11404 IEEE SENSORS JOURNAL, VOL. 23, NO.

11, 1 JUNE 2023

Diffusion-Modulated Colorimetric Sensor for

Continuous Gas Detection
Jingjing Yu , Wei Ding, Laxmi Jaishi, Chenwen Lin , Rachel Boylan , Chinmay Chandrakant Dixit,
Buddhi Sagar Lamsal, Wei He , Francis Tsow , Songxin Tan , Yue Zhou,
and Xiaojun Xian , Member, IEEE

Abstract—Due to their high sensitivity and selectivity,

low cost, and good compatibility for sensor array integra-
tion, colorimetric gas sensors are widely used in hazardous
gas detection, food freshness assessment, and gaseous
biomarker identification. However, colorimetric gas sensors
are usually designed for one-time discrete measurement
because the sensing materials are entirely exposed to ana-
lytes during the sensing process. The fast consumption of
sensing materials limits colorimetric sensors’ applications
in continuous analytes monitoring, increases the operation
complexity, and brings challenges for calibration. In this work, we reported a novel sensor design to prolong the lifetime
of colorimetric gas sensors by engineering the gas diffusion process to preserve the sensing materials. We compared
two geometries for gas diffusion control in a sensing matrix through simulation and experiment on an ammonia sensing
platform. We found that the 2-D gas diffusion geometry enabled a better sensor performance, including more stable and
higher sensitivity and a more linear response to ammonia concentration compared to 1-D gas diffusion geometry. We also
demonstrated the usability of this diffusion-modulated colorimetric sensor for continuous environmental ammonia
monitoring.
Index Terms— Air quality, ammonia, colorimetric sensor, diffusion, gas sensor, lifetime, sensitivity.

I. I NTRODUCTION analytes, portable chemical sensors, such as electrochemical

AS sensors play a critical role in hazardous gas detec- sensors [13], [14], metal–oxide sensors [15], [16], [17], pho-
G tion [1], [2], [3], gaseous biomarker detection [4],
[5], [6], and food quality monitoring [7], [8], [9], [10].
toionization detector (PID) sensors [18], [19], piezoelectrical
sensors [20], [21], and colorimetric sensors [22], [23], are
GC-MS is considered as the gold standard for gaseous highly preferred, mainly due to their small size, low cost, and
analysis, but it is bulky, expensive, and only available in simple operation [11]. Compared to other chemical sensing
research institutions [11], [12]. For the onsite detection of gas approaches, colorimetric sensors show competitive advantages
in high sensitivity and selectivity, low cost, and good com-
Manuscript received 15 January 2023; revised 13 April 2023; patibility for sensor array integration [24], [25]. Colorimetric
accepted 17 April 2023. Date of publication 24 April 2023; date of current sensors use color change to identify and quantify gaseous ana-
version 31 May 2023. This work was supported by the National Institutes
of Health under Project 5R44ES029006-03. The associate editor coordi- lytes. The color changes in colorimetric sensors originate from
nating the review of this article and approving it for publication was Prof. different physical or chemical interactions between the gaseous
Bhaskar Choubey. (Corresponding author: Xiaojun Xian.) analytes and sensing materials, such as van der Waals inter-
Jingjing Yu, Wei Ding, Laxmi Jaishi, Buddhi Sagar Lamsal, Wei He,
Songxin Tan, and Xiaojun Xian are with the Department of Elec- actions, hydrogen bonding interactions, 5-5 stacking com-
trical Engineering and Computer Science, South Dakota State plexation, Bronsted acid-base interactions, and donor–acceptor
University, Brookings, SD 57007 USA (e-mail: jingjing.yu@sdstate. interactions [23], [26]. High sensitivity and selectivity can
edu; [email protected]; [email protected]; buddhi.
[email protected]; [email protected]; songxin.tan@ be achieved through sensing materials engineering. However,
sdstate.edu; [email protected]). because sensing materials are consumed during the sensing
Chenwen Lin, Rachel Boylan, Chinmay Chandrakant Dixit, and Fran- process, colorimetric gas sensors are usually designed for
cis Tsow are with the Center for Bioelectronics and Biosensors, Arizona
State University, Tempe, AZ 85281 USA (e-mail: [email protected]; one-time discrete measurement. The sensor only provides a
[email protected]; [email protected]; [email protected]). reading of a short-term gas concentration or an averaged
Yue Zhou was with the Department of Electrical Engineering long-term gas concentration. This limits their applications in
and Computer Science, South Dakota State University, Brookings,
SD 57007 USA. He is now with the Department of Mechanical Engi- continuous analytes monitoring (continuous monitoring means
neering, The University of Texas at Dallas, Richardson, TX 75080 USA the sensor can track the dynamic concentration profile of the
(e-mail: [email protected]). analyte by providing a series of real-time concentration read-
This article has supplementary downloadable material available at
https://doi.org/10.1109/JSEN.2023.3268537, provided by the authors. ing over a long period), increases the operation complexity,
Digital Object Identifier 10.1109/JSEN.2023.3268537 and brings challenges for calibration. A reversible colorimetric

1558-1748 © 2023 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See https://www.ieee.org/publications/rights/index.html for more information.

