A method for processing a micro-silicon spherical cavity

Abstract

The application discloses a processing method of micro-silicon spherical cavities, which comprises the following steps: sequentially performing deposition etching mask, photolithography, removing part of the etching mask, removing glue and deep reactive ion etching on the side of a silicon wafer to be patterned; obtaining the micro-silicon spherical cavities through wet etching; and inputting the real-time collected images into an etching circle detection module and a roughness prediction model and an etching depth prediction model which are pre-established and trained, to obtain the top view, sectional view images of the micro-silicon spherical cavities under the current etching state and the etching related prediction values; comparing the etching related prediction values with the expected indicators, if the expected indicators are not reached, adjusting the wet etching related process parameters and continuing the wet etching; if the expected indicators are met, stopping the etching to obtain the micro-silicon spherical cavities meeting the expected indicators; the etching circle detection module is based on a Hough circle detection algorithm, the roughness prediction model adopts a BP neural network, and the etching depth prediction model adopts a convolutional neural network.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of semiconductor processing, in particular to a processing method of micro-silicon spherical cavity. BACKGROUND

[0002] Isotropic silicon etching technology has wide application prospects in the fields of acoustic lens of high-frequency focused ultrasonic transducer, sacrificial mold of MEMS resonator, mold of optical micro-lens, etc. Among them, the chemical etching solution (HNA) represented by the mixture of hydrofluoric acid (HF), nitric acid (HNO3) and acetic acid (CH3COOH) isotropically etches silicon in wet method, which has been widely used in semiconductor process. However, with the increase of the radius of the etched spherical cavity, the "lateral etching effect" that the lateral etching rate is much greater than the longitudinal etching rate will occur. At this time, the spherical cavity of bulk silicon etching no longer has good sphericity. Therefore, it is of great significance to develop an etching method and its device to weaken the lateral effect for the semiconductor process of isotropic bulk silicon etching.

[0003] In the field of semiconductor process technology, the inspection of key parameters such as the roughness and sphericity of bulk silicon etched spherical cavity and the final etched cavity radius requires laser cutting along the central axis of the cavity after etching is completed and then checking by scanning electron microscope (SEM). Due to the irreversibility of etching technology and laser cutting technology, the inspection cost of etching defects is very high and many etched spherical cavities need to be destroyed. Moreover, such inspection method can only represent the etching results of part of the silicon wafer area and cannot completely reflect the consistency of silicon cavities in different areas. Therefore, the isotropic etching process of silicon urgently needs a method that can predict key parameters such as longitudinal and lateral etching depth, sphericity, roughness, etching radius after etching. SUMMARY

[0004] The present application aims to overcome the defects of the prior art and provides a processing method of micro-silicon spherical cavity.

[0005] In order to achieve the above-mentioned purpose, the present application provides a processing method of micro-silicon spherical cavity, which comprises:

[0006] Step 1) sequentially performing deposition of etching mask, photolithography, removal of part of etching mask, degumming and deep reactive ion etching on the side of the silicon wafer to be patterned;

[0007] Step 2) wet etching to obtain micro-silicon spherical cavity;

[0008] Step 3) real-time image acquisition of the micro-silicon spherical cavity;

[0009] Step 4) The collected images are preprocessed and sequentially input into the etching circle detection module and the pre-established and trained roughness prediction model and etching depth prediction model to obtain the overhead image, cross-sectional image and etching-related prediction value of the micro-silicon spherical cavity under the current etching state.

[0010] Step 5) The etching-related prediction value is compared with the expected index. If the expected index is not reached, the wet etching-related process parameters are adjusted, and the process returns to step 2). If the expected index is met, the etching is stopped, and the micro-silicon spherical cavity meeting the expected index is obtained.

[0011] As an improvement of the above method, the step 1) comprises:

[0012] Step 1-1) The silicon wafer is cleaned or a no-clean silicon wafer is used.

[0013] Step 1-2) An etching mask is prepared by a thin film deposition method, and the thickness of the single-layer film ranges from 10 nm to 100 μm.

[0014] Step 1-3) A positive photoresist is spin-coated on the surface to be etched, and exposure and film hardening are performed.

[0015] Step 1-4) Part of the etching mask is removed.

[0016] Step 1-5) The silicon wafer is subjected to a photoresist removal treatment to remove the photoresist and form an etching window of the etching mask.

[0017] Step 1-6) Deep reactive ion etching is performed on the surface to be etched.

[0018] Step 1-7) The silicon wafer is subjected to isotropic wet etching using a wet chemical etching solution.

[0019] As an improvement of the above method, the step 1-2) of preparing the etching mask by a thin film deposition method specifically comprises:

[0020] Gold / chromium, silicon nitride, aluminum, parylene or polyimide thin films and their composite films are prepared by ion beam sputtering, low-pressure chemical vapor deposition or plasma-enhanced chemical vapor deposition as etching masks.

