Semiconductor sensor for rapid identification of gas molecules and methods of making and using the same

Abstract

The application belongs to the technical field of sensing, and particularly relates to a semiconductor sensor capable of rapidly identifying gas molecules and a preparation and use method thereof. A gas sensor based on an oxide nanorod sensitive film is obtained by vacuum glancing angle sputtering of a metal oxide nanorod array film on a low-thermal-conductivity insulating substrate plated with an interdigital electrode; the sensitive film is simultaneously used as a heating layer, the working temperature of the sensitive film is controlled by Joule heat generated by self-heating of the sensitive film, a transient response signal of the sensor under self-heating modulation is collected, a machine learning algorithm is used to extract features of different gas molecules, and gas molecule identification application is performed. Since the sensitive film is simultaneously used as a heating film, the heat conduction rate is fast, the sensor can reach a set temperature within ms, and therefore, rapid self-heating modulation can be performed for molecule identification, meeting the demand of rapid identification of target gas under the condition of rapid dynamic change of outdoor polluted gas. In addition, the preparation method is compatible with conventional micro-nano etching process, and meets the demand of uniformity and batch preparation of the sensor.

Description

Technical Field

[0001] This invention relates to the field of sensing technology, specifically to a semiconductor sensor capable of rapidly identifying gas molecules, and its preparation and usage methods. Background Technology

[0002] Metal-oxide-semiconductor (MOF) gas sensors are ideal units for building IoT gas sensors for gridded monitoring and early warning due to their advantages such as good material stability, small size, extremely low cost, silicon process compatibility, and ease of gridding. With the continuous emergence of new materials (especially two-dimensional materials and metal-organic frameworks, MOFs), the development of advanced defect characterization and control technologies, and the rapid advancement of fine nanofabrication techniques and machine learning algorithms, several key performance indicators of MOF gas sensors (such as sensitivity, selectivity, stability, moisture resistance, response and recovery time) have made significant progress in the past decade.

[0003] In particular, the selectivity bottleneck restricting the application of semiconductor sensors has been addressed by using a sensor heater for variable-temperature thermal modulation testing. This method has proven to be an efficient way to uncover the intrinsic characteristics of different analyte gas molecules, accurately identifying volatile organic compounds (VOCs) with similar structures and properties, and even isomers with the same molecular weight and functional groups. For example, Chinese invention patent CN109709163A discloses a P-type metal oxide gas sensor and its preparation and use method; and Chinese invention patent CN113030194A discloses a gas sensor array and a method for identifying three types of alcohols and their concentrations.

[0004] Although thermal modulation significantly improves the selectivity of semiconductor gas sensors, current methods primarily rely on external heaters integrated at the bottom or periphery of the sensitive layer. The time required for the external heater to heat to the set temperature (i.e., thermal relaxation time) is related to the sensor configuration (heater, insulating layer, sensitive layer position, material, size, etc.). For conventional ceramic plate heaters (alumina ceramic plates with a side length of 1.5-2 mm and a thickness of 0.25-0.3 mm), the thermal relaxation time is typically 3-5 seconds. Therefore, when the sensor performs variable-temperature thermal modulation identification, it takes at least 4-5 seconds to complete transient signal acquisition, leading to the following drawbacks: First, the thermal modulation transient signal... The long acquisition time (>4s) limits the sensor to static or quasi-static conditions (where the type and concentration of the target gas change slowly), restricting its application in situations with rapid dynamic changes in the target gas (such as outdoor or indoor environments with airflow). Secondly, the long testing time reduces the sensor's sampling rate (number of identifications per unit time), limiting its ability to track rapid changes in the target gas in gridded monitoring applications. Thirdly, existing thermal modulation methods primarily utilize heaters integrated into the sensor, requiring heater integration during sensor fabrication. This restricts the use of flexible substrates and significantly increases device power consumption and fabrication costs, hindering wireless and gridded monitoring applications. Summary of the Invention

[0005] One objective of this invention is to provide a method for fabricating a semiconductor sensor capable of rapidly identifying gas molecules, comprising the following steps:

[0006] S1, Take a piece with 0.05×0.05mm 2 -2×2mm 2 A perforated stainless steel mask is attached tightly to the surface of an insulating substrate. Interdigitated electrodes are positioned on the insulating substrate, with their locations corresponding to the perforations. The insulating substrate with the stainless steel mask attached is tilted and inserted into a magnetron sputtering cavity, with the angle α between the substrate normal and the metal sputtering particle beam being 75-95°. Simultaneously, a metal target, which can be any one of W, Sn, In, Zn, or Ti, is inserted into the magnetron sputtering cavity. The cavity is then evacuated to a vacuum of 1×10⁻⁶. -3 Below Pa, high-purity argon and oxygen are simultaneously introduced, and the pressure of the magnetron sputtering cavity is controlled at 0.2-10 Pa and the radio frequency power at 100-400 W. Metal is sputtered on the surface of an insulating substrate with a stainless steel mask attached. The sputtering time is 20-120 min. After sputtering, the vacuum in the cavity is broken, the stainless steel mask is removed, and an oxide nanorod array film is deposited on the interdigitated electrode and the surface of the insulating substrate where the interdigitated electrode is located.

