Investigation of the adsorption of gaseous aromatic compounds at surfaces coated with heptakis(6-thio-6-deoxy)-β-cyclodextrin by surface-enhanced Raman scattering

Article abstract of DOI:10.1021/jp990574u

Heptakis(6-thio-6-deoxy)-β-cyclodextrin (TCD) was prepared and chemisorbed at the surface of rough silver substrates for surface-enhanced Raman scattering (SERS). The enhanced adsorption of methyl orange and the stability of its azo form at the TCD-coated substrates proved the capability of the immobilized TCD to form inclusion complexes with organic guest molecules. This complexation was used for the detection of aromatics in the gas phase. Toluene, ethylbenzene, m-xylene, chlorobenzene, and m-dichlorobenzene were strongly adsorbed at the TCD-coated substrates, thus making possible the detection of low concentrations, as for example 15 ppm of m-dichlorobenzene. Benzene was weakly adsorbed and detectable in high concentrations only. Since the strongly adsorbed aromatics desorb very slowly, TCD-coated substrates may be applicable for the detection of slowly changing concentrations or as integrating sensors in the surveillance of chemical exposures. The capability of immobilized TCD to adsorb also polycyclic aromatic hydrocarbons was demonstrated using naphthalene.

Full text of DOI:10.1021/jp990574u

J. Phys. Chem. B 1999, 103, 4707-4713
4707
Investigation of the Adsorption of Gaseous Aromatic Compounds at Surfaces Coated with
Heptakis(6-thio-6-deoxy)-â-cyclodextrin by Surface-Enhanced Raman Scattering
Wieland Hill,* Vale´rie Fallourd, and Dieter Klockow
Institut fu¨r Spektrochemie und Angewandte Spektroskopie (ISAS), P.O. Box 101352, 44013 Dortmund, Germany
ReceiVed: February 16, 1999; In Final Form: March 30, 1999
Heptakis(6-thio-6-deoxy)-â-cyclodextrin (TCD) was prepared and chemisorbed at the surface of rough silver
substrates for surface-enhanced Raman scattering (SERS). The enhanced adsorption of methyl orange and
the stability of its azo form at the TCD-coated substrates proved the capability of the immobilized TCD to
form inclusion complexes with organic guest molecules. This complexation was used for the detection of
aromatics in the gas phase. Toluene, ethylbenzene, m-xylene, chlorobenzene, and m-dichlorobenzene were
strongly adsorbed at the TCD-coated substrates, thus making possible the detection of low concentrations, as
for example 15 ppm of m-dichlorobenzene. Benzene was weakly adsorbed and detectable in high concentrations
only. Since the strongly adsorbed aromatics desorb very slowly, TCD-coated substrates may be applicable
for the detection of slowly changing concentrations or as integrating sensors in the surveillance of chemical
exposures. The capability of immobilized TCD to adsorb also polycyclic aromatic hydrocarbons was
demonstrated using naphthalene.
Introduction
inclusion of aromatic compounds. Quantum chemical simula-
tions have shown that BCAT most probably adsorbs the
aromatics between the tert-butyl groups and not deep in the
cone.3 Therefore, larger cone- or cage-shaped recipients are
badly needed as sorbents for aromatics or even larger molecules.
Unfortunately, completely substituted thiol derivatives of larger
calixarenes are difficult to synthesize up to now.4
Submonomolecular layers of adsorbed molecules at rough
metal surfaces can easily be detected and characterized by
surface-enhanced Raman scattering (SERS). The SERS method
became attractive for the development of chemical sensors
because of this high sensitivity and because of recent improve-
ments in the instrumentation.
Another type of cone-shaped molecules, the cyclodextrins,
can also be used as receptors on the surface of chemical sensors.
Cyclodextrins exhibit larger cavities with different dimensions,
and their primary hydroxyls are more reactive and thus easier
to be chemically modified than the phenolic OH groups of
calixarenes. The cavities of the cyclodextrins are hydrophobic
and capable of forming inclusion complexes with a wide range
of guest compounds, such as acids, ions, halides, aliphatic,
alicyclic, and aromatic hydrocarbons.5-8 Thiolated cyclodextrins
have been immobilized on gold and silver surfaces and
characterized by SERS9 and surface plasmon spectroscopy.10
In the present paper, heptakis(6-thio-6-deoxy)-â-cyclodextrin
(TCD), a multithiolated derivative of â-cyclodextrin (CD), was
prepared, immobilized at silver substrates, characterized by
SERS, and used for the adsorption of gaseous mono- and
bicyclic aromatics.
Most chemical sensors measure properties such as refractive
index, mass, or fluorescence intensity, which are rarely specific
for the substances to be detected. Chemical selectivity of sensors
is mainly achieved by coating appropriate surfaces with certain
adsorbing or reacting substances. In case these coatings interact
with different adsorbates, the performance of the corresponding
sensors is seriously affected by cross-sensitivities. Additional
efforts such as recognition of adsorption patterns at differently
coated sensors forming an array are often necessary to improve
the selectivity. These measures increase considerably the
technical requirements and introduce new uncertainties, for
instance, drift phenomena of the signal patterns.