Authorized licensed use limited to: VTU Consortium. Downloaded on February 22,2024 at 06:24:54 UTC from IEEE Xplore. Restrictions apply.
YU et al.: DIFFUSION-MODULATED COLORIMETRIC SENSOR FOR CONTINUOUS GAS DETECTION 11405

sensor could be an option for continuous gas monitoring, diffusion control on a sensing matrix through simulation and
but reversibility typically is at the price of selectivity and experiment. We found that the 2-D gas diffusion geometry
sensitivity [23]. Weaker interactions between gas and sensing enabled a better sensor performance, including more stable
material would improve the reversibility but often restrict the and higher sensitivity and a more linear response to ammo-
selectivity and sensitivity and vice versa. Moreover, prolonged nia concentration compared to 1-D gas diffusion geometry.
recovery time could be another challenge to hinder continuous We also demonstrated the usability of this diffusion-modulated
detection. colorimetric sensor for continuous environmental ammonia
Considering that most colorimetric sensors are based on monitoring.
irreversible chemical reactions between analytes and sens-
ing materials, a universal approach is highly preferred to II. E XPERIMENTAL S ECTION
empower the colorimetric sensors with continuous sensing A. Materials
capability. Compared to reversible colorimetric sensors, irre- Bromophenol blue (CAS Number: 115-39-9) and ethanol
versible colorimetric sensors usually show a quick response were purchased from Sigma. Ultrapure water was produced
time but a decaying sensitivity and a limited lifetime due via the E-pure water system. The NH3 (1000 ppm, balanced
to sensing materials consumption. Different approaches have with air) was purchased from Gasco. The clean air from
been reported to maintain the stability of sensitivity and Matheson Inc. was used to dilute 1000-ppm NH3 into different
increase the lifetime of colorimetric sensors for continuous concentrations. The Whatman grade 43 filter paper (220 µm
monitoring of analytes. Qin et al. [27] have demonstrated thick, particle retention of 16 µm) was used as the sensor sub-
that by increasing the sensitivity and loading capacity of the strate. Fertilizer derived from chicken manure (includes 2.0%
colorimetric sensor, the sensor’s exposure time to the gas water soluble nitrogen and 3.0% water insoluble nitrogen)
analyte can be decreased. Thus, continuous monitoring of was manufactured by The Espoman Company. The NH3000
formaldehyde can be accomplished. A similar concept has (Forensics Detectors Company) was used as a reference sensor
also been demonstrated in high-performance carbon monoxide for fertilizer-emitted NH3 detection.
monitoring [28]. This method requires active control of the
gas sampling system, increasing the colorimetric sensor’s B. Sensor Preparation
complexity and cost. Yu et al. [29] have developed a light-
The sensing solution was prepared by dissolving 40-mg bro-
controlled method to preserve the sensing materials from
mophenol blue into 5-mL ethanol and 5-mL water (4 mg/mL).
unwanted consumption, but this method can only be applied
The pH of the sensing solution was ∼2.65. The filter paper
to colorimetric sensors that utilize photochemical reactions.
was immersed in the sensing solution for 1 min and then
Lin et al. [30] introduced a diffusion channel-based colorimet-
vacuum dried for 30 min. The concentration of bromophenol
ric sensor to localize the chemical reactions and continuously
blue coated on filter paper was around 5 mol/m3 (amount of
quantify the analyte concentrations by tracking the movement
bromophenol blue/volume of filter paper). The coated filter
of the color gradient along the diffusion channel. Although the
papers were cut into a circle or channel shape with a laser
results are promising, this diffusion-based channel approach
cutter (Universal VLS3.50). The circle diameter was ∼5 mm,
faces challenges of sensitivity drop due to the increasing
and the channel size was ∼0.8 × 5 mm. The coated circle
diffusion distance for the gases along the channel [31].
filter paper was sealed with tape (Scotch Transparent Tape),
Ammonia is a colorless gas with a strong irritating odor.
and a circular hole with a diameter of ∼0.25 mm was drilled to
Most of the ammonia was emitted from agriculture [32].
build a 2-D diffusion sensor. For the 1-D diffusion sensor, the
Among all agricultural ammonia emissions, fertilizer use-
coated paper strip was covered with tape that had an opening
related emissions have been estimated to contribute around
at one end.
10%–35% [33]. Recent studies have indicated that NH3 emis-
sion is increasing [32], [34]. This is a concern because ammo-
nia negatively impacts the environment and human health. The C. Experimental Methods
environmental effects include soil acidification, eutrophication, The sensor was placed in a test chamber (14 × 14 ×
and decreased biodiversity [35]. While for human health, 3 mm) for continuous monitoring (Fig. S5). Gas samples were
it can cause multiple adverse health effects, such as rhinorrhea, introduced into the test chamber at the flow rate of 0.55 L/min.
lung disease, and irritation of the respiratory tract, skin, and The Logitech C525 webcam was used to record the color
eyes [36]. Ammonia was also considered as a precursor to change with a framerate of 5 frames/s under white light
PM2.5, which affects the environment and human health in emitted from a white LED backlight module (1528-1790-ND,
another way [37], [38]. Monitoring ammonia is essential for Adafruit Industries LLC). The real-time average intensity of
controlling ammonia emissions, thus decreasing environmental the entire sensor area was calculated with the MATLAB
issues and protecting humans from ammonia exposure. program.
In this work, we reported a novel design to stabilize the
sensitivity of the colorimetric sensor for long-time continuous III. R ESULTS AND D ISCUSSION
ammonia monitoring by engineering the gas diffusion process. A. Design of the 2-D Diffusion-Based Colorimetric
This colorimetric sensor detects ammonia through a nearly Sensor
nonreversible chemical reaction (see Supplementary Material A key challenge for diffusion-based colorimetric sensors is
Figs. S6–S8 for detail). We compared two geometries for gas the sensitivity drops due to the increase of diffusion distance