[0007] S2. The insulating substrate coated with oxide nanorod array film is directly annealed, or sensitized with noble metal and then annealed. The annealing process involves placing the substrate at 300-400℃ in an air or oxygen atmosphere for 1-10 hours. The oxide nanorod array film is fully oxidized and crystallized to form an oxide semiconductor sensitive film. The oxide semiconductor sensitive film, interdigitated electrodes, and insulating substrate together constitute a semiconductor sensor that can quickly identify gas molecules.

[0008] Further improvements to the fabrication method of semiconductor sensors capable of rapidly identifying gas molecules:

[0009] Preferably, the insulating substrate is any one of a quartz substrate, a glass substrate, or a polyimide (PI) substrate, and has a thickness of 0.05-1 mm.

[0010] Preferably, the specific steps for performing noble metal sensitization treatment on the insulating substrate coated with the oxide nanorod array thin film sensitive layer in step S2 are as follows:

[0011] In step S1, a dual-target or multi-target magnetron sputtering cavity is selected. A noble metal target is loaded into the magnetron sputtering cavity simultaneously with a metal target. The noble metal target can be any one of Au, Pt, Pd, or Ag targets. During the sputtering of the metal target, a baffle is used to shield the noble metal target. After the metal target deposition is completed, the baffle is used to shield the metal target again. The insulating substrate is leveled, and the angle α between the substrate normal and the metal sputtering particle beam is 0-10°. At the same time, high-purity argon gas is introduced, and the pressure of the magnetron sputtering cavity is controlled to be 0.2-10 Pa. The noble metal target is sputtered at an RF power of 50-100 W for 0.5-2 min to obtain a sensitive layer of oxide nanorod array film with noble metal sensitization on the surface.

[0012] Preferably, in step S1, the flow rate of high-purity argon gas is 4-20 sccm, and the flow rate of oxygen gas is 6-20 sccm.

[0013] Preferably, the interdigitated electrode consists of two opposing toothed interdigitated electrodes, the electrode teeth of the two toothed interdigitated electrodes interlacing to form a serpentine channel, the width of the serpentine channel being 2-50 μm, and the size of the interdigitated region being 0.05 × 0.05 mm. 2 -2×2mm 2 The interdigitated electrode is made of any one of Au, Pt, or Sn-doped In2O3.

[0014] The second objective of this invention is to provide a semiconductor sensor that can quickly identify gas molecules.

[0015] The third objective of this invention is to provide a method for using a semiconductor sensor capable of rapidly identifying gas molecules. When used to identify unknown gases, the semiconductor sensor can be used alone or in combination with two or more other sensors. The method includes two steps: training and identification. The specific steps are as follows:

[0016] Step 21: Under an air background, the semiconductor sensor employs a self-heating modulation mode. By adjusting the driving voltage of the sensitive film, the self-heating power of the sensor is controlled, thereby regulating the Joule heat in the sensitive film and forming a self-heating modulation waveform. The temperature of the sensitive film is regulated to reach the set operating temperature within 0.01s. The transient resistance R of the semiconductor sensor under an air background is recorded. air (t);

[0017] Step 22: Replace the air background with target gas A at concentration C1, keep the self-heating modulation waveform unchanged, and record the transient response resistance of the semiconductor sensor under the target gas background at concentration C1. Repeat the test 3-10 times, and calculate the average value of the measured values ​​as R. gas (t) C1 ;

[0018] Step 23: Adjust the concentration of target gas A to C2, C3, C4... respectively, keeping the self-heating modulation waveform unchanged, and repeat the above operation to obtain the transient response signal set S of the semiconductor sensor to target gas A at different concentrations. gas (t)=R gas (t) / R air (t);

[0019] The transient response signals of target gas A at different concentrations are normalized. nor (t)=(S gas (t)-S min ) / (S max -S min );

[0020] Step 24: Repeat steps 22-23, replacing target gas A with other target gases, and obtain normalized transient response signal sets for different target gases at different concentrations; use supervised learning algorithms to calibrate different gas molecules in the feature space.

[0021] Step 25: Using the same self-heating modulation mode as described above, test the transient resistance R of the semiconductor sensor against the background of the unknown gas to be measured. gas (t), and then a supervised learning algorithm is used to give the judgment.

[0022] Further improvements to the use of semiconductor sensors capable of rapidly identifying gas molecules:

[0023] Preferably, in step S21, the self-heating power of the sensor is 10mW-200mW, the temperature of the control sensitive membrane changes periodically during the test, the resistance acquisition test cycle of a single thermal modulation from low temperature to high temperature is 0.5-2s, and the control sensitive membrane is gradually heated from room temperature 10-35℃ to the working temperature of 100-300℃ at a heating rate of 100-200℃ / s.

