sensors

Article
A Room Temperature VOCs Gas Sensor Based on a
Layer by Layer Multi-Walled Carbon
Nanotubes/Poly-ethylene Glycol Composite
Zitao Liu 1 , Tuoyu Yang 1 , Ying Dong 1, * and Xiaohao Wang 1,2
1 Graduate School at Shenzhen, Tsinghua University, University Town of Shenzhen, Shenzhen 518055, China;
lzt17@mails.tsinghua.edu.cn (Z.L.); yty15@mails.tsinghua.edu.cn (T.Y.); xhwang@tsinghua.edu.cn (X.W.)
2 Tsinghua-Berkeley Shenzhen Institute, Tsinghua University, University Town of Shenzhen, Shenzhen 518055,
China
* Correspondence: dongy@tsinghua.edu.cn; Tel.: +86-755-2603-2505; Fax: +86-755-2603-6356




Received: 3 August 2018; Accepted: 13 September 2018; Published: 15 September 2018


Abstract: Sensitive detection of volatile organic compounds (VOCs) is significant for environmental
monitoring and medical applications. In this work, multi-walled carbon nanotubes (MWCNTs) and
polyethylene glycol (PEG) that have good adsorption for VOCs, were sprayed layer by layer on an
interdigitated electrode (IDE) to build a sensitive VOCs gas sensor. The relative resistance change
(4R/R) when the sensor was exposed to VOCs was measured. The sensor showed high sensitivity
to acetone, ethanol, isopropanol and isoprene with fast response (110 ± 5 s) and recovery (152 ± 5 s)
at room temperature, and the lower detection limit (LDL) of the sensor reached 9 ppm. With the
micro-fabricated IDE structure, the sensor can be easily built into an electric nose for VOC recognition
and measurement.

Keywords: VOC; gas sensor; PEG; MWCNTs; room temperature

1. Introduction
The detection of VOCs in the environment or in exhaled breath has received considerable attention
and is becoming increasingly important in the development of point of care testing (POCT) devices
and Internet of Things (IoT) nodes [1–7]. The development of versatile materials enables advances
in gas sensors or electric nose detection. Unique properties, including high sensing performance
sensing, portability, and low power consumption are desired in sensors material and devices design.
Metal oxide-based sensing materials have been investigated for more than 60 years and some of
them are used in commercial gas sensing products [8]. For example, ZnO, SnO2 and WO3 -based gas
sensors can detect several kinds of VOCs (e.g., isoprene, ethanol and acetone) at ppb levels [9–11].
However, these sensors still exhibit limitations such as poor selectivity and high operating temperatures
(typically within the 300 to 500 ◦ C range). This is because they are based on chemical oxidizing or
reducing reactions between oxygen ions and the analyte, while VOCs have similar reducing reaction
energies [12], and the high temperature is a necessary condition for generating oxygen ions. Hence,
these disadvantages increase the power consumption and decrease the portability.
Polymers represent a class of VOC-sensing materials that can work at room temperature (RT).
They have good absorption ability toward VOC molecules because the porous and fibrous structure of
polymers can provide a large specific surface area, which can provide more adsorption sites. It was
reported that a polymer coated silicon micro-ring resonator device achieved successful detection of
VOCs at levels as low as 100 ppb [13]. However, this kind of sensor requires a complicated system to
operate it and is thus not suitable for compact devices or IoT nodes. A chemiresistor has the advantage
of constituting a miniaturized sensing system, and conductive polymers, carbon black polymers or

Sensors 2018, 18, 3113; doi:10.3390/s18093113 www.mdpi.com/journal/sensors

Sensors 2018, 18, x FOR PEER REVIEW 2 of 11

Sensors
operate it 18,
2018, and3113is thus not suitable for compact devices or IoT nodes. A chemiresistor has 2 ofthe
11

advantage of constituting a miniaturized sensing system, and conductive polymers, carbon black
polymers
metal or metalpolymer
nanoparticle nanoparticle polymer
composites havecomposites
been applied have been applied
as sensing materials asfor
sensing materials for
chemiresistor-based
chemiresistor-based VOC sensors [14]. Carbon nanotubes (CNTs)
VOC sensors [14]. Carbon nanotubes (CNTs) are used because of their good conductivity and are used because of their good
intrinsic
conductivity
gas and intrinsic
sensing properties gas sensing
[15,16]. Polymerproperties
are chosen[15,16].
as VOC Polymer are chosen
adsorbents to enhanceas VOC the adsorbents
sensitivity of to
enhance the sensitivity of sensors. Nanotubepolymer composites are therefore
sensors. Nanotube−polymer composites are therefore interesting topics to explore because of their interesting topics to
explore because
potential gas sensingof their potential gas sensing performance.
performance.
In addition
In addition to to the
the type
type ofof sensing
sensing material, the material
material, the material assembly
assembly method
method alsoalso affects
affects the
the sensing
sensing
properties of sensors because it has a significant influence on the final topography
properties of sensors because it has a significant influence on the final topography of the sensing of the sensing
material. Various techniques such as spray-coating, drop-coating, spin-coating,
material. Various techniques such as spray-coating, drop-coating, spin-coating, dip-coating and dip-coating and
thermal evaporation
thermal evaporation have have been
been applied
applied toto add
add sensing
sensing materials.
materials. Spray-coating
Spray-coating is is easy
easy toto operate
operate andand
it can provide an ultrathin layer. The composite film is prepared by alternatively
it can provide an ultrathin layer. The composite film is prepared by alternatively spraying MWCNT spraying MWCNT
and polymer
and polymerlayers,layers, producing
producing a composite-like
a composite-like layer layer at the interface
at the interface between between the separately
the separately deposited
deposited
materials [17].materials [17].
In this
In this work,
work, aa spraying
spraying process
process was
was applied
applied to to deposit
deposit two
two separate
separate layers
layers ofof MWCNT
MWCNT and and
polyethyleneglycol
polyethylene glycol(PEG)
(PEG) aiming
aiming to produce
to produce a composite-like
a composite-like structure
structure that can that
act can act as material
as sensing sensing
material for a chemiresistive sensor that is to be operated at room temperature.
for a chemiresistive sensor that is to be operated at room temperature. A schematic of the sensor A schematic of theis
sensor is shown in Figure 1. Gas molecules can diffuse into the interior
shown in Figure 1. Gas molecules can diffuse into the interior and are adsorbed on the surface of and are adsorbed on the
surface
the of the
sensing sensing
material material
where where
it reacts it reacts
with with theThe
the material. material.
responseThe ofresponse
the sensor ofwhen
the sensor
exposedwhen to
exposed to different concentrations of four VOCs (acetone, ethanol, isopropanol
different concentrations of four VOCs (acetone, ethanol, isopropanol and isoprene) was investigated. and isoprene) was
investigated.
Gas Gas sensing characteristics
sensing characteristics such as the lower suchdetection
as the lower limit detection limit (LDL),
(LDL), response time, response time,
recovery time,
recovery time, reversibility features
reversibility features have been analyzed. have been analyzed.