Authorized licensed use limited to: VTU Consortium. Downloaded on February 22,2024 at 06:24:54 UTC from IEEE Xplore. Restrictions apply.
11406 IEEE SENSORS JOURNAL, VOL. 23, NO. 11, 1 JUNE 2023

Fig. 1. (a) One-dimensional diffusion-based colorimetric sensor (1-D

sensor) versus (b) 2-D diffusion-based colorimetric sensor (2-D sensor). Fig. 2. Simulation results of the 2-D diffusion-based colorimetric sensor
(2-D sensor) and 1-D diffusion-based colorimetric sensor (1-D sensor).
(a) Color development (chemical reaction) along the diffusion direction
after 120 s with different reaction coefficients. The product concentration
for the gas analyte as the reaction progresses. A typical distribution along diffusion direction after 120 s in (b) 2-D sensor and
diffusion-based colorimetric sensor has a small gas inlet, (c) 1-D sensor. The change of average product concentration over the
where the gas molecules can diffuse to a porous matrix and entire sensor with the time of (d) 2-D sensor and (e) 1-D sensor.
react with coating colorimetric sensing probes [see Fig. 1(a)].
According to Fick’s first law, the gas diffusion rate (n) in the
will also increase. This will compensate for the gas concen-
sensor can be calculated as follows:
tration gradient drop and maintain a stable chemical reaction
n = N × S = −D × dc/d x × S (1) and color development rate. Fig. 1 shows the geometries and
expected optical responses of the passive colorimetric sensors
where n is the diffusion rate of gas (g/m2 ), S is the cross- with 1-D and 2-D diffusion geometries.
sectional area of the diffusion path (m2 ), N is the diffusion flux
(g/m2 /s), D is the diffusion coefficient in the sensor (m2 /s),
and dc/dx is the concentration gradient along the diffusion B. Simulation of the Sensing Behaviors of 1-D and 2-D
path. Since the concentration at the inlet and edge is constant Diffusion Systems
C0 and 0, dc/dx decreases with edge development. We used COMSOL to simulate the diffusion and reaction
In the initial stage, color development happens around the in 1-D and 2-D colorimetric sensing systems. As shown in
gas inlet. Because the diffusion distance is short, there is a Fig. 2(a), a gas inlet was introduced at the center of the 2-D
high gradient of gas concentration (dc/dx). Correspondingly, colorimetric sensing system. In contrast, the gas inlet was
there is a rapid color development if the reaction rate is located at one end of the 1-D colorimetric sensing system.
proportional to gas diffuse rate n. As time increases, the To ensure the initial cross-sectional area of the diffusion
sensing materials near the gas inlet are gradually consumed; path was the same for both sensing systems, we made the
thus, the gas analyte must diffuse further from the gas inlet circumference of the gas inlet opening for the 2-D system the
to react with fresh sensing materials along the path. As diffu- same as the width of the gas inlet opening for the 1-D system.
sion distance increases, the gas concentration gradient in the The center hole radius of the 2-D system is 0.125 mm; thus,
color-developing area decreases, and the sensor’s sensitivity the inlet width of the 1-D system is 2 × 3.14 × 0.125 mm.
drops. To address this issue, we propose compensating for the Suppose gas analyte A diffuses into the sensor through the
decreasing concentration gradient by increasing the frontier inlet and reacts with sensing material B coated on a porous
of the reaction area. Instead of introducing a 1-D diffusion substrate to produce the product C: A(g) + B(s) → C(s).
channel on the colorimetric sensing substrate [see Fig. 1(a)], Assume the chemical reaction rate depends on the concentra-
we designed a 2-D diffusion geometry with a center opening tions of reactants A and B, and then, the reaction rate R can
[see Fig. 1(b)]. be described as
For a one-dimension diffusion system [see Fig. 1(a)], the R = kC A C B = (∂CC )/∂t = −(∂C B )/∂t (2)
cross-sectional area S is a constant along the diffusion path.
In contrast, S increases along the diffusion path for the 2-D where C A , C B , and Cc are the concentrations of reactants
diffusion system [see Fig. 1(b)]. The increasing S means that A, B, and product C, t is time, and k is the reaction
the available sensing materials to react with the gas analyte rate coefficient. According to diffusion law and the law of

Authorized licensed use limited to: VTU Consortium. Downloaded on February 22,2024 at 06:24:54 UTC from IEEE Xplore. Restrictions apply.
YU et al.: DIFFUSION-MODULATED COLORIMETRIC SENSOR FOR CONTINUOUS GAS DETECTION 11407

