"This is the peer reviewed version of the following article: [FULL CITE], which has been published in
final form at [Link to final article using the DOI]. This article may be used for non-commercial purposes
in accordance with Wiley Terms and Conditions for Self-Archiving."

Deep cavitand self-assembled on Au NPs-MWCNT as highly sensitive benzene sensing
interface
Pierrick Clément, Saša Korom, Claudia Struzzi, Enrique J. Parra, Carla Bittencourt, Pablo
Ballester*, Eduard Llobet*

P. Clément, Dr. E. J. Parra, Pr. E. Llobet
MINOS-EMaS, Universitat Rovira i Virgili, Avenida Paisos Catalans 26, 43007, Tarragona,
Spain
E-mail: eduard.llobet@urv.cat
S. Korom, Prof. P. Ballester
Institute of Chemical Research of Catalonia (ICIQ), Av. Països Catalans 16, 43007,
Tarragona, Spain
E-mail: pballester@iciq.es
C. Struzzi, Dr. C. Bittencourt
ChIPS, Université de Mons, Place du Parc 23, 7000 Mons, Belgium
Keywords: quinoxaline-walled deep cavitand, multi-walled carbon nanotubes, oxygen plasma
treatment, gold nanoparticle, benzene.

The unprecedented sensitivity and partial selectivity of quinoxaline-walled thioether-legged
deep cavitand functionalized multiwall carbon nanotubes toward traces of benzene vapors is
presented. The cavitand is grafted onto gold nanoparticle (Au-NP) decorated oxygen plasma
treated multiwall carbon nanotubes (O-MWCNT) by a self-assembled monolayer process
affording a product referred to as cav-Au-MWCNT. The reported technique is suitable for the
mass production of hybrid nanomaterials at low cost. The cav-Au-MWCNT resistive gas
sensor operates at room temperature and shows an outstanding performance toward traces of
benzene vapors. The detection of 2.5 ppb of benzene in dry air is demonstrated with a limit of
detection (LOD) near 600 ppt. For the first time, it is shown that a CNT nanomaterial can
effectively sense the extremely harmful benzene molecule with higher sensitivity than toluene
or o-xylene at the trace levels. The cavitand is well suited for binding benzene, which, being
in close proximity to the MWCNT, affects its density of states (DOS) shifting the Fermi level
away from the valence band. The binding of benzene is transduced in a diminution of
1

MWCNT conductance. Furthermore, the inclusion of benzene is fully reversible at room
temperature, implying that the sensor can operate at very low power consumption.

1.

Introduction
Benzene is a non-polar six-membered ring aromatic hydrocarbon with molecular

formula C6H6, with dimensions of ca. 6.0 x 3.5 Å and a volume of 120 Å3. It belongs to
the BTEX (Benzene, Toluene, Ethylbenzene and Xylene) group of compounds. The
components of this group feature similar structures but quite different toxicological
properties. Benzene is listed among the most harmful volatile organic compounds
(VOC). It has a highly flammable and toxic vapors and is recognized as a human
carcinogen by the US Environmental Protection Agency and by the European
Commission.[1] Long term exposures to relatively low concentrations of benzene over
months or years leads to severe hemotoxic effects such as aplastic anemia and
pancytopenia and to acute non-lymphocytic leukemia.[2] In the last ten years, the
permissible exposure limit has been lowered from 10 ppm to 100 ppb.[3] According to the
Directive 2008/50/EC of the European Parliament and of the Council of May 2008, the
limit value for the annual average exposure to benzene is 5 μg m-3 (1.6 ppb).[4]
Nowadays, several methods for detecting benzene traces are in use. Most of them involve
pumping of the sample and subsequent analysis by employing colorimetric detector
tubes or gas chromatography (GC-FID, GC-MS). These methods are bulky, expensive
and do not allow implementation for a continuous monitoring of benzene traces. In the
last few years, pre-concentration methods and GC equipment have been improved in
terms of miniaturization and with a LOD reaching the ppb level for benzene. [5] However,
such systems are still limited by their long response time, high power consumption and
high cost. Alternatively, the use of portable photoionization detectors (PID) has been
reported as well, but PID devices are not selective to benzene and give a total reading for
2

VOCs. The only option to make PID more selective for benzene is to utilize a single-use,
disposable and rather expensive filter at the inlet port of the device what would result in
a dramatic cost increase of running benzene measurements. Benzene is present in the
petrochemical industry, land reclamation, petroleum coke oven operators, petrol
stations, motor vehicle repair stations, roadside works and many other industries. Their
activity may result in active exposure to benzene and would clearly benefit from
affordable, portable, highly sensitive and selective detectors able to run continuous
measurements.
The fact that benzene lacks active chemical functional group(s) renders its trace
detection a challenge by employing miniaturized sensors. For a few decades, many
studies have been focused on the use of weak and reversible interactions for the
development of receptors for the selective complexation of aromatic compounds via
“host-guest” strategy. Cram et al. pioneered host-guest studies using deep cavitands
derived from resorcin[4]arene scaffolds.[6] The shallow aromatic cavity present in the
resorcin[4]arene parent compound was further elaborated by installing bridging groups
at the upper rim. Quinoxaline-bridged resorcin[4]arene cavitands are known to bind
aromatic guests (e.g. benzene, toluene, fluorobenzene), not only in the liquid, but also in
the gas phase.[7] The attractive CH- and - interactions established between the
receptor and the included aromatic compound constitute the main driving forces
responsible for the formation of inclusion complexes. Thoden et al. modified the alkyl
substituents at the lower rim of the cavitands with thioether functions in order to anchor
the receptors on a gold surface.[8] The improvements achieved in the synthesis of the
cavitands together with the possibility to deposit them as self-assembled monolayers on
different solid substrates led to the emergence of new strategies in the design of sensors
devices.[9] Consequently, several approaches for the detection of aromatic compounds
both in liquid and gas phases have been reported. The recognition event (formation of a
3

