I. INTRODUCTION
Foods that provide nutrients essential for human survival have the disadvantage of
being easily spoiled. Protein and moisture in food provide an optimal environment
for microorganisms to reproduce, and the reproduced microorganisms degenerate food
cells [1-4]. In addition, it is difficult to control the circumstance of the entire supply chain
or storage period, and since there are many variables, food quality monitoring is
essential to keep the threat away from the table [5]. The World Health Organization reported that 600 million suffer from foodborne diseases
worldwide, with 420,000 death in 2020 [6].
The freshness of the food was evaluated by methods including gas chromatography (GC)
and mass spectroscopy (MS) [7]. However, the existing methods are ex-situ methods in which sample collection and
measurement are done separately, and they have disadvantages such as requiring the
preservation of the collected sample, reducing accuracy, and consuming a lot of time
and money. In-situ food quality monitoring systems are being actively developed to
compensate for these limitations. Among them, gas sensing food quality monitoring,
which determines whether food is spoiled through the gas generated when it decays,
has been one of the great candidates.
In particular, due to the rise of the Internet of Things (IoT), sensor technologies
such as imaging, pressure sensing [8], voice recognition, electronic tongues, and gas monitoring are regarded as irreplaceable
technology that can deliver the surrounding environment to a device and are showing
a tendency to grow rapidly. Various types of gas sensors have been studied according
to their operating principles, namely optical gas sensor [9,10], cantilever gas sensor [11], and solid-state electrochemical gas sensor [12-14]. Among them, chemoresistive gas sensors have been suggested as suitable for simple
structure, facile synthesis process, and low cost [15-18]. In addition, strong physical resistance and easy miniaturization, which are essential
properties for food quality monitoring, are accelerating research.
In this review, we introduce various semiconductor materials for food quality monitoring
and strategies to enhance gas sensing performance. To the best of the author’s knowledge,
there has been no organized review of the chemoresistive gas sensors based on various
semiconductor materials, including metal oxides, metal sulfides, carbon nanomaterials,
polymers, and their composites. We are quite sure that this review can provide a fresh
perspective on the chemoresistive gas sensors to observe food freshness in the supply
chain.
II. CHEMORESISTIVE GAS SENSORS
1. Gas Sensors in Food Quality Monitoring
The most important part of evaluating food quality through gas sensors is which type
of gas sensor is used because a type of gas sensor that is not suitable for the operation
circumstance is difficult to commercialize no matter how good the performance is.
Among the previous classifications of gas sensors, chemoresistive gas sensors have
advantages in many aspects. Optical gas sensors are challenging to miniaturize since
they are composed of optical and electrical components [9,10]. Cantilever gas sensors are vulnerable to physical impact, so their performance depends
on the operating environment [11]. Solid-state electrochemical gas sensors have trouble retaining the performance for
a long time and covering a wide range of performance temperatures [12]. Chemoresistive gas sensors based on semiconductors are suitable for food quality
monitoring due to their low mass production cost, simple electrical circuit-based
device structure, and excellent adaptability to the operating environment [15,16].
2. Operation Mechanism of Semiconductor Chemoresistive Gas Sensor
Various materials including metal, metal oxide, and polymer have been used for the
sensing materials of chemoresistive gas sensors [19-30]. The most commonly considered sensing mechanism is sensing through a change in resistance
due to the semiconducting material [31]. When the sensing material is exposed to air, the chemical adsorption of oxygen molecules
on the surface of sensing materials creates the surface receptor states (O$^{2-}$,
O$^{-}$, O$_{2}$$^{-}$). Adsorbed oxygen captures electrons at the surface, resulting
in a larger electron depletion layer (EDL) [31-35]. When the reactant gas approaches the surface of the reactant, a reaction involving
oxygen in adsorbed state occurs, and the electron density near the surface of the
reactant changes and the thickness of the EDL also changes. According to the major
charge carriers of the semiconductor device material, the resistance increases or
lowers due to the oxidation-reduction tendency of the n and p types and the reaction
gas as shown in Fig. 1 [36]. In the case of n-type, since the main charge carriers are electrons, as the amount
of adsorbed oxygen increases, the thickness of the EDL increases and the density of
the main charge carrier decreases, increasing the resistance. When the adsorbed oxygen
reacts with the reducing gas, the resistance drops, and changes in the surrounding
gas environment are detected.