conservation of mass, the mass transfer of A in the sensing

system will follow the equation
(∂C A )/∂t + ∇(−D∇C A ) = −kC A C B . (3)
The initial concentrations of A and C in the porous substrate
were set as 0, while initial concentration B was described as
1 mol/m3 . The concentration at the gas inlet was selected as
the gas concentration that the sensor was exposed to. Other
edges of the sensors were set as sealing walls (which means
zero flux). The gas diffusion rate in the porous substrate was
set as 1 × 10−5 m2 /s.
Fig. 2 shows the simulation results with different k values.
Fig. 2(a) shows the concentration profile of product C on
the sensor before and after 120 s exposure to 50 ppm gas
(0.00223 mol/m3 ). The concentration distribution depends on Fig. 3. Simulation results with different gas concentrations. Prod-
the reaction rate coefficient k. A larger k results in a sharper uct concentration over the entire sensor increases with time in
(a) 2-D sensor and (b) 1-D sensor. (c) and (d) Derivates of the product
edge. Fig. 2(b) and (c) shows the concentration distribution concentration change rate from (a) and (b).
along the diffusion direction in a numerical way for 2-D
diffusion and 1-D diffusion systems, respectively. No matter
which system it is, when k is larger, the product gradient product changing rate drops sharply for the 1-D diffusion
is higher, as indicated by the sharp edge. The relation- system [see Fig. 3(d)]. If the sensor was not fresh, the 1-D
ship between product concentration over the entire sensor diffusion system can barely distinguish between 200 and
and time for 2-D and 1-D diffusion systems are shown in 5 ppm ammonia. These results show that the 2-D diffusion
Fig. 2(d) and (e), respectively. When k ≥ 100 m3 /(s·mol), the system has stable, higher sensitivity, and better resolution than
plots overlap with each other. No matter which k is used, the 1-D diffusion system.
the product amount–time relationship of the 2-D diffusion
system is more linear than the 1-D diffusion system [R 2 C. Sensing Behaviors of 1-D and 2-D Diffusion Systems
of 0.9778 for 2-D versus R 2 of 0.6477 for 1-D when k is From Experiments
1000 m3 /(s·mol). R 2 and other detail of fitting method can To experimentally verify the sensing behaviors of 1-D and
be found in Supplementary Material], which suggests that the 2-D diffusion systems, we prepared the 1-D and 2-D diffusion-
2-D system has a more stable sensitivity. Moreover, the 2-D based colorimetric ammonia sensors using filter paper as
system shows a higher average product change rate (slope) substrates. The colorimetric sensing probe (bromophenol blue)
than the 1-D system, suggesting that the sensitivity of the for ammonia sensing was dip-coated on filter paper with a
2-D system is higher than the 1-D system. A comparison concentration of around 5 mol/m3 (amount of bromophenol
between 1-D and 2-D sensors with the same sensing area and blue/volume of filter paper). The filter paper was used as the
opening inlet also indicates more stable and higher sensitivity substrate because it can provide a matrix with suitable porosity
for 2-D sensor (Fig. S3). In conclusion, the simulation results for ammonia to diffuse. The acidic form of bromophenol blue
in Fig. 2(a) indicate that the 2-D diffusion system has a more reacted with NH3 and produced its basic form [see Fig. 4(a)],
stable and higher sensitivity than the 1-D diffusion system. causing a color change from yellow to purple blue [39].
Other simulation results with different diffusion coefficients, This color change was monitored with a homemade webcam-
different inlet sizes, and different structures were provided in MATLAB program.
the Supplementary Material. The absorbance spectra of the sensor before and after gas
We also simulated the sensor response to different concen- exposure were recorded in Fig. 4(b). The peaks at 440 and
trations of the gas analyte. Fig. 3(a) and (b) shows that the 600 nm were associated with the acid form and basic form of
product concentration over the entire sensor increases with bromophenol blue, respectively [40]. The descent of the peak
time when the 2-D and 1-D diffusion systems are exposed to at 430 nm and the rise of the peak at 600 nm were due to
different gas concentrations. The higher the analyte concen- decreased reactant and increased product concentration after
tration, the higher the product increasing rate. 2-D diffusion NH3 exposure [40]. Thus, the color of the colorimetric sensor
system shows a more linear (stable) product increasing rate at changes from yellow to purple-blue. The optical signal from
a given analyte concentration, while the product increasing the R channel of the webcam (wavelength larger than 570 nm)
rate for the 1-D diffusion system keeps decreasing toward was used for sensing signal processing (Supplementary Mate-
plateauing. To compare the product changing rate of the rial Fig. S6 for detail). Scanning electron microscope (SEM) of
2-D and 1-D diffusion systems under different concentrations, the sensor strip before and after coating and ammonia exposure
the derivatives [see Fig. 3(c) and (d)] were calculated from was characterized in Fig. 4(c)–(e). The pictures on the left
Fig. 3(a) and (b). Fig. 3(c) indicates that although the change top corner were captured under the indoor light with a phone
rate decreases with time, the 2-D diffusion system can clearly webcam. The SEM pictures showed that the filter paper has
distinguish the concentrations of 200 and 50 ppm, no matter microsize pores, and the sensor structure did not change after
how long the sensor was exposed to the analytes, but the coating and reacting with NH3 .

Authorized licensed use limited to: VTU Consortium. Downloaded on February 22,2024 at 06:24:54 UTC from IEEE Xplore. Restrictions apply.
11408 IEEE SENSORS JOURNAL, VOL. 23, NO. 11, 1 JUNE 2023