host-guest complex) was transduced in changes on optical properties (surface plasmon
resonance,[10] fluorescence spectroscopy[11]) or in mass changes in the case of resonant
devices (quartz crystal micro balance sensor,[12] or a PZT piezoelectric device[13]). For
these devices, the reported LOD for BTEX ranges from 50 to hundreds of ppm.
Resorcin[4]arene cavitands have also been used as absorbent materials in preconcentrator devices coupled to a μ-GC column for analyte separation. Furthermore,
the detection of analytes was performed by a metal oxide gas sensor or mass
spectrometer.[14] In these examples the detection limit for BTEX was found to be in the
ppb range.[15] Recently, resorcin[4]arene cavitands were covalently attached to carbon
nanotubes (CNTs) that acted as transducers[16] for conductance measurement in the
solid-liquid sensor interphase. CNTs have attracted considerable interest as
nanomaterial for sensing in solid-gas interphase.[17] They are particularly sensitive to
local chemical environment of the gas phase.[18] The functionalization of CNTs with
metal-NPs has been exploited to enhance the sensitivity of the material for benzene
sensing, reaching an LOD of about 50 ppb in dry air.[19] The metal-NP-CNT system acts
as the transduction unit of the adsorbed event in a resistive gas sensor. Indeed, the
interaction of the material with molecules of benzene in the gas phase results in an
electronic charge transfer process between the organic molecule and the metal-NP-CNT
nanomaterial. This affects the position of the Fermi energy and, hence, the conductivity
of the detection unit. The different used metals show different response towards a
variety of gas or vapor molecules, which can be used to improve the selectivity.[20]
However, the decoration of CNTs with metal nanoparticles for improving the device’s
selectivity has been implemented with limited success because these nanomaterials show
similar sensitivity to variety of aromatic compounds and heating is needed to recover the
sensor baseline.[19a]

4

Here we describe a simple experimental procedure to prepare an unprecedented type of
resistive gas sensing device that employs gold nanoparticle (Au-NP) decorated oxygenfunctionalized multiwall carbon nanotubes (O-MWCNTs). The Au-NPs are
functionalized after deposition on the O-MWCNTs with quinoxaline-walled thioetherlegged cavitand 4 (Figure 1.) leading to highly sensitive molecular recognition of benzene
vapors. Additionally, we explored the sensitivity of the sensor towards others air
pollutants such as toluene, o-xylene, carbon monoxide, nitrogen dioxide and ethanol
vapors in order to evaluate cross sensitivity. The gas sensing mechanism is discussed on
the basis of the experimental findings and the mechanism of inclusion complex
(benzene4) formation in the light of reported, structurally related systems in solutions.

Figure 1.

2.
2.1

Results and discussion
Anchoring of the cavitand on MWCNTs
The gas sensitive, hybrid nanomaterial was prepared by employing a three-step
approach. In the first step MWCNTs were treated with oxygen in plasma to create

surface oxygenated defects quoted in literature as VO2 (oxygenated vacancies) and V2O2
(oxygenated double-vacancies); these carbon nanotubes are referred to as OMWCNT.[21] In the second step, the O-MWCNTs were decorated with Au nanoparticles
by means of RF-sputtering and finally, in the third step the Au-decorated and oxygen
plasma treated MWCNTs (referred to as Au-MWCNT) were functionalized with a selfassembled monolayer of cavitand 4 to produce a desired hybrid material (referred to as
cav-Au-MWCNT). These steps are illustrated in
5

Figure 2.

The sensing properties of the cav-Au-MWCNTs when operated at room temperature
were studied by exposing them to different chemical environments (benzene, toluene, oxylene, carbon monoxide, ethanol and nitrogen dioxide) at different concentrations. For
comparison, the results obtained in the gas-sensing measurements performed on AuMWCNT sensors are also reported.
2.2

Step by step characterization

The morphology and chemical composition of the active layers were characterized using
transmission electron microscopy (TEM, Figure 3. and Figure 4.) and X-ray photoelectron
spectroscopy (XPS, Figure 5.). The homogeneous dispersion of Au nanoparticles (average
diameter  2 nm) on MWCNTs is illustrated in Figure 3. Figure 4a. and Figure 4b. show
that SAM procedure, implemented for Au-MWCNTs functionalization with the cavitand 4,
promoted aggregation of the previously deposited Au-NPs (with cluster diameters ranging
from 10 to 15 nm). The immersion of Au-MWCNTs in a chloroform solution of the cavitand
4 (0.5 mM) under mild heating (60°C), as required by the SAM technique,[17] favored both the
cavitand assembly on the Au-NPs and their aggregation on the MWCNTs. We performed
XPS analyses for O-MWCNT, Au-MWCNT, cav-Au-MWCNT and cavitand 4 samples.

Figure 3.

Figure 4.

6

In Figure 5a. we show the XPS spectra acquired for (1) O2 plasma treated MWCNTs,
(2) gold decorated O-MWCNTs and (3-4) cav-Au-MWCNTs samples. The XPS spectra
for the cavitand 4 alone (5) is included to assist in the identification of important
features in the XPS spectra of cav-Au-MWCNTs samples differing in the time used for
the construction of the SAM: 13h (3) and 24h (4). All samples contained the oxygen peak
corresponding to O1s (Figure 5a.) due to the oxygen plasma treatment experienced by
the MWCNTs. All Au-decorated samples showed the characteristic Au4d and Au4f
doublets. The presence of the cavitand 4 on cav-Au-MWCNT samples, and therefore,
the success of the functionalization protocol, was confirmed by the observation of the N1s
peak located at 399 eV. This peak is diagnostic of the presence of organic nitrogen and
appears only in the spectra of the cavitand and cav-Au-MWCNT samples (Figure 5b).
We obtained similar values of relative atomic concentrations ([N]=3% and 2% at. conc.)
for the samples treated in the SAM process for 13 and 24 h, curves (3) and (4)
respectively, indicating that 13 h are sufficient to fully functionalize the Au-MWCNTs
with cavitand 4.

Figure 5.