Figure
Figure 1.
1. Schematic
Schematic of
of the
the sprayed
sprayed MWCNTs-PEG composite chemiresistive
MWCNTs-PEG composite chemiresistive gas
gas sensor.
sensor.

2.
2. Materials
Materials and
and Methods
Methods

2.1. Materials and Characterization

2.1. Materials and Characterization
Short multi-wall carbon nanotubes (MWCNTs), poly(acrylic acid) (PAA), polyethylenimine
Short multi-wall carbon nanotubes (MWCNTs), poly(acrylic acid) (PAA), polyethylenimine
(PEI) and polyethylene glycol (PEG), acetone, ethanol, isopropanol, isoprene and sodium dodecyl
(PEI) and polyethylene glycol (PEG), acetone, ethanol, isopropanol, isoprene and sodium dodecyl
sulfate (SDS) were purchased from Aladdin, Inc. (Shanghai, China) Morphological and structural
sulfate (SDS) were purchased from Aladdin, Inc. (Shanghai, China) Morphological and structural
characterization of the MWCNTs were carried out using Field Emission Scanning Electron Microscopy
characterization of the MWCNTs were carried out using Field Emission Scanning Electron
(SU8010, FESEM, Tokyo, Japan) and a BX53M Upright Metallurgical Microscope (Olympus, Tokyo,
Microscopy (SU8010, FESEM, Tokyo, Japan) and a BX53M Upright Metallurgical Microscope
Japan). Molecular structure characterization of MWCNTs was carried out using Raman spectrometry
(Olympus, Tokyo, Japan). Molecular structure characterization of MWCNTs was carried out using
(iHR320, HORIBA Scientific, Paris, France).
Raman spectrometry (iHR320, HORIBA Scientific, Paris, France).
2.2. Gas Sensor Fabrication
2.2. Gas Sensor Fabrication
As shown in Figure 2, the interdigitated electrode (IDE) is 1 × 3 mm in size. Each finger gap
As shown in Figure 2, the interdigitated electrode (IDE) is 1 × 3 mm in size. Each finger gap is 25
is 25 µm. This type of geometry increases the contact area between the electrodes connected by the
μm. This type of geometry increases the contact area between the electrodes connected by the sensing
sensing film. The IDE was prepared on a silicon substrate with a 500 nm insulating SiO2 surface and it
film. The IDE was prepared on a silicon substrate with a 500 nm insulating SiO2 surface and it
consisted of a 4 nm layer of Cr and a 100 nm layer of Au. The two metal layers were deposited on the
 x FOR PEER REVIEW
Sensors 2018, 18, 3113 33of
of11
11

consisted of a 4 nm layer of Cr and a 100 nm layer of Au. The two metal layers were deposited on the
surface by
surface by e-beam
e-beam evaporation
evaporation and
and patterned
patterned by
by photolithography
photolithography and lift-off processes.
and lift-off The Cr
processes. The Cr layer
layer
was
was used to improve the adhesion of Au to the silicon dioxide layer. The Au layer was used as the
used to improve the adhesion of Au to the silicon dioxide layer. The Au layer was used as the
contact layer
contact layer to
to interface
interface with
with the
the sensing
sensing material.
material. The
The work
work function
function ofof Au
Au matches
matches the
the energy
energy level
level
of the
of the MWCNTs
MWCNTs so it tends
so it tends to
to form
form an
an Ohmic
Ohmic contact
contact between
between the
the electrode
electrode and
and the
the sensing
sensing material.
material.

Fig.2
Figure Optical micrograph
2. Optical micrograph of
of the
IDEIDE.

MWCNTs was wasdispersed

dispersedinin deionized
deionized water
water by ultrasonication
by ultrasonication with with SDS for
SDS (1:10) (1:10)
4 h.for
Then4 h.
3
Then 3 wtwas
wt% PEG % PEG was dissolved
dissolved in deionized
in deionized water.
water. PAA andPAA andprepared
PEI are PEI are prepared in way
in the same the same wayThe
as PEG. as
PEG. were
IDEs The IDEs were
coated by coated by spray-coating.
spray-coating. Thebetween
The distance distancethe
between
nozzlethe nozzle
of the of the
spray penspray pen and
and substrate
substrate
is 10 cm is
and10 the
cm and the air pressure
air pressure is 0.2 is 0.2 MPa.
MPa. The coated
The coated areaarea
waswas defined
defined byby a maskand
a mask andthe
the layer
layer
thickness is controlled by the spray time, which in optimized experimental conditions is 2 s for each
layer. Then the sensor was put on a hot plate at 55 ◦°C C for about 12 h to evaporate the solvent.

