Reversible temperature dependence in surface-enhanced raman scattering of 1-propanethiol adsorbed on a silver island film

Article abstract of DOI:10.1021/jp981820p

Intensity of the surface-enhanced Raman scattering (SERS) of 1-propanethiol adsorbed on a silver film prepared chemically has been measured over the 15-300 K surface temperature range. A reversible change in the SERS intensity with the temperature has been observed, with the intensity at 15 K being larger than that at 300 K by a factor of ～3. Various factors that can affect the SERS intensity such as the laser-induced change in surface morphology, adsorption and desorption, and destruction of adatoms could be eliminated as a possible cause for the observed temperature dependence. Enhancement of the Raman intensity has been calculated over the experimental temperature range based on the electromagnetic (EM) model for SERS proposed by Gersten and Nitzan. It has been found that the observed temperature dependence can be explained qualitatively with the EM model calculation.

Full text of DOI:10.1021/jp981820p

J. Phys. Chem. B 1998, 102, 7203-7209
7203
Reversible Temperature Dependence in Surface-Enhanced Raman Scattering of
1-Propanethiol Adsorbed on a Silver Island Film
Yoon Soo Pang,† Hyun Jin Hwang,‡ and Myung Soo Kim*,†
Department of Chemistry, Seoul National UniVersity, Seoul 151-742, Korea and Department of Chemistry,
Kyunghee UniVersity, Seoul 130-701, Korea
ReceiVed: April 13, 1998; In Final Form: June 23, 1998
Intensity of the surface-enhanced Raman scattering (SERS) of 1-propanethiol adsorbed on a silver film prepared
chemically has been measured over the 15-300 K surface temperature range. A reversible change in the
SERS intensity with the temperature has been observed, with the intensity at 15 K being larger than that at
300 K by a factor of ∼3. Various factors that can affect the SERS intensity such as the laser-induced change
in surface morphology, adsorption and desorption, and destruction of adatoms could be eliminated as a possible
cause for the observed temperature dependence. Enhancement of the Raman intensity has been calculated
over the experimental temperature range based on the electromagnetic (EM) model for SERS proposed by
Gersten and Nitzan. It has been found that the observed temperature dependence can be explained qualitatively
with the EM model calculation.
1. Introduction
roughness, which led them to conclude the importance of the
adatom model for SERS. The adatom model was also advocated
Observation of surface-enhanced Raman scattering (SERS)1-4
made more than 20 years ago has stimulated various research
activities in the areas of spectroscopy, surface science, etc. The
phenomenon has been found to be rather general, being
observable on rough surfaces of various metals prepared by
many different ways for a wide variety of adsorbate molecules.2
Intensive efforts have been made over the years, mainly focused
on the optical properties of the substrates and the substrate-
adsorbate interaction, to identify the origin of the enhance-
ment.3,4 Even though the exact nature of the phenomenon is
not firmly established yet, there is a general consensus that the
electromagnetic (EM) effect arising from surface plasmon
resonance plays a major role in most of the cases.5-9
Silver island films are excellent surfaces for the investigation
of the EM effect. Both experimental and theoretical evidences
suggest that the plasmon resonance on an individual island plays
an important role in causing SERS.10,11 For example, SERS
from a periodic array of uniform silver ellipsoids was in
excellent agreement with the theoretical prediction for an
isolated ellipsoid.12 However, on island films prepared by
general means or on other less orderly surfaces, EM interaction
among silver particles and the size and shape distributions of
these particles would be also important in determining the nature
and role of the plasmon resonance.
by some investigators studying SERS on electrode surfaces
immersed in electrolyte solutions at room temperature.14,15 The
adatoms were thought to be stabilized by coadsorbed anions
under this condition. Then, the irreversible loss of the SERS
activity in various processes such as the increase in the solution
temperature was taken as the evidence supporting the adatom
model. The above consideration suggests that the adatom model
would not be important for SERS on silver films prepared (or
stored) under room conditions in the absence of appropriate
anions.
In this paper, we report our recent observation of a reversible
temperature dependence in SERS of 1-propanethiol adsorbed
on a silver film prepared chemically. It will be shown that this
temperature dependence can be explained qualitatively within
the framework of the EM model by considering the temperature
dependence of the surface optical property.
2. Experimental Section
The chemical procedure to prepare the silver film is as
follows.16 Thirty milliliters of 2% AgNO3 solution was mixed
with 1-1.5 mL of 5% NaOH to produce dark brown AgOH
precipitate. Ammonium hydroxide solution was added dropwise
until the precipitate dissolved and the solution became transpar-
ent. Then, 9 mL of 10% D-glucose solution was added as a
reducing agent in the presence of a sapphire window. The latter
was chosen as the cryogenic substrate for its large thermal
conductivity at low temperature. The substrate was precleaned
with nitric acid to remove remaining silver film and was
sonicated successively in several solvents before chemical
deposition. A silver film prepared as above was dipped in 10
mM KCl solution for 3-4 min and rinsed with water. This
treatment improved the surface quality of the film such as the
mechanical stability and reproducibility of SERS activity. Then,
the film was dipped in 1 mM 1-propanethiol solution in
methanol for several minutes. The remaining solution on the
film was removed by sufficient rinsing in water. 1-Propanethiol
Temperature dependence of SERS intensity has been studied
by some investigators mainly as an attempt to identify the
specific features on the surface contributing to SERS.13-15 For
example, Pockrand and Otto13 observed SERS of pyridine
adsorbed on silver films prepared by evaporation at 120 K and
reported that the SERS intensity decreased irreversibly and rather
abruptly as the film was heated above 230 K. Such a decay of
the SERS activity was attributed to the annealing of atomic scale
* To whom correspondence should be addressed. Fax: +82-2-889-1568.