[0024] Preferably, the supervised learning algorithm is a linear discriminant analysis (LDA), a convolutional neural network (CNN), or a multilayer perceptron neural network.

[0025] Preferably, the unknown gas to be tested is one of the inorganic gases NO2, O3, NH3, and H2S, or one of the volatile organic compounds formaldehyde, ethanol, benzene, toluene, and xylene, with a concentration of 1-300 ppm.

[0026] The advantages of this invention compared to the prior art are as follows:

[0027] 1) The technical solution of this invention uses a self-heating sensitive film instead of a traditional integrated external heater for thermal modulation. The sensitive layer also serves as the heating layer. When the size of the sensitive area is similar, the resistance relaxation time of the heating process can be significantly reduced from 3-5s for traditional integrated external heaters to less than 1ms. This allows the sensor to achieve rapid temperature modulation within a short time (0.5-2s), collect sufficient molecule characteristics of the target gas, and achieve rapid and accurate identification of multiple gas molecules. The significant reduction in thermal modulation identification time helps to obtain a high sampling rate and realize rapid dynamic monitoring of pollutant gases in situations with dynamic changes in pollutant gases.

[0028] 2) The self-heating device of this invention is simple to fabricate. A tightly packed oxide nanorod array film with adjustable thickness (150-2000 nm) can be obtained on an interdigitated electrode substrate using a multi-target magnetron sputtering cavity via vacuum grazing angle sputtering. The target material can be switched to a noble metal target for sputtering a noble metal sensitization layer, resulting in a highly sensitive oxide semiconductor gas sensor. The interconnected oxide nanorods form conductive channels, serving as both the sensitive and heating layers. The substrate is an insulating glass substrate coated with conductive indium tin oxide interdigitated electrodes, exhibiting high thermal conductivity, low heat dissipation, and low power consumption. No external heater design or integration is required. Furthermore, this invention is compatible with flexible substrates, and the magnetron sputtering process is compatible with conventional micro-nano etching processes, meeting the requirements for uniform mass production of sensors. The sensor of this invention enables rapid and accurate identification of various gases under self-heating modulation, solving the problems of relatively long acquisition time for transient signals in existing thermally modulated sensors, high power consumption, and complex device fabrication.

[0029] 3) Preferably, the thickness of the insulating substrate is less than 1 mm, which can reduce the dissipation of Joule heat generated by self-heating, promote effective heating of the sensitive film, and achieve low power consumption; preferably, the area of ​​the interdigitated region is less than 1 × 1 mm. 2 This helps to reduce the dissipation of self-heating Joule heat and lower power consumption, so as to ensure that the sensor can heat up and cool down quickly at a lower power consumption. Attached Figure Description

[0030] Figure 1 (a) is a schematic diagram of the oxide semiconductor sensitive film prepared by vacuum grazing sputtering in Example 1; (b) is an electron microscope image of the morphology of the WO3 semiconductor sensitive film; (c) is an electron microscope image of the structure of the WO3 semiconductor sensitive film; (d) is a structural diagram of the sensitive film, interdigitated electrode, substrate and mask of the WO3 semiconductor sensor prepared in Example 1.

[0031] Figure 2 (a) is an electron microscope image of the morphology of the Pd-WO3 semiconductor sensitive film sensitized by noble metal Pd in ​​Example 2; (b) is an electron microscope image of the structure of the Pd-WO3 semiconductor sensitive film.

[0032] Figure 3 The self-heating temperature waveform and sensor transient resistance curve are shown in the examples and comparative examples.

[0033] Figure 4 Figure 1 shows the self-heating characteristics of the WO3 nanorod sensing film; Figure (a) shows the relationship between the self-heating peak temperature and the self-heating power; (b) shows the relaxation characteristics of the resistance (temperature) of the device under 3V heating; (c) shows the steady-state response curves of the device to 1ppm NO2 gas under different self-heating powers; and (d) shows the relationship between the sensor's response to 1ppm NO2 and the self-heating power.

[0034] Figure 5 Normalized transient response curves of WO3 nanorod array films and Pd-sensitized WO3 nanorod array films to various target gases (NO2, O3, NH3, C3H9N, CH2O, C2H5OH) at 30-215℃ (10-200mW).

[0035] Figure 6 The results of principal component analysis (PCA) of 12 gas molecules using the WO3 semiconductor sensor in Example 1 were obtained under isothermal testing and thermal modulation waveforms.

[0036] Figure 7 The results of linear discriminant analysis (LDA) of the Pd-WO3 semiconductor sensor in Example 2 for 12 gas molecules were obtained under isothermal testing and thermal modulation waveforms.

[0037] Figure 8The WO3 semiconductor sensor in Example 1 uses all features for 1 second and half features for 0.5 seconds to identify the confusion matrix of 12 gas molecules using a multilayer perceptron neural network (MLP) under self-heating; where (a) is the discrimination index using 1 second features under isothermal modulation; (b) is the discrimination index using 1 second features under self-heating thermal modulation; and (c) is the discrimination index using 0.5 second features under self-heating thermal modulation.