2.3. Gas
2.3. Gas Sensing
Sensing Experiments
Experiments
The sensor
The sensor was was placed
placed in in aa test
test chamber
chamber (~50(~50 mLmL volume)
volume) and and its
its resistance
resistance waswas monitored
monitored
continuously through a voltage divider method. The sensor was connected
continuously through a voltage divider method. The sensor was connected to a metal resistor to a metal resistor whose
whose
resistance is
resistance is about
about 1010 times
times ofof the
the sensor’s
sensor’s resistance.
resistance. The The source
source was
was aa NI NI PXI-4132 source unit
PXI-4132 source unit
(National
(National Instruments, Austin, TX, USA) NI and the supply voltage is 5 V, and the voltage of the
Instruments, Austin, TX, USA) NI and the supply voltage is 5 V, and the voltage of the
sensor was
sensor was acquired.
acquired. The The signal
signal was
was delivered
delivered to to the
the digital
digital interface (NI, PXIe-6368)
interface (NI, PXIe-6368) connected
connected to to
PC through
PC through a shielded cable.
a shielded The data
cable. The data acquisition
acquisition waswas performed
performed by by the
the user
user software
software programmed
programmed
under LabVIEW
under LabVIEW (NI) (NI)environments
environmentswith withthe sampling
the sampling rate of 5ofHz.
rate AndAnd
5 Hz. the data werewere
the data smoothed with
smoothed
their neighboring points by the Loess method to suppress rapidly-varying
with their neighboring points by the Loess method to suppress rapidly-varying noise. noise.
To obtain
To obtain thethe sensor
sensor response,
response, thethe resistance
resistance waswas acquired
acquired as as aa reference
reference baseline
baseline before
before the
the
sensor was exposed to VOC gas. Then the sensor was cleaned with dry air for 15 min and and
sensor was exposed to VOC gas. Then the sensor was cleaned with dry air for 15 min afterafter
that
that
it wasit was
testedtested
uponupon exposure
exposure to alternating
to alternating ca. 3 ca.
min3 cycles
min cycles
of VOCof VOC
and dry andair,dryrespectively.
air, respectively.
The
The ethanol and acetone vapors were generated by passing carrier dry air
ethanol and acetone vapors were generated by passing carrier dry air through a bubbler which was through a bubbler which
was placed in°C a 25 ◦ C constant water bath and then diluted with carrier dry air. Each flow was
placed in a 25 constant water bath and then diluted with carrier dry air. Each flow was controlled
controlled
using using programmable
programmable mass flowmass flow controller
controller (MFC, Sevenstar
(MFC, Sevenstar Inc.,Beijing,
Inc.,Beijing, China). Considering
China). Considering that the
that the saturated vapor pressure of the VOC fluctuated with temperature
saturated vapor pressure of the VOC fluctuated with temperature or pressure fluctuations, a precise or pressure fluctuations,
a precise concentration
concentration of VOC was of VOC was generated
generated by dilutionbyfrom dilution from theconcentration
the standard standard concentration
VOC (100 ppm, VOC
(100
HUATEppm,GAS HUATE Inc.,GAS Inc., Shenzhen,
Shenzhen, China). China). The concentration
The concentration of the ofdiluted
the diluted
VOCVOC was was calculated
calculated by
by Equation
Equation (1): (1):
F1
Cdiluted = C0 𝐹1 (1)
C𝑑𝑖𝑙𝑢𝑡𝑒𝑑 = 𝐶01 + F2
F (1)
𝐹1 + 𝐹2
where 𝐹1 and 𝐹2 are flow rate of VOC and carrier dry air representatively, and 𝐶0 is 100 ppm here.
The gas sensor testing platform is schematically represented in Figure 3. The sensor response of each
test is calculated using Equation (2):
Sensors 2018, 18, 3113 4 of 11

Sensors
where2018, 18, x FOR
F1 and PEER
F2 are REVIEW
flow rate ofVOC and carrier dry air representatively, and C0 is 100 ppm4 here.
of 11
The gas sensor testing platform is schematically represented in Figure 3. The sensor response of each
test is calculated using Equation (2): 𝑅𝑣𝑜𝑐 − 𝑅0
𝑆(%) = × 100 (2)
𝑅0
Rvoc − R0
where 𝑅𝑣𝑜𝑐 and 𝑅0 are the resistanceS(of = sensor in ×
%)the the100
presence of diluted VOC and dry air (2)
R0
respectively. All the experiments have been carried out under room temperature and atmospheric
where Rvoc and R0 are the resistance of the sensor in the presence of diluted VOC and dry
conditions.
air respectively. All the experiments have been carried out under room temperature and
atmospheric conditions.

Schematicdiagram
Figure3.3.Schematic
Figure diagramof
ofthe
theexperimental
experimentalsetup
setupfor
forgas
gassensing
sensingmeasurements.
measurements.

3. Results and Discussion

3. Results and Discussion
3.1. Device Fabrication
3.1. Device Fabrication
The FESEM image of the as prepared MWCNTs layer is shown in Figure 4a. The short and curved
The FESEM image of the as prepared MWCNTs layer is shown in Figure 4a. The short and
nanotubes formed a porous network. The MWCNTs seem like agglomerating. This may be due to the
curved nanotubes formed a porous network. The MWCNTs seem like agglomerating. This may be
flow of the solvent evaporation. Figure 4b shows a FESEM image of the PEG/MWCNT composite
due to the flow of the solvent evaporation. Figure 4b shows a FESEM image of the PEG/MWCNT
layers. Spreading of PEG on the nanotubes network was observed.
composite layers. Spreading of PEG on the nanotubes network was observed.
Figure 4c shows a cross-section of a MWCNTs and PEG composite layer. It was confirmed that the
PEG/MWCNT composite consisted of a layer of MWCNTs and a layer of PEG, and MWCNT and PEG
were mixed at the interface. The thickness of the MWCNTs layer was about 330.5 nm and the thickness
of the PEG layer was approximately 482.3 nm. Figure 4d shows an optical microscopy image of the
PEG/MWCNTs composite layers. From FESEM and the optical microscopy image it was confirmed that
PEG was coated on the MWCNTs network. The morphology of the sensing material has an important
role in determining a sensor’s performance. The crystalline structure and amorphous structure
influence the VOC adsorption ability of material. With adsorbed VOC molecules, the nanostructure
geometry and grain size vary and alter the electrical properties. The performances of the sensor, such
as sensitivity, selectivity and response time, are directly related to the adsorption behaviors.
Sensors 2018, 18, 3113 5 of 11
Sensors 2018, 18, x FOR PEER REVIEW 5 of 11