[0021] As an improvement of the above method, the pattern center of the etching mask in step 1-2) is circular and has a large opening, and various arrangements of circular, annular or wavy patterns and combinations thereof gradually decrease in diameter, ring width or opening away from the center of the circle.

[0022] As an improvement of the above method, the step 1-6) specifically comprises: etching the silicon wafer by deep reactive ion etching based on Bosch process, and using the loading effect of deep reactive ion etching related to etching aspect ratio to etch the center region with larger opening deeper and the surrounding region with smaller opening shallower.

[0023] As an improvement of the above method, the etching related prediction value comprises: longitudinal and transverse etching depth, sphericity, surface roughness and etching radius of the micro-silicon spherical cavity in the current etching state.

[0024] As an improvement of the above method, the processing process of the etching circle detection model is: extracting the etching circle boundary from the pre-processed image by Hough circle detection algorithm, and then calculating the sphericity and etching radius of the etching circle in the current etching state.

[0025] As an improvement of the above method, the input of the roughness prediction model is the feature value corresponding to the pre-processed image, and the output is the top view image of the micro-silicon spherical cavity and the surface roughness of the micro-silicon spherical cavity in the current etching state; the feature value includes energy, contrast, entropy, homogeneity, difference and correlation of the image; the roughness prediction model comprises:

[0026] a BP neural network for outputting the top view image of the micro-silicon spherical cavity; and

[0027] a surface roughness calculation module for calculating the surface roughness according to the top view image of the micro-silicon spherical cavity.

[0028] As an improvement of the above method, the input of the etching depth prediction model is the pre-processed image, and the output is the cross-sectional image of the micro-silicon spherical cavity and the etching depth prediction value of the micro-silicon spherical cavity in the current etching state, and the etching depth prediction model comprises:

[0029] a convolutional neural network comprising an input layer, a convolutional layer, a pooling layer, a convolutional layer, a pooling layer, a full connection layer and an output layer connected in sequence; for outputting the cross-sectional image of the micro-silicon spherical cavity; and

[0030] an etching depth prediction value calculation module for calculating the longitudinal and transverse etching depth according to the cross-sectional image of the micro-silicon spherical cavity.

[0031] As an improvement of the above method, the method further comprises a training step of the roughness prediction model and the etching depth prediction model; specifically comprising:

[0032] The sample after etching is cut along the center line, and the longitudinal and transverse etching depths, sphericity, roughness, etching radius, top view and cross-sectional view of the sample after cutting are obtained by using a scanning electron microscope, a laser three-dimensional imaging system and an optical profiler, and the parameter results are correspondingly established with the imaging results of the etching process to form a training set;

[0033] The training set is sequentially input into the roughness prediction model, and the roughness prediction model meeting the training requirements is obtained by adjusting the model parameters;

[0034] The training set is sequentially input into the etching depth prediction model, and the etching depth prediction model meeting the training requirements is obtained by adjusting the model parameters.

[0035] Compared with the prior art, the advantages of the present application are:

[0036] 1. The present application improves the transverse etching effect of the silicon etching spherical cavity by combining dry deep reactive ion etching with wet isotropic etching in advance;

[0037] 2. The prior art can detect etching parameters only after etching is completed, and the present application can predict the longitudinal and transverse etching depths, sphericity, surface roughness, etching radius and other etching results of the micro-silicon spherical cavity in the current etching state in real time during etching;

[0038] 3. The present application can adjust the etching parameters (such as etching temperature, etching liquid ratio, stirring speed and mode) according to the real-time prediction results to adjust the etching process to achieve the expected etching parameter results;

[0039] 4. The present application can obtain the key etching parameters such as longitudinal and transverse etching depths, sphericity, surface roughness and etching radius after etching without damaging the micro-silicon spherical cavity. BRIEF DESCRIPTION OF DRAWINGS

[0040] Figure 1 It is a cross-sectional view of a silicon wafer after the mask pattern is completed;

[0041] Figure 2 It is a schematic diagram of an etching mask with a circular opening in the center and a first, second, third and fourth layer of ring width gradually decreasing from the inside to the outside;

[0042] Figure 3 It is a schematic diagram of an etching mask with a circular opening in the center and a first, second, third and fourth layer of ring width gradually decreasing from the inside to the outside;

[0043] Figure 4 It is a schematic diagram of an etching mask with a circular opening in the center and a first, second, third and fourth layer of ring width gradually decreasing from the inside to the outside;

[0044] Figure 5 A schematic diagram of an etching mask with a circular opening in the center and gradually decreasing ring width from the inner to the outer third layer and the first, second, and fourth layers being annular;

[0045] Figure 6 A schematic diagram of an etching mask with a circular opening in the center and gradually decreasing ring width from the inner to the outer second and third layers and the first and fourth layers being annular;

[0046] Figure 7 A schematic diagram of an etching mask with a circular opening in the center and gradually decreasing ring width from the inner to the outer first layer and the second and third layers being annular;