The resistive sensor consist of a hybrid nanomaterial, which comprises both recognition
and transducer elements, embedded in a standard electronic device. With a cavity depth
in the order of 8.3 Å, a quinoxaline-bridged cavitand 4 can completely include one
BTEX molecule and form a 1:1 host-guest complex mainly stabilized by π-π and CH-π
interactions.[23] The Au-MWCNTs are selected as an integral part of the resistive sensor.
They provide sites where the thioether groups of the cavitand 4 can be anchored. The
7

oxygen plasma treatment resulted in the presence of oxygenated defects on the outer
wall of MWCNTs. Such defects help gripping, nucleating and stabilizing Au-NPs.
Conversely, pristine carbon nanotubes shows very weak interactions with the Au-NPs by
establishing interactions with the p-orbitals of the sp2 carbons of the network.[24]
Theoretical studies have shown that Au atoms get trapped at VO2 defects (oxygenated
vacancies), which are the most abundant in oxygen plasma treated MWCNTs compared
to V2O2 defects (oxygenated di-vacancies), with a binding energy that is 0.55 eV higher
than the binding energy of Au decorated pristine MWCNT.[25] DFT calculations
indicated that the presence of oxygen atoms at the functionalized site reduces the
HOMO (Highest Occupied Molecular Orbitals)–LUMO (Lowest Unoccupied Molecular
Orbitals) gap to 0.82 eV. This reduction can be explained by the higher density of states
near the Fermi level, arising from the overlap of the 2p electrons of the O atoms and the
p electron system of the nanotube.[26] MWCNTs can either be metallic or
semiconducting depending on the axial chirality of the individual shells and depending
on the intershell interaction. A detailed description of their conductance is rather
complex, but the main contribution to charge conduction near the Fermi energy level is
given by the outer tube. Mats of MWCNTs, such as those employed in the present
experiment, consist of a mixture of metallic and semiconducting tubes. Macroscopically,
these mats behave as mild p-type semiconductors since their conductance increases or
decreases upon adsorption of electron-accepting or donating molecules, respectively.[19b,
25]

Furthermore, according to DFT calculations, the decoration with Au-NPs slightly

perturbs the band structure of MWCNTs causing a small shift of the Fermi level energy
toward lower energies, which is equivalent to a p-doping of the tubes (i.e. there is a small
electronic charge transfer from the tube to the Au-NP).[27] Finally, since the MWCNT
mat consists of defective nanotubes,[28] its resistance is mostly influenced by the
resistance of individual nanotubes and not by the inter-nanotube or the electrode8

nanotube junctions.[29] Upon formation of the host-guest complex, BTEXcavitand 4
(molecular recognition event), and by assuming the existence of a close proximity
between the walls of the cavitand 4 and the surface of the MWCNT, we are prone to
speculate that the overall resistance of the MWCNT mat will be influenced (transducer
function).
2.3

Gas sensing properties

Four cav-Au-MWCNT sensors and two Au-MWCNT sensors were employed for the
study of the gas sensing capabilities. Au-MWCNT sensors were fabricated employing
the same conditions to those used for cav-Au-MWCNT sensors, but the last step (i.e. the
functionalization with cavitand 4 via SAM) was performed just with pure chloroform to
induce the aggregation of the Au-NPs on the same level. Each measurement was
repeated at least three times. The typical response and recovery cycles of a cav-AuMWCNT sensor towards increasing concentrations of benzene in dry air are shown in
Figure 6a. The signal to noise ratio for the sensor response was high. The response and
recovery cycles were recorded while the sensor was operated at room temperature and
the sensing process was demonstrated to be reversible. During the recovery phase, pure
dry air was flown resulting in full baseline resistance recovery. The measurement period
lasted for about six months in which not significant changes in the baseline resistance
nor in sensitivity were observed. In short, the sensor response is reversible and not
affected by low-term drift. The typical calibration curves for the cav-Au-MWCNT
sensor exposed to benzene, toluene, o-xylene, ethanol and carbon monoxide are depicted
in Figure 6a. This sensor showed significantly higher sensitivity to benzene than to the
other tested pollutants. At 100 ppb level of benzene, the sensor is seven and thirty times
more responsive than to toluene or o-xylene, respectively. Furthermore, the slope of the
calibration curve (i.e. sensitivity) was calculated for the two lowest tested concentrations
9

and was found to be much higher in the case of benzene (4.2 × 10-3 % ppb-1) than for
toluene (1.1 × 10-4 % ppb-1) and o-xylene (3.3 × 10-5 % ppb-1). Therefore, the sensor
should be able to detect with high sensitivity traces of benzene in the presence of toluene
and/or o-xylene. The responses to ethanol and carbon monoxide were significantly lower.
In summary, the sensor is partially selective toward benzene and a possible strategy for
further enhancing selectivity (beyond the scope of this paper) would be to use an array
of sensors with different sensitive layers together with a pattern recognition engine.[30]
Taken together, these results demonstrated an unprecedented high sensitivity of the cavAu-MWCNT sensor for benzene. The different species whose sensing is reported in
Figure 6b. are electron donors. The overall resistance of the cav-Au-MWCNT mat
increased when exposed to any of these species. Macroscopically, this implies that our
hybrid nanomaterial retains the p-type semiconductor behavior observed for bare or
Au-MWCNT mats.

Figure 6.

As often encountered in gas sensors, sensitivity and dynamics of response can be further
enhanced by increasing the gas flow rate (Figure 7a.). When the flow is increased, layer
gradient that defines a profile of laminar flow attenuated in the boundary layer at the
sensor surface. This allowed a better diffusion of benzene towards the sensing layer, and
results in an increase of the concentration of the inclusion complexes BTEXcavitand 4
on the surface of MWCNTs mats and therefore a higher sensor response was
observed.[31] Up to 2.5 ppb of benzene in air can easily be detected at 400 mL min-1
(Figure 7b.). To the best of our knowledge, this is the first system that can detect such a
low level of benzene in the gas phase by employing a CNT-based material.
10

Figure 7.