SERS provides vibrational spectra, which are quite charac-
teristic of the detected molecules and which can be used for
their reliable identification by comparison with reference spectra.
Therefore, SERS offers the capability for chemical multisub-
stance sensing because the interesting analytes that are adsorbed
at a coating can be identified by their vibrational spectra.
This concept has been successfully demonstrated for the
detection of aromatics in water and in the gas phase at substrates
coated with p-tert-butylcalix[4]arenetetrathiol (BCAT).1,2 The
aromatics were fully reversibly adsorbed at BCAT-coated SERS
substrates, and therefore, the SERS intensities from these
substrates gave a measure of the present bulk concentrations of
the aromatics. The limits of detection were comparatively high,
which was probably due to low adsorption enthalpies.1 The
BCAT cone is actually somewhat too narrow for an optimum
Experimental Section
Chemicals. Zinc powder (purum), methanesulfonyl chloride
(puriss., g99%), bromine (puriss.), CD (purum, g99% HPLC),
potassium thioacetate (purum, g98%), N,N-dimethylformamide
(DMF: puriss., g99.5% GC; anhydrous, H2O ) 0.01%) and
methanol (puriss., g99.5%; anhydrous, H2O ) 0.01%) were
supplied by Fluka. Chloroform (g99%) was supplied by Merck.
High-purity water with an electrical conductivity below 0.07
µS/cm was produced by a commercial water purification system
(Millipore Milli-Q) equipped with an additional filter for
organics. Potassium thioacetate was recrystallized twice in water
to remove fluorescent impurities.
* Corresponding author. Fax: +49-231-1392120. E-mail: hill@
isas-dortmund.de.
10.1021/jp990574u CCC: $18.00 © 1999 American Chemical Society
Published on Web 05/14/1999

4708 J. Phys. Chem. B, Vol. 103, No. 22, 1999
Hill et al.
Figure 1. Structure and dimensions of the TCD molecule.
described by Bello et al.14 A chromium layer with a thickness
of 2 nm was evaporated in a vacuum onto the substrates to
improve the adhesion between glass and silver. A silver layer
with a thickness of 75 nm was evaporated immediately after
the deposition of chromium. TCD-coated substrates were
generated by immersion of metal-coated slides into 0.8 mM
aqueous solutions of TCD for 2 h. After being coated with the
TCD, the substrates were rinsed with pure water, and then
immersed for 10 min in a 10 mM aqueous solution of KI, which
is known to increase the SERS intensities by an additional
“chemical” enhancement.15 After being rinsed with water and
dried, the slides were ready for use.
Preparation of Test Gas Mixtures. Schemes of the measur-
ing cell for time-resolved investigations and of the apparatus
used for test gas generation and flow control were published
elsewhere.2 Special care was taken to avoid contact of the test
gases with materials other than stainless steel and glass. Gaseous
nitrogen with 99.999% purity was produced by evaporation of
liquid nitrogen. After pressure reduction, traces of water and
organic impurities were removed by molecular sieve and
activated carbon filters. Then the carrier gas stream passed a
flow controller operating down to flow rates of 1 mL/min. A
four-port valve (Valco C4WP) allowed the gas to flow either
directly to the measuring cell or through a glass flask for
evaporation of aromatic substances. The 280 mL evaporation
flask was equipped with a septum for the injection of liquid
aromatics and contained a glass-coated magnetic stirrer for
continuous mixing of the gas. The gas concentrations were
calculated according to the ideal gas equation for the actual room
temperature of 25 °C, assuming complete evaporation of the
liquids and uniform mixing of the vapors with the carrier gas.
A soap bubble flow meter was used to measure flow rates
downstream the measuring cell. Gas lines connecting the valve,
flask, and cell had an overall length of 10 cm and an inner
diameter of 1 mm.
Figure 2. Scheme of the reaction pathway of the TCD preparation.
Chemical Preparations. TCD (Figure 1) was prepared from
CD via its bromine and thioacetate derivatives (Figure 2).
Methanesulfonyl bromide CH3SO2Br was freshly prepared from
methanesulfonyl chloride, zinc powder, and bromine as de-
scribed by Sieber.11 Composition and purity of the vacuum-
distilled product were confirmed by mass spectrometry. Heptakis-
(6-bromine-6-deoxy)-â-cyclodextrin (BrCD) was prepared from
CD and methanesulfonyl bromide according to the procedure
described by Takeo.12
The TCD was prepared from BrCD along a reaction pathway
similar to that proposed by Chung13 but with some modifications
of the procedure: 0.9 g of potassium thioacetate were dissolved
in 10 mL of DMF. This solution was added dropwise at room
temperature and under nitrogen atmosphere to a stirred solution
of 1.7 g of dry BrCD in 30 mL of DMF. After further stirring
for 24 h, the yellowish solution was poured into a stoichiometric
amount of 5% KOH in methanol and stirred for 5 h at 40 °C
under a nitrogen atmosphere. Some product precipitated during
the addition of the methanolic solution and during stirring. The
suspension was then cooled in ice water, mixed under steady
cooling with 100 mL of ice water, and neutralized with small
quantities of 2 M HCl. The resulting mixture was stored
overnight in a refrigerator and then centrifuged. The precipitate
was cleaned twice by washing with small amounts of pure water
and centrifugation. The product was dried for 5 days at room
temperature in a vacuum over P4O10. CD, BrCD, and TCD were
characterized by NMR spectroscopy.