E-mail: myungsoo@plaza.snu.ac.kr.
† Seoul National University.
‡ Kyunghee University.
S1089-5647(98)01820-3 CCC: $15.00 © 1998 American Chemical Society
Published on Web 08/21/1998

7204 J. Phys. Chem. B, Vol. 102, No. 37, 1998
Pang et al.
Figure 1. Schematic diagram for the low-temperature Raman instru-
mentation: A, argon ion laser; B, He/Ne laser; C, sample holder; D,
holographic notch filter; E, laser line filter; F1 and F2, shutters, G, CCD
detector; L1, L2, and L3, optical assembly.
was purchased from TCI and was used without further purifica-
tion. All the chemicals otherwise specified were reagent grade,
and triply distilled water was used throughout.
A schematic of Raman instrumentation is shown in Figure 1.
The sample was mounted on a holder attached to a helium
cryostat (Janis Research CCS-600) inside a vacuum chamber
with inner diameter of 6 in. The chamber was evacuated by a
turbomolecular pump to the base pressure lower than 5 × 10-7
Torr. A liquid nitrogen cold trap was installed inside the
chamber to remove volatile materials when needed. Sample
temperature could be controlled within (0.1 K in the range of
10-300 K. The sample was irradiated with the 514.5 nm line
of an argon ion laser (Spectra Physics model 164-06) or with
the 632.8 nm line of a He/Ne laser (Research Electro-Optics
model LHRP 2001). Typical laser power at the sample position
was 50 mW for Ar+ laser and 10 mW for He/Ne laser. Raman
scattering was collected at 90° and was analyzed using a
spectrometer equipped with a CCD detector (the spectrograph
stage of a SPEX 1877E spectrometer). Collimation and
collection lens assembly consisting of lenses with 127 and 700
mm focal lengths was used. To improve the light throughput
and thus to minimize the data acquisition time, the filter stage
of the SPEX 1877E spectrometer was replaced with a holo-
graphic filter (Kaiser Optical System). With this modification,
a factor of ca. 10 improvement in the signal-to-noise ratio was
achieved. This modification along with use of shutters that
synchronize excitation laser irradiation and Raman measurement
was critical in performing the present experiments. This is
because a severe laser-induced irreversible damage of the sample
can occur after an extended period of laser irradiation (to be
discussed). Slit width of the spectrometer was set at the spectral
resolution of 7 cm-1. A Spex DM3000R data system was used
for data acquisition and processing. Slit width correction for a
Raman bandwidth was made using the following relation.17
Figure 2. SER spectrum of 1-propanethiol on silver film (a) at 300
K, 1 atm (air) and those obtained in a vacuum at (b) 300 K, (c) 200 K,
(d) 100 K, and (e) 15 K.
TABLE 1: Assignmentsa and Spectral Datab for Some
Vibrational Modes of 1-Propanethiolate in a Basic Aqueous
Solution (OR) and on the Silver Surface (SERS)
SERS
ORS
15 K
300 K
assignment
372 (9)
650 (11)
727 (14)
788 (11)
893 (16)
1032 (10)
375 (14)
630 (19)
702 (17)
777 (10)
894 (12)
1026 (11)
377 (14)
623 (24)
697 (20)
778 (10)
894 (13)
1025 (11)
CCC deformation (T)
CS stretching (G)
CS stretching (T)
CH2 rocking (G)
CH2 rocking (T)
CC stretching (T)
a Taken from refs 18 and 19. T and G denote trans and gauche
conformations with respect to the C(1)-C(2) axis. b Positions and
bandwidths (fwhm) in cm-1. Bandwidths are given in parentheses.
to each other except for some minor variations (to be explained).
Also, they are similar to the SER spectra obtained in the silver
sol solution reported by Joo et al.18 To reiterate the spectral
interpretation in the above work, 1-propanethiol adsorbs dis-
sociatively by losing its thiol proton, namely, as 1-propane-
thiolate, and the adsorption occurs via its sulfur atom. Two
different conformations, namely, gauche (G) and trans (T) with
respect to the C(1)-C(2) axis, are possible on the surface with
variable intensity ratio determined by environmental factors such
as the surface coverage. Vibrational assignments for the major
Raman bands investigated here have been taken from the above
and are listed in Table 1. In the SER spectrum (Figure 2a)
obtained at 300 K without evacuation, several small peaks that
do not find their counterparts in other SER spectra appear. These
peaks disappeared after several hours of evacuation, as can be
seen in Figure 2b, indicating that these are not due to
chemisorbed 1-propanethiolate. Weakly bound molecules or
δ ) δ′[1 - (S/δ′)2]
(1)
Here δ and δ′ are the corrected and measured full widths at
half-maxima (fwhm). S is the spectral slit width.