Even at very low benzene levels, the response of the sensor was highly reproducible
(Figure 8.). Thanks to the low levels of noise in the response signals, the theoretical LOD
for the sensors was calculated to be 600 ppt, which corresponds to a sensor response
three times higher than the level of the noise.[32] The responses to benzene provided by
the Au-MWCNT sensor were tested for comparison. The Au-MWCNT sensor was not
responsive to benzene vapor levels under 60 ppb in dry conditions. Even at this
concentration (60 ppb), the response was low. This observation is in good agreement
with theoretical findings that predict a very weak binding energy and the lack of charge
transfer between the benzene molecule and the Au-MWCNTs system. These results
supported the idea that cavitand 4, in junction with Au-NPs and MWCNTs, were
necessary for the high sensitivity measured for benzene.

Figure 8.

2.4

Gas sensing mechanism

In Figure 9., we portray simplified scheme of the plausible mechanisms for benzene
sensing and recovery. We propose that the interior of cavitand 4, anchored to the
surface of cav-Au-MWCNT, is initially occupied by an air molecule (e.g. nitrogen). This
status of the system is referred to as N2cav-Au-MWCNT. The inclusion of guest in the
cavity of 4 is a dynamic process allowing a constant chemical exchange, between free
and included air molecules. When the N2cav-Au-MWCNT system is exposed to air
contaminated with benzene vapors, some of the cavitands bind a benzene molecule.
11

Cavitand 4 shows higher affinity for benzene than nitrogen owing to the establishment
of additional interactions (CH-π and π-π) between the benzene and the walls of the
cavitand. In solution, the exchange process is referred to as “hostage-exchange
mechanism” and occurs via significant conformation changes involving moving from the
vase to the kite conformation of the cavitand, a process with an energy barrier of 11.6
kcal mol-1.[33] The kite conformation of the N2cav-Au-MWCNT allows an easy access of
a guest molecule (benzene) to the shallow cavity of the cavitand and substitution of the
previously bound guest molecule (N2).[33b] Upon guest-exchange, the cavitand’s walls fold
back to vase conformation producing a new host-guest inclusion complex (referred to as
benzenecav-Au-MWCNT). In this way, two inclusion complexes are present on the
surface of cav-Au-MWCNT, one with an included nitrogen molecule and the other with
a benzene molecule (Figure 9., marked with *).

Figure 9.

To explain the transduction mechanism we hypothesized two plausible types of
communication between the cavitand and the MWCNT (Figure 10.). The first type of
communication would involve gold mediation between the π-electron cloud(s) of the
quinoxaline wall(s) from the cavitand and the π-electron clouds of MWCNT (referred to
as π-Au-π communication). The second type of communication would exclude gold
mediation and would occur between cavitand molecules located on the edge of Au-NPs
whose quinoxaline wall(s) are in direct contact with the MWCNT surface allowing π-π
stacking interaction (referred to as π-π communication). In both cases, the π-electron
density of the quinoxaline walls of the cavitand must be affected by the nature of the
included guest molecule (benzene or air molecule). Upon inclusion of a benzene molecule,
which also possesses delocalized π-electrons, additional π-π interactions can be
12

established between benzene and the cavitand wall. As a consequence, the -electron
density of the cavitand walls is modified. This electron density changes are translated
onto the π-electron clouds of the MWCNT through a charge-transfer/electron-transfer
mechanisms. This results in hole depletion and decrease in the conductance of MWCNT,
experimentally measured as an increase in sensor resistance. Conversely, the stream of
pure air modifies the equilibria between benzenecav-Au-MWCNT and N2cav-AuMWCNT. The formation of the latter inclusion complex recovers the initial state of the
sensor and therefore the baseline resistance of the sensor.

Figure 10.

The lower response displayed by the sensor (reduced increase of resistance) towards the
more electron-rich toluene and o-xylene molecules, which based on the sensing
mechanism proposed above, must imply a reduction of the electron-donating capabilities
of their inclusion complexes with the cavitand 4, is not clear to us. On the one hand,
published binding experiments demonstrated that toluene and o-xylene, in fact, bind
more strongly than benzene inside the cavitand 4, both in the liquid and gas phase.[15a,
23]

On the other hand, the dipole moments of benzene, toluene and o-xylene are 0, 0.36

and 0.64 D, respectively. We speculate that the lack of dipole moment exhibited by
benzene could be somehow responsible of the different electron donating properties
assigned to its host-guest complex (benzenecav-Au-MWCNT). Clearly, the verification
of this hypothesis must await the results of further experimental and theoretical studies
that are beyond the scope of the present work.
Nitrogen dioxide gas, belongs to the so-called NOx and is a toxic gas released from
combustion facilities and automobiles.[34] Thus, it is present in atmosphere and its effect has
to be studied on our sensors. The adsorption of NO2 was reported to occur either over Au-NP
13

or over oxygenated defects and exhibits strong response.[35],[27] The first-principle modeling of
gas adsorption on Au-MWCNT revealed that NO2, a polar molecule with partial positive
charge localized on the nitrogen atom and partial negative charge divided among oxygen
atoms, is attracted to Au-NPs by its nitrogen atom. The computed binding energies and bond
lengths showed that NO2 strongly interacts with Au-NPs and that the molecule accepted a
significant amount of electronic charge from the Au-MWCNT system. This hypothesis is
consistent with the well-known electron accepting nature of this molecule.[27] Quantum
electron transport calculations revealed that the adsorption of a single NO2 molecule on the
surface of the Au-NP resulted in a remarkable decrease of the Fermi energy level of the AuMWCNT system.[27] This is consistent with the decrease in resistance that we observed for
Au-MWCNT mats in the presence of NO2.
The response of both cav-Au-MWCNT and Au-MWCNT sensors to different
concentrations of NO2 was also studied. The calibration curve of NO2 detection for a cavAu-MWCNT and Au-MWCNT sensors operated at room temperature can be found in
the Supporting Information. The response towards NO2 displayed by cav-Au-MWCNT
sensor is about two times lower than that exhibited by the Au-MWCNT sensor. However,
NO2 cross-sensitivity of cavitand-functionalized sensors remains noticeable. The cav-AuMWCNT may remain somewhat responsive to NO2 because this molecule can be
adsorbed directly onto Au-NPs. However, the important decrease in the response
toward NO2 observed for cav-Au-MWCNTs in comparison to that of Au-MWCNTs
indicated that the presence of the cavitand significantly reduced the affinity of Au-NPs
to adsorb NO2 molecules. From a practical point of view, if the detection of benzene
should be performed in an environment in which the presence of trace levels of nitrogen
dioxide are likely, a detector comprised of two sensors, namely one cav-Au-MWCNT

