.14, and V = 5.6 for Q98, and Z1 = 15.8 cm, ZC = 2 cm, x0e = 72 lm,
x1p = 501 lm, m = 48.5, and V = 7.9 for LSCAS-Ce.
(a)
(a)
(b)
(b)
Fig. 3. Experimental results for the phosphate QX glass: (a) shows the thermal
diffusivities and (b) the parameter hTM/Pe obtained from tting to the nite, Eq. (6),
and semi-innite, Eq. (7), models for different thicknesses.
experimental data for the thicker sample while for L = 0.42 mm the
semi-innite model does not t well to the data.
The thermal diffusivities and the parameters hTM obtained from
the regressions are shown in Fig. 3. Both models retrieve approximately the same values as the thickness is increased, although the
semi-innite model over predicts the thermal diffusivity and the
parameter hTM as the thickness is reduced. Note that the thick sample falls into the semi-innite approximation; hTM and D are
approximately the same as the ones obtained using the nite model. The gray areas show the region within the acceptable values for
D and hTM from the nite analises, and the average parameters are
presented in Table 1.
The same behavior is also observed for Q98 and LSCAS-Ce
glasses. Fig. 4 shows the thermal diffusivity and the parameter
hTM/Pe for different thicknesses of the samples. It is evident that
as the thickness is reduced to values lower than L = 1.5 mm the
semi-innite approximation cannot be used to analyze the experimental data. In fact, it does not pose major problems because of
low optical absorption materials the nite modeling is available.
However, for materials not following the LAM approximation only
Fig. 5. Experimental results for LSCAS-Cr and SLMn glasses: (a) shows the thermal
diffusivities and (b) the parameter hTM/Pe obtained from tting to the semi-innite
BLM model, Eq. (4), for different thicknesses. Dashed and dotted lines show the
average parameters obtained from the thick samples. The experimental parameters
for both samples are Z1 = 5.75 cm, ZC = 6.4 cm, x0e = 80 lm, x1p = 200 lm, m = 6.25,
and V = 0.89.
a semi-innite description is possible and the thickness of the sample has to be chosen according to the modeling approximation.
Fig. 5 displays the results for the soda-lime and LSCAS-Cr samples with different thicknesses. D and hTM/Pe were obtained from
the experimental transients tted to I(t) with Eq. (4). Dashed and
dotted lines show the average parameters obtained from the thick
samples. The averaged values are also presented in Table 1. It can
be seen that even for samples with thickness around L = 1 mm the
semi-innite approximation gives values of D and hTM/Pe close to
the ones of the thicker samples. This lower thickness limit of validity of the semi-innite BLM model compared to the semi-innite
LAM can be explained by the fact that in samples with high optical
absorption coefcient, the contribution to the TM phase shift is
mainly supercial. The attenuation of light along the sample thickness makes the inner layers to contribute less to the deformation
prole. The same does not happen for very low optical absorbing
materials, where the sample absorbs practically evenly within
the material.
1133