The development of the gas concentrations after switching
the valve was controlled by placing a photoionization detector
(PID) downstream the measuring cell. A few seconds after
opening the evaporation flask, the aromatics reached the PID,
leading to about 80% of the final signal within 2 s and
developing the full signal within 5-10 s. This delay can be
neglected, considering the integration time of SERS measure-
ments of 1 min. The decrease of concentrations was also
controlled by the PID. The decay was similarly fast as the
increase, and no noticeable desorption from cell materials could
be observed. It should be mentioned that worse results were
obtained using a cell made essentially of Teflon.
Preparation of Substrates. SERS substrates were prepared
by spin-coating glass slides with 150 µL of a 5% aqueous
suspension of 0.3 µm alumina particles according to a method

Adsorption of Aromatic Compounds
J. Phys. Chem. B, Vol. 103, No. 22, 1999 4709
TABLE 1: Proton and 13C NMR Shifts (ppm) of
The low concentrations of test gases applied in the determi-
nation of the limits of detection were produced in separate glass
flasks by two dilution steps. First, 10-100 µL of a liquid
aromatic substance was injected with a microliter syringe
through a septum into a first glass flask with a volume of 1.127
L and evaporated there under constant stirring with a glass-
coated magnetic stirrer. With a second syringe proper amounts
of the gas from the first flask were injected into a second flask
with a volume of 1.124 L. The gas in the second flask was also
stirred.
â-Cyclodextrin and Its Derivatives
H1
H2
H3
H4
H5
H6a/b
3.63
4.01/3.65
3.18/2.77
SH
CD
BrCD
TCD
4.83
4.98
4.92
3.28
3.37
3.34
3.63
3.65
3.60
3.34
3.34
3.33
3.55
3.83
3.66
2.10
C1
C2
C3
C4
C5
C6
CD
BrCD
TCD
102.0
102.0
102.0
72.2
71.9
72.1
72.9
72.0
72.4
81.3
84.5
84.8
71.8
70.9
71.9
59.7
34.1
25.7
NMR Spectroscopy. All NMR measurements were carried
out on a JEOL-GX-400 NMR spectrometer equipped with a 5
mm 13C-1H dual probe-head. Measurement frequencies were
was sparingly soluble in water and soluble in pyridine and
DMF: mp ∼270-280 °C (decomp), [R]D +90° (c 1, DMF),
Beilstein test negative. The most delicate issue of the preparation
was to avoid hydrolysis. The high pH during the cleavage of
the thioester led apparently to strong hydrolysis of the cyclo-
dextrin skeleton if the reaction was performed in water as
proposed elsewhere.13 Cleavage of the thioester in anhydrous
methanol increased the yield significantly.
1
400 MHz for H and 100 MHz for 13C. Spectral widths of 5
1
and 24 kHz were used for the one-dimensional H and 13C
spectra, respectively. A quantity of 32K data points were
1
acquired for H and 13C spectra. Exponential broadenings of
0.31 and 1.47 Hz were applied to the free induction decays
(FIDs) prior to the Fourier transformation for 1H and 13C spectra,
respectively. Absolute value two-dimensional 1H-1H-correlated
(COSY) spectra and 13C-1H-correlated (HETCOR) spectra were
acquired using standard pulse sequences.
COSY. The spectral width was 2.1 kHz in both dimensions;
1024 data points in F2 and 512 data points in F1 were acquired
with subsequent zero filling to 1K in the second dimension.
An unshifted sine filter was applied in both dimensions prior
to the Fourier transformation.
Proton and 13C NMR shifts of CD, BrCD, and TCD are listed
in Table 1. The measured NMR shifts for CD are in good
agreement with literature data.16 Considerable shifts of H6 and
C6 signals after the preparations and weak influences on the
other shifts confirm the desired selective substitution of the
primary hydroxyls and the conservation of the cyclodextrin
skeleton. The location of the thiol groups is further confirmed
by the appearance of H6a/b-SH cross-peaks and the absence of
other cross-peaks of the SH signal in the 1H-1H-correlated two-
dimensional spectrum. The 13C shifts of C6 were lowered by
25.6 ppm owing to substitution of OH by Br and by 8.4 ppm
owing to substitution of Br by SH. Both values agree well with
literature data for alkanes (28.3 and 8.9 ppm, respectively17)
and thus confirm the desired type of substitution. The ratio of
the integrals of the SH signal and the anomeric (H1) proton
signal in the one-dimensional proton NMR spectrum was found
to be 1.1:1.0, which shows that every glucose unit carries an
SH substituent. The 1H NMR spectra of BrCD and TCD showed
that these samples also contained DMF with signals at 2.71 and
2.89 ppm. The integral ratio of the fairly well isolated NMR
signals of the H1 protons and the DMF signal at 2.89 ppm was,
after correction for the different numbers of protons per
molecule, 1.00:1.07. For BrCD, the analogous ratio was 1.15:
1.00. Both these findings account for a TCD/DMF ratio close
to unity. Since physisorbed DMF was certainly removed during
the vacuum-drying, there remained probably 1:1 inclusion
complexes of BrCD or TCD on the one hand and DMF used as
solvent during the preparation on the other hand. This inclusion
confirmed the preservation of complexing capabilities of the
CD after derivatization.