[0038] Figure 9 The identification performance of the WO3 semiconductor sensor in Example 1 for three gases NO2, C3H9N and C3H6O under different relative humidity (a), and the test stability of the electrical response signals of four gases NH3, NO2, C2H5OH and C3H6O over 1 month, 2 months and 3 months (b). Detailed Implementation

[0039] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0040] Example 1

[0041] This embodiment provides a method for fabricating a semiconductor sensor capable of rapidly identifying gas molecules, specifically including the following steps:

[0042] Step 1: Using insulating glass coated with an indium tin oxide conductive layer as a substrate, interdigitated electrodes are etched by laser etching. Each interdigitated electrode consists of two opposing toothed interdigitated electrodes, with the teeth of the two electrodes interlacing to form a serpentine channel. The width of the serpentine channel is 10 μm, and the interdigitated area is 1 × 1 mm. 2 The interdigitated electrodes are made of indium tin oxide; the process is repeated to obtain multiple insulating glass substrates with indium tin oxide interdigitated electrodes.

[0043] Step 2: Take a stainless steel mask with an opening 1.5 mm long and 1 mm wide, and attach it tightly to the surface of the indium tin oxide (ITO) glass substrate, with the opening corresponding to the location of the interdigitated electrodes on the ITO substrate. Tilt the ITO glass substrate with the stainless steel mask attached into the magnetron sputtering cavity, ensuring the angle α between the substrate normal and the metal sputtering particle beam is 86°. Simultaneously, place a 2-inch diameter tungsten disk as the metal target into the magnetron sputtering cavity and evacuate the cavity to 1 × 10⁻⁶. -3Below Pa, 15 sccm of high-purity argon and 7 sccm of high-purity oxygen were simultaneously introduced. The pressure of the magnetron sputtering cavity was controlled at 1.33 Pa and the radio frequency power at 100 W. Metal was sputtered on the surface of an indium tin oxide glass substrate with a stainless steel mask attached. The sputtering time was 120 min. After sputtering, the vacuum in the cavity was broken and the stainless steel mask was removed. A 1400 nm thick WO3 nanorod array film was deposited on the interdigitated electrodes and the surface of the indium tin oxide glass substrate where the interdigitated electrodes were located.

[0044] Step 3: After deposition, the vacuum in the cavity is broken, and the indium tin oxide glass substrate with nanorod array film obtained in step 2 is placed in a tube furnace and annealed in air at 400°C for 2 hours to fully oxidize and crystallize to form a WO3 semiconductor sensitive film. The WO3 semiconductor sensitive film, interdigitated electrodes and indium tin oxide glass substrate together constitute a WO3 semiconductor sensor.

[0045] like Figure 1 As shown in (a), this is a schematic diagram of the vacuum grazing angle sputtering preparation of the WO3 semiconductor sensitive film according to the present invention. Figure 1 (d) is a schematic diagram of the structure of the WO3 semiconductor sensitive film, interdigitated electrode, mask and substrate prepared by vacuum grazing angle sputtering; Figure 1 Images (b) and (c) are electron micrographs of the morphology of the WO3 semiconductor sensing film, showing uniformly arranged columnar nanorods with a large surface-to-volume ratio, suitable for gas sensing. The nanorods are connected to electrodes to form a self-heating conductive channel.

[0046] Example 2

[0047] This embodiment provides a method for fabricating a semiconductor sensor capable of rapidly identifying gas molecules, specifically including the following steps:

[0048] Step 1: Select the indium tin oxide glass substrate with interdigitated electrodes prepared in Step 1 of Example 1.

[0049] Step 2: A stainless steel mask with a 1.5mm long and 1mm wide square opening is attached tightly to the surface of an indium tin oxide (ITO) glass substrate with interdigitated electrodes. The square opening corresponds to the position of the interdigitated electrodes. The metal target W, the noble metal target Pd, and the ITO glass substrate with interdigitated electrodes are then placed into a dual-target magnetron sputtering cavity. The cavity is evacuated to a vacuum level of 1×10⁻⁶. -3Below Pa; a 2-inch diameter tungsten disk was placed at a distance of 87 mm from the indium tin oxide glass substrate as a sputtering target. The angle α between the normal of the indium tin oxide glass substrate and the metal sputtering particle beam was 86°. The noble metal target Pd was shielded by a baffle. The tungsten target was reactively sputtered in a mixed gas of 15 sccm argon and 7 sccm oxygen. The pressure of the magnetron sputtering cavity was controlled at 1.33 Pa, the sputtering power was 100 W, and the deposition time was 120 min. A 1400 nm WO3 nanorod array film was deposited at the mask opening. The nanorod array film deposited on the interdigitated electrode and the indium tin oxide glass substrate where the interdigitated electrode is located was obtained.

