Room-temperature ultra-sensitive trimethylamine and ethylenediamine assays for fish freshness detection based on an imine-covalent organic framework

Abstract

Covalent organic frameworks (COFs) are next-generation materials with pre-designed π-electronic backbones and highly ordered topologies that hold promise for use in fast gas sensing detection. Inspired by this, we synthesised imine COFs (COF_RT) using melamine and terephthaldehyde with the aid of a catalyst (Sc(OTf)₃) at room temperature. Based on density functional theory calculations, the COF_RT material was prepared with a stronger adsorption capacity for trimethylamine and ethylenediamine. Results showed that the COF_RT sensor exhibited excellent selectivity for trimethylamine and ethylenediamine and high responses (up to 3079.2% and 1540.3%, with detection limits as low as 28 ppb and 405 ppb, respectively), and it was able to complete the response recovery cycle within 18–26 s. In addition, the sensor was successfully used for monitoring fish freshness in real environments. Compared with solvothermal methods, the detection method using the COF_RT sensor was more energy efficient and offered a wider range of applications. In addition, this study provides valuable insights into the development of amine gas sensors using imine-based COFs and paves the way for the development of room-temperature synthesis of COFs materials as chemical gas sensors.

---

Introduction

The detection of harmful and toxic gases is becoming more and more important because of the increasing environmental pollution and the emphasis on healthcare. Amines are indispensable raw materials in the chemical, pharmaceutical and petroleum industries, and trimethylamine (TMA) is one of the very common amine analogue compounds.¹ TMA, as a toxic gas, poses serious threats to the environment and human health. Once the body inhales, consumes or absorbs this gas through the skin, it can cause headache, nausea, and coughing.^2,3 The concentration of TMA is considered to be an effective indicator for evaluating the quality of seafood.^4,5 Hence, high-performance TMA gas sensors are in high demand for applications in food testing, healthcare, and other fields. In addition, ethylenediamine (EDA) is a common toxicant in many factories, and even brief exposure of the human body to EDA can cause serious health effects, and in severe cases, it can even damage the liver and kidneys.⁶ The National Institute for Occupational Safety and Health (NIOSH) recommends occupational exposure EDA limits of not more than 10 ppm.⁷ Traditional instrumental analytical methods, such as gas chromatography and mass spectrometry, play an important role in the identification and detection of amines, but they are usually expensive, time-consuming, and can only be operated by trained personnel.⁸ In such cases, chemical sensors are inexpensive and reliable devices that can offer quick responses.^9–11 Moreover, most sensors currently employed for detecting TMA and EDA are fluorescence sensors and quartz crystal microbalance (QCM) sensors.^6,7,12–14 Rianjanu et al.¹² reported the preparation of citric acid-doped polyvinyl acetate (PVAc/CA) nanofibres on QCM chips, functioning as a highly sensitive and selective TMA gas sensor. Under TMA exposure conditions, the PVAc/CA nanofibre sensor demonstrated a detection sensitivity of 85.4 Hz ppm⁻¹ and an LOD of 19 ppb. There are currently very few chemical gas sensors in the market capable of detecting TMA.^4,11,15 Moreover, most of the current TMA sensor materials are prepared from metal oxides, which means that most of the sensors need high temperatures to operate, which is very unfavourable for real-time monitoring. In addition, today there is an extreme lack of EDA sensors in the market, and there is an urgent need to find suitable sensing materials to develop gas sensors that can detect TMA and EDA at room temperature (RT).

Considering the current diversification of sensing materials, covalent organic frameworks (COFs), which have a crystalline porous skeleton covalently connected by organic repeating units, combine the advantages of designability, periodic pore structure, environmental stability, and a large number of active sites.^16–18 COFs have been synthesized by solvothermal, ionothermal, and solvent-free mechanochemical syntheses, evaporation-assisted conversion, solid–gas interfacial reaction, and microwave-assisted and room temperature synthesis methods.¹⁹ The most common synthesis methods are solvothermal and ionothermal, but the preparation usually takes 2–9 days and requires high temperatures (80–120 °C). In contrast, room-temperature synthesis is simpler, but few materials have been successfully prepared.^20–23 Niu et al.²² constructed a series of imine-COFs (COF3) at RT that exhibited NH₃ sensing ability (response of up to 25% for 100 ppm NH₃). There are few examples of COFs being used to prepare chemical gas sensors²¹ and even fewer sensors for measuring TMA and EDA at RT. In conclusion, there is a need to develop COF-based gas sensors for fast, reliable, non-destructive, and real-time detection of trace TMA and EDA at RT.

In this study, imine-COFs (COF_RT) were synthesised from melamine and terephthalaldehyde (TA) by room-temperature synthesis with the aid of Sc(OTf)₃ catalyst (Scheme S1). Among them, the structure of COF_RT was prepared by taking advantage of the slow room-temperature reaction and the low molecular weight of melamine to prepare melamine encapsulated with COFs (Scheme 1). This structure can improve the utilisation of the surface of the sensing material to some extent. COF_RT was used as a gas-sensing material to detect TMA and EDA at RT. The COF_RT sensor exhibits good selectivity with a high response of 3079.2% and 1540.3% for 500 ppm TMA and EDA, respectively. The response/recovery time of one cycle was established to be 26 s, which enables real-time detection. The excellent sensing performance of the COF_RT sensor for TMA and EDA (adsorption energies up to −1.75 eV and −1.57 eV, respectively) is consistent with the conclusions of the density functional theory calculations. In addition, the sensor was successfully used to monitor the freshness of fish in a real environment. This work provides a useful reference for the preparation of real-time COF gas-sensitive materials and further development of COF-based gas sensors.