Figure 4. FESEM images of (a) MWCNTs on silicon dioxide; (b) PEG/MWCNTs on silicon dioxide;
Figure 4. FESEM images of (a) MWCNTs on silicon dioxide; (b) PEG/MWCNTs on silicon dioxide; (c)
(c) the cross section of MWCNTs and PEG composite layer. (d) Optical microscope images of MWCNTs
the cross section of MWCNTs and PEG composite layer. (d) Optical microscope images of MWCNTs
and PEG composite layer.
and PEG composite layer.
Figure 5 shows the Raman spectra of the MWCNTs and PEG/MWCNTs on silicon dioxide.
Figure 4c shows a cross-section of a MWCNTs and PEG composite layer. It was confirmed that
Prominent peaks were observed around 1350, 1580 and 2700 cm−1 corresponding to the D, G and G0
the PEG/MWCNT composite consisted of a layer of MWCNTs and a layer of PEG, and MWCNT and
bands, respectively [18,19]. The intensity ratio of the D and G bands (ID /IG ) from the Raman spectra
PEG were mixed at the interface. The thickness of the MWCNTs layer was about 330.5 nm and the
of the MWCNTs and PEG/MWCNTs are 0.8442 and 0.8802, respectively. PEG/MWCNTs has a higher
thickness of the PEG layer was approximately 482.3 nm. Figure 4d shows an optical microscopy
ID /IG value, which indicates a PEG coating on the MWCNTs. A higher value of the ID /IG is related to
image of the PEG/MWCNTs composite layers. From FESEM and the optical microscopy image it was
a high quantity
Sensors 2018, 18, xofFOR
structural defects which provide more adsorption sites for VOCs.
PEER REVIEW 6 of 11
confirmed that PEG was coated on the MWCNTs network. The morphology of the sensing material
has an important role in determining a sensor’s performance. The crystalline structure and
amorphous structure influence the VOC adsorption ability of material. With adsorbed VOC
molecules, the nanostructure geometry and grain size vary and alter the electrical properties. The
performances of the sensor, such as sensitivity, selectivity and response time, are directly related to
the adsorption behaviors.
Figure 5 shows the Raman spectra of the MWCNTs and PEG/MWCNTs on silicon dioxide.
Prominent peaks were observed around 1350, 1580 and 2700 cm−1 corresponding to the D, G and G′
bands, respectively [18,19]. The intensity ratio of the D and G bands (ID/IG) from the Raman spectra
of the MWCNTs and PEG/MWCNTs are 0.8442 and 0.8802, respectively. PEG/MWCNTs has a higher
ID/IG value, which indicates a PEG coating on the MWCNTs. A higher value of the ID/IG is related to
a high quantity of structural defects which provide more adsorption sites for VOCs.
The atomic composition of the short MWCNTs was examined by Energy Dispersive
Spectroscopy (EDS). Samples were deposited onto silicon dioxide substrates. The results are
summarised in Table 1. Short MWCNTs have been oxidized so they contain lots of oxygen species.

Figure 5. Raman
Figure 5. Raman spectra of MWCNTs
spectra of MWCNTs andand
PEG/MWCNTs onon
PEG/MWCNTs silicon dioxide
silicon recorded
dioxide from
recorded 1000
from to to
1000
3000 −1 −1
cmcm with
3000 withcalculated
calculatedIDI/I G ratios.
D/IG

Table 1. The atomic composition of the short MWCNTs.

Spectrum Composition
C 32.14
O 67.86
Sensors 2018, 18, 3113 6 of 11

The atomic composition of the short MWCNTs was examined by Energy Dispersive Spectroscopy
(EDS). Samples were deposited onto silicon dioxide substrates. The results are summarised in Table 1.
Figure 5. Raman spectra of MWCNTs and PEG/MWCNTs on silicon dioxide recorded from 1000 to
Short MWCNTs have been oxidized so they contain lots of oxygen species.
3000 cm with calculated ID/IG ratios.
−1

Table 1. The atomic composition of the short MWCNTs.

Table 1. The atomic composition of the short MWCNTs.
Spectrum Composition
Spectrum Composition
C C 32.14 32.14
O O 67.86 67.86
TotalTotal 100 100

Figure 6 shows linear increase in current-voltage characteristic

characteristic of the composite film, which
indicates it works as a resistor.

Figure 6. Current-voltage characteristic curve of PEG/MWCNTs.

PEG/MWCNTs.

3.2. Sensor
3.2. Sensor Response
Response
The responses
The responses ofof the
the sensors
sensors with
with different
different sensitive materials to
sensitive materials to different
different concentrations of
concentrations of
acetone and
acetone and ethanol
ethanol are
are shown
shown inin Figure
Figure 7.
7. The
The sensor
sensor based
based on
on MWCNTs
MWCNTs and and PEG
PEG composite has
composite has
the highest sensitivity to the two vapors. The carboxyl group of the PAA polymer can
the highest sensitivity to the two vapors. The carboxyl group of the PAA polymer can combine withcombine with
NH33 to
NH to form
form aa new
new chemical
chemical bond.
bond. Amino
Amino functional
functional groups
groups of
of the
the PEI
PEI molecule
molecule can
can undergo
undergo aa
nucleophilic addition reaction. The PEG chain contains an oxygen atom to form
nucleophilic addition reaction. The PEG chain contains an oxygen atom to form a hydrogen a hydrogen bondbond
[20].
These properties make the gas-sensitive properties of the four polymers different. The properties
[20]. These properties make the gas-sensitive properties of the four polymers different. The properties of
PEG
of PEGareare
more likely
more to produce
likely to producehigher sensitivity
higher to VOCS.
sensitivity to VOCS.
The sensor responses of the MWCNTs and PEG/MWCNTs upon exposure to 20% to 50% diluted
VOC vapors and to water vapor at room temperature are shown in Figure 8a,b, respectively. It can
be noticed from Figure 8 that PEG/MWCNTs on fabrics has higher response and smaller shift in the
baseline when compared to naked MWCNTs. The sensor response to water vapor is obvious, implying
that water vapor should be filtered out to obtain accurate VOC measurements. The responses of the
MWCNTs and PEG/MWCNT sensors upon exposure to the four VOCs (isopropanol, acetone, ethanol
and isoprene), at different concentrations (10, 20, 50, 70, 100 ppm) at room temperature are shown
in Figure 9a,b, respectively. From Figure 9a, the linear correlation coefficient values of the fitting
curves of isopropanol, acetone, ethanol and isoprene, are 0.965, 0.980, 0.949 and 0.967, respectively.
From Figure 9b the linear correlation coefficient values are 0.981, 0.999, 0.913 and 0.997, respectively.
This result indicates that sensor works linearly at low concentrations of isopropanol, acetone and
isoprene, and it can be noticed that PEG/MWCNTs on fabrics has better linearity to isopropanol,
acetone and isoprene when compared to naked MWCNTs. From Figure 9a, the slope values of
the fitting curves of isopropanol, acetone and isoprene, are 0.0572, 0.0245 and 0.0372, respectively.
Sensors 2018, 18, 3113 7 of 11
(a) (b)