3. Results and Discussion
A. General SER Spectral Feature. The SER spectra of
1-propanethiol adsorbed on the silver island film obtained with
514.5 nm excitation at various surface temperatures are shown
in Figure 2. Overall patterns of these spectra are very similar

SERS of 1-Propanethiol Adsorbed on a Silver Film
J. Phys. Chem. B, Vol. 102, No. 37, 1998 7205
overlayers may be responsible for these. Another noticeable
spectral difference between parts a and b of Figure 2 is in the
G/T intensity ratio of the CS stretching modes at 625 (G) and
697 (T) cm-1. Removal of the contaminants mentioned above
seems to increase the fraction of the G conformer on the surface.
Also can be seen in Figure 2 is that the G/T ratio decreases as
the temperature is lowered. This is because the T form is a
little bit more stable than the G form on the surface.19 The
G/T ratio was also found to depend on the dipping condition.
For example, dipping the film in 1 × 10-6 M solution of
1-propanethiol resulted in the G/T ratio of 4.5, which is larger
by an order of magnitude than that in Figure 2b. This is in
agreement with our previous observation of the concentration
dependence in the sol solution.18
Figure 3. Laser-induced decay of the SERS intensity of the CS
stretching vibration at 300 K (+), 100 K (2), 70 K (O), and 40 K (9).
50 mW of a 514.5 nm laser was irradiated continuously.
Table 1 lists the positions and widths of major vibrational
bands in the SER spectra obtained at 15 and 300 K together
with those in the ordinary Raman (OR) spectrum of 1-propane-
thiolate ion in basic aqueous solution. Comparing the values
in the OR and SER spectra, one finds that the CS stretching
modes undergo the most remarkable change upon surface
adsorption. In the previous investigations on SERS of mer-
captans, this was taken as evidence for the participation of the
sulfur atom in the adsorption process, namely, chemisorption
via formation of metal-sulfur bond(s).18,20 Also noticeable in
the table is that the positions and widths of these bands show
noticeable temperature dependence. For example, the CS
stretching mode of the gauche conformer shifted gradually from
623 to 630 cm-1 and its width changed from 24 to 19 cm-1 as
the surface temperature was lowered from 300 to 15 K. A simi-
lar trend was observed for the CS stretching mode of the trans
conformer. Temperature dependences of other bands were much
less or were hardly detectable. We also investigated SERS of
1-propanethiolate with 632.8 nm excitation. The positions and
widths of the major bands were the same as in the 514.5 nm
excitation within the experimental error, as expected.
One final concern in the present study is the irreversible decay
of a SERS signal due to the damage caused by laser illumination
over an extended period of time. This laser damage reported
for SERS studies on metal film and electrode surfaces is thought
to occur due to surface heating and photochemical effect of
laser.32,33 We have investigated the laser-induced decay of the
SERS intensity in this work, the main purpose of the study being
to reduce the amount of decay as much as possible. The decay
rate was found to depend on various factors such as laser power,
wavelength, surface temperature, etc. Higher power, shorter
wavelength, and the presence of air or oxygen resulted in faster
decay. Laser-induced decays with 50 mW (at sample position)
of 514.5 nm excitation at several surface temperatures in a
vacuum are shown in Figure 3. It is seen that the decay is
slower at lower temperature, indicating that the decrease in
SERS activity has something to do with the surface diffusion
of the silver atoms. The decay was quite slow at low
temperature, say at 40 K, more than 80% of SERS signal
remaining after 30 min of continuous laser irradiation. At room
temperature, however, the decay rate was uncomfortably fast.
Hence, we installed a mechanical shutter between the laser and
sample such that laser was irradiated on the sample only during
the spectral acquisition period, which was usually 1 s. The total
laser irradiation time for the measurement of temperature
dependence of SERS intensity to be described below did not
exceed a few tens of seconds.
B. Temperature Dependence of SERS Intensity. The SER
spectra of 1-propanethiolate on the silver island film were
obtained at various surface temperatures between 15 and 300
K. The area of a band was evaluated as its intensity. To
compensate for the minor temperature-dependent variation in
conformational equilibrium, the intensities of the CS stretching
vibrations of the G (630 cm-1) and T (702 cm-1) forms were
summed. For other modes, only the T form bands were
considered because correction for the conformational equilibrium
was found not to be important. The solid curve in Figure 4a
shows the temperature-dependent variation of the CS stretching
vibration obtained at 514.5 nm excitation. This trace represents
one batch of experiments in which the surface temperature was
lowered from 300 to 15 K and then increased back to 300 K.