Scheme 1 Reaction process of melamine with TA to form COF_RT encapsulating the melamine structure.

Experimental

Materials and reagents

2,4,6-Triamino-1,3,5-triazine (melamine), terephthalaldehyde (TA), scandium(III) trifluoromethanesulfonate (Sc(OTf)₃) and dimethyl sulfoxide (DMSO) were purchased from McLean Reagent Company Co., Ltd, Shanghai, China. Acetone and diethyl ether were purchased from Sinopharm Chemical Reagent Co., Ltd, Beijing, China.

Preparation of COF_RT

First, 1.8 g of TA was added to 25 mL of DMSO and stirred until the TA was completely dissolved. Then, 1.13 g of melamine was added to 25 mL of DMSO and stirred vigorously with a magnetic stirrer for 6 h. After mixing the two solutions, 0.25 g of Sc(OTf)₃ was added and allowed to stand for 20 h. The prepared sample was then rinsed three times with acetone and ether, respectively, to remove unreacted chemicals and solvents. Finally, the obtained material was dried under vacuum (100 °C, 24 h). The sample obtained was named COF_RT.

Characterization

The surface and internal morphology of the COF_RT samples were investigated by field emission scanning electron microscopy (FESEM, Zeiss sigma300, Germany), transmission electron microscopy (TEM, FEI, Tecnai F, USA), X-ray powder diffraction (XRD, Bruker D8 Advance with CuKα radiation, Shimadzu XRD-6100, Japan) and FTIR (FTIR, Thermo Nicolet iS5, USA). X-ray photoelectron spectroscopy (XPS, ESCALAB 250Xi, USA) was used to determine the structure and functional group composition of the samples. The specific surface area and pore size of COF_RT were determined by measuring the N₂ adsorption–desorption isotherms (ASAP 2050, Micromeritics, Inc., USA), and the Brunauer–Emmett–Teller (BET) surface area was measured using a porosity analyzer. Gas sensing tests were carried out on a CGS–MT gas sensing measurement system (Beijing Sinoagg Co., Ltd).

Gas sensor fabrication and sensing measurements

The process used for the preparation of the COF_RT sensor is as follows: COF_RT was placed in an agate mortar, several drops of acetonitrile were added, and the mixture was ground gently. The mixture was then applied to the sensors (the sensor substrate (13 mm × 7 mm, 0.5 mm thick) was uniformly coated with platinum–palladium fork-finger electrodes with electrode widths and spacing of 0.2 mm) and dried at 25 °C for about 24 h to form the sensor films, named COF_RT sensors. The preparation of the target gas and the principle of gas testing were similar to previous reports.^9–20 The gas sensitivity test was performed with a CGS-MT multifunctional detection station at room temperature, and the RH% in the testing room was maintained at 35% ± 2% with an air conditioning system. The response to gas is defined as Response = ΔI/I_air = (I_gas − I_air)/I_air × 100%, where I_air and I_gas are the currents in air and target gas, respectively. The response time is defined as the time to reach 90% of the stabilized response value, and the recovery time is defined as the time to reach within 10% of the initial response value.

Results and discussion

Morphology and structure of the COF_RT compound

SEM characterization tests were performed on the samples to better understand their morphology. Fig. 1a–c shows the SEM images of COF_RT samples under different magnifications. It can be seen that COF_RT retains its blocky structure with many holes on the surface. This kind of morphology indicates that the sample has a good specific surface area. The holes on the surface of the sample help transport the target gas molecules. To further validate the SEM images, TEM images of the samples were also examined. The same COF_RT high magnification TEM image shows the lattice structure (Fig. 1d–f). As can be seen in Fig. 1f, the lattice stripes represent the lattice stripes of melamine, so it is reasonable to assume that the melamine has not yet been fully consumed to the extent that it exhibits a structure in which the melamine is encapsulated by COFs. This structure can improve the utilisation of the material surface and thus improve the gas-sensitive sensing performance.

Fig. 1 SEM (a)–(c) and TEM (d)–(f) images of samples at different magnifications of COF_RT, inset in (e) is the electron diffraction pattern corresponding to (f). XRD patterns (g), FTIR spectra (h), XPS survey spectrum (i) and XPS spectrum with high-resolution spectra (j)–(l) of COF_RT.

The crystal structure and purity of the synthesized composite microspheres were analyzed by XRD. The XRD spectrum of COF_RT is shown in Fig. 1g. The COF_RT samples all showed a crystal structure, and most of the diffraction peaks matched well with the characteristic peaks of melamine. The 25.8°, 27.3°/28.5° and 29.5° peaks correspond to the (−3 0 1), (−3 1 1) and (3 1 0) diffraction peaks, respectively.^24,25 The presence of melamine impurity peaks in COF_RT indicates that the formation of condensation polymers is incomplete, and that there is still unreacted melamine. This also verifies our conjecture that the micrometer structure of COF wrapped melamine is formed through effective structural design to promote the interfacial reaction. In addition, the absence of the heterogeneous peaks of COF_RT indicates an amorphous state, which confirms the formation of an amorphous polymer network, in accordance with literature reports.²⁵ It has been shown that a three-dimensional conductive network is formed in disordered imine-connected COFs samples, and that the imine-connected polymer framework remains Π-conjugated, which facilitates the passage of gas molecules through the potential barriers. This also provides a basis for the excellent gas-sensitizing properties of COF_RT.