[0047] Figure 8 A schematic diagram of an etching mask with a circular opening in the center and gradually decreasing ring width from the inner to the outer first and second layers and the third and fourth layers being annular;

[0048] Figure 9 A schematic diagram of an etching mask with a circular opening in the center and gradually decreasing ring width from the inner to the outer first, second, and third layers and the fourth layer being annular;

[0049] Figure 10 A cross-sectional and plan view of a silicon wafer after deep reactive ion etching is completed;

[0050] Figure 11 A schematic diagram of the method deployment composition of the present application;

[0051] Figure 12 A cross-sectional view of a silicon wafer to be etched;

[0052] Figure 13 A cross-sectional view of a silicon wafer to be etched after deposition of an etching mask;

[0053] Figure 14 A cross-sectional view of a silicon wafer to be etched after photolithography is completed;

[0054] Figure 15 A plan view of a silicon wafer to be etched after photolithography is completed;

[0055] Figure 16 A cross-sectional view of a silicon wafer to be etched after partial removal of the etching mask;

[0056] Figure 17 A plan view of a silicon wafer to be etched after partial removal of the etching mask;

[0057] Figure 18 A cross-sectional view of a silicon wafer to be etched after removal of the photoresist;

[0058] Figure 19 A plan view of a silicon wafer to be etched after removal of the photoresist;

[0059] Figure 20 Figure 4 is a cross-sectional view of a silicon wafer after deep reactive ion etching of a region to be etched.

[0060] Reference signs

[0061] 1. Etching mask 2. Silicon wafer 3. Photoresist

[0062] 4. Fluorinated ethylene propylene (FEP) cover plate 5. Etching solution 6. Polytetrafluoroethylene

[0063] 7. Magnetic sub 8. Stirring device 9. High-speed high-definition video camera

[0064] 10. Computer DETAILED DESCRIPTION

[0065] The purpose of the present application is to provide a method for improving the lateral etching effect and predicting key parameters such as longitudinal and lateral etching depth, sphericity, roughness, and etching radius after etching based on machine learning. Deep reactive ion etching is used to pattern and etch a deep cavity in the region to be etched to reduce the lateral effect. During the etching process, a high-speed high-definition video camera is used to continuously capture images of the etching sample during the etching process above the etching device. Then the obtained images are transmitted to the computer. The neural network based on machine learning is compared with the pre-trained verification set to predict the key parameters such as longitudinal and lateral etching depth, sphericity, roughness, and etching radius of the etching sample in real time.

[0066] To achieve the above-mentioned purpose, the technical solutions of the present application are as follows:

[0067] First, Au / Cr, silicon nitride, parylene, polyimide, and their composite films are prepared as etching masks on the surface of the silicon wafer to be etched by ion beam sputtering, low pressure chemical vapor deposition (LPCVD), plasma enhanced chemical vapor deposition (PECVD), and other thin film deposition methods. The thickness of the single layer film ranges from 10 nm to 100 μm. Then the above-mentioned mask or composite mask is patterned. First, photoresist is applied to the side of the silicon wafer to be patterned, then exposure and hardening are performed. After hardening, one or more of chemical etching, dry etching, and lift-off process are used to remove part of the etching mask to complete the patterning of the etching mask. Finally, one or more of wet process (such as acetone), plasma stripping process, and ashing process are used to remove the photoresist to form the etching window of the etching mask. One or more of chemical wet etching, deep reactive ion etching, and dry etching are used to etch the etching mask. The cross-sectional view of the silicon wafer after the etching mask is patterned is shown in Figure 1 and the top view is shown in Figure 2The formed silicon cavity mask pattern is characterized by a circular center and various complex arrangements of circular, annular, wavy patterns and combinations thereof, with the opening gradually decreasing in diameter, ring width, or opening away from the center, Figures 2-9 is a top view schematic diagram of the silicon wafer after the pattern is completed on the etching mask. Figure 2 is a schematic diagram of an etching mask with a circular center opening and a first, second, third, and fourth layer of annular rings with gradually decreasing ring widths from the inside out. Figure 3 is a schematic diagram of an etching mask with a circular center opening and a first layer of annular rings with gradually decreasing ring widths from the inside out.

[0068] Figure 4 is a schematic diagram of an etching mask with a circular center opening and a first, third, and fourth layer of annular rings with gradually decreasing ring widths from the inside out. Figure 5 is a schematic diagram of an etching mask with a circular center opening and a first, second, and fourth layer of annular rings with gradually decreasing ring widths from the inside out. Figure 6 is a schematic diagram of an etching mask with a circular center opening and a first, fourth layer of annular rings with gradually decreasing ring widths from the inside out.

[0069] Figure 7 is a schematic diagram of an etching mask with a circular center opening and a first, second, and fourth layer of annular rings with gradually decreasing ring widths from the inside out. Figure 8 is a schematic diagram of an etching mask with a circular center opening and a first, second, and fourth layer of annular rings with gradually decreasing ring widths from the inside out. Figure 9 is a schematic diagram of an etching mask with a circular center opening and a first, second, and fourth layer of annular rings with gradually decreasing ring widths from the inside out.