Concomitant increase and decrease of the band intensity are
noticeable. More important is the fact that the SERS intensity
was restored to the original value after a cooling-heating cycle,
which was confirmed by many duplicate experiments. Minor
intensity fluctuation was not reproducible. Similar reversible
temperature dependence was also observed at 632.8 nm excita-
tion as shown in the same figure. To check the mode specificity
of the temperature dependence, similar studies have been done
for other vibrations also. Figure 4b compares the temperature
Vibrational relaxation of molecular adsorbates on metal
surfaces has been much investigated over the years,21-28 even
though most of the experimental and theoretical studies have
been focused on diatomic species such as CO on the well-
defined crystal surfaces. Temperature dependence of the
position and width of a vibrational band is thought to be an
especially useful probe in this study. For example, vibrational
energy relaxation via excitation of electron-hole pairs, which
is important for high-frequency vibrations, results in temperature
independence of vibrational bands.21-23 Temperature-dependent
change in width but not in position is taken as the evidence for
the vibrational energy relaxation via multiphonon emission.23,24
On the other hand, vibrational dephasing by way of anharmonic
coupling to a low-frequency mode which undergoes resonant
energy exchange with substrate modes is suspected when both
the position and width show noticeable temperature depen-
dence.25,26 Other relaxation processes are also possible for
polyatomic adsorbates such as intramolecular dephasing and
energy relaxation.29-31 In the case of 1-propanethiolate ad-
sorbed on the silver island film investigated here, it is possible
that the inhomogeneous broadening arising from the surface
inhomogeneity especially at the adsorption site also contributes
to the overall band broadening. Hence, some understanding
on the inhomogeneous broadening would be required before
detailed studies on vibrational relaxation are carried out by
analyzing SERS band shapes. It seems to be clear, however,
that the major relaxation mechanism for the CS stretching mode
is different from those for other bands as indicated by their
different temperature dependences.

7206 J. Phys. Chem. B, Vol. 102, No. 37, 1998
Pang et al.
modes affected by different mechanisms display similar tem-
perature dependence as observed. Finally, an explanation may
be searched for in the surface enhancement mechanism. One
does not have to invoke the adatom model for SERS here
because the method for preparation of silver film adopted in
this work precludes the existence of adatoms on the surface.
Also, reversibility of the temperature dependence is a strong
evidence against SERS mediated by adatoms in this case. A
charge-transfer model34,35 is not likely to provide an adequate
explanation either because various models presented so far do
not involve terms that would be affected by temperature. This
leaves the EM model for SERS as the potential source of
explanation for the observed reversible temperature dependence.
Recently, Leung and co-workers36 carried out temperature-
dependent calculation of the EM enhancement in SERS using
Gersten and Nitzan’s model5 and showed that the SERS effect
can be sustained at elevated temperatures (∼900 K). In the
classical treatment by Gersten and Nitzan, the EM enhancement
for a molecule with polarizability R located at a distance H from
a hemispheroid (semimajor axis a, semiminor axis b) with
dielectric constant ꢀ is given by
4
(1 - ꢀ)ê0Q′1(ê1)
1
R )
1 +
(2)
|
|
4
ꢀQ1(ê0) - ê0Q′1(ê0)
|1 - Γ|
Figure 4. (a) Temperature-dependent change in SERS intensity of the
CS stretching mode (G + T) of 1-propanethiol adsorbed on the silver
surface in a cooling-heating cycle of 300 K f 15 K f 300 K.
Excitation laser wavelengths are 514.5 (s) and 632.8 (‚‚‚) nm.
Normalized to the intensity at 300 K. (b) Intensity changes for the CCC
deformation (375 cm-1, ‚]‚), CH2 rocking (894 cm-1, -O-), CC
stretching (1026 cm-1, ‚-4-‚), and CS stretching (630 cm-1 + 702
cm-1, -9-) vibrations when the surface was cooled from 300 to 15
K. 514.5 nm excitation.
with
R
Γ )
+
4( fê1)3
(2n + 1)Pn(ê0) P′ (ê0) [Q′ (ê1)]2
2R[ꢀ - 1]
n
n
′
(3)
∑
f 3
ꢀP′n(ê0) Qn(ê0) - Pn(ê0) Q′n(ê0)
n
dependences of the CCC deformation (T, 375 cm-1), CS
stretching (G + T), CH2 rocking (T, 894 cm-1), and CC
stretching (T, 1026 cm-1) in a given batch of experiments. It is
seen that all the bands display virtually the same temperature
dependence. Reversibility of the temperature dependence was
also observed for all the bands, even though not shown in Figure
4b.
Here, f, ê0, and ê1 are defined as follows.
f ) (a2 - b2)1/2
ê0 ) a/f
(4)
(5)
One can think of various factors that may lead to variation
in Raman scattering intensity as the temperature changes. The
first and most obvious is the temperature-dependent change in
the surface coverage such as the thermal desorption of the
adsorbate at elevated temperature. The way the experiment was
performed (300 K-15 K-300 K) in the high-vacuum system
and reversibility of the intensity variation readily remove such
a possibility. Second, one may suspect that surface condensation
and evaporation of contaminants such as air and water may
contribute in some way to the observed intensity variation.