In order to understand the chemical structure, the samples were examined by FTIR spectroscopy (Fig. 1h). The peak at 3348 cm⁻¹ corresponds to N–H.^26,27 The C–H bending and stretching vibration peaks occur at 1170 cm⁻¹ (olefinic C–H stretching from imine).²⁸ The peak at 1446 cm⁻¹ represents the C=C (in the benzene rings). The distinct C=N vibration bands at 1554–1660 cm⁻¹ indicate the occurrence of imine condensation, confirming the successful preparation of the imine-based COFs.²⁷ Additionally, the signal at 1734 cm⁻¹ (C=O) can be assigned to the residual aldehyde groups (mainly located at the margin of the 2D COF plane).²⁹ These characteristic peaks ascertained that the COF was successfully synthesized. The XPS tests were also performed to further verify the elemental composition of the samples. The total XPS of the COF_RT samples is shown in Fig. 1i. COF_RT is composed of three elements: C 1s (∼284.8 eV), N 1s (∼395.6 eV), and O 1s (∼528.4 eV). From Fig. 1j, it can be seen that the C 1s levels are categorized into three types: C–C/C=C, C=N/C–N and O–C=O.^20,30 The N 1s can be categorized at 398.8, 399.7, and 400.1 eV as pyridinic N (N₁), C=N, and pyrrolic N (N₃) (Fig. 1k).^27,31 The main component of O 1s are C–O and chemisorbed oxygen (O_C) (Fig. 1l).^32,33 The XPS results are consistent with the FTIR results, further demonstrating the successful synthesis of the imine COFs. In addition, the Brunauer–Emmett–Teller (BET) surface areas and pore sizes of COF_RT are also shown in Fig. S1. The BET surface area of COF_RT is 10.31 m² g⁻¹, and the average pore size is 5.03 nm. Large pore sizes are more conducive to the transport of gas molecules.

Gas sensing measurements

Further investigations established the gas sensitivity performance of the COF_RT sensor as well as its selectivity to the target gases. Fig. 2a shows the gas sensitivity and selectivity of the COF_RT sensor at RT to 500 ppm of TMA, EDA, aniline (C₆H₇N), ethyl acetate (C₄H₈O₂), 1-methyl-2-pyrrolidone (C₅H₉NO), toluene (C₇H₈), hexane (C₆H₁₄), trimethylene (C₉H₁₂), ammonia (NH₃), acetonitrile (C₂H₃N), formaldehyde (CH₂O), dimethyl sulfoxide (C₂H₆OS), ethanol (C₂H₆O), ozone (O₃), methanol (CH₄O), benzaldehyde (C₇H₆O), and acetone (C₃H₆O) at 11% RH, 64% RH and 95% RH. As can be seen from Fig. 2a, COF_RT only responds to six gases. Consequently, a correlation radar chart was plotted for the response values of these six gases (Fig. 2b). Additionally, the response value line chart, response time, and recovery time statistics are plotted for the six target gases (Fig. 2c). Fig. 2a shows the response of COF_RT to 500 ppm of the different gases. The highest response of 3079.22% was obtained for TMA, followed by EDA (1540.35%), with lower responses for the other gases. The COF_RT sensor response to TMA is twice as high as that to EDA, six times as high as the response to C₆H₇N, and 4–23 times as high as the response to the other interfering gases. Additionally, the COF_RT sensor exhibits favourable selectivity towards trimethylamine and ethylenediamine compared to other amine gases (NH₃ and C₆H₇N). This also shows the good selectivity of the COF_RT sensor to TMA and EDA. The results show that COF_RT sensors are very promising for the detection of TMA and DEA gases. The response/recovery time is another criterion for evaluating the goodness of the sensor. From the analysis of Fig. 2c, the COF_RT sensor can complete one response/recovery cycle for the target gases within 4–26 s. Primarily due to the holes in the surface of the COF_RT material that help to accelerate gas transport, the COF_RT sensor has a single response recovery cycle of 18 s for TMA and 26 s for EDA, which is faster than many TMA and EDA sensors on the market.^12,24 The COF_RT sensors are thus ideal for real-time monitoring of TMA and EDA.

Fig. 2 (a) Sensing curves of COF_RT sensors to 500 ppm of TMA, EDA, C₆H₇N, C₄H₈O₂, C₅H₉NO, C₇H₈, C₆H₁₄, C₉H₁₂, NH₃, C₂H₃N, CH₂O, C₂H₆OS, C₂H₆O, O₃, CH₄O, C₇H₆O, and C₃H₆O at 11% RH, 64% RH and 95% RH at RT. (b) Radar plot of response values and (c) response value line chart, response time statistics and recovery time statistics of COF_RT to 500 ppm of TMA, EDA, C₆H₇N, C₄H₈O₂, C₅H₉NO and C₇H₈.