HETCOR. Spectral widths of 2.1 kHz for 1H and 11 kHz for
13C were used. Quantitites of 4K data points in F2 and 256 data
points in F1 were acquired with subsequent zero filling to 8K
in F2 and 512 points in F1. Prior to the Fourier transformation,
an exponential window of 2.7 Hz line broadening and an
unshifted sine window were applied in F2 and F1, respectively.
The delay times τ ) 3.57 ms and δ ) 1.80 ms were adjusted
1
to a J(C,H) coupling constant of 140 Hz.
1
Assignment of the H chemical shifts was obtained from
COSY analysis, starting with the anomeric H1 proton signal that
was easily identified because of the absence of a second COSY
correlation signal and the high 13C chemical shift value of its
HETCOR cross-peak. The HETCOR spectrum was then used
to correlate 1H assignments with 13C shifts in a straightforward
manner.
SERS Measurements. SERS spectra were taken with a
DILOR XY triple monochromator with subtracting dispersion
of the first two stages. A CW Ti:sapphire laser (Coherent 890)
operating at a wavelength of 702 nm served as the excitation
source. The laser radiation was filtered by a grating monochro-
mator for removal of spontaneous emission background. A
Wright Instruments nitrogen-cooled CCD camera with a 298
× 1152 pixel EEV 88131 chip served as a detector. The Raman
spectra were measured in backscattering geometry using f/1.0
collection optics. To prevent heating of the surface, which might
cause desorption of adsorbates or degradation of SERS sub-
strates, a cylinder lens with f ) 200 mm was used to generate
a line focus with a length of 300 µm and a width of 30 µm at
the substrate. The laser power at the sample was about 50 mW.
Raman spectra of CD, BrCD, and TCD are shown in Figure
3. The spectrum of CD exhibits CHx stretching bands around
2900 cm-1 and numerous bands in the fingerprint region, which
cannot be assigned unambiguously. The spectrum of BrCD
exhibits two strong bands at 615 and 683 cm-1, which is in the
expected range for C-Br stretching vibrations in two different
conformations.18 The spectrum of TCD does not show any
C-Br stretching bands but a strong S-H stretching band at
2574 cm-1. The band at 667 cm-1 in the TCD spectrum may
be related to the C-S stretching vibration; however, a definite
assignment is not possible, since this band is close to one of
the strongest bands of DMF being also present in the sample.
Results and Discussion
Preparation of TCD. BrCD was obtained in a yield of 75%
as a white, water-insoluble powder, which was soluble in
pyridine and DMF: mp ∼180 °C (decomp), [R]D +100° (c 1,
DMF) (lit.: mp 205-206 °C decomp, [R]D +98° (c 1,
pyridine),12 Beilstein test for halogen strongly positive. TCD
was obtained in a yield of 22% as a white-grayish powder, which
Both NMR and Raman data gave strong evidence for the
quantitative occurrence of the desired substitutions and the
successful preparation of TCD.

4710 J. Phys. Chem. B, Vol. 103, No. 22, 1999
Hill et al.
Figure 3. Raman spectra of â-cyclodextrin (CD) and its bromine-
(BrCD) and thioderivatives (TCD). The lower trace represents the SERS
spectrum of TCD adsorbed at the surface of a silver SERS substrate
(TCD/Ag). Spectra were corrected for broadband background and
shifted upward to enhance the clarity of the presentation.
Figure 5. Raman spectra of saturated MO solutions at pH ) 7 (a)
and pH ) 3 (b). SERS spectra of bare silver substrates immersed in
saturated MO solution at pH ) 7 (c) and pH ) 3 (d). SERS spectrum
of a TCD-coated substrate immersed in 20 µM MO solution at pH )
3 (e). Spectra were corrected for broadband background and shifted
vertically to enhance the clarity of the presentation.