opening length as the 2-D sensor for better comparison of

the two sensors). Different ammonia concentrations from 1 to
200 ppm were tested. The color development started near
the gas inlet after NH3 exposure [see Fig. 5(a)]. The higher
concentrations led to rapid and remarkable color develop-
ment. Low-concentration ammonia (1–25 ppm) was tested
for 1 h, and high-concentration ammonia (50–200 ppm) was
tested until the developed color got close to the edge of the
sensor. The absorbance–time relationships of 2-D and 1-D
sensors were recorded and displayed in Fig. 5(b) and (e). The
derivatives of these curves are shown in Fig. 5(c) and (f).
The 1-D sensor’s slope of absorbance decreased rapidly with
time. In contrast, the 2-D sensor’s slope of absorbance was
much more stable than the 1-D sensor. Since absorbance is
Fig. 4. Reaction of NH3 and bromophenol blue. (a) Chemical reaction
proportional to the concentration of product from the chemical
equation. (b) Absorbance spectra of the sensor before and after NH3 reaction, the experimental results were consistent with the
exposure. SEM pictures of (c) pristine filter paper, (d) filter paper coated simulation results.
with bromophenol blue, and (e) coated filter paper after ammonia expo-
sure. The inserted pictures in the top-left corner are relative pictures
The real-time absorbance change rates (slope of absorbance)
captured by a phone camera under the indoor light. in Fig. 5(c) and (f) for different ammonia concentrations were
calculated. The mean values with errors were plotted along
with ammonia concentrations to build the calibration plot,
as shown in Fig. 5(d) and (g). The small error bars in Fig. 5(d)
indicated the better stability of the 2-D sensor’s sensitivity.
The following equations linearly fit the data for two types of
sensors.
1) Two-dimensional sensor

Absorbance change rate (a.u./s)

= 2.434 × 10−6 (a.u./s/ppm) × concentration (ppm).
(4)

2) One-dimensional sensor

Absorbance change rate (a.u./s)

= 1.563 × 10−6 (a.u./s/ppm) × concentration (ppm).
(5)

The 2-D sensor’s calibration plot had higher R 2 and a larger

slope, suggesting that the 2-D sensor had higher sensitivity
and better accuracy for gas sensing than the 1-D sensor.

D. Two-Dimensional Diffusion-Based Colorimetric

Sensors for Continuous Quantitative NH3 Detection
We also tested the response of the same sensor to differ-
Fig. 5. One-dimensional sensor versus 2-D sensor for long-time ent concentrations of ammonia. As shown in Fig. 6(a), the
ammonia detection. (a) NH3 sensor exposed to different concentrations sensor was continuously exposed to varying concentrations
of NH3 . Images were captured before and after NH3 gas exposure.
The absorbance–time plots of (b) 2-D sensor and (e) 1-D sensor. The of NH3 . The concentration of NH3 gas increased from 1 to
absorbance of the entire sensor was monitored. The gray shade areas in 200 ppm and then decreased back to 1 ppm. The real-
the figures indicate the air purging period. (c) and (f) Derivative of plots time concentrations of ammonia can be calculated from the
in (b) and (e). The average smooth of every 100 data points was applied
to absorbance and derivative data for (c) and (f). Calibration plots of derivative of absorbance [see Fig. 6(b)] and calibration plots
(d) 2-D sensors and (g) 1-D sensors. [(4) and (5)]. The calculated concentrations of 1-D and 2-D
sensors were plotted in Fig. 6(c) and compared with the actual
concentration of NH3 . Compared to the 1-D sensor’s response,
As shown in Fig. 5(a), the gas inlet for the 2-D sensor the response of the 2-D sensor was more reproducible and
was at the center of the sensor with a radius of 0.125 mm. had a better correlation with the actual NH3 concentration.
The gas inlet for the 1-D sensor was located on the top end These results indicated that the 2-D sensor could continuously
with a width of ∼0.8 mm (≈0.25 × 3.14 mm, the same monitor NH3 for a long time.

Authorized licensed use limited to: VTU Consortium. Downloaded on February 22,2024 at 06:24:54 UTC from IEEE Xplore. Restrictions apply.
YU et al.: DIFFUSION-MODULATED COLORIMETRIC SENSOR FOR CONTINUOUS GAS DETECTION 11409