Another indicator of gas sensor performance is humidity resistance. The response curves of the COF_RT sensor to 11%, 64% and 95% RH are shown in Fig. 2a. It can be seen that the COF_RT sensor does not respond to 11% RH. The COF_RT sensor response to 64% and 95% RH is 101.4% and 10,000%, respectively. The immunity of the COF_RT sensor to humidity is significantly better than that of the other COF-based gas sensors. The humidity in a normal factory environment is controlled at about 50% RH, so COF_RT can meet the needs of actual TMA detection under normal humidity. Furthermore, in order to assess the limit of detection (LoD) of the COF_RT sensor for TMA, Fig. 3a and c illustrate the response curves of the COF_RT sensor for 50, 100, 200, 300, 400 and 500 ppm TMA and EDA at RT. The response of the COF_RT sensor at 50, 100, 200, 300, 400 and 500 ppm TMA was 736.95%, 1230.1%, 1622.82%, 2236.88%, 2735.33% and 3069.32%, respectively. The response of the COF_RT sensor at 50, 100, 200, 300, 400 and 500 ppm EDA is 153.44%, 415.79%, 599.81%, 1020.18%, 1227.17% and 1523.03%, respectively. Fig. 3b and d show the linear fitting curves of the COF_RT sensor response to different concentrations of TMA. Based on LoD = 3S_D/m, where S_D is the standard deviation of the noise of the response curve (S_D denotes the baseline data for the COF_RT sensor response in air, calculated using the root mean square deviation from 30 data points, S_D = 0.0027) and m is the slope.³¹ The m values of the COF_RT response curves for TMA and EDA at different concentrations were 0.05 and 0.02, respectively. The LoDs of the COF_RT sensor for TMA and EDA were calculated to be approximately 165 ppb and 405 ppb, respectively. The NIOSH has recommended short-term airborne exposure limits of 15 ppm for TMA and 10 ppm for EDA.^7,25 This indicates that the theoretical detection limits of the COF_RT gas sensor for TMA and EDA are well below the concentration ranges recommended by the NIOSH. It shows that the COF_RT sensor is expected to detect TMA and EDA in practical applications. At the same time, repeatability and long-term stability are also indicators that are used to evaluate the performance of sensors. Fig. 3e shows the response recovery cycle plot of the COF_RT sensor against 500 ppm TMA for 9 times. The average value of the COF_RT response to 500 ppm TMA is 3056.83%. The fluctuation of the response of the COF_RT sensor to TMA is measured to be 1%, which shows good response and recovery characteristics, indicating that the COF_RT sensor has good repeatability. After 15 or 30 days, the COF_RT sensor could still complete 3 response/recovery cycles (Fig. 3f), and the response to 500 ppm of TMA only decreased by 3.2% and 11.2%, respectively, showing good long-term stability. The experimental results further demonstrate the potential value of the COF_RT sensor in TMA detection applications.

Fig. 3 (a) Dynamic response curves, (b) fitting curves of the COF_RT sensor for different concentrations of TMA. (c) Dynamic response curves, (d) fitting curves of the COF_RT sensor for different concentrations of EDA. (e) Dynamic response curves of 9 response/recovery cycles of the COF_RT sensor to 500 ppm TMA and (f) dynamic response curves of the COF_RT sensor to TMA at a concentration of 500 ppm for response/recovery cycles after 0, 15 and 30 days.

To further evaluate the TMA and EDA gas-sensing properties of the COF_RT sensor composites, various TMA and EDA gas sensors based on different materials are summarized and compared in Table 1. Compared with MXene@Au DN hydrogel,³⁵ V₂O₃–Cu₂O,³⁶ MoO₃/V₂O₅,³⁷ and WS₂/PA,³⁸ the COF_RT sensor has significant advantages in terms of response/recovery time and gas performance. Furthermore, most of the TMA gas sensors need to be used at high temperatures, whereas the COF_RT sensor can be employed at room temperature to monitor TMA and EDA. In addition, the COF_RT sensor has a detection limit of 280 ppb, which is lower than most of the TMA sensors on the market. Given that there are only a few examples of EDA gas sensors on the market, this work provides food for thought for developing a market for EDA gas sensors.

Table 1 Comparison of the present TMA and EDA sensor based on COF_RT with other sensors

Sensing material	Analyte/concentration (ppm)	Response	T Res/TRec (s)	Temp. (°C)	LoD (ppm)	Ref.
WS2/PAA	TMA/1	6	107/194	0	0.02	38
PbMoO4/MoO3	TMA/20	33.2	12/18.8	133	0.0013	34
MXene@Au DN hydrogel	TMA/10	∼7%	107/193	RT	5	35
V2O3-Cu2O	TMA/50	∼15	62/128	30	3	36
MoO3/V2O5	TMA/0.5	∼52%	54/58	RT	0.02	37
Au/WO3 nanosheets	TMA/25	217.12	∼8/6	300	0.5	39
MoO3/rGO composites	EDA/100	843.7	∼8/357	RT	0.235	15
COFRT	TMA/500	3087.5%	∼16.1/1.6	RT	0.028	This work
	EDA/500	1540.3%	24.5/1.4	RT	0.405