[0050] Step 3: Level the indium tin oxide glass substrate with stainless steel mask prepared in Step 2. The angle α between the substrate normal and the metal sputtering particle beam is 0°. High-purity argon gas is introduced. The metal target is shielded with a baffle. A 2-inch diameter Pd disk is used as the sputtering target. The pressure of the magnetron sputtering cavity is controlled at 0.67 Pa, the sputtering power is 100 W, and the deposition time is 40 s. A 2.5 nm thick Pd layer is deposited at the mask opening. The stainless steel mask is removed to obtain the interdigitated electrode substrate with oxide nanorod array film after noble metal Pd sensitization treatment.

[0051] Step 4: After deposition, the vacuum in the cavity is broken, and the sensitized interdigitated electrode substrate obtained in step 3 is placed in a tube furnace and annealed in air at 400°C for 2 hours to fully oxidize and crystallize to form a Pd-WO3 semiconductor sensitive film. The Pd-WO3 semiconductor sensitive film, interdigitated electrode and indium tin oxide glass substrate together constitute a Pd-WO3 semiconductor sensor.

[0052] Depend on Figure 2 As can be seen, the present invention uses grazing-angle magnetron sputtering to grow a nanorod array on an interdigitated electrode and an indium tin oxide glass substrate coated with the interdigitated electrode. The nanorods have a diameter of 50-200 nm, the top diameter of the nanorods is smaller than that of the root, the roots of adjacent nanorods are closely connected, and there are nanometer gaps (50-200 nm) in the middle and top. The nanorods are polycrystalline and have a rough outer surface (not atomically smooth). The thickness of the nanorod array film is 0.5-5 μm.

[0053] Example 3

[0054] This embodiment provides a method for using a semiconductor sensor capable of rapidly identifying gas molecules, specifically including the following steps:

[0055] The WO3 semiconductor sensor prepared in Example 1 was placed in a gas-sensitive test chamber. Under a flowing dry air atmosphere, the temperature was controlled by controlling the driving voltage applied to both ends of the interdigital electrodes of the sensor. The applied voltage range for the WO3 semiconductor sensor was 50-140V, the applied power range was 10-200mW, and the test cycle for a single thermal modulation was 1s.

[0056] The WO3 semiconductor sensor underwent a self-heating test, and the self-heating temperature waveform is shown below. Figure 3 As shown, the specific test results are as follows: Figure 4 As shown, by Figure 4 (a) It can be seen that under the self-heating test, the self-heating applied power is linearly correlated with the actual temperature of the sensor. Under an applied power of 200mW, the actual temperature distribution of the sensor substrate is as follows: Figure 4 (a) As shown in the illustration, the sensor temperature gradually decreases from the center outwards due to heat diffusion to the electrodes. The thermal relaxation time of the WO3 semiconductor sensor is as follows: Figure 4 As shown in (b), ultrafast current-voltage (IV) testing reveals that the thermal relaxation time of the WO3 semiconductor sensor is 3-5 ms. The gas-sensing response characteristics of the WO3 semiconductor sensor in Example 1 to 1 ppm NO2 were tested under different bias voltages (different self-heating power consumptions), as shown below. Figure 4 As shown in (c), by Figure 4 (c) It can be seen that as the applied power across the sensor increases, the sensor baseline resistance gradually decreases, which is consistent with the semiconductor characteristics of the sensor, indicating that the self-heating effect begins to appear and the sensor temperature gradually rises. Simultaneously, the gas-sensing response to NO2 also gradually increases. The gas-sensing response curve of the WO3 semiconductor sensor to 1 ppm NO2 as a function of power is shown below. Figure 4 As shown in (d), by Figure 4 (d) It can be seen that the gas-sensitive response of the sensor first increases and then decreases as the applied power increases, and the maximum response value appears when the applied heating power is 200mW.

[0057] Example 4

[0058] This embodiment provides a self-heating modulation mode, which specifically includes the following steps:

[0059] Step 1: Place the WO3 semiconductor sensor prepared in Example 1 in a gas-sensitive test chamber. Under a flowing dry air atmosphere, control the temperature by controlling the driving voltage applied to both ends of the interdigital electrode of the sensor. The applied voltage range for the WO3 semiconductor sensor is 50-140V, the applied power is 10-200mW, and the test cycle for a single thermal modulation is 1s.

[0060] The Pd-WO3 semiconductor sensor prepared in Example 2 was placed in a gas-sensitive test chamber. Under a flowing dry air atmosphere, the temperature was controlled by controlling the voltage applied to both ends of the interdigital electrode of the sensor. The applied voltage range for the Pd-WO3 semiconductor sensor was 10-34V, the applied power was 10-200mW, and the test cycle for a single thermal modulation was 1s.