organic molecules was tested using aqueous solutions of the
azo dye methyl orange (MO). Such azo dyes are known to
change their structure, thus resulting in marked changes of their
vibrational spectra with changing pH and also with complexation
with cyclodextrins in aqueous solutions.9 Earlier attempts to
observe SERS from mixtures of MO and Ag sols failed.9,20
Inclusion complexes of MO with 6-(2-mercaptoethylamino)-6-
deoxy-â-CD (MEACD), another CD modified by thiol groups,
immobilized at colloidal silver were detected by Maedo and
Kitano using surface-enhanced resonance Raman scattering
(SERRS).9
In the work presented here, the adsorption of MO at bare
metal substrates immersed in aqueous MO solutions was
measurable at the saturation concentration of about 0.61 mM,21
whereas practically no SERS bands of MO were detected from
a 30-fold diluted (20 µM) solution. The SERS spectra (Figure
5, spectra c and d) were quite similar to the Raman spectra of
dissolved MO (Figure 5, spectra a and b) and changed also
considerably with the pH of the solution as the Raman spectra
did. Especially, the disappearance of the strong bands of the
phenyl-Nd stretching vibration at 1143 cm-1 and the NdN
stretching vibration at 1446 cm-1 as well as the appearance of
the CdC stretching vibration band at 1620 cm-1 22 showed the
presence of the azonium form under acidic conditions in solution
and at the surface, whereas the azo form is present under neutral
conditions in the dissolved and the adsorbed state.
Figure 4. SERS intensity of the C-S stretching band of TCD at 725
cm-1 as a function of time during the coating process.
Coating of SERS Substrates with TCD. After being coated
with TCD, the substrates show SERS spectra as given in Figure
3. The complete absence of the S-H stretching band in this
spectrum confirms the expected chemisorption due to reaction
of the thiol groups with the metal surface. Obviously, the major
part of the thiol groups reacted with the silver surface, resulting
in a multiple attachment of each TCD molecule.
The 498 cm-1 band is located close to a strong unassigned
band of CD and its derivatives. Most fingerprint bands are
comparatively weak, whereas a strong C-S stretching band
appears at 725 cm-1. The appearance of a single C-S stretching
band at a position typical for aliphatic thiols in the trans
conformation19 shows that the sulfur groups were attached in a
uniform conformation.
The SERS intensity of the C-S stretching band was also
analyzed in situ during the coating process to obtain information
on the velocity of adsorption (Figure 4). More than 50% of the
final intensity was reached within 10 min, and the coating was
practically completed after half an hour. The observed curve
should represent the relative surface coverage (θ) as a function
of time (t), and it could be shown that it fits well with a rate of
adsorption proportional to the uncovered portion of the surface
(1 - θ):
θ(t) ) θmax (1 - e-t/T
)
TCD coatings on SERS substrates resulted in distinct changes
of the adsorption of MO. First, intense SERS spectra of MO
were obtained even for substrates immersed in 20 µM solutions,
where no adsorption was detectable with bare metal substrates.
Second, the SERS bands of MO at the coated substrate were
characteristic for the azo form at pH ) 7 as well as at pH ) 3
with a characteristic response time of T ) 8.7 min (curve in
Figure 4).
Inclusion Complexes at the Surface. The capability of TCD-
coated SERS substrates to form inclusion complexes with

Adsorption of Aromatic Compounds
J. Phys. Chem. B, Vol. 103, No. 22, 1999 4711
Figure 7. Time dependencies of the SERS intensities of the bands of
adsorbed m-dichlorobenzene at 997 cm-1 (*), m-xylene at 999 cm-1
(+), and benzene at 992 cm-1 (×). The intensities were normalized
with respect to temporal changes of the intensity of the TCD band at
725 cm-1. The substrates were exposed for 30 min to the following
concentrations: 9.8% benzene, 7% m-xylene, and 1.4 % m-dichloro-
benzene. The lines were calculated by fitting exponential decay curves
to the adsorption and desorption processes.
Figure 6. SERS spectra of toluene (a) and m-xylene (c) adsorbed at
TCD-coated substrates. Raman spectra of toluene (b) and m-xylene (d)
are given for comparison. Spectra were corrected for broadband
background and shifted vertically to enhance the clarity of the
presentation.
(Figure 5, spectrum e), thus indicating that no protonation took
place under acidic conditions. Third, MO was so strongly
adsorbed at TCD-coated surfaces that it was not removable by
rinsing with pure water. Rinsing with 10 mM NaCl solutions
removed the MO and resulted in substrates that were capable
of adsorbing MO again. All these observations proved that there
was a strong, but still reversible adsorption of MO with
immobilized TCD, preventing protonation of the azo moiety as
it is the case for inclusion complexes of MO with CD in
solution.9 Unlike in the investigation on MEACD-coated
substrates,9 the presence of reference spectra from uncoated
substrates allows the exclusion of influences of the metal surface
on the protonation behavior and thus provides additional
evidence for inclusional complexation at the TCD-coated
surfaces.
Adsorption of Gaseous Aromatics. SERS bands of adsorbed
aromatic substances were already observed at uncoated sub-
strates at high gas concentrations close to the vapor pressures.