14

sensor and one Au-MWCNT sensor could be used for benzene level correction. Further
details on this strategy are given in the Supporting Information.
Ambient moisture plays an important role in the response of chemiresistors. [36]
Therefore, a new set of measurements was perfomed in which trace concentrations of
benzene, toluene or o-xylene in a flow of dry synthetic air were humidified to 10%, 25%
and 60% of relative humidity (R.H.) Taking into account that the pressure was 1 atm
and the temperature 22 ºC, the relative humidity tested corresponds to levels that
ranged between 2,630 and 15,980 ppm of water, i.e. 3 to 5 orders of magnitude higher
than the concentrations of the aromatic VOCs present in the mixtures. The presence of
moisture affected the sensitivity of the cavitand-Au-MWCNT sensor toward VOCs.
Response and recovery cycles for benzene under humid conditions together with
calibration curves can be found in the Supporting Information. According to these
results, the sensitivity of the sensor toward benzene (i.e. slope of the calibration curve) is
4.20 % ppm-1 under dry conditions and changed to 1.41 % ppm-1 @ 10% R.H., 0.48 %
ppm-1 @ 25% R.H and 0.45 % ppm-1 @ 60% R.H. Taken together, these results
indicated that the presence of moisture affected significantly the limit of detection
(LOD) of the sensor for benzene. While the benzene LOD remains below 20 ppb for R.H.
up to 25%, it rises to about 50 ppb at 60% R.H. In case of toluene or o-xylene, the
presence of moisture resulted in the cav-Au-MWCNT sensor not displaying any
measurable response for both aromatic VOCs up to at concentrations of 5,000 ppb (i.e.,
the partial selectivity is not destroyed by the presence of humidity). Two possible
strategies can be exploited to overcome the loss in benzene sensitivity caused by ambient
moisture. The first option would involve dehumidification by employing inexpensive
filters containing polymers such as polyacrylate, polypyrrole or sodium polyacrylate
salts, which selectively absorb polar compounds (i.e. water) from the input gas flow
leaving non-polar compounds such as benzene unaffected. This strategy has been
15

implemented in solid-phase micro-extraction for quantitative gas or liquid
chromatography analysis of environmental samples[37] or in hand-held photo-ionization
gas.[38] The second option would involve sensor redesign taking into consideration that
the oxygen plasma treatment of MWCNTs used to generate defects at their surface for
anchoring Au nanoparticles makes CNTs more hydrophilic. Even if the employed
cavitand has a hydrophobic character, oxygenated defects present on the surface of
carbon nanotubes can be responsible for sensor moisture cross-sensitivity. Alternative
methods for achieving Au decoration of CNT sidewalls have been reported, which
accounts on the hydrophobic nature of pristine, non-defective CNTs[39] what could be
taken into advantage for reducing effect of ambient moisture to the sensor’s response.[40]

3.

Conclusion

In conclusion, a simple technique for functionalize the multiwall carbon nanotubes, in
view of designing a gas sensor with a superior performance, has been introduced. A
quinoxaline-walled thioether-legged cavitand 4 is attached onto oxygen plasma treated
Au-NP decorated MWCNTs. The technique is suitable for the mass production of
described hybrid sensing nanomaterial at low production costs, allowing cost-effective
commercialization. The cavitand-functionalized MWCNT sensor shows unprecedented
high sensitivity toward low levels of benzene in dry air at trace levels. The detection of
2.5 ppb is demonstrated experimentally and a theoretical LOD of 600 ppt was calculated.
Furthermore, sensor response towards toluene and o-xylene is significantly lower,
clearly showing the strong response for benzene, what is the first example of such
selective carbon nanotube based material. This was possible thanks to the molecular
recognition introduced by quinoxaline-bridged cavitand 4. The sensor shows significant
cross-sensitivity to NO2, but this can be compensated by combining a cav-Au-MWCNT
16

sensor with an Au-MWCNT sensor, since the latter is more sensitive to NO2 and while
insensitive to benzene levels below 60 ppb. The cav-Au-MWCNT sensor response
toward aromatic VOCs diminishes as the relative humidity in the gas flow increases. It
remains responsive to benzene traces in presence of humidity but it becomes insensitive
to toluene or o-xylene in the same range of concentrations. Possibilities to avoid the
effect of moisture could be the dehumidification of the gas flow by employing a
inexpensive filter, or keep the hydrophobic character of MWCNTs by using an
alternative route to the oxygen plasma treatment used to anchor gold nanoparticles.
Finally, it is worth mentioning that both the detection and the recovery of the baseline
are performed at room temperature, which implies that these sensors can operate at
very low power consumption. This makes the sensor suitable for being integrated in
hand-held portable analyzers, wearable detectors and semi-passive radio frequency
identification tags with sensing capabilities or in the nodes of wireless sensor networks
with a wide range of potential applications in environmental monitoring, workplace
safety or medical devices, among others.

4.