[0070] A method for processing micro-silicon spherical cavities, specifically comprising the following steps:

[0071] 1. Deep reactive ion etching of the silicon wafer. Due to the different sizes of the openings, the deep reactive ion etching and the etching depth ratio related load effect are used to etch the center of the spherical cavity deeper, the surrounding opening smaller, and the etching depth shallower. The effect of deep reactive ion etching is shown in Figure 10 .

[0072] 2. Subsequently, the silicon wafer is subjected to isotropic wet etching using HNA solution. The deployment diagram is shown in Figure 11 . The top cover of the etching device is made of perfluoroethylene propylene copolymer (FEP) which is corrosion-resistant and has good light transmittance, and the remaining four sides are made of polytetrafluoroethylene material. The bottom of the etching device is connected to a magnetic stirring device with heating function, and a high-speed high-definition industrial camera is placed above the top cover. A high-speed high-definition camera is used to take real-time images of the etching area during the etching process.

[0073] 3. Using a high-speed high-definition industrial camera to collect the image of the etching area of the silicon wafer, and performing image preprocessing on the obtained image: using an image preprocessing program to perform grayscale conversion, median filtering and grayscale enhancement processing on the image.

[0074] 4. Using an etching circle detection module to perform Hough circle detection algorithm on the processed image to extract the etching circle boundary and calculate the sphericity, etching radius and other parameters of the current etching circle.

[0075] 5. The roughness prediction model is based on BP neural network, and the characteristic values such as energy, contrast, entropy, homogeneity, difference and correlation of the processed image are taken as inputs of the model, and the output is the top view image of the micro-silicon spherical cavity and the roughness prediction value.

[0076] 6. The etching depth prediction model is based on convolutional neural network, which includes input layer, convolutional layer, pooling layer, convolutional layer, pooling layer, fully connected layer and output layer from input to output. The input layer is image data, and the output is the cross-sectional image of the micro-silicon spherical cavity and the etching depth prediction.

[0077] The data analysis program module compares the parameters obtained from the circle detection module and the roughness prediction model and the etching depth prediction model with the expected etching parameters, and adjusts the process parameters of wet etching in real time and determines the etching endpoint.

[0078] Previously, the roughness prediction model and the etching depth prediction model have been established and trained.

[0079] The input of the roughness prediction model is the characteristic value corresponding to the preprocessed image, and the output is the top view image of the micro-silicon spherical cavity and the surface roughness of the micro-silicon spherical cavity under the current etching state; the characteristic value includes the energy, contrast, entropy, homogeneity, difference and correlation of the image; the roughness prediction model comprises:

[0080] BP neural network, used to output the top view image of the micro-silicon spherical cavity;

[0081] A surface roughness calculation module is used to calculate the surface roughness according to the top view image of the micro-silicon spherical cavity.

[0082] The input of the etching depth prediction model is the preprocessed image, and the output is the cross-sectional image of the micro-silicon spherical cavity and the etching depth prediction value of the micro-silicon spherical cavity under the current etching state, and the etching depth prediction model comprises:

[0083] The convolutional neural network comprises an input layer, a convolutional layer, a pooling layer, a convolutional layer, a pooling layer, a full connection layer and an output layer connected in sequence; is used for outputting a cross-sectional image of the microsilicon spherical cavity; and the etching depth prediction value calculation module is used for calculating the longitudinal and transverse etching depths according to the cross-sectional image of the microsilicon spherical cavity.

[0084] The specific training steps are as follows:

[0085] 1. The sample after etching is cut, and the scanning electron microscope, laser three-dimensional imaging system and optical profiler are used to obtain the key parameters of the sample, such as longitudinal and transverse etching depths, sphericity, roughness and etching radius, and the results are corresponded with the imaging results of the etching process to establish a complete database.

[0086] 2. The database is classified, 70% is used as a training set, and 30% is used as a verification set to train the roughness prediction model and the depth prediction model. The trained model is used to predict the roughness and etching depth.

[0087] 3. The sample after etching is cut, and the scanning electron microscope, laser three-dimensional imaging system and optical profiler are used to obtain the key parameters of the sample, such as longitudinal and transverse etching depths, sphericity, roughness and etching radius, and the results are corresponded with the imaging results of the etching process to establish a complete database.

[0088] 4. The training set is input into the roughness prediction model in sequence, and the roughness prediction model meeting the training requirements is obtained by adjusting the model parameters;

[0089] 5. The training set is input into the etching depth prediction model in sequence, and the etching depth prediction model meeting the training requirements is obtained by adjusting the model parameters.

[0090] The technical solutions of the present application will be described in detail below in combination with the drawings and examples.