Fig. 6. One-dimensional and 2-D sensors for continuous NH3 mon-

itoring. (a) Absorbance changes with time. The gray shades indicate
different concentrations of NH3 exposure. The darker the shade, the
higher concentration. (b) Derivatives of absorbance curves in (a).
(c) Comparison of NH3 concentration monitored from 1-D sensors, 2-D Fig. 7. Two-dimensional sensor array for NH3 monitoring. (a) Color
sensors, and actual NH3 concentration. (d) Absorbance change rate change of 2-D sensor array after 20 min exposure to 50 ppm NH3 .
(slope of absorbance over time) versus NH3 concentration. The circles in the images indicated the locations of each 2-D sensing
area in the array. (b) Absorbance change of each 2-D sensing area was
recorded in real time. Absorbance was recorded from each indicated
circle region, as shown in (a). (c) Derivative of each absorbance curve
The slope of the absorbance–time curve during each NH3 in (b).
exposure period was calculated and plotted along with the
concentration in Fig. 6(d). The responses of the 2-D sensor
(red circles) to the same concentration of NH3 during two
separate exposures were very close to each other. These data
were linearly fit with the following equation with a higher R 2
of 0.9718:
Absorbance change rate (a.u./s)
= 3.248×10−6 (a.u./s/ppm) × concentration (ppm). (6)
While for the 1-D sensor (blue squares), the responses to the
same concentration of NH3 during two separate exposures
were not as consistent as the 2-D sensors, leading to a poor
R 2 of 0.6732. These results indicated that the 1-D sensor is
unsuitable for long-time continuous ammonia detection.
Fig. 8. Two-dimensional diffusion-based colorimetric sensor for environ-
mental NH3 detection. (a) Blank bottle and a bottle filled with fertilizer.
E. Reproducibility of the 2-D Diffusion-Based (b) Real-time response of 2-D sensor for blank air sample and fertilizer-
released NH3 sample detection. (c) Color development of 2-D sensor
Colorimetric Sensors after different gas sample exposures. (d) Correlation of the NH3 con-
The reproducibility of the 2-D diffusion-based colorimetric centrations readings from the 2-D sensor and reference NH3 sensor.
NH3 sensors was evaluated with a sensing array. The array
was created by creating multiple gas inlets on the same
sensor chip. The 2-D NH3 sensor array was exposed to Air Quality Standards set by Environmental Protection Agency
50 ppm NH3 , and the color development from each 2-D (EPA) (O3 : 0.07 ppm 8 h, SO2 : 0.5 ppm 3 h, and CO: 9 ppm
sensor in the array was recorded by webcam, as shown in 8 h).
Fig. 7. The areas with color development had very similar
sizes [see Fig. 7(a)]. The absorbance change rate (slope of F. Two-Dimensional Diffusion-Based Colorimetric Sensor
absorbance over 20 min) only had minimal variation (13% for Environmental NH3 Detection
error), as shown in Fig. 7(b). Fig. 7(c) showed that the slope of The 2-D colorimetric NH3 sensor was used to test the NH3
absorbance became stable after ∼5-min NH3 exposure, which released from organic fertilizer. NH3 released from fertilizer
had similar results as Fig. 5(c). These test results suggested is an important air pollutant, especially in agricultural states,
that the 2-D diffusion-based colorimetric NH3 sensor had such as South Dakota. The fertilizer was put in a closed
good reproducibility. We also checked the selectivity of the container in our test, as shown in Fig. 8(a). One bottle was
NH3 sensor with common air pollutants: O3 (3.8 ppm), SO2 blank, and the other bottle was filled with organic fertilizer
(18.6 ppm), and CO (50 ppm) (see Fig. S9), which shows that derived from chicken manure. The sampling gas was gen-
our NH3 sensor did not show significant responses to these erated by running ambient air through the bottles and then
interferents at the concentrations higher than National Ambient introduced to the detection chamber of the 2-D sensor. The

Authorized licensed use limited to: VTU Consortium. Downloaded on February 22,2024 at 06:24:54 UTC from IEEE Xplore. Restrictions apply.
11410 IEEE SENSORS JOURNAL, VOL. 23, NO. 11, 1 JUNE 2023