Figurevalues
The slope 7. Sensor
are responses for various
0.0672, 0.0623 sensitive
and 0.0620, layer and vapors.
respectively, The 9b.
in Figure vapors corresponding
These to the the
values represent
figures are (a) acetone; (b) ethanol.
sensitivity of the sensor to VOCs. It is clear that PEG/MWCNTs sensor has better sensitivity when
compared
Sensors 2018, to
18,naked MWCNTs.
x FOR PEER REVIEW 7 of 11
The sensor responses of the MWCNTs and PEG/MWCNTs upon exposure to 20% to 50% diluted
VOC vapors and to water vapor at room temperature are shown in Figures 8a,b, respectively. It can
be noticed from Figure 8 that PEG/MWCNTs on fabrics has higher response and smaller shift in the
baseline when compared to naked MWCNTs. The sensor response to water vapor is obvious,
implying that water vapor should be filtered out to obtain accurate VOC measurements. The
responses of the MWCNTs and PEG/MWCNT sensors upon exposure to the four VOCs (isopropanol,
acetone, ethanol and isoprene), at different concentrations (10, 20, 50, 70, 100 ppm) at room
temperature are shown in Figure 9a,b, respectively. From Figure 9a, the linear correlation coefficient
values of the fitting curves of isopropanol, acetone, ethanol and isoprene, are 0.965, 0.980, 0.949 and
0.967, respectively. From Figure 9b the linear correlation coefficient values are 0.981, 0.999, 0.913 and
0.997, respectively. This result indicates that sensor works linearly at low concentrations of
isopropanol, acetone and isoprene, and it can be noticed that PEG/MWCNTs on fabrics has better
linearity to isopropanol, acetone and isoprene when compared to naked MWCNTs. From Figure 9a,
the slope values of the fitting (a) curves of isopropanol, acetone and isoprene, (b) are 0.0572, 0.0245 and
0.0372, respectively. The slope values are 0.0672, 0.0623 and 0.0620, respectively, in Figure 9b. These
Figure 7.
Figure 7. Sensor
Sensor responses
responses forfor various
various sensitive
sensitive layer
layer and
and vapors.
vapors. The
The vapors
vapors corresponding
corresponding to
to the
the
values represent
figures are
figures are (a)
the sensitivity
(a) acetone;
acetone; (b)
of the sensor to VOCs. It is clear that PEG/MWCNTs sensor has better
(b) ethanol.
ethanol.
sensitivity when compared to naked MWCNTs.
The sensor responses of the MWCNTs and PEG/MWCNTs upon exposure to 20% to 50% diluted
VOC vapors and to water vapor at room temperature are shown in Figures 8a,b, respectively. It can
be noticed from Figure 8 that PEG/MWCNTs on fabrics has higher response and smaller shift in the
baseline when compared to naked MWCNTs. The sensor response to water vapor is obvious,
implying that water vapor should be filtered out to obtain accurate VOC measurements. The
responses of the MWCNTs and PEG/MWCNT sensors upon exposure to the four VOCs (isopropanol,
acetone, ethanol and isoprene), at different concentrations (10, 20, 50, 70, 100 ppm) at room
temperature are shown in Figure 9a,b, respectively. From Figure 9a, the linear correlation coefficient
values of the fitting curves of isopropanol, acetone, ethanol and isoprene, are 0.965, 0.980, 0.949 and
0.967, respectively. From Figure 9b the linear correlation coefficient values are 0.981, 0.999, 0.913 and
0.997, respectively. This result indicates that sensor works linearly at low concentrations of
isopropanol, acetone and isoprene, and it can be noticed that PEG/MWCNTs on fabrics has better
linearity to isopropanol, acetone(a) and isoprene when compared to naked(b) MWCNTs. From Figure 9a,
the slope
Figure values
Figure 8. of
8. Sensor the fitting
Sensor response curves
response comparison of
comparison of isopropanol,
of (a)
(a) MWCNTs acetone
MWCNTsand
and(b) and isoprene,
(b)PEG/MWCNTs
PEG/MWCNTs forforare 0.0572,
various
various 0.0245
VOCs
VOCs and and
and
0.0372,
Sensors respectively.
2018,
water
water TheREVIEW
18, x FOR PEER
vapors.
vapors. slope values are 0.0672, 0.0623 and 0.0620, respectively, in Figure 9b. These
8 of 11
values represent the sensitivity of the sensor to VOCs. It is clear that PEG/MWCNTs sensor has better
sensitivity when compared to naked MWCNTs.

9. Fitting
Figure 9.
Figure Fitting curves
curves of
of the
the (a)
(a) MWCNTs
MWCNTsand
and(b)
(b)PEG/MWCNTs
PEG/MWCNTs sensor response as aa function
function of
of
(a) (b)
VOCs
VOCs concentration
concentration (10–100
(10–100ppm).
ppm).
Figure 8. Sensor response comparison of (a) MWCNTs and (b) PEG/MWCNTs for various VOCs and
According
water to the definition, the lower detection limit (LDL) was estimated by extrapolating the
vapors.
△R/R versus concentration curve to 3 σ/S (σ is the standard deviation of R, which was 0.178 here, and
S is the slope of the fitting curve), and the LDL of the PEG/MWCNTs sensor to isopropanol, acetone
and isoprene were 7.946, 8.571 and 8.613 ppm, respectively. The calculated LDL was not accurate
enough as the gas line system was not perfectly accurate. But due to the high linear correlation
 Figure 9. Fitting curves of the (a) MWCNTs and (b) PEG/MWCNTs sensor response as a function of
Sensors 2018, 18, 3113 8 of 11
VOCs concentration (10–100 ppm).