[0091] Example 1

[0092] Example 1 of the present application proposes a processing method of a microsilicon spherical cavity, which specifically comprises the following steps:

[0093] 1. Prepare the substrate

[0094] The <1 0 0> N-type silicon wafer is cleaned by a standard cleaning method, first boiled and cleaned by using an acidic cleaning solution and an alkaline cleaning solution respectively, then cleaned by using deionized water, and finally dried by using nitrogen. Figure 12 The cross-sectional image of the etching region of the silicon wafer.

[0095] 2. Deposition of etching mask

[0096] Au / Cr composite film was prepared on the surface of silicon wafer by ion beam sputtering as etching mask, in which the thickness of Au mask is 120 nm and the thickness of Cr mask is 50 nm. The cross section of the wafer after deposition of etching mask is shown in Fig. 1. Figure 13

[0097] 3. Photolithography

[0098] The positive photoresist was spin-coated on the side of silicon wafer to be etched by using a spin coater, in which the rotation speed of the spin coater was set to 3000 rpm and the spin-coating time was 30 seconds. After spin-coating, the wafer was pre-baked at 80 °C for 30 minutes. Then, the wafer was subjected to photolithography using a soda-chrome plate with negative pattern as mask and developed. Finally, the wafer was post-baked at 100 °C for 30 minutes to harden the photoresist. Figure 14 The cross section of the wafer after photolithography is shown in Fig. 2. Figure 15 The plan view of the wafer after photolithography is shown in Fig. 3.

[0099] 4. Removing part of the etching mask

[0100] Part of the etching mask was removed by using one or more of the following processes: chemical wet etching, dry etching, pulsed laser or lift-off.

[0101] In this embodiment, part of the etching mask was removed by chemical wet etching. The mixed solution of iodine and potassium iodide was used to remove part of the Au etching mask. After etching, the wafer was rinsed with deionized water for 10 minutes and then dried with nitrogen. The cerous nitrate etching solution was used to etch the Cr mask. After etching, the wafer was rinsed with deionized water for 10 minutes and then dried with nitrogen. Figure 16 The cross section of the wafer after removing part of the etching mask is shown in Fig. 4. Figure 17 The plan view of the wafer after removing part of the etching mask is shown in Fig. 5.

[0102] 5. Stripping

[0103] The wafer was subjected to stripping by using one of the following processes: acetone wet process, plasma stripping process or ashing process to remove the photoresist and form the etching window of the etching mask.

[0104] In this embodiment, the wafer was soaked in acetone solution for 5 minutes to dissolve the photoresist. After soaking in acetone, the wafer was soaked in anhydrous ethanol solution for 10 minutes. After soaking, the surface of the wafer was gently wiped with alcohol cotton. Then, the wafer was rinsed with deionized water for 10 minutes and then dried with nitrogen. Figure 18 The cross section of the wafer after stripping is shown in Fig. 6. Figure 19 The plan view of the wafer after stripping is shown in Fig. 7.

[0105] 6. Deep reactive ion etching

[0106] ​Deep reactive ion etching was performed on the surface to be etched, with a central aperture etching depth of 70 μm. Figure 20 This is a cross-sectional view of the area to be etched on the silicon wafer after deep reactive ion etching.

[0107] 7. Wet etching

[0108] The wet chemical etching solution is one of the following combinations:

[0109] Hydrofluoric acid (HF), nitric acid (HNO3), and water (H2O); or

[0110] Hydrofluoric acid (HF), nitric acid (HNO3), and acetic acid (CH3COOH); or

[0111] Hydrofluoric acid (HF), nitric acid (HNO3), and isopropanol (IPA).

[0112] In this embodiment, isotropic wet etching of silicon wafers is performed using an HNA solution. The volume ratio of acetic acid:hydrofluoric acid:nitric acid in the HNA wet etching solution is 1:2:7, with acetic acid concentration of 99.99%, hydrofluoric acid concentration of 49%, and nitric acid concentration of 69%. The silicon wafer and its fixture are then placed... Figure 11 The etching apparatus shown is equipped with etching liquid, and magnetic stirring is activated. The heating temperature is set to 50°C. During the etching process, a high-speed, high-definition camera images the etched area in real time. The computer inputs the expected etching parameters, and the etching circle detection module calculates the sphericity and etching radius of the etching circle under the current etching state. A roughness prediction model is used to obtain a top-view image of the micro-silicon spherical cavity and its surface roughness. An etching depth prediction model is used to obtain a cross-sectional image of the micro-silicon spherical cavity and the predicted etching depth of the micro-silicon spherical cavity under the current etching state. Based on the key parameters of the silicon wafer under the current etching state, such as longitudinal and transverse etching depth, sphericity, roughness, and etching radius, the process parameters of the wet etching are adjusted in real time by comparing the calculated results with the expected etching parameters. Finally, the etching stop point is determined, completing the isotropic wet etching.

[0113] It should be noted that the steps in this example correspond to... Figures 11-20 The schematic diagram is also applicable to the following embodiments, and will not be repeated here.