---

Monitoring the freshness of fish

Monitoring the freshness of fish is becoming increasingly important as people become more aware of food safety. Currently, fish and other aquatic products are prone to deterioration during transportation or market sale, resulting in economic losses. In addition, accidental ingestion of spoiled fish can cause foodborne poisoning in humans, seriously affecting the health of consumers.⁴⁰ According to previous reports, fish spoilage releases TMA.^41,42 To evaluate the potential of the COF_RT sensor to monitor fish-freshness, a homemade test system was used to detect TMA released during the storage of 80 g of fish for 7 days at about 25 °C, as shown in Fig. 4a. Fig. S2 shows photographs of basa fish at different storage periods for the COF_RT sensor. It was found that the physical properties of the fish (color and softness) changed with storage time, with the flesh darkening in color and the muscle tissue becoming softer (indicating the onset and exacerbation of deterioration of the fish). Fig. 4b–i show the response-recovery transient curves of the COF_RT sensors for balsa fish (recorded 0–7 days). The images show that the signal of the gas sensor to the released volatile gases varies more as the storage period increases. In addition, the high humidity content of the fish may lead to humidity variations throughout the test, resulting in unreliable output signals. Therefore, humidity values were monitored during the testing process. First, the precise relative humidity of the target air stream coming out of the bottle containing the balsa fish fillets was determined using a commercial humidity sensor (the humidity at the time of the test is shown in the graphs of Fig. 4b–i). It can be seen that the response of the COF_RT sensor to the fish samples from 0–7 days can reach 14 040.23%, 77 824.7%, 118 114.08%, 189 577.77%, 395 322.65%, 848 736.45%, 7 651 871.65%, and 9 947 774.87%, respectively. As shown in Fig. 2a, the COF_RT sensor exhibits a response value of 10 000% at 95% relative humidity. Consequently, the measured fish freshness response values in high-humidity environments demonstrate a significant disparity compared to those in low-humidity conditions. In addition, under high humidity conditions, the concentration of TMA in 80 g of salsa fish reached 1000 ppm on day 0 and 13 888 ppm on day 1, based on the fitted curves, which showed that the concentration of TMA increased with time. Moreover, the response of the COF_RT sensor to the TMA released from the salsa fish increased rapidly from the sixth day onwards, which corresponds to the fact that the fish started to become unshapely, decomposing into a puddle, and the sensor electrodes became rusty on the sixth day, as can be seen from Fig. S2. This may be due to the sharp increase in spoilage of the fish in the presence of a large number of microorganisms, and the fact that we used thawed balsa, so the freshness of the fish before freezing was not known. The day 0 test was conducted when the fish was already completely thawed and the surface water was wiped dry. On this basis, it has also been shown in the literature that fish are in the early spoilage stage on days 3 and 4, and completely spoiled on days 5 and 8.⁴¹ This experiment further demonstrates the potential of the COF_RT sensor for practical applications, because it is possible to detect the level of TMA released and assess the freshness of the fish meat.

Fig. 4 (a) Block diagram of the fish freshness evaluation system. (b)–(i) Response-recovery curves of the COF_RT sensor to the spoilage volatile gas of fish stored for 0–7 days at 25 °C.

With the continuous elevation of quality standards for seafood by governments and industries, ongoing innovation in spoilage detection technologies for seafood products is increasingly prevalent. The primary compound released during fish spoilage is TMA, and most current research approaches determine fish freshness by measuring its concentration (TMA is a primary marker of fishy odour).^{2,4–6,12–14,32,34} Current mainstream techniques for detecting seafood spoilage include gas chromatography, electronic noses, microbial testing, and pH measurement. Compared to other detection technologies, COF_RT sensors demonstrate superior response and recovery speeds relative to other TMA gas sensors on the market. Furthermore, the COF_RT sensor's high sensitivity enables real-time monitoring, offering significant potential for application in seafood safety surveillance. However, this sensor is not without limitations: it exhibits extreme sensitivity to environmental fluctuations and is susceptible to interference from other gases.

In summary, the COF_RT sensors offer significant potential for seafood freshness analysis due to their high sensitivity and rapid response recovery. However, interference from humidity also limits the practical application scope to a certain extent. This aspect can be addressed in subsequent research by applying hydrophobic coatings to the sensor surface, thereby mitigating the impact of humidity during testing.

Possible sensing mechanism

In this study, the highly gas-sensitive characteristics of the COF_RT materials can be attributed to several aspects. To further investigate the sensing mechanism of the COF_RT sensor for TMA and EDA, we performed density-functional theory (DFT) calculations using the CP2K, details of which are given in the SI. We chose a structural fragment of an imine-linked COF ring as a model molecule to explore the interaction of TMA and EDA molecules with imine-linked COF. We optimised the TMA and EDA model molecules with a single molecule of the imine-linked COFs. A cyclic molecule consisting of imine bonds and benzene was used as the COF model. Two complex systems, COF model molecule/TMA and COF model molecule/EDA, were calculated. Fig. 5 shows the geometries of the optimized COF model molecule alone and with adsorbed TMA or EDA molecules. From the optimized geometries, it is observed that there is a bond between the TMA or EDA and the benzene ring between the C and N atoms (circled in red in Fig. 5). Table 2 lists the calculated adsorption energies and the shortest atom-to-atom distances. The shortest atom-to-atom distances between the COF molecule and TMA and EDA are 1.744 Å and 1.674 Å, respectively. This distance represents the covalent bond between C and N. The adsorption energies of COF molecules on TMA and EDA are −1.74 eV and 1.57 eV, respectively. From the data shown in Table 2, it is clear that the adsorption energy of the COF model molecule/TMA system is higher than that of the COF model molecule/EDA system. This may be the key reason for the higher inductive selectivity of the COF_RT sensor for TMA compared to EDA. It has been reported that the sensing phenomenon arises due to the hydrogen bonding process between NH₃ and the imine bond, and any COFs containing imine bonds should have NH₃ sensing capability.²² Both TMA and EDA are ammonia derivatives, so imine-COFs have the ability to detect TMA/EDA. During COF_RT detection of TMA or EDA, hydrogen bonds may form between TMA/EDA and the imine bond. This hydrogen bonding interaction is reversible; upon removal of TMA/EDA, the corresponding charge transfer phenomenon ceases. Consequently, the COF_RT sensor unit can be reused for TMA/EDA identification. This is why the COF_RT sensor unit can be reused to recognize TMA/EDA. This hydrogen bonding is also the reason why the COF_RT sensor has a higher response value to TMA and EDA than to other interfering gases.²²

Fig. 5 (a) Structural view of the optimized COF_RT material (red circles show simulated TMA and EDA optimized adsorption sites), (b) optimized cell expansion; (c) and (d) two views of the COF interacting with TMA and EDA, respectively (red circled part indicate the interaction bond between C and N).