According to the definition, the lower detection limit (LDL) was estimated by extrapolating the
4△R/R
R/Rversus
versusconcentration
concentrationcurve curve toto
3 σ/S
3 σ/S (σ is
(σthe standard
is the deviation
standard deviationof R,ofwhich
R, whichwas was
0.1780.178
here,here,
and
S is the slope of the fitting curve), and the LDL of the PEG/MWCNTs
and S is the slope of the fitting curve), and the LDL of the PEG/MWCNTs sensor to isopropanol, sensor to isopropanol, acetone
and isoprene
acetone were 7.946,
and isoprene were 8.571 and
7.946, 8.613
8.571 andppm,8.613respectively. The calculated
ppm, respectively. LDL wasLDL
The calculated not accurate
was not
enough as
accurate the gas
enough linegas
as the system was not
line system wasperfectly
not perfectlyaccurate. ButBut
accurate. due dueto to
thethehigh
highlinear
linear correlation
coefficientvalues,
coefficient values,thethecalculated
calculated LDLsLDLs
make makesense.sense.
These These measurements
measurements as shown as shown
in Figure in 8b
Figure 8b
present
apresent
shift inathe
shift in the baseline,
baseline, which influences
which influences the estimation
the estimation of the
of the lower lower detection
detection limit (LDL). limitThis
(LDL). This
baseline
baseline
shift shift is universally
is universally present for present
the gasforsensor
the gasand sensor and the influence
the influence on the LDL on the LDL is acceptable.
is acceptable. The
The results
results implied
implied that thethat the composite
composite provided provided large numbers
large numbers of active
of active sites tosites
VOC tomolecules
VOC molecules and the and the
VOC
VOC molecules
molecules can diffuse
can diffuse into the into the composite,
composite, wherewhere they induced
they induced polymer polymer
swelling,swelling,
whichwhich
resulted resulted
in the
in the conductive
conductive MWCNTs MWCNTsnetwork network being reconstructed
being reconstructed and thusandincreasing
thus increasing the contact
the contact resistance
resistance of the
of the MWCNT
MWCNT network. network.
Figure 10a shows the response recovery curve of the sensor exposed exposed to 50% diluted diluted acetone for
five cycles. Although
Althoughaabaseline
baselinedrift was
drift wasobserved,
observed, thethe
reversibility of the
reversibility response
of the is acceptable.
response The
is acceptable.
response
The of the
response sensor
of the sensorupon uponexposure
exposure to to
lowlowconcentrations
concentrationsofofVOCsVOCsisisnot notstable
stable atat the
the beginning
and then the response
response becomes
becomes stable. The response of the sensor to 100 ppm acetone dropped by
cycles, as shown
half after 100 cycles, shown in FigureFigure 10b.
10b. The introduction
introduction of dry air air allows
allows the the VOCs
VOCs adsorbed
adsorbed
on the sensitive material to be desorbed, thereby enabling reversible reversible detection
detection using using thethe gas
gas sensor.
sensor.
However, the effects of light, force, water vapor and heat may cause polymer
However, the effects of light, force, water vapor and heat may cause polymer aging, and in the process aging, and in the
process
of of detecting
detecting gases, resistance
gases, resistance heat andheat and swelling
swelling of VOCs of VOCs
and water and water
vapor arevapor are inevitable.
inevitable. In a
In a practical
application, the sensor should work in a section with good long-term stability, and the device canthe
practical application, the sensor should work in a section with good long-term stability, and be
device can
pre-aged be pre-aged
during during the
the production production process.
process.

Figure 10. (a) Reversibility

10. (a) Reversibility test
test for PEG/MWCNTs
PEG/MWCNTs upon upon exposure
exposure to
to 50% diluted acetone; (b)
Degradation of
of the sensor
sensor after
after 100
100 cycles
cycles upon
upon exposure
exposure to
to 100
100 ppm
ppm of
of acetone.
acetone.

The response and recovery time of the sensor are defined as the time taken for the sensor to
reach the 90% saturation response value and the time taken for recovery to 10%, respectively. It was
observed that the response/recovery time of the sensor was different when it was exposed to different
concentrations of VOCs. It was about 60 ± 5 and 110 ± 5 s respectively for 50% diluted acetone
vapor and this was increased to 110 ± 5 and 152 ± 5 s, respectively, for 100 ppm acetone. This may
be attributed to the lower possibility of contact between the VOC molecules and sensing material at
lower concentration.
The normalized sensor response under exposure to 100 ppm of different gases is shown in
Figure 11. This result indicates that the selectivity of the sensor is limited. This may be attributed
to the similar properties of the tested VOCs and the fact their adsorption is universal. Compared to
other MWCNT-based sensors or carbon material-based sensors reported in the literature, this work
has achieved good LDL and sensitivity at room temperature. The comparison is shown in Table 2.
Furthermore, due to the miniaturization of the sensor size and the simplicity of resistance measurement,
lower concentration.
The normalized sensor response under exposure to 100 ppm of different gases is shown in Figure
11. This result indicates that the selectivity of the sensor is limited. This may be attributed to the
similar properties of the tested VOCs and the fact their adsorption is universal. Compared to other
MWCNT-based
Sensors 2018, 18, 3113sensors or carbon material-based sensors reported in the literature, this work 9 ofhas
11
achieved good LDL and sensitivity at room temperature. The comparison is shown in Table 2.
Furthermore, due to the miniaturization of the sensor size and the simplicity of resistance
this kind of sensor
measurement, this would
kind ofbesensor
easilywould
integrated into a integrated
be easily sensor array and
into a micro-preconcentrator
a sensor array and a micro-to
achieve better discrimination and sensitivity [3,21,22].
preconcentrator to achieve better discrimination and sensitivity [3,21,22].