However, the fact that the temperature dependence looks
monotonic, not abrupt, argues against such a possibility. Third,
one may try to find a proper explanation from the intrinsic
characteristics of Raman scattering. For example, population
of the ground vibrational level increases as the temperature
decreases, which results in the increase of Raman intensity. This
is a minor effect, however, especially for the high-frequency
modes observed and can be ignored. Other than the above,
mode intensity of ordinary Raman scattering is known to be
fairly constant as long as the interaction between the mode and
the surrounding is weak. In fact, we measured the Raman
scattering intensity of the CS stretching mode of silver 1-pro-
panethiolate salt also and found it to be essentially invariant
with temperature. The observed temperature dependence is not
likely to be related to the relaxation mechanism either because
ê1 ) (a + H)/f
(6)
Pn and Qn are Legendre functions of the first and second kinds,
respectively, and Pn′ and Qn′ represent their derivatives. Equation
2 shows that the EM enhancement can be extremely large when
the following condition is satisfied for the real part of ꢀ.
ꢀQ1(ê0) - ê0Q′1(ê0) ) 0
(7)
This is the condition for a surface plasmon resonance of the
spheroid. Another requirement for a large enhancement is that
the imaginary part of ꢀ be small at the resonance frequency
satisfying eq 7.
Calculation of the surface temperature effect on the EM
enhancement requires knowledge of the temperature depend-
ences of the molecular polarizability (R) and the dielectric
constant (ꢀ) of the metal, the experimental data of which are
not available over a wide temperature range. Leung and co-
workers36 assumed that R be temperature-invariant. To calculate
ꢀ at various temperatures, the expression in the modified Drude
model presented by Ujihara was utilized.37
ꢀ ) (n + ik)2 ) 1 - ωp2/ω(ω + iωc)
Here, n and k are the index of refraction and the extinction
(8)

SERS of 1-Propanethiol Adsorbed on a Silver Film
J. Phys. Chem. B, Vol. 102, No. 37, 1998 7207
coefficient, respectively. ω is the angular frequency of the
electromagnetic wave, and ωp and ωc represent the plasmon
frequency and collision frequency of the metallic electrons,
respectively. Rearranging eq 8, the following simultaneous
equations were obtained.
ωp2
n2 - k2 ) 1 -
(9)
ω2 + ω2c
ω2pωc
ω(ω2 + ω2c)
2nk )
(10)
Small change in ωp with temperature was ignored, and the
temperature dependence of ωc was taken from Ujihara’s work.37
z4 dz
ez - 1
θD/T
ωc ) KT5
(11)
∫
0
Here θD is the Debye temperature and K is a constant. Then,
ωp and K were determined by solving eqs 9-11 with the room-
temperature data for n and k at a given ω. With this K, ωc at
higher temperature was calculated using eq 11, which, in turn,
was used together with ωp to evaluate ꢀ. Ujihara37 showed that
ωc in eq 11 is an increasing function of T, i.e., the probability
of electron-phonon scattering increases because the phonon
population increases with T. This results in the increase in the
imaginary part of the dielectric constant (eq 10) with T, leading
to decrease in the EM enhancement. More specifically, k is
nearly invariant with temperature in the visible to infrared range,
while n is an increasing function of T. The method of Leung
and co-workers is not adequate, however, to explain the factor
of 3-4 increase in the EM enhancement near absolute zero
temperature compared to that at room temperature observed in
this work. As T approaches zero, ωc in eq 11 and the imaginary
part of ꢀ in eq 10 approach zero, which results in infinite EM
enhancement when light is in resonance with the surface
plasmon frequency. Such a singular behavior at absolute zero
occurs because relaxation mechanisms other than the electron-
phonon scattering were ignored in the above approach. Hence,
a modification is needed to evaluate ꢀ at low temperature.
The absorptivity (A) of a metal can be expressed with n and
k as follows.38
Figure 5. Temperature dependences of the (a) real and (b) imaginary
parts of the dielectric function of silver metal calculated with the method
described in the text. Light wavelengths are 514.5 (s) and 632.8 (‚‚‚)
nm.
transition contribution may be the most troublesome because
visible wavelengths have been used to obtain SER spectra in
this work, which would require the use of experimental data.
Considering eqs 13 and 14 and the discussion made above, it
is proposed here to use the following expression for the
absorptivity.
z4 dz
ez - 1
θD/T
A ) A0 + KT5
(15)
∫
0
From the measurement of the absorptivity of silver at 4.2 K
reported by Biondi,41 A0 can be taken as 0.012 and 0.007 at
514.5 and 632.8 nm, respectively. The constant K in eq 15
can be determined by comparing with the absorptivity (eq 12)
calculated with the experimental n and k (n ) 0.130, k ) 3.05
at 514.5 nm42 and n ) 0.14, k ) 4.02 at 632.8 nm43) at room
temperature. With A0 and K fixed, eq 15 can be used to evaluate
A at any temperature. As was mentioned earlier, k of silver
evaluated with eqs 9-11 in the visible region was rather
insensitive to the temperature. In particular, temperature
dependence of k was very weak at room temperature and below.