Table 2 Adsorption energy (E_a) and equilibrium distance (d) of TMA and EDA adsorbed on the imine-linked COF model molecule

System	E a (eV)	d (Angstrom)
a E a is defined as Ea = ECOF/TMA – ECOF – ETMA, where E is the calculated thermal enthalpies. The same calculation is used for EDA. b d is defined as the shortest atom-to-atom distance between the imine-linked COF model molecule and TMA or EDA.
COF model molecule/TMA	−1.74	1.744
COF model molecule/EDA	−1.57	1.674

---

From the structural design point of view, due to the small molecular weight of melamine and the slow reaction rate at room temperature, we first allowed melamine to form a covalent organic framework with TA, and then formed micrometer structures surrounded by COFs around the melamine (Fig. 6a). According to the SEM image representation (similar to a sphere), ideally 9 × 10⁵ spheres can be lined up on the upper layer of the sensor chip, and the surface exposure ratio could reach about 12.6. It is well known that most of the gas sensing process occurs at the surface and interface of the sensing material. Therefore, the structure of the melamine encapsulated by the COF can effectively increase the sensing area and thus improve the sensing response.⁴³

Fig. 6 Schematic of the COF_RT sensor (a) with its TMA and EDA sensing mechanism (b).

A schematic of the overall process occurring during the COF_RT sensor operation is shown in Fig. 6b. The COF_RT sensor shows an increasing response to the current of TMA/EDA (a reducing gas) and is a typical n-type semiconductor. The specific gas sensing characteristics of n-type semiconductors are closely related to the chemical adsorption of target gas molecules and the resulting changes in sensor resistance. When the COF_RT sensor is placed in air, a large number of oxygen molecules are adsorbed on the surface of COF_RT and the surface electronic states are redistributed. These redistribution regions are often referred to as space charge regions, and they determine the frequency of electron transfer. According to Morrison's gas sensing model, adsorbed oxygen molecules capture a large number of electrons from the COF_RT sensor crystal, generating chemically adsorbed oxygen ions and thereby forming a charge depletion layer. The increased thickness of the depletion layer leads to a heightened potential barrier height, which, in turn, results in reduced carrier concentration and conductivity. Eqn (1) and (2) represent the overall reaction process:²¹

O₂ → O₂ (ads) (1)

O₂ (ads) + e⁻ → O₂⁻ (2)

when the COF_RT sensor is placed in a TMA/EDA gas environment, the TMA/EDA molecules and adsorbed O₂⁻ react to release trapped electrons back into the conduction band (Fig. 6b). This results in a narrowing of the depletion layer and an increase in carrier concentration and conductivity. The reaction proceeds as shown in eqn (3)³² and eqn (4):¹⁵

4N(CH₃)₃ + 21O₂ (ads)⁻ → 2N₂ + 18H₂O + 12CO₂ + 21e⁻ (3)

C₂H₈N₂ + 4O₂ (ads)⁻ → 2CO₂ + 4H₂O + N₂ + 4e⁻ (4)

In addition, the nitrogen atoms in TMA have a higher cloud density due to the electron assignment of the –CH₃ group in TMA. As well as the TMA being more basic and reducing, it is easier to lose electrons, which means that more electrons are transferred from the TMA to the COF_RT sensor, resulting in a change in the concentration of COF_RT carriers. As a result, the resistance of the COF_RT sensor device changes, generating a sensing signal. The above three factors contribute to the better response of COF_RT sensors to TMA and EDA.

Conclusion

This study successfully synthesised an imine covalent organic framework (COF_RT) composed of melamine and terephthalaldehyde at room temperature using Sc(OTf)₃ as a catalyst. By utilising the covalent bonds and hydrogen bonding interactions formed between COF_RT and amine gases (TMA and EDA), COF_RT achieves relatively stable detection of TMA and EDA. COF_RT exhibits high response values, rapid response-recovery, and low detection limits for TMA and EDA. Furthermore, the COF_RT sensor has been successfully applied to assess fish freshness, demonstrating its potential for practical market applications in this field. Subsequently, the sensing mechanism of COF_RT towards TMA and EDA was analysed. Density functional theory was employed to validate the role of COF_RT in detecting TMA and EDA. This study provides a reference for amine gas sensor detection based on imine-type COFs and lays the foundation for developing further COF-based chemical gas sensors.

Author contributions

Weiyu Zhang: conceptualization, writing–original draft, investigation, formal analysis, data curation, writing – review & editing, funding acquisition. Qihua Sun: writing – review & editing, supervision, methodology, funding acquisition, investigation. Ping Hu: Data curation, writing – review & editing, methodology, investigation. Ning Tian: writing – review & editing, methodology. Zhaofeng Wu: writing – original draft, supervision, project administration, resources, funding acquisition.

Conflicts of interest

There are no conflicts to declare.

Data availability

The datasets supporting this article have been uploaded as part of the supplementary information (SI). Supplementary information is available. See DOI: https://doi.org/10.1039/d5tc03161c.