[0114] Example 2

[0115] Embodiment 2 of the present invention proposes a method for fabricating a micro-silicon spherical cavity, which specifically includes the following steps:

[0116] 1. Prepare the substrate

[0117] Select <100>N-type no-clean silicon wafers. Figure 12 This is a cross-sectional view of the area to be etched on the silicon wafer.

[0118] 2. Deposition and etching mask

[0119] A silicon nitride film was deposited on the surface of a silicon wafer using low pressure chemical vapor deposition, with a thickness of 1 μm. The cross-section after deposition of the etch mask is shown in Figure 13

[0120] 3. Photolithography

[0121] A positive photoresist was spin-coated on the side of the silicon wafer to be etched using a spin coater, with a spin speed of 3000 rpm and a spin time of 30 seconds. After spin coating, a pre-bake was performed at a temperature of 80°C for 30 minutes. Subsequently, photolithography was performed using a soda-chrome mask with a negative pattern as the mask, and the photoresist was developed. Finally, a post-bake was performed to harden the photoresist, at a temperature of 100°C for 30 minutes. Figure 14 The cross-section of the silicon wafer after photolithography, showing the area to be etched, Figure 15 is shown in the plan view of the corresponding area.

[0122] 4. Partial removal of the etch mask

[0123] The etch mask was partially removed using one or more of a chemical wet etching process, a dry etching process, a pulsed laser process, or a lift-off process.

[0124] In this example, the silicon nitride mask was removed using reactive ion etching (RIE), Figure 16 The cross-section of the silicon wafer after partial removal of the etch mask, showing the area to be etched, Figure 17 is shown in the plan view of the corresponding area.

[0125] 5. Stripping

[0126] The silicon wafer was subjected to a stripping process to remove the photoresist and form an etch window in the etch mask, using an acetone wet process, a plasma stripping process, or an ashing process.

[0127] In this example, the silicon wafer was immersed in an acetone solution for 5 minutes to dissolve the photoresist, and then immersed in an anhydrous ethanol solution for 10 minutes. After the immersion process, the surface of the silicon wafer was gently wiped with an alcohol cotton, and then rinsed with deionized water for 10 minutes. After the rinsing process, the silicon wafer was dried with nitrogen. Figure 18 The cross-section of the silicon wafer after stripping, showing the area to be etched, Figure 19 is shown in the plan view of the corresponding area.

[0128] 6. Deep reactive ion etching

[0129] Deep reactive ion etching was performed on the surface to be etched, with a center opening etching depth of 70 μm. Figure 20 The cross-section of the silicon wafer after deep reactive ion etching, showing the area to be etched.

[0130] 7. Wet etching

[0131] ​The wet chemical etching solution is one of the following combinations:

[0132] Hydrofluoric acid (HF), nitric acid (HNO3), and water (H2O); or

[0133] Hydrofluoric acid (HF), nitric acid (HNO3), and acetic acid (CH3COOH); or

[0134] Hydrofluoric acid (HF), nitric acid (HNO3), and isopropanol (IPA).

[0135] In this embodiment, isotropic wet etching of silicon wafers is performed using an HNA solution. The volume ratio of acetic acid:hydrofluoric acid:nitric acid in the HNA wet etching solution is 2:2:6, with acetic acid concentration of 99.99%, hydrofluoric acid concentration of 49%, and nitric acid concentration of 69%. The silicon wafer and its fixture are then placed... Figure 11 The etching apparatus shown is used to pour in etching liquid and activate magnetic stirring. During the etching process, a high-speed, high-definition camera images the etched area in real time. The computer inputs the expected etching parameters, and the etching circle detection module calculates the sphericity and etching radius of the etching circle under the current etching state. A roughness prediction model is used to obtain a top-view image of the micro-silicon spherical cavity and its surface roughness. An etching depth prediction model is used to obtain a cross-sectional image of the micro-silicon spherical cavity and the predicted etching depth of the micro-silicon spherical cavity under the current etching state. Based on the key parameters of the silicon wafer under the current etching state, such as longitudinal and transverse etching depth, sphericity, roughness, and etching radius, the process parameters of the wet etching are adjusted in real time by comparing the calculated results with the expected etching parameters. Finally, the etching stop point is determined, completing the isotropic wet etching.

[0136] Example 3

[0137] Embodiment 3 of the present invention proposes a method for fabricating a micro-silicon spherical cavity, which specifically includes the following steps:

[0138] 1. Prepare the substrate

[0139] <1 1 1> N-type no-clean silicon wafers are selected. Figure 12 This is a cross-sectional view of the area to be etched on the silicon wafer.

[0140] 2. Deposition and etching mask

[0141] A parylene / aluminum / parylene composite film was prepared on a silicon wafer surface using a parylene deposition machine and ion beam sputtering as an etching mask. The two parylene mask layers were 10 μm thick, and the aluminum film was 100 nm thick. The cross-section after deposition and etching is shown below. Figure 13 As shown.