At TCD-coated substrates, several of the aromatic compounds
could be detected at considerably lower concentrations (Figure
6). Enhanced adsorption at immobilized TCD was observed for
benzene, toluene, ethylbenzene, m-xylene, chlorobenzene, and
m-dichlorobenzene, whereas o- and p-isomers of xylene and
dichlorobenzene were not detectable at these substrates. Reten-
tion times at CD-coated gas chromatographic columns did not
indicate such a strong stereoselectivity of the adsorption.23
Therefore, the missing detection of o- and p-isomers may be
related to features of the SERS substrates, for example, the
presence of silver and sulfur atoms close to the cyclodextrin
cavity. The absence of SERS bands of these substances may
also be related to a flat orientation of adsorbed molecules and
the action of surface selection rules of SERS.24
the same positions as the Raman bands of the liquid sub-
stances.25 Figure 6 shows as an example the SERS spectra of
toluene and m-xylene, which exhibit the strong, broad TCD band
at 725 cm-1 and the bands of toluene at 784 and 1002 cm-1
and of m-xylene at 724 and 999 cm-1, respectively. SERS bands
of the other substances were observed at the following posi-
tions: 992 cm-1 for benzene, 1005 and 1032 cm-1 for
ethylbenzene, 1002 and 1022 cm-1 for chlorobenzene, and 998
cm-1 for m-dichlorobenzene. The positions of these SERS bands
depend clearly on the substituents so that they are suited for an
unambigious identification of the adsorbed aromatic compound
within the investigated group.
A further advantage of the TCD-coated substrates was that
the TCD SERS bands at 498 and 725 cm-1 (Figures 5e and 6)
can serve as an internal standard in the evaluation of band
intensities of the aromatic adsorbates. By this means, SERS band
intensities can be used as a measure for the surface concentration
of adsorbates that is independent of the actual SERS enhance-
ment of the respective substrate and the experimental conditions
applied.
Time Dependencies of Adsorption. Investigations of the
response of SERS intensities from aromatic adsorbates at TCD-
coated substrates on concentration changes revealed that only
benzene was desorbed completely in a short time (Figure 7),
whereas portions of the benzene derivatives remained at the
surface for considerable periods after switching to a pure
nitrogen atmosphere (see Figure 7 for m-xylene and m-
dichlorobenzene). Owing to this slow desorption, the SERS
intensities were not a pure measure of the present gas concen-
tration. However, the slow desorption may be useful for
accumulative detection purposes as required, for instance, in
exposure measurements or the detection of slowly changing low
concentrations.
The SERS spectra of TCD-coated substrates with adsorbed
aromatic compounds are apparently a pure superposition of the
SERS bands of TCD and bands of the aromatics at practically

4712 J. Phys. Chem. B, Vol. 103, No. 22, 1999
Hill et al.
TABLE 2: Limits of Detection of Gaseous Aromatic
Substances at TCD-Coated SERS Substrates after Exposure
Times of 10 min (* after 24 h Exposure)
substance
limit of detection (ppm)
benzene
m-xylene
10 000
1400
toluene
400
250
15*
not detected
not detected
chlorobenzene
m-dichlorobenzene
o- and p-dichlorobenzene
o- and p-xylene
Analyses of time dependencies at different concentrations
showed that fast and slowly desorbing adsorbate portions can
be observed at higher concentrations of the derivatives (Figure
7, m-xylene), whereas at lower concentrations the slowly
desorbing adsorbate dominates (Figure 7, m-dichlorobenzene).
Benzene and the fast desorbing adsorbates formed at higher
concentrations of the derivatives adsorbed and desorbed with
response times below 2 min. The response times of the slowly
desorbing adsorbate portion were determined by exponential
fits for m-dichlorobenzene to be 6.5 min for the adsorption at
a concentration of 1.4% and about 2 h for the desorption in the
initial period. Later, the response time increased further and
reached about 30 h after a few days. Maybe the response in the
initial period was superposed by residuals of a faster desorbing
portion of the adsorbate. The large ratio of the velocities of
adsorption and desorption at the investigated concentration
promises that clearly lower gas concentrations may be applicable
without decreasing the surface coverage clearly.
At gas concentrations close to the vapor pressures, very
similar SERS intensities were observed for the fast and the
slowly desorbing adsorbate portions of the benzene derivatives.
This points to similar numbers of the two different adsorption
sites, as would be the case with the adsorption of two aromatic
molecules per TCD cavity. One molecule may be strongly
adsorbed deep inside the cavity by interactions including one
or both substituents. Adsorption and desorption at this site may
also be slowed by an activation barrier at the opening of the
cavity. The second molecule of the derivatives and all benzene
molecules were weakly adsorbed, as indicated by the high
equilibrium gas concentrations and without a high barrier as
perceptible from the fast response. These weakly adsorbed
portions may be located closer to the rim of the cavity, at
unfavorable adsorption sites left after adsorption of a first
molecule, or they may not interact efficiently with the TCD
cavity because of missing substituents. However, a detailed
analysis of the adsorption sites and their interactions with the
aromatic molecules was not possible on the basis of the present
SERS data.