Experimental section

The cavitand 4 was synthesized following a reported procedure (Error! Reference source
not found..).[9c, 10a] The synthesis of 4 was accomplished in four steps: (1) acid catalyzed
condensation of 10-undecenal with resorcinol to yield the resorcin[4]arene framework 1;
(2) TMS protection of the OH groups in the upper rim of the resorcin[4]arene to
afforded cavitand 2; (3) addition of 1-decanethiol to the terminal alkenes of the legs of
cavitand 2 with concomitant cleavage of the TMS protecting groups afforded cavitand 3.
Finally, (4) base promoted coupling of 3 with 2,3-dichloroquinoxaline produced cavitand
4 (details on the synthesis are reported in the Support information).
17

The chemical compositions of O-MWCNT, Au-MWCNT, and cav-Au-MWCNT samples
were analyzed using X-ray photoelectron spectroscopy (XPS), VERSAPROBE PHI 5000
from Physical Electronics, equipped with a Monochromatic Al Kα X-Ray. The energy
resolution was 0.7 eV. For the compensation of built up charge on the sample surface
during the measurements, a dual beam charge neutralization composed of an electron
gun (~1 eV) and the Argon Ion gun (≤10 eV) was used.
MWCNTs were obtained from NanocylTM (3101 grade). They were grown by chemical
vapor deposition with purity higher than 95%. Carbon nanotubes were up to 1.5 μm
long and 9.5 nm in outer diameter. They underwent an oxygen plasma treatment to
clean them from amorphous carbon, to promote their dispersion in an appropriate
solvent and to create reactive sites (i.e. oxygenated vacancies, O-MWCNTs) in which
metal nanoparticles can nucleate. O-MWCNTs were suspended in chloroform (0.5%
w/w) and sonicated during 30 min to uniformly disperse them in the solution, which led
to a reproducible CNT density. The prepared suspension was air-brushed on an alumina
substrate that comprised of 10x10 mm screen-printed, Pt-interdigitated electrodes with
gap of 500 μm between electrodes. During the airbrushing, the substrate was kept at
100 °C for achieving fast evaporation of the solvent coupled with the resistance
monitoring of the device until it reached the defined value of 5 kΩ. This strategy, which
is very similar to the one reported by Zilberman et al.,[41] enables us to control both the
density of the CNT coating and the amount of O-MWCNTs deposited, ensuring device
to device reproducibility. To ensure the complete removal of the solvent and promote
adhesion of the MWCNT mat to the substrate, sensors were heated at 150 °C for 2 h.
The final resistance of thermally-treated sensors was a few hundred of ohms.
The next step was decoration of O-MWCNTs with Au-NPs achieving by RF sputtering
process conducted at 13.56 MHz with a power of 30 W under Ar plasma at room
temperature, under 0.1 Torr, for 10 s. For the formation of the self-assembled
18

monolayer of the deep cavitand, sensor substrates were immersed in a solution (20 mL)
of the cavitand (0.5 mM). The SAM process was conducted at 60 °C for a period of
maximum 24 h. These conditions allowed a reversible adsorption in order to have wellordered assembly of the quinoxaline-walled cavitand (4).[8a] Finally, the sensor
substrates were cooled to room temperature, rinsed with pure chloroform and dried at
50 °C during 30 min.
The gas sensing properties of the sensors were measured using a miniaturized Teflon
chamber (35 mL). Computer-controlled mass flow meters (Bronkhorst hi-tech 7.03.241)
and calibrated gas bottles were used (NO2, CO, ethanol, benzene, toluene, o-xylene all
diluted in pure air from Praxair), and pure air from Air Products was used to obtain
different concentrations. A continuous flow (100, 200 or 400 mL/min) was used
throughout measurements. The flow were humidified to 10%, 25% and 60% R.H. by
employing an Environics Series 4000 gas mixing system (Environics Inc., Tolland, CT,
USA). Once sensors were placed inside the test chamber, they were connected to a
multimeter interface, which allowed the real-time reading of the resistance. Sensor
response is defined as SR% = (Rgas - Rair) * 100 / Rair. Throughout the whole testing
period sensors were always operated at room temperature (25 ºC). Given the low
concentration levels tested, the measurement rig was checked to rule out the presence of
contamination in mass-flows and tubing by performing a set of control GC/MS tests
(run before, during and after gas measurements).

19

[1]

a) United States Environmental Protection Agency, Drinking Water Contaminants,

http://water.epa.gov/drink/contaminants/index.cfm#Organic, May, 2009; b) European
Commission, Air Quality Standards,
http://ec.europa.eu/environment/air/quality/standards.htm, accessed: Feb 16, 2015.
[2]

a) R. Baan, Y. Grosse, K. Straif, B. Secretan, F. El Ghissassi, V. Bouvard, L.

Benbrahim-Tallaa, N. Guha, C. Freeman, L. Galichet, The lancet oncology 2009, 10, 1143; b)
A. FassÒ, G. Arduino, Encyclopedia of Environmetrics 2012; c) D. Dougherty, S. Garte, A.
Barchowsky, J. Zmuda, E. Taioli, Toxicology letters 2008, 182, 7; d) R. E. Hester, R. M.
Harrison, Volatile organic compounds in the atmosphere, Royal Society of Chemistry, 1995;
e) A. P. DeCaprio, CRC Critical Reviews in Toxicology 1999, 29, 283.
[3]

The National Institute for Occupational Safety and Health, NIOSH Pocket guide for

chemical hazards: Benzene, www.cdc.gov/niosh/npg/npgd0049.html, accessed: Apr 4, 2011
[4]

Official Journal of the European Union,50/EC of the European Parliament and of the

Council of 21 May 2008 on ambient air quality and cleaner air for Europe,
http://ec.europa.eu/environment/air/quality/legislation/existing_leg.htm, accessed: Apr 14,
2008.
[5]

a) J. Sun, F. Guan, D. Cui, X. Chen, L. Zhang, J. Chen, Sensors and Actuators B:

Chemical 2013, 188, 513; b) C. Liaud, N. Nguyen, R. Nasreddine, S. Le Calvé, Talanta 2014,
127, 33; c) R.-S. Jian, Y.-S. Huang, S.-L. Lai, L.-Y. Sung, C.-J. Lu, Microchemical Journal
2013, 108, 161.
[6]

D. J. Cram, S. Karbach, H. E. Kim, C. B. Knobler, E. F. Maverick, J. L. Ericson, R. C.

Helgeson, Journal of the American Chemical Society 1988, 110, 2229.
[7]

a) M. Vincenti, E. Dalcanale, P. Soncini, G. Guglielmetti, Journal of the American

Chemical Society 1990, 112, 445; b) M. Vincenti, E. Dalcanale, J. Chem. Soc., Perkin Trans.
2 1995, 1069.