Fig. 1. Schematic illustrations of gas sensing mechanism of chemoresistive gas sensors.
Three basic factors describe the performance of gas sensors based on semiconductor
materials: receptor function, transducer function, and utility factor [33]. First, the receptor function influences the reaction between the grain of the sensing
material and gas and determines how sensitive and selective sensing is. The receptor
function can be improved by attaching a catalyst to the surface to improve gas adsorption
and reaction. The transducer function occurs at the contact between grains or particles
of the sensing material and indicates whether the particle's reaction determines the
resistance of the entire device. The double Schottky barrier between particles is
determined according to the change of EDL generated on the surface. Lastly, the utility
factor is a factor determined by gas diffusion and refers to how widely the reactant
gas contacts the surface of the sensor. The more porous the structure of the reactant,
the greater the resistance change can be induced because the thickness can be fully
utilized.
III. GAS SENSORS FOR FOOD QUALITY MONITORING
1. Gas Sensors for Fruit Quality Monitoring
The Volatile Organic Compounds (VOCs) are released in the process of ripening and
rotting fruits. The VOCs can give information on fruits' quality, such as freshness
[37-40]. In the past, the Gas Chromatography-Mass Spectrometry (GC-MS) technique was conventionally
used. By using this method, it is possible to infer the molecular weight of the analysis
component enabling the accurate qualitative analysis of the sample. However, it is
hard to be used broadly because the equipment is expensive and requires a long time
to analyze a single sample [41-43]. Fruits release VOCs as secondary metabolites during biological interactions and
non-biological stress reactions [44]. As an alternative to this conventional method, semiconductor-based chemo-resistive
gas sensors are widely used due to their fast response, small volume, and high sensitivity
[45-47]. Jeong et al. investigated ethylene sensors based on SnO$_{2}$ coated with nanoscale
catalytic Cr$_{2}$O$_{3}$ overlayer [48]. Ethylene is one of the most emitted VOCs when the fruits are in the process of ripening
or rotting. Therefore, chemoresistive gas sensors that can effectively detect low
concentrations of ethylene are essential for fruit quality monitoring. The gas sensing
mechanism is based on a surface reaction between the target gas and the ionized surface
oxygen. As shown in Fig. 2(a), the SnO$_{2}$ sensor was fabricated by the screen-printing method. Subsequently,
a Cr$_{2}$O$_{3}$ catalytic overlayer was coated on the SnO$_{2}$ sensor using an
electron-beam (e-beam) evaporator (Fig. 2(a)). They prepared 0.05-, 0.3-, and 0.6-um-thick Cr$_{2}$O$_{3}$ overlayer on the SnO$_{2}$
sensing film. Then, they figured out 0.3 Cr$_{2}$O$_{3}$- SnO$_{2}$ specimen is the
most suitable for the ethylene sensing material by using Scanning Electron Microscope
(SEM), Electron Probe X-ray Micro Analyzer (EPMA), Transmission Electron Microscope
(TEM), and X-ray diffraction (XRD). Fig. 2(b) shows the gas sensing properties of SnO$_{2}$ with a thickness of 9 um (left), SnO$_{2}$
with a thickness of 22 um (middle), and 0.3 Cr$_{2}$O$_{3}$- SnO$_{2}$ with a thickness
of 21 um (right), respectively. As shown in the right-sided graph in Fig. 2(b), the 0.3 Cr$_{2}$O$_{3}$- SnO$_{2}$ sensor shows the highest selectivity for ethylene
compared to other sensors. The sensing transients of the sensor to 0.1-2.5 ppm ethylene
at 375 $^{\circ}$C were measured to show high reactivity, as shown in Fig. 2(c). Also, the sensor exhibited high reliability upon exposure to 2.5 ppm ethylene (Fig. 2(c)). Meanwhile, the detection limit of ethylene was measured to be 24 ppb. Also, the
sensor exhibited outstanding long-term stability over 15 days when exposed to 2.5
ppm ethylene. The fruits were placed inside the open chamber at room temperature (20
$^{\circ}$C~ 25 $^{\circ}$C), where the relative humidity was 35-55%. The response
to ethylene emitted from various fruits was monitored for 15 days. The ethylene response
of the sensor was increased as the ripening proceeded, and these experimental results
show that the 0.3 Cr$_{2}$O$_{3}$- SnO$_{2}$ sensor can be used to estimate the freshness
and ripening degree of fruits. In particular, there was a noticeable difference in
the sensor reactivity on the first day and the ninth day (Fig. 2(d)). Consequently, the ethylene gas sensor comprised of a SnO$_{2}$ sensing layer coated
by a Cr$_{2}$O$_{3}$ overlayer shows great gas sensing performances, which can be
efficiently used in estimating the ripeness of fruits.