[0142] 3. Photolithography

[0143] A positive photoresist is spin-coated on the side of the silicon wafer to be etched using a spin coater with a rotation speed of 3000 rpm for 30 seconds. After spin-coating, the wafer is pre-baked at 80°C for 30 minutes. Then, a negative photoresist patterned soda-chrome mask is used as a mask for photolithography and development. Finally, the wafer is post-baked at 100°C for 30 minutes to harden the photoresist. Figure 14 A cross-sectional view of the silicon wafer after the photoresist is removed. Figure 15 A top view of the corresponding region.

[0144] 4. Partial removal of the etching mask

[0145] The etching mask is partially removed using one or more of the following processes: chemical wet etching, dry etching, pulsed laser, or lift-off.

[0146] In this example, the upper parylene mask is removed using a pulsed laser. After removal, the wafer is rinsed with deionized water for 10 minutes and then dried with nitrogen. The aluminum mask is etched using an aluminum etching solution. After etching, the wafer is rinsed with deionized water for 10 minutes and then dried with nitrogen. Then, the lower parylene mask is removed using a pulsed laser. After removal, the wafer is rinsed with deionized water for 10 minutes and then dried with nitrogen. Figure 16 A cross-sectional view of the silicon wafer after the etching mask is partially removed. Figure 17 A top view of the corresponding region after the etching mask is partially removed.

[0147] 5. Photoresist removal

[0148] The photoresist is removed using one of the following methods: acetone wet process, plasma stripping, or ashing. This forms an etching window in the etching mask.

[0149] In this example, the wafer is soaked in an acetone solution for 5 minutes to dissolve the photoresist. After soaking in acetone, the wafer is placed in an anhydrous ethanol solution for 10 minutes. After soaking, the wafer surface is gently wiped with alcohol cotton. Then, the wafer is rinsed with deionized water for 10 minutes and then dried with nitrogen. Figure 18 A cross-sectional view of the silicon wafer after the photoresist is removed. Figure 19 A top view of the corresponding region after the photoresist is removed.

[0150] 6. Deep reactive ion etching

[0151] The surface to be etched is subjected to deep reactive ion etching, with a central opening etching depth of 70 μm. Figure 20 A cross-sectional view of the silicon wafer after deep reactive ion etching.

[0152] 7. Wet etching

[0153] The wet chemical etching solution is one of the following combinations:

[0154] Hydrofluoric acid (HF), nitric acid (HNO3), and water (H2O); or

[0155] Hydrofluoric acid (HF), nitric acid (HNO3), and acetic acid (CH3COOH); or

[0156] Hydrofluoric acid (HF), nitric acid (HNO3), and isopropanol (IPA).

[0157] In this embodiment, isotropic wet etching of silicon wafers is performed using an HNA solution. The volume ratio of acetic acid:hydrofluoric acid:nitric acid in the HNA wet etching solution is 1:3:6, with acetic acid concentration of 99.99%, hydrofluoric acid concentration of 49%, and nitric acid concentration of 69%. The silicon wafer and its fixture are then placed... Figure 11 The etching apparatus shown is used to pour in etching liquid and activate magnetic stirring. During the etching process, a high-speed, high-definition camera images the etched area in real time. The computer inputs the expected etching parameters, and the etching circle detection module calculates the sphericity and etching radius of the etching circle under the current etching state. A roughness prediction model is used to obtain a top-view image of the micro-silicon spherical cavity and its surface roughness. An etching depth prediction model is used to obtain a cross-sectional image of the micro-silicon spherical cavity and the predicted etching depth of the micro-silicon spherical cavity under the current etching state. Based on the key parameters of the silicon wafer under the current etching state, such as longitudinal and transverse etching depth, sphericity, roughness, and etching radius, the process parameters of the wet etching are adjusted in real time by comparing the calculated results with the expected etching parameters. Finally, the etching stop point is determined, completing the isotropic wet etching.

[0158] Overview:

[0159] The method of this invention is a process for isotropic wet etching of silicon in MEMS, which improves the lateral etching effect in real time based on machine learning. It reduces the lateral etching effect of the silicon spherical cavity by combining dry deep reactive ion etching with isotropic wet chemical etching, and predicts the etching effect in real time based on machine learning, including key parameters such as longitudinal and lateral etching depth, sphericity, surface roughness, and etching radius after etching. The desired effect is achieved by adjusting the etching parameters in real time.