[0061] Testing the resistance signal R of WO3 semiconductor sensor and Pd-WO3 semiconductor sensor against an air background air，n (t), where n represents the first and second gas sensors, and t represents time;

[0062] Step 2: Test the self-heating thermal modulation response signals of the WO3 semiconductor sensor and the Pd-WO3 semiconductor sensor under a set concentration of nitrogen dioxide. Repeat the acquisition at least 5 times to obtain the thermal modulation resistance signal set {R} of the n-channel gas sensors under different concentrations of nitrogen dioxide. gas，n (conc,i,t)}, where conc is the target gas concentration, i represents the i-th test sampling, t represents time, and R gas，n (conc,i,t) represents the measured thermal modulation resistance signal;

[0063] Step 3: Perform sensitivity and normalization preprocessing on the self-heating thermal modulation resistor signal set. For S gas (conc,i,t)=R gas，n (conc,i,t) / R air，n (t) is used to obtain the sensitivity signal of each gas sensor array when measuring the target gas for the i-th time at a set concentration. Then, the sensitivity signal is normalized, i.e., S nor (conc,i,t)=(S gas (conc,i,t)-S gas (conc,i,t) min ) / (S gas (conc,i,t) max -S gas (conc,i,t) min ), where S in the formula gas (conc,i,t) max S gas (conc,i,t) min These are the maximum and minimum values ​​of the target gas sensitivity for the i-th measurement at that concentration;

[0064] Step 4: Use linear discriminant analysis to establish a mapping between different concentrations of nitrogen dioxide and the input normalized sensitivity signal, and complete the supervised learning training of nitrogen dioxide.

[0065] Step 5: Repeat steps 1-4 to obtain supervised learning training for ozone, ammonia, trimethylamine, formaldehyde, ethanol, acetone, benzene, toluene, o-xylene, m-xylene, and para-xylene, respectively. The difference is that when calculating the sensitivity signal, for oxidizing gases, it is S. gas (conc,i,t)=R gas，n (conc,i,t) / R air , n (t), for reducing gases: S gas (conc,i,t)=R air，n (t) / R gas，n (conc,i,t);

[0066] Step 6: When the target gas of unknown type and concentration enters the gas-sensitive test chamber, linear discriminant analysis or multilayer sensor neural network analysis is performed based on its self-heating thermal modulation response signal to predict the specific gas type.

[0067] The WO3 semiconductor sensor prepared in Example 1 and the Pd-WO3 semiconductor sensor prepared in Example 2 were subjected to self-heating thermal modulation tests. After signal preprocessing as described in Example 4, the normalized response characteristic results are as follows: Figure 5 As shown, by Figure 5 It can be seen that the normalized response features extracted in 1 second reflect the inherent characteristics of the target gas. Most of the feature signals gradually increase with time (applied power), but the response features of different gases differ in the rate of increase and the occurrence of extreme points, indicating that the self-heating test has uncovered the molecular characteristics of different gases.

[0068] Comparative Example

[0069] This comparative example provides a constant temperature self-heating method. The specific steps are the same as in Example 4, except that: the voltage applied to the WO3 semiconductor sensor prepared in Example 1 is 140V and the applied power is 200mW; the voltage applied to the Pd-WO3 semiconductor sensor prepared in Example 2 is 34V and the applied power is 200mW.

[0070] like Figure 6 The image shows the principal component analysis (PCA) results of 12 gas molecules using the WO3 semiconductor sensor from Example 1 under isothermal testing and thermal modulation waveforms. Figure 7 The image shows the linear discriminant analysis (LDA) results of the Pd-WO3 semiconductor sensor in Example 2 for 12 gas molecules under isothermal testing and thermal modulation waveforms. Figure 6 , 7The comparison shows that the WO3 semiconductor sensor in Example 1 and the Pd-WO3 semiconductor sensor in Example 2 are confused and difficult to distinguish in the two-dimensional feature space under constant temperature self-heating; while the 12 gas molecules in the two semiconductor sensors are completely separated in the two-dimensional feature space under self-heating thermal modulation, and the features of different gases do not overlap, indicating that thermal modulation of the sensor can obtain the subtle differences of the gas molecules to be measured.

[0071] Figure 8 This is the confusion matrix of a multilayer perceptron neural network (MLP) for recognizing 12 gas molecules using the WO3 semiconductor sensor of Example 1 under isothermal self-heating conditions, employing all features in 1 second and half features in 0.5 seconds. Where (a) is the discrimination index of the WO3 semiconductor sensor using 1 second of features under isothermal modulation; (b) is the discrimination index of the WO3 semiconductor sensor using 1 second of features under self-heating thermal modulation; and (c) is the discrimination index of the WO3 semiconductor sensor using 0.5 seconds of features under self-heating thermal modulation. Figure 8 It can be seen that the WO3 semiconductor sensor in Example 1 achieves a recognition accuracy close to 100% using a 1-second feature under self-heating thermal modulation, demonstrating higher discrimination than the isothermal self-heating sensor using a 1-second feature. Furthermore, a recognition accuracy comparable to the example using a 1-second feature can be obtained using a 0.5-second feature, indicating that real-time ultrafast detection of multiple gases within 0.5 seconds or even shorter is achievable.

[0072] Example 5

[0073] This embodiment provides a self-heating modulation mode under different relative humidity air backgrounds, and tests the recognition performance of the WO3 semiconductor sensor of Embodiment 1 for three gases NO2, C3H9N and C3H6O under different relative humidity conditions.