Fig. 2. (a) Process used to fabricate sensors; (b) Gas-sensing properties of the thin SnO$_{2}$ sensor (thickness: ~9 um, left graph), thick SnO$_{2}$ sensor (thickness: ~22 um, middle graph), 0.3 Cr$_{2}$O$_{3}$-SnO$_{2}$ sensor (thickness: ~21 um, right graph); (c) Dynamic gas-sensing transients of the 0.3 Cr$_{2}$O$_{3}$-SnO$_{2}$ sensor (right graph), six repeated measurements of sensing properties of the sensor to 2.5 ppm ethylene at 375 $^{\circ}$C); (d) Gas sensing characteristics of the 0.3 Cr$_{2}$O$_{3}$-SnO$_{2}$ sensor for five different fruits[45].
Esser et al. reported a chemoresistive gas sensor based on a mixture of single-walled
carbon nanotubes (SWNTs) and copper complex 1 [49]. The copper complex 1 is shown in Fig. 3(a). The SWNTs and copper complex 1 were drop-casted between the gold electrodes, and
the resistance change of the sensor to ethylene was measured. The gas sensing mechanism
of SWNTs largely relies on their electronic surroundings that affect adsorption of
gas molecules and subsequently charge transfer to the molecules. The SWNTs show high
sensitivity, and the copper (1) complex shows high selectivity to ethylene. To obtain
both high sensitivity and high selectivity to ethylene, these two materials were mixed.
The ratio of copper 1 complex to SWNT carbon atoms was almost 1:6. According to the
mixing ratio, the mixture of SWNTs and copper complex 1 was named 1-SWNT. The 1-SWNT
sensor showed high sensitivity to ethylene, as shown in Fig. 3(b). The graph on the top of Fig. 3(b) shows the relative response of 1-SWNT sensors to 0.5, 1, 2, 5, 20, and 50 ppm ethylene
diluted with nitrogen gas. Compared to the response of pristine SWNT, which is not
mixed with copper, the 1-SWNT sensor exhibited 40 times higher sensitivity to 20 ppm
ethylene. The graph at the bottom of Fig. 3(b) displays the average responses of three different sensors. Despite a slight error
in the 1-SWNT sensor, the error was within an acceptable range, and the sensitivity
of the sensor was proportional to the ethylene concentration. Then, the 1-SWNT sensor
was used to compare the ethylene emission from five common fruits such as banana,
avocado, apple, pear, and orange. The responses of 1-SWNT sensors to five different
fruits are shown at the top of Fig. 3(c). The intensities are shown in relation to the response to 20 ppm ethylene and were
normalized to 100 g of each fruit. The ethylene concentration emitted from banana,
avocado, apple, and pear was over 20 ppm, while the concentration was below 20 ppm
for oranges. Especially, the banana showed a high concentration of ethylene emission,
which was more than 8 times higher than 20 ppm ethylene. During the ripening process,
the ethylene emission tended to decrease gradually, as shown at the bottom of Fig. 3(c). It is expected that the ripeness of fruits can be determined by measuring the amount
of ethylene using the 1-SWNT sensor.