Figure 11. Radar plots of the response for four kinds of VOCs.

Table
Table 2. Comparisonofofthis
2. Comparison thiswork
workwith
withvarious
various carbon
carbon nanotube-based
nanotube-based gasgas sensors
sensors reported
reported in
in the
the literature.
literature.
Sensing
Sensing Operating
OperatingConcentration Sensitivity
Concentration SensitivityResp/Recov.
Resp/Recov. LDL
LDL (ppm) Ref.
Ref.
Materials (◦ C) (°C) (ppm)(ppm) (ppm−(ppm
Materials Temp.Temp.
1 ) −1
) TimeTime
(s) (s) (ppm)
PSE/MWCNTs
PSE/MWCNTs RT RT 50–50050–500 0.00007
0.00007 – -- 5050 [17]
[17]
Au/CNTs
Au/CNTs RT RT 50–80050–800 - - – -- 5050 [23]
[23]
PVA/MWCNTs
PVA/MWCNTs RT RT 100–500100–500 0.00005
0.00005 26/3026/30 100100 [24]
[24]
ZnO/Graphene 280 10–10,000
ZnO/Graphene 280 10–10,000 0.123 0.123 2/1 2/1 <10<10 [25]
[25]
PEO/MWCNTs RT 43–356 0.00022 275/300
PEO/MWCNTs RT 43–356 0.00022 275/300 4343 [26]
[26]
IO-POSS/CNTs RT 1.5–7 0.0025 -- 1.5 [27]
IO-POSS/CNTs RT 1.5–7 0.0025 – 1.5 [27]
PEG/MWCNTs RT 10–10,00 0.0006 110/152 <10 This work
PEG/MWCNTs RT 10–1000 0.0006 110/152 <10 This work

4. Conclusions
4. Conclusions
A chemiresistive VOC sensor based on MWCNTs and PEG composite layers was designed and
A chemiresistive VOC sensor based on MWCNTs and PEG composite layers was designed and
fabricated, and tested on different concentrations of four kinds of VOC (acetone, ethanol, isopropanol
fabricated, and tested on different concentrations of four kinds of VOC (acetone, ethanol, isopropanol
and isoprene). The sensor parameters such as LDL, reversibility, response and recovery time were
and isoprene). The sensor parameters such as LDL, reversibility, response and recovery time were
analyzed. This sensor has shown good sensitivity and has the advantages of operating at room
analyzed. This sensor has shown good sensitivity and has the advantages of operating at room
temperature, low power consumption and easy fabrication process. With the miniaturized sensor
temperature, low power consumption and easy fabrication process. With the miniaturized sensor size
size and the simplicity of resistance measurement, this kind of sensor would be easy to integrate into
and the simplicity of resistance measurement, this kind of sensor would be easy to integrate into a
a sensor array and it could be used for portable e-nose applications.
sensor array and it could be used for portable e-nose applications.
Author Contributions: Tuoyu Yang and Ying Dong contributed to the conception of the study and the
Author Contributions:
preparation T.Y. andZitao
of the manuscript. Y.D.Liu
contributed to Yang
and Tuoyu the conception
performedofthe
theexperiments
study and and
the preparation of and
data analysis, the
manuscript.
composed the Z.L.manuscript;
and T.Y. performed the experiments
Ying Dong and Xiaohaoand data analysis,
Wang and composed
helped perform the manuscript;
the analysis Y.D. and
with constructive
X.W. helped perform the analysis with constructive discussions.
discussions.
Funding: This work is supported by Shenzhen Science and Technology Research and Development Funds
(No: JCYJ20160411164305110 and No: JCYJ20150831192244849). The authors also acknowledge the support from
the National High Technology Research and Development Plan of China (2015AA043505).
Conflicts of Interest: The authors declare no conflicts of interest.

References
1. Boots, A.W.; Bos, L.D.; van der Schee, M.P.; van Schooten, F.J.; Sterk, P.J. Exhaled molecular fingerprinting
in diagnosis and monitoring: Validating volatile promises. Trends Mol. Med. 2015, 21, 633–644. [CrossRef]
[PubMed]
Sensors 2018, 18, 3113 10 of 11