Figure 8. Concentration dependencies of the SERS intensities of the
bands of adsorbed benzene at 992 cm-1 (9), chlorobenzene at 1002
cm-1 (b), and m-dichlorobenzene at 997 cm-1 (2). The intensities were
normalized with respect to temporal changes of the intensity of the
TCD band at 725 cm-1. The lines were calculated by fitting a linear
function to the benzene data and exponential decay curves to the data
of the derivatives.
with m-dichlorobenzene, for which 15 ppm were detectable after
24 h of exposure (Table 2).
Concentration Dependencies. In principle, determinations
of gas concentration or exposure should be possible on the basis
of SERS intensities in equilibrium and measurements as a
function of time. The quantitative evaluation of these measure-
ments requires calibration curves, which may be obscured by
the presence of two adsorbate portions with different temporal
response and affinity to the surface.
The concentration curves shown in Figure 8 were taken after
successive application of pure nitrogen and stepwise increasing
concentrations of the respective substances for 5 min in each
case. The strongly adsorbed substituted aromatics were not
completely desorbed during the intermediate flushing with
nitrogen, but this procedure was the only possibility of obtaining
concentration curves in a reasonable time.
The curve for benzene increases linearly up to comparatively
high concentrations, whereas the other curves show clear
saturation effects already at low concentrations. These differ-
ences appearently correlate with the presence of the strongly
adsorbed portion. The weakly adsorbed benzene cannot saturate
the surface even at the highest concentrations applied. The
strongly adsorbed species of the chlorosubstituted aromatics
seem to cover a considerable portion of the TCD cavities already
after admission of comparatively low concentrations.
Polycyclic Aromatic Hydrocarbons (PAHs). PAHs are
rather hazardous compounds, even at very small concentrations.
Attempts to adsorb traces of PAHs from solution or the gas
phase at metal substrates and to detect them by SERS failed,
probably owing to the absence of groups attaching to the metal
surface. PAHs in methanol were detectable after evaporation
of the solvent on a SERS substrate.26
It should be noted that a fast desorption of the strongly
adsorbed portions was achieved by rinsing the substrates with
water. After drying, these substrates were again capable of
adsorbing aromatic compounds.
Limits of Detection. The limits of detection for gaseous
aromatic compounds were determined by measuring SERS
spectra after an exposure to the respective concentration for 10
min (Table 2). As expected from the absence of a strongly
adsorbed portion, benzene showed a clearly higher limit of
detection than the other compounds. Benzene derivatives were
detectable at concentrations of several 100 ppm with a distinct
effect of the substituents on the limits of detection.
Even lower limits of detection can be reached for substances
forming slowly desorbing adsorbates at the surface by the
application of longer exposure times. This was demonstrated

Adsorption of Aromatic Compounds
J. Phys. Chem. B, Vol. 103, No. 22, 1999 4713
Conclusions
TCD has been prepared from CD and chemisorbed at silver
SERS substrates by the reaction of its thiol functions with the
metal surface. The capability of surface-bound TCD to form
inclusion complexes has been proven by the adsorption of MO
and the stability of its azo form at the surface at low pH. The
TCD-coated substrates have been used for the detection and
identification of low concentrated aromatics in the gas phase.
Most detected aromatics have formed two differently adsorbed
adsorbate portions. The stronger adsorbed portion has made the
detection of concentrations in the ppm range possible but with
comparatively large response times. Experiments with naph-
thalene have shown that this polycyclic aromatic compound can
also be adsorbed at TCD-coated surfaces.
Figure 9. SERS spectrum of naphthalene at a TCD-coated substrate
(a). The Raman spectrum of solid naphthalene (b) is given for
comparison.
Acknowledgment. The authors thank Dr. J. Lambert, ISAS
Dortmund, for the measurement and analysis of NMR spectra
and Dr. B. Wehling for his interest and support. This work was
financially supported by the European Community (Network
ENVIROTRACE within the HCM program), the Ministerium
fu¨r Wissenschaft und Forschung des Landes Nordrhein-West-
falen, and the Bundesministerium fu¨r Bildung, Wissenschaft,
Forschung und Technologie, Germany.
The size of the TCD cavity seemed to be sufficient for the
inclusion of some of the smallest PAH molecules. Therefore,
first experiments for the detection of gaseous PAHs at TCD-
coated SERS substrates were made with naphthalene and
anthracene containing two and three aromatic rings, respec-
tively. Clear SERS bands of naphthalene were observed at TCD-
coated substrates after storing them for 30 min in a nitrogen
atmosphere above solid naphthalene (Figure 9), where a
concentration around 100 ppm should be present according to
the vapor pressure of the substance. Owing to the appearance
of all strong naphthalene Raman bands at practically the same
positions in the SERS spectrum, a reliable identification of the
substance was possible. The signal-to-noise ratio of the two
strongest bands suggests that even lower concentrations should
be detectable.
References and Notes
(1) Hill, W.; Wehling, B.; Gibbs, C. G.; Gutsche, C. D.; Klockow, D.
Anal. Chem. 1995, 67, 3187.
(2) Wehling, B.; Hill, W.; Klockow, D. Int. J. EnViron. Anal. Chem.,
in press.