Fig. 3. (a) A mixture of single-walled carbon nanotubes (SWNTs) and copper complex 1 is drop-cast between gold electrodes. The change in resistance to ethylene exposure is measured; (b) Relative responses of 1-SWNT sensor to 0.5, 1, 2, 5, 20, and 50 ppm ethylene diluted with nitrogen gas and pristine SWNT to 20 ppm ethylene (top), average responses of 3 different sensors (bottom); (c) Responses of 1-SWNT sensor to 100 g of 5 different fruit relative to 20 ppm ethylene (top), ethylene emission of 5 different fruits over several weeks[46].
2. Gas Sensors for Meat Quality Monitoring
Due to the changes in lifestyle in modern society, the proportion of online purchases
of meat is increasing. Since most of the meat quality is determined in the delivery
process, it is important to efficiently manage meat spoilage that occurs in the process
of delivering meat [50-52]. In this context, many studies have been conducted in that the meat quality monitoring
system based on chemoresistive gas sensors can monitor changes in meat quality in
real-time during delivery and selectively detect gases mainly emitted from meat [53-55]. During the meat spoilage process, Volatile Basic Nitrogen (VBN) gases are mainly
emitted, and the degree of meat spoilage can be determined by measuring the concentration
of VBN with chemoresistive gas sensors [56,57]. Spoilage micro-organisms and natural enzymes in the meat break down proteins and
produce VBN compounds [58]. Liu et al. reported naphthyl end-capped terthiophene-based chemoresistive gas sensors
to monitor meat spoilage [59]. They compared 5-(naphthalen-1-yl)-2,2’:5’,2’’-terthiophene (NA-3T) and 5,5’’-di(naphthalen-1-yl)-2,2’:5’,2’’-terthiophene
(NA-3T-NA), which are suitable materials for detecting Biogenic Amines (BAs). There
are plenty of shallow charge traps below (above) the charge conduction level for electrons
(holes). These traps allow the thermal activation and hopping of charge carriers,
resulting in charge transportation. NA-3T-NA showed high sensitivity to trimethylamine
(TMA) compared to NA-3T and for aromatic BAs such as dopamine, histamine, tryptamine,
and tyramine. Especially, the NA-3T-NA sensor showed good sensing properties to TMA,
including high sensitivity, low detection limit ({\textless} 22 ppm), and short recovery
time (23 s). The response of the NA-3T-NA based sensor to various aromatic BAs is
shown in Fig. 4(a), which includes dopamine (R=2.3), histamine (R=2.8), tryptamine (R=2.6), and tyramine
(R=3.4). Also, the response to TMA was significantly high (R=4.1). On the other hand,
the response to aliphatic BAs such as cadaverine, putrescine, hexanediamine, spermidine,
and spermine was marginal. Fig. 4(b) shows that NA-3T-NA based sensors have a short recovery time, which is within 23
s to the level of 95 % resistance change. The reactivity of the NA-3T-NA based sensor
changes linearly with TMA concentration, as shown in Fig. 4(c). The intensity of the sensor went up with an increasing concentration of TMA with
a linear relationship over a concentration range from 2110 to 16880 ppm, as shown
in the inset of Fig. 4(d). Based on the outstanding gas selectivity to TMA, the real-time meat quality monitoring
test was conducted. Fig. 4(e) shows the response to the vapors emitted from various raw meat samples, such as pork,
chicken, and fish which were stored at different temperatures for 0-5 days. Compared
to 4 ℃ samples, 25 ℃ samples show a high response. As time passed from 0 days to 5
days, the responses of the sensor tended to increase in all samples. The NA-3T-NA
sensor can be used for an in-situ evaluation of meat freshness by monitoring the concentration
of relevant volatile BAs.
Fig. 4. (a) Responses of the NA-3T-NA based sensor to various gas; (b) Response and recovery time of NA-3T-Na based sensor to TMA; (c) A reusability of the sensor to TMA at 25 ℃; (d) The signal intensity of the NA-3T-NA based sensor at different TMA concentrations. Inset: a plot of the differences in the response versus the concentrations of TMA; (e) Responses of the sensor to the vapors above various meat samples. Reproduced with permission from[56]Copyright © 2019 John Wiley & Sons.