Acknowledgements

This research was funded by the National Natural Science Foundation of China (21964016, 52563033), the Xinjiang Natural Science Fund for Distinguished Young Scholars (2022D01E37), the Xinjiang Tianshan Talent Project (2024TSYCCX0007), the Tianshan Talent Training Project-Xinjiang Science and Technology Innovation Team Program (2023TSYCTD0012), the Xinjiang University Outstanding Doctoral Student Innovation Program (XJU2023BS027), the Autonomous Region “Tianchi Talent” Introduction Program Youth Doctoral Program (5105240151p) and the Natural Science Foundation of Xinjiang Uygur Autonomous Region (2024D01C237). Additionally, we extend our gratitude to the China Association for Science and Technology's Youth Talent Support Programme, Doctoral Student Special Scheme.

References

1. S. Yu, J. Dong, H. Wang, S. Li, H. Zhu and T. Yang, J. Mater. Chem. A, 2022, 10, 25752,  10.1039/d2ta90267b.
2. X. Zhai, Y. Sun, S. Cen, X. Wang, J. Zhang, Z. Yang, Y. Li, X. Wang, C. Zhou, M. Arslan, Z. Li, J. Shi, X. Huang, X. Zou, Y. Gong, M. Holmes and M. Povey, Food Hydrocolloid., 2022, 133, 107989,  DOI:10.1016/j.foodhyd.2022.107989.
3. S. Chen, X. Duan, C. Liu, S. Liu, P. Li, D. Su, X. Sun, Y. Guo, W. Chen and Z. Wang, J. Hazard. Mater., 2024, 467, 133672,  DOI:10.1016/j.jhazmat.2024.133672.
4. P. Wang, Z. Zheng, X. Cheng, L. Sui, S. Gao, X. Zhang, Y. Xu, H. Zhao and L. Huo, J. Mater. Chem. A, 2017, 5, 19846–19856,  10.1039/c7ta06392j.
5. X. Huang, M. Li, X. Xu, H. Chen, H. F. Ji and S. Zhu, Biosens. Bioelectron., 2011, 30, 140–144,  DOI:10.1016/j.bios.2011.09.001.
6. A. E. D. A. Bekhit, S. G. Giteru, B. W. B. Holman and D. L. Hopkins, Comput. Rev. Food Sci. F, 2021, 20, 3620–3666,  DOI:10.1111/1541-4337.12764.
7. Y. Ke, Y. Liu, B. Zu, D. Lei, G. Wang, J. Li, W. Ren and X. Dou, Angew. Chem., Int. Ed., 2022, 61, 202203358,  DOI:10.1002/ange.202203358.
8. P. M. Chuang, Y. J. Tu and J. Y. Wu, Sens. Actuators, B, 2022, 366, 131967,  DOI:10.1016/j.snb.2022.131967.
9. Z. Qin, Z. Wu, Q. Sun, J. Sun, M. Zhang, T. Shaymurat, C. Lv and H. Duan, Chem. Eng. J., 2023, 462, 142203,  DOI:10.1016/j.cej.2023.142203.
10. P. F. M. Pereira, P. H. de Sousa Picciani, V. Calado and R. V. Tonon, Trends Food Sci. Tech., 2021, 118, 36–44,  DOI:10.1016/j.tifs.2021.08.036.
11. L. X. Ou, M. Y. Liu, L. Y. Zhu, D. W. Zhang and H. L. Lu, Nano-Micro Lett., 2022, 14, 206,  DOI:10.1007/s40820-022-00956-9.
12. A. Rianjanu, R. Aflaha, N. I. Khamidy, M. Djamal, K. Triyana and H. S. Wasisto, Mater. Adv., 2021, 2, 3705–3714,  10.1039/d1ma00152c.
13. R. Aflaha, L. Katriani, A. H. As’ari, N. L. I. Sari, A. Kusumaatmaja, A. Rianjanu, R. Roto and K. Triyana, MRS Commun., 2023, 13, 664–672,  DOI:10.1557/s43579-023-00409-3.
14. A. Rianjanu, M. A. A. P. Rayhan, M. Aulya, R. Aflaha, K. Triyana, T. Taher and H. S. Wasisto, Chem. Select., 2024, 9, e202304694,  DOI:10.1002/slct.202304694.
15. Y. Yusran, H. Li, X. Guan, D. Li, L. Tang, M. Xue, Z. Zhuang, Y. Yan, V. Valtchev and S. Qiu, Adv. Mater., 2020, 32, 1907289,  DOI:10.1002/adma.201907289.
16. X. Liu, D. Huang, C. Lai, G. Zeng, L. Qin, H. Wang, H. Yi, B. Li, S. Liu and M. Zhang, Chem. Soc. Rev., 2019, 48, 5266–5302,  10.1039/c9cs00299e.
17. J. L. Segura, S. Royuela and M. M. Ramos, Chem. Soc. Rev., 2019, 48, 3903–3945,  10.1039/c8cs00978c.
18. W. Zhang, S. Liu, Q. Sun, N. Tian and Z. Wu, Nano Res., 2023, 17, 162–195,  DOI:10.1007/s12274-023-6027-x.