20

[8]

a) E. U. Thoden van Velzen, J. F. Engbersen, D. N. Reinhoudt, Journal of the

American Chemical Society 1994, 116, 3597; b) E. U. Thoden van Velzen, J. F. Engbersen, P.
J. de Lange, J. W. Mahy, D. N. Reinhoudt, Journal of the American Chemical Society 1995,
117, 6853; c) E. U. Thoden van Velzen, J. F. Engbersen, D. N. Reinhoudt, Synthesis 1995,
1995, 989.
[9]

a) F. Dickert, R. Sikorski, Materials Science and Engineering: C 1999, 10, 39; b) Y.

Yamakoshi, R. R. Schlittler, J. K. Gimzewski, F. Diederich, Journal of Materials Chemistry
2001, 11, 2895; c) V. A. Azov, P. J. Skinner, Y. Yamakoshi, P. Seiler, V. Gramlich, F.
Diederich, Helvetica chimica acta 2003, 86, 3648.
[10]

a) E. B. Feresenbet, M. Busi, F. Ugozzoli, E. Dalcanale, D. K. Shenoy, Sensor Letters

2004, 2, 186; b) R. Capan, Z. Özbek, H. Göktaş, S. Şen, F. Ince, M. Özel, G. Stanciu, F.
Davis, Sensors and Actuators B: Chemical 2010, 148, 358; c) M. Erdoğan, R. Capan, F.
Davis, Sensors and Actuators B: Chemical 2010, 145, 66; d) Y. Liu, T. Taira, M. C. Young,
D. Ajami, J. Rebek Jr, Q. Cheng, R. J. Hooley, Langmuir 2011, 28, 1391; e) A. Hassan, A.
Ray, A. Nabok, F. Davis, Sensors and Actuators B: Chemical 2001, 77, 638; f) E. B.
Feresenbet, E. Dalcanale, C. Dulcey, D. K. Shenoy, Sensors and Actuators B: Chemical 2004,
97, 211.
[11]

O. B. Berryman, A. C. Sather, J. Rebek Jr, Organic letters 2011, 13, 5232.

[12]

a) T. Weiss, K. Schierbaum, U. T. van Velzen, D. Reinhoudt, W. Göpel, Sensors and

Actuators B: Chemical 1995, 26, 203; b) J. Hartmann, J. Auge, R. Lucklum, S. Rösler, P.
Hauptmann, B. Adler, E. Dalcanale, Sensors and Actuators B: Chemical 1996, 34, 305; c) J.
Hartmann, P. Hauptmann, S. Levi, E. Dalcanale, Sensors and Actuators B: Chemical 1996, 35,
154; d) R. Paolesse, C. Di Natale, S. Nardis, A. Macagnano, A. D'Amico, R. Pinalli, E.
Dalcanale, Chemistry-A European Journal 2003, 9, 5388; e) O. Mermer, S. Okur, F. Siimer,
C. Ozbek, S. Sayin, M. Yilmaz, Acta Physica Polonica-Series A General Physics 2012, 121,
240.
21

[13]

M. Ferrari, V. Ferrari, D. Marioli, A. Taroni, M. Suman, E. Dalcanale, Sensors and

Actuators B: Chemical 2004, 103, 240.
[14]

a) F. Bianchi, A. Bedini, N. Riboni, R. Pinalli, A. Gregori, L. M. Sidisky, E. Dalcanale,

M. Careri, Analytical chemistry 2014; b) R. Pinalli, T. Barboza, F. Bianchi, C. Massera, F.
Ugozzoli, E. Dalcanale, Supramolecular Chemistry 2013, 25, 682.
[15]

a) F. Bianchi, R. Pinalli, F. Ugozzoli, S. Spera, M. Careri, E. Dalcanale, New Journal

of Chemistry 2003, 27, 502; b) S. Zampolli, I. Elmi, F. Mancarella, P. Betti, E. Dalcanale, G.
Cardinali, M. Severi, Sensors and Actuators B: Chemical 2009, 141, 322; c) S. Zampolli, P.
Betti, I. Elmi, E. Dalcanale, Chemical Communications 2007, 2790; d) F. Bianchi, M.
Mattarozzi, P. Betti, F. Bisceglie, M. Careri, A. Mangia, L. Sidisky, S. Ongarato, E. Dalcanale,
Analytical Chemistry 2008, 80, 6423.
[16]

a) M. Dionisio, J. M. Schnorr, V. K. Michaelis, R. G. Griffin, T. M. Swager, E.

Dalcanale, Journal of the American Chemical Society 2012, 134, 6540; b) F. Wang, T. M.
Swager, Journal of the American Chemical Society 2011, 133, 11181; c) L. Kong, J. Wang, F.
Meng, X. Chen, Z. Jin, M. Li, J. Liu, X.-J. Huang, Journal of Materials Chemistry 2011, 21,
11109.
[17]

J. M. Schnorr, D. van der Zwaag, J. J. Walish, Y. Weizmann, T. M. Swager, Advanced

Functional Materials 2013, 23, 5285.
[18]

E. Llobet, Sensors and Actuators B: Chemical 2013, 179, 32.

[19]

a) P. Clément, I. Hafaiedh, E. Parra, A. Thamri, J. Guillot, A. Abdelghani, E. Llobet,

Carbon 2014, 78, 510; b) R. Leghrib, A. Felten, F. Demoisson, F. Reniers, J.-J. Pireaux, E.
Llobet, Carbon 2010, 48, 3477.
[20]

R. Leghrib, E. Llobet, Analytica chimica acta 2011, 708, 19.

[21]

I. Suarez-Martinez, C. Bittencourt, X. Ke, A. Felten, J. Pireaux, J. Ghijsen, W. Drube,

G. Van Tendeloo, C. Ewels, Carbon 2009, 47, 1549.

22

[22]

D. J. Cram, M. E. Tanner, C. B. Knobler, Journal of the American Chemical Society

1991, 113, 7717.
[23]

P. Soncini, S. Bonsignore, E. Dalcanale, F. Ugozzoli, The Journal of Organic

Chemistry 1992, 57, 4608.
[24]

Y. Zhang, N. W. Franklin, R. J. Chen, H. Dai, Chemical Physics Letters 2000, 331, 35.

[25]

J.-C. Charlier, L. Arnaud, I. Avilov, M. Delgado, F. Demoisson, E. Espinosa, C. P.