real-time color change was recorded and shown in Fig. 8(b). [7] X.-W. Huang et al., “Determination of pork spoilage by colorimetric
The data in Fig. 8(b) and images in Fig. 8(c) indicated that gas sensor array based on natural pigments,” Food Chem., vol. 145,
pp. 549–554, Feb. 2014.
the 2-D sensor only responded to gas samples generated [8] S. Matindoust, M. Baghaei-Nejad, M. H. S. Abadi, Z. Zou, and
from the bottle with fertilizer. The calculated concentration L.-R. Zheng, “Food quality and safety monitoring using gas sensor
of NH3 from Fig. 8(b) and (6) was plotted in Fig. 8(d). The array in intelligent packaging,” Sensor Rev., vol. 36, no. 2, pp. 169–183,
Mar. 2016.
concentration readings from the 2-D sensor were compared [9] E. Schaller, J. O. Bosset, and F. Escher, “‘Electronic noses’ and
with the readings from the reference NH3 sensor. Because the their application to food,” LWT-Food Sci. Technol., vol. 31, no. 4,
commercial NH3 sensor had a low resolution (1 ppm) and pp. 305–316, Apr. 1998.
detection limit (>1 ppm), its reading was stepwise. However, [10] Y. Chen et al., “Low cost smart phone diagnostics for food using paper-
based colorimetric sensor arrays,” Food Control, vol. 82, pp. 227–232,
the 2-D colorimetric NH3 sensor and reference NH3 sensor Dec. 2017.
showed very similar ammonia concentration profiles. The gas [11] D.-D. Lee and D.-S. Lee, “Environmental gas sensors,” IEEE Sensors J.,
samples from the bottle filled with fertilizer had a higher vol. 1, no. 3, pp. 214–224, Oct. 2001.
[12] K. A. Mirica, J. G. Weis, J. M. Schnorr, B. Esser, and T. M. Swager,
NH3 concentration due to the NH3 emission from fertilizer “Mechanical drawing of gas sensors on paper,” Angew. Chem., vol. 124,
derived from chicken manure. The results indicated that the no. 43, pp. 10898–10903, Oct. 2012.
2-D diffusion-based colorimetric NH3 sensor could be used [13] M. Khan, M. Rao, and Q. Li, “Recent advances in electrochemical
sensors for detecting toxic gases: NO2 , SO2 and H2 S,” Sensors, vol. 19,
for environmental ammonia monitoring. no. 4, p. 905, Feb. 2019.
[14] R. Baron and J. Saffell, “Amperometric gas sensors as a low
IV. C ONCLUSION cost emerging technology platform for air quality monitoring appli-
We have demonstrated that the sensing behavior of col- cations: A review,” ACS Sensors, vol. 2, no. 11, pp. 1553–1566,
Nov. 2017.
orimetric sensors can be modulated through gas diffusion
[15] C. Wang, L. Yin, L. Zhang, D. Xiang, and R. Gao, “Metal oxide gas
engineering. A typical 1-D diffusion-based colorimetric sensor sensors: Sensitivity and influencing factors,” Sensors, vol. 10, no. 3,
has a long lifetime but will experience a dramatic sensitivity pp. 2088–2106, Mar. 2010.
decrease after a long-time detection, because as diffusion dis- [16] H. C. Ji, W. Zeng, and Y. Q. Li, “Gas sensing mechanisms of metal
oxide semiconductors: A focus review,” Nanoscale, vol. 11, no. 47,
tance increases, the analyte concentration gradient in the color pp. 22664–22684, Oct. 2019.
developing area decreases, and the sensitivity of the sensor [17] A. Dey, “Semiconductor metal oxide gas sensors: A review,” Mater.
drops. By introducing a 2-D diffusion geometry, the decreasing Sci. Eng. B, Solid State Mater. Adv. Technol., vol. 229, pp. 206–217,
Mar. 2018.
analyte concentration gradient can be compensated with the [18] S. O. Agbroko and J. Covington, “A novel, low-cost, portable PID sensor
increasing available sensing probes, thus leading to a stable for the detection of volatile organic compounds,” Sens. Actuators B,
sensitivity. Our simulation and experimental results prove that Chem., vol. 275, pp. 10–15, Dec. 2018.
the 2-D diffusion-based colorimetric NH3 sensor has more [19] W. Xu et al., “New understanding of miniaturized VOCs
monitoring device: PID-type sensors performance evaluations in
stable and higher sensitivity. The 2-D diffusion-based colori- ambient air,” Sens. Actuators B, Chem., vol. 330, Mar. 2021,
metric sensor also shows a more linear response to different Art. no. 129285.
concentrations of NH3 . The 2-D sensor can continuously [20] G. Korotcenkov, “Materials for piezoelectric-based gas sensors,” in
Handbook of Gas Sensor Materials. New York, NY, USA: Springer,
monitor ppm NH3 with reproducible results. The usability of 2013, pp. 307–328.
the 2-D diffusion-based colorimetric sensor for environmental [21] K. Liu and C. Zhang, “Volatile organic compounds gas sensor based
NH3 monitoring has been demonstrated by detecting the NH3 on quartz crystal microbalance for fruit freshness detection: A review,”
Food Chem., vol. 334, Jan. 2021, Art. no. 127615.
released from organic fertilizers. This novel 2-D diffusion- [22] A. Azzouz, K. Vikrant, K.-H. Kim, E. Ballesteros, T. Rhadfi, and
based colorimetric sensor design can be adapted for other gas A. K. Malik, “Advances in colorimetric and optical sensing for gaseous
analytes monitoring and easily integrated with sensor array volatile organic compounds,” TrAC—Trends Anal. Chem., vol. 118,
pp. 502–516, Sep. 2019.
technology for air quality monitoring, biomarkers detection,
[23] Z. Li, J. R. Askim, and K. S. Suslick, “The optoelectronic nose:
food freshness evaluation, and hazardous chemical detection. Colorimetric and fluorometric sensor arrays,” Chem. Rev., vol. 119, no. 1,
pp. 231–292, Jan. 2019.
R EFERENCES [24] K. Kawamura, K. Miyazawa, and L. Kent, “The past, present and future
[1] A. Mirzaei, S. G. Leonardi, and G. Neri, “Detection of hazardous volatile in tube- and paper-based colorimetric gas detectors,” AppliedChem,
organic compounds (VOCs) by metal oxide nanostructures-based gas vol. 1, no. 1, pp. 14–40, Aug. 2021.
sensors: A review,” Ceram. Int., vol. 42, no. 14, pp. 15119–15141, [25] S. H. Cho, J. M. Suh, T. H. Eom, T. Kim, and H. W. Jang, “Colorimetric
Nov. 2016. sensors for toxic and hazardous gas detection: A review,” Electron.
[2] H. Wang, W. P. Lustig, and J. Li, “Sensing and capture of toxic and Mater. Lett., vol. 17, no. 1, pp. 1–17, Jan. 2021.
hazardous gases and vapors by metal–organic frameworks,” Chem. Soc. [26] M. C. Janzen, J. B. Ponder, D. P. Bailey, C. K. Ingison, and K. S. Suslick,
Rev., vol. 47, no. 13, pp. 4729–4756, Mar. 2018. “Colorimetric sensor arrays for volatile organic compounds,” Anal.
[3] F. Karagulian et al., “Review of the performance of low-cost sensors for Chem., vol. 78, no. 11, pp. 3591–3600, May 2006.
air quality monitoring,” Atmosphere, vol. 10, no. 9, p. 506, Aug. 2019. [27] X. Qin, R. Wang, F. Tsow, E. Forzani, X. Xian, and N. Tao,
[4] K.-H. Kim, S. A. Jahan, and E. Kabir, “A review of breath analysis “A colorimetric chemical sensing platform for real-time monitoring of
for diagnosis of human health,” TrAC—Trends Anal. Chem., vol. 33, indoor formaldehyde,” IEEE Sensors J., vol. 15, no. 3, pp. 1545–1551,
pp. 1–8, Mar. 2012. Mar. 2015.
[5] R. P. Arasaradnam, J. A. Covington, C. Harmston, and C. U. Nwokolo, [28] C. Lin et al., “High performance colorimetric carbon monoxide sensor
“Review article: Next generation diagnostic modalities in for continuous personal exposure monitoring,” ACS Sensors, vol. 3,
gastroenterology–gas phase volatile compound biomarker detection,” no. 2, pp. 327–333, Feb. 2018.
Alimentary Pharmacol. Therapeutics, vol. 39, no. 8, pp. 780–789, [29] J. Yu, X. Qin, D. Wang, C. Lin, X. Xian, and N. Tao, “Light-controlled
Apr. 2014. configurable colorimetric sensing array,” Anal. Chem., vol. 91, no. 10,
[6] A. Amann et al., “The human volatilome: Volatile organic compounds pp. 6632–6637, May 2019.
(VOCs) in exhaled breath, skin emanations, urine, feces and saliva,” [30] C. Lin et al., “Gradient-based colorimetric sensors for continuous gas
J. Breath Res., vol. 8, no. 3, Jun. 2014, Art. no. 034001. monitoring,” Anal. Chem., vol. 90, no. 8, pp. 5375–5380, Apr. 2018.