Use of eq 15 in the present approach is equivalent to adding a
small zero-point term to ωc in eq 11. Sample calculations with
eqs 9-11 after adding a small zero-point term to eq 11 resulted
in a k that was not sensitive to T in the visible range. Hence,
k was assumed to be temperature-independent, and n was
evaluated as a function of temperature using eq 12. The real
and imaginary parts of ꢀ at 514.5 and 632.8 nm calculated as a
function of temperature are shown in Figure 5. It is seen in
the figure that the real part of ꢀ is negative and nearly constant
in the temperature range of 10-300 K, while the imaginary
part decreases as the temperature goes down. Namely, larger
EM enhancement is expected at lower temperature. It is to be
mentioned that addition of the zero-point term made here is in
general agreement with the modified model reported by Chiang
et al. very recently.44 In the latter work, a constant term was
4n
A )
(12)
(n + 1)2 + k2
The intraband-transition absorptivity of metals in the near-
infrared is related to the electron-phonon relaxation time (τ)
by the Mott-Zener formula.
2
ωpτ
A )
(13)
The expression for the effective relaxation time derived by
McKay and Rayne is as follows.39
1
τ
1 2
T
θD
z4 dz
5
θD/T
)
+ 4
(14)
∫
z
0
( )
[
]
τ0
5
e - 1
The 2/5 term is due to spontaneous emission of a phonon by a
photon-exited electron, while the temperature-dependent term
corresponds to phonon absorption and stimulated emission
processes. There are other contributions to the temperature-
invariant part of 1/τ and A.40 Among these, the interband-

7208 J. Phys. Chem. B, Vol. 102, No. 37, 1998
Pang et al.
Figure 6. Calculated electromagnetic enhancement for SERS at (a)
514.5 and (b) 632.8 nm excitation as a function of the a/b ratio at the
surface temperature of 10 K (s), 100 K (- - -), 200 K (‚‚‚), and 300
K (-‚-).
Figure 7. (a) Relative electromagnetic enhancement for SERS at 514.5
nm calculated as a function of temperature with the a/b ratio of 2.10
(s), 3.12 (-‚‚-), 3.22 (- - -), 3.32 (-‚-), 3.42 (‚‚‚), 3.52 (+++),
and 4.65 (4). The temperature dependence averaged over a uniform
distribution of the a/b ratio (1-6) is shown as circles (-O-).
Enhancement is normalized to the value at 300 K. (b) Similar data at
632.8 nm with the a/b ratio of 3.5 (s), 4.65 (-‚‚-), 4.75 (- - -), 4.85
(-‚-), and 4.95 (‚‚‚). The temperature dependence averaged over a
uniform distribution of the a/b ratio (1-6) is shown as circles (-O-).
added to the electron-phonon frequency according to the
Holstein model.45 In addition, contribution from the electron-
electron scattering to ωc was also considered. At room
temperature and below in the visible range, this term46 makes
a virtually constant contribution to ωc because (kBT)2 , (pω/
2π)2. Hence, the model by Chiang et al. would show a similar
functional dependence on T as the present one. Then, the main
difference between the two approaches lies in the fact that the
experimental values for A0, n, and k have been utilized here
while theoretical models in addition to the experimental data
were used by Chiang and co-workers to evaluate the temperature
dependence of ωc.
Several parameters, namely, a, b, H, and R, are needed to
calculate the EM enhancement factor (R). The molecular
polarizability of 9.73 Å3 for 1-propanethiol was taken from ref
47. An order of magnitude change in R did not seriously affect
the calculated enhancement (∼1%), however. Even though R
decreases monotonically but rapidly with H, the calculated
temperature dependence is hardly affected by the latter. Hence,
H of 1 nm was used in accordance with Leung and co-workers.36
Concerning the geometrical parameters (a and b) of the spheroid,
not the magnitudes of each parameter but the ratio (a/b) has
been found to affect the calculated enhancement. This is
because the surface plasmon resonance frequency is determined
by this ratio. In the present calculation, a was set to 50 nm, by
referring to the atomic force microscopic study of chemically
prepared silver film reported by Pan and Phillips.48 Our own
measurement of the island radius for the silver films prepared
in this work with scanning electron microscopy displayed a
broad distribution in the range of 30-80 nm. Figure 6 shows
the changes in the EM enhancements with 514.5 and 632.8 nm
excitations as functions of the a/b ratio at several surface
temperatures. The maximum enhancement occurs at the a/b
ratios of 3.32 and 4.85 with 514.5 and 632.8 nm excitations,
respectively, which correspond to the surface plasmon resonance
condition. Also, the temperature-dependent change in the EM
enhancement is more rapid at these a/b ratios than at other
values. To demonstrate this point more clearly, Figure 7 plots
the EM enhancement as a function of temperature at several
a/b ratios. It is seen that the increase in the calculated EM