Liu et al. reported the chemoresistive gas sensor based on single-walled carbon nanotube
(SWCNT)/metallo-porphyrin composites [60]. The magnitude of the sensor response could be improved through the changes in the
oxidation state of the metal and the electron-withdrawing characteristic of the porphyrinato
ligand. SWCNTs can be functionalized covalently or noncovalently with other materials
to pass on sensitivity or selectivity for the desired analyte. Porphyrins are attractive
compounds for functionalizing SWCNTs because their aromatic core can noncovalently
bind to the SWCNTs' walls. Fig. 5(a) shows the response of the sensor composed of various materials with exposure to NH$_{3}$.
By using the Co(tpp)ClO$_{4}$-SWCNT which exhibited high selectivity to NH$_{3}$,
the meat spoilage monitoring was conducted for 4 days, as shown in Fig. 5(b). The measurements were made with 4 types of meats such as cod, salmon, chicken, and
pork at both 4 ℃ and 22 ℃. For the samples measured at 4 ℃, the sensor showed no increase
in response for 4 days. However, for the samples measured at 22 ℃, the response of
the sensors increased over time gradually. Overall, the gas sensor based on SWCNT-metalloporphyrin
composites can be an excellent candidate for monitoring meat spoilage by measuring
the concentration of volatile BAs.
Fig. 5. (a) Responses of the sensors based on porphyrin-SWCNT composites to various concentration of NH$_{3}$ for 30 s; (b) Responses of the sensor to 30 s exposures of emitted vapors from various meat samples stored at RT (22 ℃) and 4 ℃. Reproduced with permission from[57]Copyright © 2015 John Wiley & Sons.
3. Gas Sensors for Fish Quality Monitoring
Fish is the most representative food whose freshness directly determines the taste.
For that reason, freshness maintenance of the fish is a major challenge, and several
studies have proceeded to improve the maintenance [61-65]. After the death of fish, the microorganisms on the fish surface increase exponentially
and permeate into the tissues. The microorganisms spoiled fish and emitted various
kinds of gas such as trimethylamine (TMA), dimethylamine (DMA), ammonia, and CO [66,67]. Mostly, TMA gas was formed by bacteria such as Shewanella putrifaciens, Aeromonas
spp., psychrotolerant Enterobacteriacceae, P.phosphoreum, and Vibrio spp. Fish use
trimethylamine oxide (TMAO) as osmoregulant to avoid dehydration in marine environments.
In such an environment, the bacteria produced TMA gas to obtain energy by reducing
TMAO [65].
Zhu et al. reported the TMA gas sensor synthesized by the amphiphilic perylene diimide
derivative (1,6,7,12-tetra-chlorinated perylene-N-(2-hydroxyethyl)-N-hexyl amine-3,4,9,10-tetracarboxylic
bisimide, TC-PDI) and CdS [68]. As the heterojunction of organic-inorganic compounds formed, the gas sensitivity
and selectivity were enhanced. The crystal structures of CdS were either cubic or
hexagonal so that it could simplify the decoration of the microstructure and construct
the stable heterostructures. Fig. 6(a) is the SEM image of TC-PDI/CdS, which has micro belts structure with 1${\mu}$m width
and over a hundred micrometers in length. The heterostructures were formed by the
Cd$^{2+}$ that built the Cd${-}$O formation between the Cd$^{2+}$center and the hydroxyethyl
groups of TC-PDI. The Cd${-}$O coordination bonds were balanced with hydrophobic interactions
between side chains and induced the belt-like gas was injected into the TC-PDI/CdS,
the resistance was increased significantly. In case the various concentration of TMA
was exposed from 1 to 100 ppm, the resistance curve exhibited a good linear relationship.
By following Dua’s method, TC-PDI/CdS sensor could detect 200 ppb of TMA. The long-term
stability of TC-PDI/CdS was also excellent. After storing for over 30 days in the
air at room temperature, there was no significant difference in the gas sensing properties.