[0160] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to the embodiments, those skilled in the art should understand that modifications or equivalent substitutions to the technical solutions of the present invention do not depart from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for processing micro-silicon spherical cavity, the method comprising: Step 1) sequentially performing deposition of etching mask, photolithography, removal of part of etching mask, stripping and deep reactive ion etching on the side of the silicon wafer to be patterned; Step 2) performing isotropic wet etching on the silicon wafer by using wet chemical etching solution to obtain micro-silicon spherical cavity; Step 3) performing real-time image acquisition on the micro-silicon spherical cavity during wet etching; Step 4) preprocessing the collected images and sequentially inputting the etching circle detection module and the pre-established and trained roughness prediction model and etching depth prediction model to obtain the top view image, cross-sectional image and etching related prediction value of the micro-silicon spherical cavity under the current etching state; Step 5) comparing the etching related prediction value with the expected index, if the expected index is not reached, adjusting the wet etching related process parameters and returning to Step 2); if the expected index is met, stopping etching to obtain the micro-silicon spherical cavity meeting the expected index; The etching related prediction value of Step 4) includes the longitudinal and transverse etching depth, sphericity, surface roughness and etching radius of the micro-silicon spherical cavity under the current etching state; wherein the sphericity and etching radius are extracted and calculated by Hough circle detection algorithm; the surface roughness is calculated according to the top view image of the micro-silicon spherical cavity; and the longitudinal and transverse etching depth is calculated according to the cross-sectional image of the micro-silicon spherical cavity.

2. The method of claim 1, wherein the micro-silica spherical cavity is formed by a method comprising: Step 1) comprises: Step 1-1) cleaning the silicon wafer or using a non-washing silicon wafer; Step 1-2) preparing the etching mask by thin film deposition method, the thickness of the single layer film ranges from 10 nm to 100 μm; Step 1-3) spin coating positive photoresist on the etching surface and performing exposure and film hardening; Step 1-4) removing part of the etching mask; Step 1-5) stripping the silicon wafer to remove the photoresist and form the etching window of the etching mask; Step 1-6) performing deep reactive ion etching on the etching surface.

3. The method for fabricating a micro-silicon spherical cavity according to claim 2, characterized in that, Step 1-2) preparing the etching mask by thin film deposition method, specifically: Preparation of gold / chromium, silicon nitride, aluminum, parylene or polyimide thin film and their composite film as etching mask by ion beam sputtering, low pressure chemical vapor deposition or plasma enhanced chemical vapor deposition.

4. The method of claim 2 or 3, wherein the microsilica spherical cavity is formed by a method comprising: The pattern center of the etching mask of Step 1-2) is circular and the opening is large, and various arrangements of circular, annular or wavy patterns and their combinations gradually decrease in diameter, ring width or opening away from the center.

5. The method for fabricating a micro-silicon spherical cavity according to claim 2, characterized in that, Step 1-6) specifically includes: etching the silicon wafer by deep reactive ion etching based on Bosch process, using the loading effect related to etching aspect ratio of deep reactive ion etching to etch deeper in the central region with larger opening and shallower in the surrounding region with smaller opening.

6. The method of claim 1, wherein the micro-silica spherical cavity is formed by a method comprising: The processing process of the etching circle detection model is: according to the preprocessed image, the etching circle boundary is extracted by Hough circle detection algorithm, and then the sphericity and etching radius of the etching circle under the current etching state are calculated. ​ 7. The method for fabricating a micro-silicon spherical cavity according to claim 1, characterized in that, The input of the roughness prediction model is the feature value corresponding to the preprocessed image, and the output is the top view image of the micro-silicon spherical cavity and the surface roughness of the micro-silicon spherical cavity under the current etching state. The feature value includes energy, contrast, entropy, homogeneity, difference and correlation of the image. The roughness prediction model comprises: a BP neural network for outputting the top view image of the micro-silicon spherical cavity; and a surface roughness calculation module for calculating the surface roughness according to the top view image of the micro-silicon spherical cavity. The input of the etching depth prediction model is the preprocessed image, and the output is the cross-sectional image of the micro-silicon spherical cavity and the etching depth prediction value of the micro-silicon spherical cavity under the current etching state.

8. The method of claim 1, wherein the micro-silica spherical cavity is formed by a method comprising: The etching depth prediction model comprises: ​ a convolutional neural network comprising an input layer, a convolutional layer, a pooling layer, a convolutional layer, a pooling layer, a fully connected layer and an output layer connected in sequence, for outputting the cross-sectional image of the micro-silicon spherical cavity; and an etching depth prediction value calculation module for calculating the longitudinal and transverse etching depths according to the cross-sectional image of the micro-silicon spherical cavity.

9. The method for fabricating a micro-silicon spherical cavity according to claim 1, characterized in that, The method further comprises a training step of the roughness prediction model and the etching depth prediction model, specifically comprising: cutting the sample along the center line after etching is completed, and obtaining the longitudinal and transverse etching depths, sphericity, roughness, etching radius, top view and cross-sectional image after cutting of the sample by using a scanning electron microscope, a laser three-dimensional imaging system and an optical profiler, and establishing corresponding training samples of the parameter results and the imaging results of the etching process to form a training set; inputting the training set into the roughness prediction model in sequence, and obtaining the roughness prediction model meeting the training requirements by adjusting the model parameters; inputting the training set into the etching depth prediction model in sequence, and obtaining the etching depth prediction model meeting the training requirements by adjusting the model parameters.

Images