enhancement as the temperature is lowered from 300 to 15 K
is larger than the experimental observation as the a/b ratio
approaches the value for the resonance condition. At the a/b
ratio of 3.32 satisfying exact resonance with 514.5 nm light,
the EM enhancement at 15 K is ∼300 times larger than that at
300 K. For a better match with the experimental data, one may
then suppose that the a/b ratio of the prepared silver island film
is a little bit different from 3.32, maybe close to 3.12 or 3.52
as indicated in Figure 7a. However, these a/b ratios result in
temperature-invariant EM enhancement (Figure 7b) at 632.8 nm
in complete disagreement with the experimental data. The
calculated temperature dependences of the EM enhancement at
632.8 nm (Figure 7b) display similar patterns, and the same
argument applies. Then, the only way to reconcile the calculated
results with the experimental ones is to suppose that the silver
film prepared chemically would consist of islands with a wide
distribution of the a/b ratio, which is very likely. We have
attempted the temperature-dependent calculations assuming a
uniform distribution of the a/b ratio over the range of 1-6. The
results at 514.5 and 632.8 nm excitations are included in parts
a and b of Figure 7, respectively. The calcuated temperature
effect on SERS from islands with a uniform shape distribution
is not as drastic as from islands with the a/b ratio satisfying the
surface plasmon resonance condition. However, the temperature
effect on the former is still larger than that in the experimental
data by more than an order of magnitude. One plausible
explanation for such a discrepancy is to suppose that the

SERS of 1-Propanethiol Adsorbed on a Silver Film
J. Phys. Chem. B, Vol. 102, No. 37, 1998 7209
References and Notes
distribution of the a/b ratio is severely asymmetrical, favoring
small values. Since the temperature dependence of the EM
enhancement gets weaker rapidly as the a/b ratio moves away
from the optimum, such a distribution can result in a reduced
temperature dependence as observed in the present work.
Preparation of silver island films with well-defined microscopic
structure and measurement of SERS activity over a wide
temperature range would be useful in this regard. Another
explanation for the above discrepancy may be found in the
possible inaccuracy of the optical data used in the calculation.
For example, a factor of 2-3 increase in the temperature-
independent term of the absorptivity (A0 in eq 15) leads to more
than an order of magnitude decrease in the EM enhancement
at low temperature. Accurate measurement of the optical
properties of the prepared silver island films over a wide
temperature range would be useful. One finally may suppose
that the mismatch reflects the deficiencies of the original EM
theory. For example, Leung and Tse49 reported that consider-
ation of a nonlocal effect50 on the spherical silver islands results
in orders of magnitude decrease of the SERS enhancement at
frequencies close to the surface plasmon frequency. It would
be interesting to see if the same effect can explain the moderate
temperature dependence observed in this work. Regardless of
the above difficulties, however, it seems that the EM model
provides a qualitative explanation for the experimental observa-
tion that the SERS intensity increases as the surface temperature
is lowered.
(1) Surface Enhanced Raman Scattering; Chang, R. K., Furtak, T. E.,
Eds.; Plenum: New York, 1982.
(2) Pockrand, I. Surface Enhanced Raman Vibrational Studies at Solid/
Gas Interfaces; Springer: Berlin, 1984.
(3) Moskovits, M. ReV. Mod. Phys. 1985, 57, 783.
(4) Otto, A.; Mrozek, I.; Grabhorn, H.; Akemann, W. J. Phys.:
Condens. Matter 1992, 4, 1143.
(5) Gersten, J.; Nitzan, A. J. Chem. Phys. 1980, 73, 3023.
(6) Kerker, M.; Wang, D.-S.; Chew, H. Appl. Opt. 1980, 19, 3373.
(7) Blatchford, C. G.; Campbell, J. R.; Creighton, J. A. Surf. Sci. 1982,
120, 435.
(8) Weitz, D. A.; Garoff, S.; Gersten, J. I.; Nitzan, A. J. Chem. Phys.
1983, 78, 5324.
(9) Hallmark, V. M.; Campion, A. J. Chem. Phys. 1986, 84, 2942.
(10) Chen, C. Y.; Burstein, E. Phys. ReV. Lett. 1980, 45, 1287.
(11) Bergman, J. G.; Chemla, D. S.; Liao, P. F.; Glass, A. M.; Pinczuk,
A.; Hart, R. M.; Olson, D. H. Opt. Lett. 1981, 6, 33.
(12) Liao, P. F.; Bergman, J. G.; Chemla, D. S.; Wokaun, A.; Melngailis,
J.; Hawryluk, A. M.; Economou, N. P. Chem. Phys. Lett. 1981, 82, 355.
(13) Pockrand, I.; Otto, A. Solid State Commun. 1981, 38, 1159.
(14) Macomber, S. H.; Furtak, T. E. Solid State Commun. 1983, 45,
267.
(15) Ha, D.-H.; Kim, J.-J. Phys. ReV. B 1988, 38, 12704.
(16) Boo, D. W.; Oh, W. S.; Kim, M. S.; Kim, K.; Lee, H.-C. Chem.
Phys. Lett. 1985, 120, 301.
(17) Tanabe, K.; Hiraishi, J. Spectrochim. Acta 1980, 36A, 341.
(18) Joo, T. H.; Kim, K.; Kim, M. S. J. Phys. Chem. 1986, 90, 5816.