Fig. 6(b) shows the application to detect a hairtail for 72 hours. The response was linearly
increased with the storage time. Fig. 6(c) depicts the selectivity of the TC-PDI/CdS to 20 ppm of TMA and 100 ppm of other gases,
including acetone, benzene, ammonia, ethanol, and NO$_{2}$. TC-PDI/CdS exhibited the
highest response to TMA among the gases. Considering the application in a real situation,
the fishes were stored in a high relative humidity atmosphere Fig. 6(d) shows the response to 60 ppm of TMA in various RH atmospheres from 30 to 100% RH.
The response to TMA was maintained even in high humid atmosphere verifying the capability
in practical applications. These results showed that the TC-PDI/CdS is suitable for
real-time fish freshness monitoring.
Fig. 6. (a) SEM images of TC-PDI/CdS microbelts; (b) Responses of the TC-PDI/CdS sensor to hairtail storage time at room temperature; (c) Response to TC-PDI/CdS sensor to 20 ppm TMA and 100 ppm of various gases at room temperatures; (d) The responses of TC-PDI/CdS sensor to 60 ppm TMA under different humidity. Reproduced with permission from[66]Copyright © 2019 ACS Publications.
Tonezzer et al. reported NH$_{3}$ sensor using tin oxide nanowire and analyzed the
quantitative assessment of trout fish spoilage [69]. Fig. 7(a) is the SEM image of SnO$_{2}$nanowires obtained by chemical vapor deposition (CVD).
The average diameter of SnO$_{2}$nanowires was 40-80 nm. The thin morphology and potential
barrier in the nanowires led to the improvement of gas sensing performance. Fig. 7(b) shows the dynamic resistance of the sensor at three different temperatures (200,
250, and 300 ℃). The resistance of the SnO$_{2}$nanowires sensor was constant in the
air at three temperature conditions. When NH$_{3}$ gas was flushed into the chamber,
the resistance dropped sharply. The resistance was rapidly returned to the initial
state when the pure air was injected. The resistance variation was linearly changed
depending on the NH$_{3}$ concentration. When the working temperature was increased,
the SnO$_{2}$nanowires also showed a higher response to NH$_{3}$. NH$_{3}$ gas is
a reducing gas that reduces the resistance of n-type semiconductors. When the SnO$_{2}$nanowires
were exposed to air, oxygen molecules were adsorbed and changed to O$^{-}$or O$^{2-}$.
The adsorbed oxygen dragged the electrons from the SnO$_{2}$nanowires. When NH$_{3}$
molecules were exposed to SnO$_{2}$ nanowires, they reacted with the adsorbed oxygen
and induced the oxygen molecules to release electrons into the SnO$_{2}$nanowires.
As the number of electrons in the sensor increased, the resistance of the SnO$_{2}$nanowires
diminished. Fig. 7(c) shows the sensor response and bacterial total viable count (TVC) in fish at room
temperature over 60 hours. The response was increased at all temperatures gradually.
The TVC also increased similarly so that it could be considered for measuring the
freshness of fish. The dashed horizontal green line indicates the threshold that the
shelf life of fish was ended. The threshold of shelf life was about 22 h and 40 min
of storage at room temperature. Fig. 7(d) shows the sensor response as a function of the logarithm of TVC. The response showed
the linear property at all temperatures with Pearson’s correlation coefficients over
0.99 in all cases. The graph shows that the sensor response is suitable to measure
the TVC. Fig. 7(e) is a principal component analysis (PCA) that could maximize the interpretability
of the sensing properties. The PCA graph was divided by a group of the TVC value.
Each group was arranged in a zigzag line that goes down to 6-7, then goes up to 8,
and goes down to 10 repeatedly. The PCA results demonstrated that the SnO$_{2}$ nanowire
sensor was suitable for analyzing the freshness of fish in various stages of the degradation
process.
Fig. 7. (a) SEM images of the SnO$_{2}$ nanowires grown by CVD; (b) Dynamic resistance at three temperature values during the injection of different concentrations of ammonia; (c) Sensor response (solid symbols, left scale) and bacterial population (green open circles, right scale) in fresh trout fish kept at room temperature (25 ℃) over a period of 60 hours; (d) Double-blind measurements of the sensor response as a function of the total viable count in rainbow trout samples; (e) PCA plot of random samples of rainbow trout[67].