[0074] Step 1: Place the WO3 semiconductor sensor prepared in Example 1 in a gas-sensitive test chamber. Under flowing dry or humid air atmospheres (0%, 40%, and 80% relative humidity, respectively), control the temperature by adjusting the voltage applied to the interdigital electrodes of the sensor. The applied voltage range for the WO3 semiconductor sensor is 50-140V, the applied power is 10-200mW, and the temperature modulation test time is 1s. Measure the resistance signal R of the WO3 semiconductor sensor against an air background. air，m (t), where m represents the first to third relative humidity background air, and t represents time;

[0075] Step 2: Test the self-heating thermal modulation response signal of the WO3 semiconductor sensor under a set concentration of nitrogen dioxide. Repeat the acquisition at least 5 times to obtain the thermal modulation resistance signal set {R} of the WO3 semiconductor gas sensor under different relative humidity air backgrounds and different concentrations of nitrogen dioxide. gas，m(conc,i,t)}, where conc is the target gas concentration, i represents the i-th test sampling, t represents time, and R gas，m (conc,i,t) represents the measured thermal modulation resistance signal;

[0076] Step 3: Perform sensitivity and normalization preprocessing on the self-heating thermal modulation resistor signal set. For S gas (conc,i,t)=R gas，m (conc,i,t) / R air，m (t) is used to obtain the sensitivity signal of the WO3 semiconductor gas sensor when measuring the target gas for the i-th time at a set concentration. Then, the sensitivity signal is normalized, i.e., S nor (conc,i,t)=(S gas (conc,i,t)-S gas (conc,i,t) min ) / (S gas (conc,i,t) max -S gas (conc,i,t) min ), where S in the formula gas (conc,i,t) max S gas (conc,i,t) min These are the maximum and minimum values ​​of the target gas sensitivity for the i-th measurement at that concentration;

[0077] Step 4: Use linear discriminant analysis to establish a mapping between different relative humidity air backgrounds and different concentrations of nitrogen dioxide and the input normalized sensitivity signal, and complete the supervised learning training of nitrogen dioxide.

[0078] Step 5: Repeat steps 1-4 to obtain supervised learning training for trimethylamine and acetone, respectively. The difference is that when calculating the sensitivity signal, for oxidizing gases, the value is S. gas (conc,i,t)=R gas，m (conc,i,t) / R air，m (t), for reducing gases: S gas (conc,i,t)=R air，m (t) / R gas，m (conc,i,t);

[0079] Step 6: When an unknown type and concentration of target gas enters the gas-sensitive test chamber under different relative humidity air backgrounds, linear discriminant analysis or multilayer sensor neural network analysis is performed based on its self-heating thermal modulation response signal to predict the specific gas type.

[0080] Test results are as follows Figure 9As shown in (a); by Figure 9 It can be seen that the complete separation of the three gases in the two-dimensional space of PCA proves the feasibility of molecular recognition in the context of humid air in the embodiment. Figure 9 (b) Demonstrates monitoring of the electrical response signals of four gases over January, February, and March. The absence of gas molecule overlap in the PCA two-dimensional space indicates good long-term stability of the embodiment. The surface self-heating thermal modulation sensor exhibits good moisture resistance and stability, meeting the requirements for target gas identification and detection in real-world (humidity) environments.

[0081] Those skilled in the art should understand that the above descriptions are merely several specific embodiments of the present invention, and not all embodiments. It should be noted that many modifications and improvements can be made by those skilled in the art, and all modifications or improvements not exceeding the scope of the claims should be considered within the protection scope of the present invention.