19. Y. Li, X. Guo, X. Li, M. Zhang, Z. Jia, Y. Deng, Y. Tian, S. Li and L. Ma, Angew. Chem., 2020, 132, 4197–4204,  DOI:10.1002/ange.201916360.
20. W. Zhang, Q. Sun, Y. Zhu, J. Sun, Z. Wu and N. Tian, ACS Sens., 2024, 9, 3262–3271,  DOI:10.1021/acssensors.4c00613.
21. F. Niu, Z. W. Shao, J. L. Zhu, L. M. Tao and Y. Ding, J. Mater. Chem. C, 2021, 9, 8562–8569,  10.1039/d1tc01662h.
22. W. Zhang, W. Wang, S. Liu, N. Tian, Q. Sun and Z. Wu, Appl. Surf. Sci., 2024, 649, 159114,  DOI:10.1016/j.apsusc.2023.159114.
23. H. Shanavaz, B. P. Prasanna, S. Archana, M. K. Prashanth, F. A. Alharthi, R. Zhou, M. S. Raghu, B. H. Jeon and K. Y. Kumar, J. Energy Storage, 2023, 67, 107561,  DOI:10.1016/j.est.2023.107561.
24. P. Najmi, N. Keshmiri, M. Ramezanzadeh, B. Ramezanzadeh and M. Arjmand, ACS Appl. Mater. Interfaces, 2022, 14, 54141–54156,  DOI:10.1021/acsami.2c14542.
25. W. Liu, T. Aldahri, B. Liang and L. Lv, Sep. Purif., 2020, 251, 117285,  DOI:10.1016/j.seppur.2020.117285.
26. X. Lin, P. Peng, J. Guo and Z. Xiang, Chem. Eng. J., 2019, 358, 427–434,  DOI:10.1016/j.cej.2018.09.185.
27. Z. Yang, A. Mei, W. Chen, Z. Wang, H. Guo and Y. Liu, Sens. Actuators, B, 2023, 392, 134051,  DOI:10.1016/j.snb.2023.134051.
28. O. S. Gonsalves and P. R. Nemade, Chemosphere, 2024, 351, 141346,  DOI:10.1016/j.chemosphere.2024.141246.
29. C. S. Diercks and O. M. Yaghi, Science, 2017, 355, 6328,  DOI:10.1126/science.aal1585.
30. S. Zuo, S. Zhu, J. Wang, W. Liu and J. Wang, Appl. Catal., B, 2022, 301, 120783,  DOI:10.1016/j.apcatb.2021.120783.
31. Z. Qin, Z. Wu, Q. Sun, J. Sun, M. Zhang, F. Chen, D. Zhang, C. Lv and H. Duan, Carbon, 2023, 213, 118297,  DOI:10.1016/j.carbon.2023.118297.
32. B. Ran, P. Hu, J. Sun, J. Fang, Q. Sun, J. Wang, Y. Zhu, N. Tian, Z. Wu and H. Duan, J. Hazard. Mater., 2024, 480, 136289,  DOI:10.1016/j.jhazmat.2024.136289.
33. F. D. Zhang, K. W. Liu, H. L. Li, S. H. Cui, D. Z. Zhang, J. B. Zeng and Z. F. Yan, ACS Appl. Mater. Interfaces, 2022, 14, 24610–24619,  DOI:10.1021/acsami.2c04722.
34. X. Li, L. Jin, A. Ni, L. Zhang, L. He, H. Gao, P. Lin, K. Zhang, X. Chu and S. Wang, ACS Appl. Mater., 2022, 14, 30182–30191,  DOI:10.1021/acsami.2c06749.
35. V. Mounasamy, G. K. Mani, D. Ponnusamy, K. Tsuchiya, A. K. Prasad and S. Madanagurusamy, Ceram. Int., 2019, 45, 19528–19533,  DOI:10.1016/j.ceramint.2019.06.074.
36. X. Sui, D. Zhang, J. Wang, M. Tang, H. Xia and Z. Wang, J. Alloys Compd., 2023, 968, 172104,  DOI:10.1016/j.jallcom.2023.172104.
37. S. H. Park, B. Y. Kim, Y. K. Jo, Z. Dai and J. H. Lee, Sens. Actuators, B, 2020, 309, 127805,  DOI:10.1016/j.snb.2020.127805.
38. D. Z. Zhang, D. D. Zhou, H. Y. Mi, Z. J. Wang, P. F. Zhang and G. S. Xi, Sens. Actuators, B, 2024, 417, 136188,  DOI:10.1016/j.snb.2024.136188.
39. C. Zhao, J. Shen, S. Xu, J. Wei, H. Liu, S. Xie, Y. Pan, Y. Zhao and Y. Zhu, Food Chem., 2022, 392, 133318,  DOI:10.1016/j.foodchem.2022.133318.
40. X. Liu, K. Chen, J. Wang, Y. Wang, Y. Tang, X. Gao, L. Zhu, X. Li and J. Li, Food Chem., 2020, 307, 125580,  DOI:10.1016/j.foodchem.2019.125580.
41. R. S. Andre, L. A. Mercante, M. H. M. Facure, R. C. Sanfelice, L. Fugikawa-Santos, T. M. Swager and D. S. Correa, ACS Sens., 2022, 7, 2104–2131,  DOI:10.1021/acssensors.2c00639.
42. P. Hao, Z. Lin, P. Song, Z. Yang and Q. Wang, Ceram. Int., 2020, 46, 9363–9369,  DOI:10.1016/j.ceramint.2019.12.194.
43. C. Yuan, J. Ma, Y. Zou, G. Li, H. Xu, V. V. Sysoev, X. Cheng and Y. Deng, Adv. Sci., 2022, 9, 2203594,  DOI:10.1002/advs.202203594.

---