(19) Oh, S. T. Ph.D. Thesis; Seoul National University, 1998.
(20) Lee, S. B.; Kim, K.; Kim, M. S.; Oh, W. S.; Lee, Y. S. J. Mol.
Struct. 1993, 296, 5.
(21) Persson, B. N. J.; Zaremba, E. Phys. ReV. B 1985, 31, 1863.
(22) Crljen, Z.; Langreth, D. C. Phys. ReV. B 1987, 35, 4224.
(23) Tobin, R. G. Surf. Sci. 1987, 183, 226.
In summary, we observed reversible increase of the SERS
intensity of 1-propanethiolate adsorbed on the silver island film,
as the surface temperature was lowered. Various factors that
can affect SERS intensity such as laser damage and change in
surface coverage could be eliminated as the possible causes for
the temperature dependence. The adatom model for SERS,
which has often been invoked to explain the temperature
dependence of SERS intensity, has been found inadequate also.
Temperature dependence of SERS intensity has been calculated
after a low-temperature modification of the model based on the
electromagnetic enhancement of SERS, which was originally
developed by Leung et al. to account for the high-temperature
SERS activity. The results are in qualitative agreement with
the experimental data. A better test of the present model, which
will also be a test for the validity and limitation of the EM
model for SERS, necessitates the use of the dielectric constant
of silver measured over a wide temperature range. Another
requirement is the temperature-dependent SERS measurement
on silver island films prepared with well-defined a/b ratio. On
such a film, the present results indicate that the SERS intensity
at cryogenic temperature can be larger by some orders of
magnitude than at room temperature when the laser frequency
is tuned to the plasmon frequency. Then, a truly gigantic surface
enhancement (1010-1012) of Raman scattering may be achieved,
which can find many practical applications.
(24) Gadzuk, J. W.; Luntz, A. C. Surf. Sci. 1984, 144, 429.
(25) Persson, B. N. J.; Ryberg, R. Phys. ReV. B 1985, 32, 3586.
(26) Persson, B. N. J.; Hoffmann, F. M. J. Electron Spectrosc. Relat.
Phenom. 1987, 45, 215.
(27) Nitzan, A.; Persson, B. N. J. J. Chem. Phys. 1985, 83, 5610.
(28) Jones, T. S.; Holloway, S.; Gadzuk, J. W. Surf. Sci. 1987, 184,
L421.
(29) Malik, I. J.; Agrawal, V. K.; Trenary, M. J. Chem. Phys. 1988, 89,
3861.
(30) Harris, A. L.; Rothberg, L.; Dubois, L. H.; Levinos, N. J.; Dhar,
L. Phys. ReV. Lett. 1990, 64, 2086.
(31) Harris, A. L.; Rothberg, L.; Dhar, L.; Levinos, N. J.; Dubois, L.
H. J. Chem. Phys. 1991, 94, 2438.
(32) Sobocinski, R. L.; Pemberton, J. E. Langmuir 1988, 4, 836.
(33) Kudelski, A.; Bukowska, J. Spectrochim. Acta 1995, 51A, 573.
(34) Adrian, F. J. J. Chem. Phys. 1982, 77, 5302.
(35) Lombardi, J. R.; Birke, R. L.; Lu, T.; Xu, J. J. Chem. Phys. 1986,
84, 4174.
(36) Leung, P. T.; Hider, M. H.; Sanchez, E. J. Phys. ReV. B 1996, 53,
12659.
(37) Ujihara, K. J. Appl. Phys. 1972, 43, 2376.
(38) Arnold, G. S. Appl. Opt. 1984, 23, 1434.
(39) McKay, J. A.; Rayne, J. A. Phys. ReV. B 1976, 13, 673.
(40) Sparks, M.; Loh Jr., E. J. Opt. Soc. Am. 1979, 69, 847.
(41) Biondi, M. A. Phys. ReV. 1956, 102, 964.
(42) Lynch, D. W.; Hunter, W. R. Handbook of Optical Constants of
Solids; Palik, E. D., Ed.; Academic Press: New York, 1985.
(43) Ward, L. The Optical Constants of Bulk Materials and Films, 2nd
ed.; IOP Publishing: Bristol and Philadelphia, PA 1994.
(44) Chiang, H.-P.; Leung, P. T.; Tse, W. S. J. Chem. Phys. 1998, 108,
2659.
(45) Holstein, T. Phys. ReV. 1954, 96, 535.
(46) See eq 5 of ref 44.
(47) Srivastava, S. L.; Dwivedi, D. C. Indian J. Pure Appl. Phys. 1981,
19, 1026.
(48) Pan, D.; Phillips, D. L. Chem. Phys. Lett. 1997, 275, 227.
(49) Leung, P. T.; Tse, W. S. Solid State Commun. 1995, 95, 39.
(50) Fuchs, R.; Claro, F. Phys. ReV. B 1987, 35, 3722.
Acknowledgment. This work was supported by the Basic
Science Research Fund of Ministry of Education, Republic of
Korea, and by the Specified Basic Research Fund of Korea
Science and Engineering Foundation (KOSEF).