2. Chang, J.E.; Lee, D.S.; Ban, S.W.; Oh, J.; Jung, M.Y.; Kim, S.H.; Park, S.; Persaud, K.; Jheon, S. Analysis of
volatile organic compounds in exhaled breath for lung cancer diagnosis using a sensor system. Sens. Actuators
B Chem. 2018, 255, 800–807. [CrossRef]
3. Gregis, G.; Sanchez, J.B.; Bezverkhyy, I.; Guy, W.; Berger, F.; Fierro, V.; Bellat, J.P.; Celzard, A. Detection and
quantification of lung cancer biomarkers by a micro-analytical device using a single metal oxide-based gas
sensor. Sens. Actuators B Chem. 2018, 255, 391–400. [CrossRef]
4. Greiter, M.B.; Keck, L.; Siegmund, T.; Hoeschen, C.; Oeh, U.; Paretzke, H.G. Differences in exhaled gas
profiles between patients with type 2 diabetes and healthy controls. Diabetes Technol. Ther. 2010, 12, 455–463.
[CrossRef] [PubMed]
5. Chang, Y.; Tang, N.; Qu, H.; Liu, J.; Zhang, D.; Zhang, H.; Pang, W.; Duan, X.X. Detection of volatile organic
compounds by self-assembled monolayer coated sensor array with concentration-independent fingerprints.
Sci. Rep. 2016, 6, 23970. [CrossRef] [PubMed]
6. Benammar, M.; Abdaoui, A.; Ahmad, S.H.; Touati, F.; Kadri, A. A Modular IoT Platform for Real-Time Indoor
Air Quality Monitoring. Sensors 2018, 18, 581. [CrossRef] [PubMed]
7. Szulczyński, B.; G˛ebicki, J. Currently commercially available chemical sensors employed for detection of
volatile organic compounds in outdoor and indoor air. Environments 2017, 4, 21. [CrossRef]
8. Moseley, P.T. Progress in the development of semiconducting metal oxide gas sensors: A review.
Meas. Sci. Technol. 2017, 28, 082001. [CrossRef]
9. Güntner, A.T.; Pineau, N.J.; Chie, D.; Krumeich, F.; Pratsinis, S.E. Selective sensing of isoprene by Ti-doped
ZnO for breath diagnostics. J. Mater. Chem. B 2016, 4, 5358–5366. [CrossRef]
10. Güntner, A.T.; Koren, V.; Chikkadi, K.; Righettoni, M.; Pratsinis, S.E. E-nose sensing of low-ppb formaldehyde
in gas mixtures at high relative humidity for breath screening of lung cancer? Acs Sens. 2016, 1, 528–535.
[CrossRef]
11. Righettoni, M.; Tricoli, A.; Pratsinis, S.E. Si: WO3 sensors for highly selective detection of acetone for easy
diagnosis of diabetes by breath analysis. Anal. Chem. 2010, 82, 3581–3587. [CrossRef] [PubMed]
12. Li, B.; Santhanam, S.; Schultz, L.; Jeffries-El, M.; Iovu, M.C.; Sauvé, G.; Cooper, J.; Zhang, R.; Revelli, J.C.;
Kusne, A.G.; et al. Inkjet printed chemical sensor array based on polythiophene conductive polymers.
Sens. Actuators B Chem. 2007, 123, 651–660. [CrossRef]
13. Fu, D.; Chung, J.; Liu, Q.; Raziq, R.; Kee, J.S.; Park, K.; Valiyaveettil, S.; Lee, P. Polymer coated silicon
microring device for the detection of sub-ppm volatile organic compounds. Sens. Actuators B Chem. 2018,
257, 136–142. [CrossRef]
14. Okazaki, M.; Maruyama, K.; Tsuchida, M.; Tsubokawa, N. A novel gas sensor from poly (ethylene
glycol)-grafted carbon black. Responsibility of electric resistance of poly (ethylene glycol)-grafted carbon
black against humidity and solvent vapor. Polym. J. 1999, 31, 672. [CrossRef]
15. Salehi-Khojin, A.; Khalili-Araghi, F.; Kuroda, M.A.; Lin, K.Y.; Leburton, J.P.; Masel, R.I. On the sensing
mechanism in carbon nanotube chemiresistors. ACS Nano 2010, 5, 153–158. [CrossRef] [PubMed]
16. Siegal, M.P.; Yelton, W.G.; Overmyer, D.L.; Provencio, P.P. Nanoporous carbon films for gas microsensors.
Langmuir 2004, 20, 1194–1198. [CrossRef] [PubMed]
17. Lorwongtragool, P.; Sowade, E.; Dinh, T.N.; Kanoun, O.; Kerdcharoen, T.; Baumann, R.R. Inkjet printing
of chemiresistive sensors based on polymer and carbon nanotube networks[C]//Systems. In Proceedings
of the 2012 9th International Multi-Conference on Signals and Devices (SSD), Chemnitz, Germany, 20–23
March 2012; pp. 1–4.
18. Rajavel, K.; Lalitha, M.; Radhakrishnan, J.K.; Senthilkumar, L.; Rajendra Kumar, R.T. Multiwalled carbon
nanotube oxygen sensor: Enhanced oxygen sensitivity at room temperature and mechanism of sensing.
ACS Appl. Mater. Interfaces 2015, 7, 23857–23865. [CrossRef] [PubMed]
19. Rajavel, K.; Dinesh, M.; Saranya, R.; Kumar, R.R. Enhanced vacuum sensing performance of multiwalled
carbon nanotubes: Role of defects and carboxyl functionalization. RSC Adv. 2015, 5, 20479–20485. [CrossRef]
20. Zhao, B.; Hu, H.; Yu, A.; Perea, D.; Haddon, R.C. Synthesis and characterization of water soluble
single-walled carbon nanotube graft copolymers. J. Am. Chem. Soc. 2005, 127, 8197–8203. [CrossRef]
[PubMed]
21. Jimenez-Cadena, G.; Riu, J.; Rius, F.X. Gas sensors based on nanostructured materials. Analyst 2007, 132,
1083–1099. [CrossRef] [PubMed]
Sensors 2018, 18, 3113 11 of 11

22. Lipatov, A.; Varezhnikov, A.; Wilson, P.; Sysoev, V.; Kolmakov, A.; Sinitskii, A. Highly selective gas sensor
arrays based on thermally reduced graphene oxide. Nanoscale 2013, 5, 5426–5434. [CrossRef] [PubMed]
23. Tasaltin, C.; Basarir, F. Preparation of flexible VOC sensor based on carbon nanotubes and gold nanoparticles.
Sens. Actuators B Chem. 2014, 194, 173–179. [CrossRef]
24. Maity, D.; Rajavel, K.; Kumar, R.T.R. Polyvinyl alcohol wrapped multiwall carbon nanotube (MWCNTs)
network on fabrics for wearable room temperature ethanol sensor. Sens. Actuators B Chem. 2018, 261, 297–306.
[CrossRef]
25. Zhang, H.; Cen, Y.; Du, Y.; Ruan, S. Enhanced acetone sensing characteristics of ZnO/graphene composites.
Sensors 2016, 16, 1876. [CrossRef] [PubMed]
26. Chuang, P.K.; Wang, L.C.; Kuo, C.T. Development of a high performance integrated sensor chip with a
multi-walled carbon nanotube assisted sensing array. Thin Solid Films 2013, 529, 205–208. [CrossRef]
27. Nag, S.; Sachan, A.; Castro, M.; Choudhary, V.; Feller, J.F. Spray layer-by-layer assembly of POSS
functionalized CNT quantum chemo-resistive sensors with tuneable selectivity and ppm resolution to
VOC biomarkers. Sens. Actuators B Chem. 2016, 222, 362–373. [CrossRef]

© 2018 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC BY) license (http://creativecommons.org/licenses/by/4.0/).