Authorized licensed use limited to: VTU Consortium. Downloaded on February 22,2024 at 06:24:54 UTC from IEEE Xplore. Restrictions apply.
YU et al.: DIFFUSION-MODULATED COLORIMETRIC SENSOR FOR CONTINUOUS GAS DETECTION 11411

[31] C. Lin, Z. Du, N. Tao, and D. Wang, “Gradient-based colorimetric array Chenwen Lin, photograph and biography not available at the time of
sensor for continuous monitoring of multiple gas analytes,” ACS Sensors, publication.
vol. 6, no. 2, pp. 439–442, Dec. 2020.
[32] S. N. Behera, M. Sharma, V. P. Aneja, and R. Balasubramanian, “Ammo-
nia in the atmosphere: A review on emission sources, atmospheric
chemistry and deposition on terrestrial bodies,” Environ. Sci. Pollut. Res.,
vol. 20, no. 11, pp. 8092–8131, Nov. 2013. Rachel Boylan, photograph and biography not available at the time of
[33] S. Bittman, J. R. Brook, A. Bleeker, and T. Bruulsema, “Air quality, publication.
health effects and management of ammonia emissions from fertilizers,”
in Air Quality Management. Dordrecht, The Netherlands: Springer,
2014, pp. 261–277.
[34] J. X. Warner, R. R. Dickerson, Z. Wei, L. L. Strow, Y. Wang, and
Q. Liang, “Increased atmospheric ammonia over the world’s major Chinmay Chandrakant Dixit, photograph and biography not available
agricultural areas detected from space,” Geophys. Res. Lett., vol. 44, at the time of publication.
no. 6, pp. 2875–2884, Mar. 2017.
[35] Y. Jiang, A. Deng, S. Bloszies, S. Huang, and W. Zhang, “Nonlinear
response of soil ammonia emissions to fertilizer nitrogen,” Biol. Fertility
Soils, vol. 53, no. 3, pp. 269–274, Apr. 2017.
[36] P. T. Arasu, A. L. Khalaf, S. H. A. Aziz, M. H. Yaacob, and Buddhi Sagar Lamsal, photograph and biography not available at the
A. S. M. Noor, “Optical fiber based ammonia gas sensor with carbon time of publication.
nanotubes sensing enhancement,” in Proc. IEEE Region Symp. (TEN-
SYMP), Cochin, India, Jul. 2017, pp. 1–4.
[37] R. W. Pinder, P. J. Adams, and S. N. Pandis, “Ammonia emission
controls as a cost-effective strategy for reducing atmospheric particulate
Wei He, photograph and biography not available at the time of
matter in the eastern United States,” Environ. Sci. Technol., vol. 41,
publication.
no. 2, pp. 380–386, Jan. 2007.
[38] A. N. Hristov et al., “Review: Ammonia emissions from dairy farms and
beef feedlots,” Can. J. Animal Sci., vol. 91, no. 1, pp. 1–35, Mar. 2011.
[39] K. Schmitt et al., “Colorimetric sensor for bad odor detection using
automated color correction,” in Proc. 8th Smart Sens., Actuators, MEMS, Francis Tsow, photograph and biography not available at the time of
Barcelona, Spain, 2017, pp. 348–357. publication.
[40] V. Nand and M. J. Ellwood, “A simple colorimetric method for deter-
mining seawater alkalinity using bromophenol blue,” Limnol. Oceanogr.,
Methods, vol. 16, no. 7, pp. 401–410, Jul. 2018.

Jingjing Yu received the B.S. degree in chemistry from Fuzhou Uni- Songxin Tan, photograph and biography not available at the time of
versity, Fuzhou, China, in 2014, and the Ph.D. degree in analytical publication.
chemistry from Nanjing University, Nanjing, China, in 2019.
She is a Postdoctoral Research Associate with the Department
of Electrical Engineering and Computer Science, South Dakota State
University, Brookings, SD, USA. Her current research interests include
environmental air quality monitor and mobile health devices. Yue Zhou, photograph and biography not available at the time of
publication.

Wei Ding, photograph and biography not available at the time of

publication. Xiaojun Xian (Member, IEEE) is an Assistant Professor with the Depart-
ment of Electrical Engineering and Computer Science, South Dakota
State University, Brookings, SD, USA. His research interests include
Laxmi Jaishi, photograph and biography not available at the time of from exploring advanced sensing materials, developing wearable sens-
publication. ing technologies, to building integrated wearable healthcare systems.

Authorized licensed use limited to: VTU Consortium. Downloaded on February 22,2024 at 06:24:54 UTC from IEEE Xplore. Restrictions apply.