(3) Hill, W. Unpublished results.
(4) Gutsche, C. D. Texas Christian University, Fort Worth, private
communication.
(5) Schlenk, H.; Sand, D. M. J. Am. Chem. Soc. 1961, 83, 2312.
(6) Cramer, F.; Saenger, W.; Spatz, H.-C. J. Am. Chem. Soc. 1967,
89, 14.
(7) Amado, A. M.; Moreira da Silva, A. M.; Ribeiro-Claro, P. J. A.;
Teixeiras-Dias, J. J. C. J. Raman Spectrosc. 1994, 25, 599.
Anthracene could not be detected in the atmosphere above
the solid substance, even after overnight storage. This might
be related to gas concentrations of the substance in the low ppb
range owing to a vapor pressure being about 4 orders of
magnitude smaller than that of naphthalene. Substrates previ-
ously stored above anthracene were no more capable of a
subsequent adsorption of naphthalene, whereas this capability
was recovered by rinsing with water and drying. The latter
observation points to a coverage with anthracene that did not
result in detectable SERS intensities. Possibly, these larger
molecules could not be properly included in the TCD cavities
and were, therefore, adsorbed at a larger distance from the metal
surface, where the enhancement factors can be clearly lower.
(8) Barreto, W. J.; Santos, P. S.; Rubin, J. C. Vib. Spectrosc. 1993, 6,
87.
(9) Maeda, Y.; Kitano, H. J. Phys. Chem. 1995, 99, 487.
(10) Weisser, M.; Nelles, G.; Wohlfart, P.; Wenz, G.; Mittler-Neher, S.
J. Phys. Chem. 1996, 100, 17893.
(11) Sieber, G. Liebigs Ann. Chem. 1960, 631, 180.
(12) Takeo, K. Staerke 1974, 26, 111.
(13) Chung, C. Ph.D. Thesis, Iowa State University, Ames, IA, 1991.
(14) Bello, J. M.; Stokes, D. L.; Vo-Dinh, T. Appl. Spectrosc. 1989,
43, 1325.
(15) Wehling, B.; Hill, W.; Klockow, D. J. Mol. Struct. 1995, 349, 117.
(16) Gidley, M. J.; Bociek, S. M. J. Am. Chem. Soc. 1988, 110, 3820.
(17) Kalinowski, H.-O.; Berger, S.; Braun, S. 13C-NMR-Spektroskopie;
Thieme: Stuttgart, 1984.
(18) Dollish, F. R.; Fateley, W. G.; Bentley, F. F. Characteristic Raman
Frequencies of Organic Compounds; J. Wiley & Sons: New York, 1974.
(19) Bryant, M. A.; Pemberton, J. E. J. Am. Chem. Soc. 1991, 113, 8284.
(20) Lee, P. C.; Meisel, D. J. Phys. Chem. 1982, 86, 3391.
(21) D’Ans, J.; Lax, L. Taschenbuch fu¨r Chemiker und Physiker, Vol.
II Organische Verbindungen, Springer: Berlin, 1983.
(22) Lin-Vien, D.; Colthup, B.; Fateley, W. G.; Grasselli, J. G. The
Handbook of Infrared and Raman Frequencies of Organic Molecules;
Academic Press: Boston, 1991.
(23) Ande´ra, L.; Smolkova´-Keulemansova´, E. J. Inclusion Phenom.
1987, 5, 397.
Aromatic Compounds in Water. Finally, it should be
mentioned that most investigated aromatic compounds were not
detectable at SERS substrates immersed in their aqueous
solutions. m-Dichlorobenzene was the only substance that was
slowly adsorbed and detected at a concentration of 0.4 mM.
Cyclodextrins are known to form inclusion complexes with
highly concentrated aromatics in water.27 An investigation of
the retention of various compounds at cyclodextrin-coated
columns for gas chromatography showed that water vapor
affects the inclusional complexation strongly.23 Therefore, it has
to be assumed that competitive adsorption of water prevents
the adsorption of most aromatic compounds at low concentra-
tions. This effect may be useful in the detection of gaseous
aromatics, since it reduces not only the surface coverages but
also accelerates the desorption as was shown in the earlier
investigation.23 Maybe a simultaneous or intermediate admission
of water vapor in defined concentrations would allow control
of sensitivity and response time of TCD-coated chemical SERS
sensors or regeneration of them quickly.
(24) Creighton, J. A. In AdVances in Spectroscopy; Clark, R. J. H.,
Hester, R. E., Eds.; Wiley: New York, 1988; Vol. 16, p 37.
(25) Schrader, B.; Meier, W. DMS Raman/IR Atlas of Organic Com-
pounds; Verlag Chemie: Weinheim, 1974.
(26) Matejka, P.; Stavek, J.; Volka, K.; Schrader, B. Appl. Spectrosc.
1996, 50, 409.
(27) Szejtli, J. Cyclodextrins and Their Inclusion Complexes; Akade´mia
Kiado: Budapest, 1982.