Ewels, A. Felten, J. Guillot, R. Ionescu, Nanotechnology 2009, 20, 375501.
[26]

T. Prasomsri, D. Shi, D. E. Resasco, Chemical Physics Letters 2010, 497, 103.

[27]

Z. Zanolli, R. Leghrib, A. Felten, J.-J. Pireaux, E. Llobet, J.-C. Charlier, ACS Nano

2011, 5, 4592.
[28]

J.-C. Charlier, Accounts of chemical research 2002, 35, 1063.

[29]

A. Salehi-Khojin, F. Khalili-Araghi, M. A. Kuroda, K. Y. Lin, J.-P. Leburton, R. I.

Masel, ACS nano 2010, 5, 153.
[30]

a) B. Wang, J. C. Cancilla, J. S. Torrecilla, H. Haick, Nano letters 2014, 14, 933; b) G.

Konvalina, H. Haick, Accounts of chemical research 2013, 47, 66.
[31]

S. Kline, W. Reynolds, F. Schraub, P. Runstadler, Journal of Fluid Mechanics 1967,

30, 741.
[32]

D. MacDougall, W. B. Crummett, Analytical Chemistry 1980, 52, 2242.

[33]

a) D. M. Rudkevich, G. Hilmersson, J. Rebek, Journal of the American Chemical

Society 1998, 120, 12216; b) D. M. Rudkevich, G. Hilmersson, J. Rebek, Journal of the
American Chemical Society 1997, 119, 9911; c) P. J. Skinner, A. G. Cheetham, A. Beeby, V.
Gramlich, F. Diederich, Helvetica Chimica Acta 2001, 84, 2146.
[34]

United States Environmental Protection Agency, Nitrogen dioxide,

http://www.epa.gov/air/nitrogenoxides/.
[35]

J. Li, Y. Lu, Q. Ye, M. Cinke, J. Han, M. Meyyappan, Nano Letters 2003, 3, 929.

[36]

K. Balasubramanian, K. Kern, Advanced Materials 2014, 26, 1154.
23

[37]

J. Pawliszyn, Handbook of Solid State Microextraction, Elsevier, London, UK, 2012.

[38]

Humidity filtering tubes for PID detectors,

http://www.rae.nl/files/pdf/FeedsEnclosure-Humidity_Tube_II_LR_051507.pdf, accessed:
April the 30th, 2015.
[39]

T. Sainsbury, T. Ikuno, D. Okawa, D. Pacile, J. M. Frechet, A. Zettl, The Journal of

Physical Chemistry C 2007, 111, 12992.
[40]

X. Guo, Advanced Materials 2013, 25, 3397.

[41]

Y. Zilberman, U. Tisch, G. Shuster, W. Pisula, X. Feng, K. Müllen, H. Haick,

Advanced Materials 2010, 22, 4317.

24

Figure 1. Energy-minimized (MM3 as implemented in Scigress v3.0) structure of cavitand 4
in vase conformation. Included molecule is omitted and cavitand legs -(CH2)9-S-(CH2)9-CH3
are presented as methyl groups for clarity.

Figure 2. Schematic representation of the preparation of the cav-Au-MWCNT material:
(1) oxygen plasma treatment; (2) Au-RF-sputtering (formation of Au-nucleus); (3) AuRF-sputtering (growth of Au-nucleus and formation of Au-NP); (4) self-assembly of the
cavitand 4 monolayer on the Au-NP surface by dipping the material in a chloroform (S)
solution of the cavitand 4; (5) solvent removal (an air molecule, e.g. nitrogen, replaces
chloroform (S) molecule from the cavitand interior).[22] Nitrogen and chloroform
25

molecules are presented as CPK models. Hydrogen atoms are omitted for clarity.
Symbol  stands for “included in”. Note: we hypothesize that upon chloroform removal,
the cavitand legs are no longer solvated and the molecule collapses on the Au-MWCNT
surface.

Figure 3. Typical TEM image of Au-MWCNTs.

26

Figure 4. Typical TEM image of: a) cav-Au-MWCNTs and b) zoom-in of the selected area.

27

Figure 5. a) Survey spectra acquired for each step of the MWCNTs modification (1-4)
and for the cavitand 4 (5). b) a zoom-in of N1s core level spectra are plotted as acquired
from cavitand 4 (5) and cav-Au-MWCNT samples (3 and 4).

28

Figure 6. a) A typical cav-Au-MWCNT sensor resistance response in function of benzene
concentration in air; b) The relative sensor response for different gas contaminants. In all
experiments, we employed a gas flow of 200 mL min-1.

29

Figure 7. a) cav-Au-MWCNT sensor resistance response in function of benzene
concentration in air at 400 mL min-1, and b) the relative sensor response for benzene at
different gas flows.

30

Figure 8. The cav-Au-MWCNT sensor resistance response in function of two benzene
concentrations (2.5 and 20 ppb) exposed to sensor in a random order.

31

Figure 9. Proposed benzene sensing and recovery mechanism. Benzene and nitrogen
molecules are shown as CPK models. Hydrogen atoms are omitted for clarity. Notes: * crucial structures in the sensing process.

32

Figure 10. Representation of two proposed types of communication between the cavitand 4
and the Au-MWCNT. Benzene molecules are presented as CPK models. Hydrogen atoms are
omitted for clarity.

33