4. Others
Besides fruits, meats, and fish, the chemoresistive gas sensors can be used to detect
various foods like dairy [70], eggs [71], wines [72], oils [73], and spices [74]. For example, milk, one of the dairy products, cannot be digested by most adults
in natural conditions. Thus, it is necessary to add adulterations to milk. However,
various side effects depend on the amount or type of adulterations. Accordingly, Pirsa
et al. reported the gas sensor to detect adulteration in milk by sensing the volatile
compounds using a poly-pyrrole-ZnO (PPy-ZnO) fiber gas sensor [75]. The PPy-ZnO was synthesized by doping ZnO nanoparticles on a polymer via the chemical
polymerization method. The adulterants with NaClO, H$_{2}$CO$_{3}$, citric acid, and
NaHCO$_{3}$were detected by the PPy-ZnO. As the volatile compounds were released from
adulterants in milk, they were extracted from milk via the headspace method. After
that, the volatile compounds gases were injected into the PPy-ZnO gas sensor and analyzed
by response surface methodology (RSM). The PPy-ZnO sensor showed a linear response
to NaClO, H$_{2}$CO$_{3}$, citric acid, and NaHCO$_{3}$. By using a statistical model
based on the adulterant, the PPy-ZnO gas sensor could discriminate the type or the
number of adulterants based on the positive and negative peaks. Hence, the PPy-ZnO
fiber could help to classify the various kinds of milk adulterants.
Eggs have sulfur-containing amino acids producing ppb level of H$_{2}$S gas under
the action of microorganisms. Guo et al. reported the H$_{2}$S gas sensor based on
SnSe$_{2}$/WO$_{3}$ composite for evaluating the spoilage of eggs at room temperature
[71]. The WO$_{3}$was decorated on the SnSe$_{2}$nanosheet via a one-step ultrasound method.
While metal oxide semiconductor gas sensors are vulnerable to high humidity, the SnSe$_{2}$/WO$_{3}$gas
sensor showed high endurance to high RH. Furthermore, The H$_{2}$S response was enhanced
at high RH levels because of the water-activated electron transfer. For the experimental
process, the glass bottles contained fresh eggs, the spoiled eggs were prepared, and
the SnSe$_{2}$/WO$_{3}$ gas sensor was put in each bottle. The resistance of the sensor
in all bottles was decreased, but the variation of resistance in the bottle of the
spoiled egg was much higher. The SnSe$_{2}$/WO$_{3}$ gas sensor could be used for
the portable and non-olfactory contact sensor to monitor the quality of eggs.
Wine is one of the alcoholic drinks which is consumed all over the world. For the
taste of wine, it is crucial to monitor the ripening process elaborately. While the
wine is ripened, volatile sulfur compounds (VSCs) are released such as H$_{2}$S and
methyl mercaptan (MM). Lim et al. reported the CuO/CuFe$_{2}$O$_{4}$ nanopattern chemoresistitve
gas sensor that could detect H$_{2}$S and MM dually. The nanopattern was composed
of aligned 1D CuO/CuFe$_{2}$O$_{4}$ using direct-write near field electrospinning
(NFES). The response of the CuO/CuFe$_{2}$O$_{4}$nanopattern structure was much higher
than the thin-film or nanofiber structure. The CuO/CuFe$_{2}$O$_{4}$gas sensor showed
high selectivity to H$_{2}$S at 200 ℃ and MM at 400 ℃. At 200 ℃, the gas response
to H$_{2}$S and MM of pure Fe$_{2}$O$_{3}$was low, but the Cu$_{4}$Fe showed a higher
response to H$_{2}$S. This phenomenon was associated with the phase transformation
of p-type CuO to metallic CuS when the sensor was exposed to H$_{2}$S. By the H$_{2}$S
gas, CuS formed the nanoscale p-n junction and led to the enhancement of gas response.
On the contrary, the mechanism of gas sensing was changed significantly at 400 ℃.