Claims

1. A method of using a semiconductor sensor capable of rapidly identifying gas molecules, characterized in that, When used to identify unknown gases, this semiconductor sensor can be used alone or in combination of two or more channels, and includes two steps: training and identification. The specific steps are as follows: Step 21: Under an air background, the semiconductor sensor employs a self-heating modulation mode. By adjusting the driving voltage of the sensitive film, the self-heating power of the sensor is controlled, thereby regulating the Joule heat in the sensitive film and forming a self-heating modulation waveform. The temperature of the sensitive film is regulated to reach the set operating temperature within 0.01s. The transient resistance R of the semiconductor sensor under an air background is recorded. air (t); Step 22: Replace the air background with target gas A at concentration C1, keep the self-heating modulation waveform unchanged, and record the transient response resistance of the semiconductor sensor under the target gas background at concentration C1. Repeat the test 3-10 times, and calculate the average value of the measured values ​​as R. gas (t) C1 ; Step 23: Adjust the concentration of target gas A to C2, C3, C4... respectively, keeping the self-heating modulation waveform unchanged, and repeat the above operation to obtain the transient response signal set S of the semiconductor sensor to target gas A at different concentrations. gas (t)=R gas (t) / R air (t); The transient response signals of target gas A at different concentrations are normalized. nor (t)=(S gas (t)-S min ) / (S max -S min ); Step 24: Repeat steps 22-23, replacing target gas A with other target gases, and obtain normalized transient response signal sets for different target gases at different concentrations; use supervised learning algorithms to calibrate different gas molecules in the feature space. Step 25: Using the same self-heating modulation mode as described above, test the transient resistance R of the semiconductor sensor against the background of the unknown gas to be measured. gas (t), and then a supervised learning algorithm is used to give a judgment; This semiconductor sensor is fabricated using the following method: S1, Take a piece with 0.05×0.05mm 2 -2×2mm 2 A perforated stainless steel mask is attached tightly to the surface of an insulating substrate. Interdigitated electrodes are positioned on the insulating substrate, with their locations corresponding to the perforations. The insulating substrate with the stainless steel mask attached is tilted and inserted into a magnetron sputtering cavity, with the angle α between the substrate normal and the metal sputtering particle beam being 75-95°. Simultaneously, a metal target, which can be any one of W, Sn, In, Zn, or Ti, is inserted into the magnetron sputtering cavity. The cavity is then evacuated to a vacuum of 1×10⁻⁶. -3 Below Pa, high-purity argon and oxygen are simultaneously introduced, and the pressure of the magnetron sputtering cavity is controlled at 0.2-10 Pa and the radio frequency power at 100-400 W. Metal is sputtered on the surface of an insulating substrate with a stainless steel mask attached. The sputtering time is 20-120 min. After sputtering, the vacuum in the cavity is broken, the stainless steel mask is removed, and an oxide nanorod array film is deposited on the interdigitated electrode and the surface of the insulating substrate where the interdigitated electrode is located. S2. The insulating substrate coated with oxide nanorod array film is directly annealed, or sensitized with noble metal and then annealed. The annealing process involves placing the substrate at 300-400℃ in an air or oxygen atmosphere for 1-10 hours. The oxide nanorod array film is fully oxidized and crystallized to form an oxide semiconductor sensitive film. The oxide semiconductor sensitive film, interdigitated electrodes, and insulating substrate together constitute a semiconductor sensor that can quickly identify gas molecules.

2. The method of using the semiconductor sensor capable of rapidly identifying gas molecules according to claim 1, characterized in that, The insulating substrate is any one of a quartz substrate, a glass substrate, or a polyimide (PI) substrate, with a thickness of 0.05-1 mm.

3. The method of using the semiconductor sensor capable of rapidly identifying gas molecules according to claim 1, characterized in that, The specific steps for performing noble metal sensitization treatment on the insulating substrate coated with the oxide nanorod array thin film sensitive layer in step S2 are as follows: In step S1, a dual-target or multi-target magnetron sputtering cavity is selected. A noble metal target is loaded into the magnetron sputtering cavity simultaneously with a metal target. The noble metal target can be any one of Au, Pt, Pd, or Ag targets. During the sputtering of the metal target, a baffle is used to shield the noble metal target. After the metal target deposition is completed, the baffle is used to shield the metal target again. The insulating substrate is leveled, and the angle α between the substrate normal and the metal sputtering particle beam is 0-10°. At the same time, high-purity argon gas is introduced, and the pressure of the magnetron sputtering cavity is controlled to be 0.2-10 Pa. The noble metal target is sputtered at an RF power of 50-100 W for 0.5-2 min to obtain a sensitive layer of oxide nanorod array film with noble metal sensitization on the surface.

4. The method of using the semiconductor sensor capable of rapidly identifying gas molecules according to claim 1, characterized in that, In step S1, the flow rate of high-purity argon gas is 4-20 sccm, and the flow rate of oxygen gas is 6-20 sccm.

5. The method of using the semiconductor sensor capable of rapidly identifying gas molecules according to claim 1, characterized in that, The interdigitated electrode consists of two opposing toothed interdigitated electrodes. The electrode teeth of the two toothed interdigitated electrodes interlace to form a serpentine channel. The width of the serpentine channel is 2-50 μm, and the size of the interdigitated region is 0.05 × 0.05 mm. 2 -2×2mm 2 The interdigitated electrodes are made of any one of Au, Pt, or Sn-doped In2O3.

6. The method of using the semiconductor sensor capable of rapidly identifying gas molecules according to claim 1, characterized in that, In step S21, the self-heating power of the sensor is 10mW-200mW. The temperature of the control sensitive membrane changes periodically during the test. The resistance acquisition test cycle of a single thermal modulation from low temperature to high temperature is 0.5-2s. The control sensitive membrane is gradually heated from room temperature (10-35℃) to the working temperature (100-300℃), with a heating rate of 100-200℃ / s.

7. The method of using the semiconductor sensor capable of rapidly identifying gas molecules according to claim 1, characterized in that, The supervised learning algorithm is linear discriminant analysis, convolutional neural network, or multilayer perceptron neural network.

8. The method of using the semiconductor sensor capable of rapidly identifying gas molecules according to claim 1, characterized in that, The unknown gas to be tested is one of the inorganic gases NO2, O3, NH3, and H2S, or one of the volatile organic compounds formaldehyde, ethanol, benzene, toluene, and xylene, with a concentration of 1-300 ppm.

Images