The high temperature caused CuS to be oxidized to CuO. The Cu$_{4}$Fe sensors showed
higher responses to MM. It can be seen that the CuO/CuFe$_{2}$O$_{4}$ junction formation
is an effective strategy for the highly selective detection of MM. The difference
in the degree of oxidation shows high selectivity to H$_{2}$S and MM. Overall, the
dual-mode gas sensing in this study is expected to facilitate the real-time monitoring
of the wine ripening.
IV. CHALLENGES AND PERSPECTIVE
Chemoresistive gas sensors have proven to be highly successful in monitoring food
quality by sensitive and selective gas detection. However, chemoresistive gas sensors
continue to face significant problems such as humidity resistance, which is one of
the most important properties for monitoring food quality. Poor humidity resistance
is a flaw that originated from the sensing mechanism of chemoresistive gas sensors.
Since chemoresistive gas sensors are the devices whose operating principle is that
the resistance changes through the change of oxygen adsorption on the surface, when
the surrounding humidity is high, oxygen adsorption is interrupted by H$_{2}$O molecules
and lowering sensitivity and stability. A method for preventing the absorption of
water molecules without interfering with the adsorption of oxygen on the surface is
required to improve humidity resistance. A typical method is to apply a humidity-resistant
material, including PDMS, to the surface to prevent water molecule adhesion. Also,
heterojunction between sensing material and humidity-resistant material can be considered
as an expected effect. The study of new humidity-resisting materials and their combination
with existing sensing materials remains an important challenge in securing the humidity
stability of chemoresistive gas sensors.
Another issue of food quality monitoring based on chemoresistive gas sensors is high
operation temperature due to relatively low sensitivity and selectivity. According
to the sensing mechanism of chemoresistive gas sensors, the more oxygen is adsorbed,
and the more electrons are taken away, the larger the area of EDL generated on the
sensing material surface. In the case of semiconductor materials, since the mobility
of the main charge carriers increases as the temperature increases, more electrons
can participate in the reaction, which is usually operated at a high temperature (100-400
$^{\circ}$C) to promote EDL generation and amplify resistance changes. However, since
food quality monitoring has to be conducted in a distribution environment rather than
a high temperature, or even at a low temperature of 0 $^{\circ}$C or less, the low
mobility of charge carriers stands out. In order to overcome the inherent limitations
of semiconductor materials, methods to improve the three basic factors introduced
above are used. To improve the utility factor, the sensing material is nanostructured
to increase the surface area and allow more oxygen to reach the surface, leading to
wider EDL generation. The receptor function can be improved by inducing more oxygen
adsorption and gas molecule adsorption through catalyst integration or heterojunction.
This enables easy food quality monitoring through chemoresistive gas sensors with
high reactivity and selectivity without limiting the operating environment.
V. CONCLUSIONS
In this article, we review the chemoresistive gas sensors based on semiconductor materials
for food quality monitoring. As online food purchases increased in the pandemic era,
the food quality monitoring system, which detects gas occurring in the food distribution
process, has become particularly important. The Gas Chromatography-Mass Spectrometry
(GC-MS) technique, which has been mainly used in the past, has several drawbacks,
such as high-priced equipment and a time-consuming measurement system. However, the
chemoresistive gas sensors have advantages such as fast response, small size, and
high reactivity compared to conventional techniques. Various materials such as metal
oxide, metal sulfide, carbon nanomaterials, polymers, and their composites are widely
used to detect the gases emitted during the food ripening process. In this work, foods
were categorized into 4 groups: fruit, meat, fish, and others (milk, egg, and wine).
Different types of gases are emitted from fruits (ethylene), meats (TMA), and fishes
(TMA, DMA, and NH$_{3}$). Moreover, diverse gases such as VOCs, H$_{2}$S, and VSCs
also appear in milk, egg, and wine. Although there are some challenges, such as poor
humidity stability and high operation temperature, many researches are being conducted
to overcome these limitations and improve the gas sensing performance. Through overcoming
the drawbacks, chemoresistive gas sensors are highly anticipated to be developed into
an innovative technology for the next generation of food quality monitoring.