Silicon wafer TTV control method based on process optimization

Abstract

The invention discloses a silicon wafer TTV control method based on process optimization, and particularly relates to the technical field of semiconductors, and the method comprises the steps: S1, employing a metal-pollution-free alkali corrosion process; s2, a single-face grinding process is introduced; s3, the regional thickness difference in the silicon wafer polishing process is detected in real time through a dial indicator; and S4, collecting the process parameters, the equipment state and the silicon wafer TTV detection data in the steps S1 to S3. According to the silicon wafer TTV control method based on process optimization, core process parameters such as S1 corrosion time, S2 grinding removal amount, S3 polishing medium pressure and constant-temperature water temperature are collected in a full-dimension mode, the equipment operation state and silicon wafer TTV detection data are recorded synchronously, and comprehensive and accurate data support is provided for model establishment; according to the method, a final silicon wafer TTV prediction model is constructed, and the TTV value of a new batch of silicon wafers can be accurately predicted; and according to the deviation between a prediction result and an actual detection value, parameters of each process are reversely corrected, and a closed loop of parameter input-prediction-verification-correction is formed.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of semiconductor technology, in particular to a silicon wafer TTV control method based on process optimization. BACKGROUND

[0002] In the semiconductor industry, as the core substrate for carrying integrated circuits, the total thickness variation (TTV) of silicon wafers is a key indicator that determines the yield and reliability of downstream chip manufacturing. Whether it is lithography alignment for logic chips, heat dissipation performance for power devices, or stacked packaging for memory chips, all require high precision in silicon wafer TTV.

[0003] Especially with the advancement of semiconductor processes to 7nm and below, the industry has increasingly stringent requirements for silicon wafer TTV control precision and mass production stability. However, existing processes have obvious bottlenecks: the pre-treatment stage uses ordinary alkali etching process with high metal ion content, the etching solution is not selected according to the type of silicon wafer, and the temperature and time control is rough, resulting in large etching removal deviation; the grinding process does not subdivide the initial TTV, and the fixed parameters are prone to excessive grinding or insufficient correction, and the thickness monitoring is lagging; the polishing process lacks real-time thickness difference pressure adjustment, and the temperature control precision of the fixed disc is low, resulting in large thermal deformation; meanwhile, there is a lack of data feedback and closed loop in each link, resulting in large batch TTV fluctuations, which restricts the development of silicon wafers towards high precision.

[0004] Therefore, a silicon wafer TTV control method based on process optimization is needed. SUMMARY

[0005] The main purpose of the present application is to provide a silicon wafer TTV control method based on process optimization, which can effectively solve the problems raised in the above.

[0006] To achieve the above purpose, the technical solution adopted by the present application is:

[0007] A silicon wafer TTV control method based on process optimization, comprising:

[0008] S1: using a metal-pollution-free alkali etching process, by controlling the type of etching solution, etching temperature and time, achieving uniform etching removal of 3-5um, and improving the initial flatness of the silicon wafer;

[0009] S2: introducing a single-face grinding process, setting a grinding removal of 5-10um according to the initial TTV value of the silicon wafer after S1 treatment, and accurately adjusting the thickness distribution of the silicon wafer through a single-face grinding machine;

[0010] S3: real-time detecting the regional thickness difference of the silicon wafer during polishing by using a millionth meter, dynamically adjusting the corresponding area pressure of the polishing machine, and combining constant temperature water control to realize the negative taper of the silicon wafer in the longitudinal direction;

[0011] S4: Collecting the process parameters, equipment state and silicon wafer TTV detection data of S1-S3, establishing a data feedback model, continuously optimizing the process parameters, and improving the TTV control stability.

[0012] Preferably, the alkali etching process in S1 comprises:

[0013] S11: Selecting the etching solution according to the type of silicon wafer, selecting TMAH etching solution for P-type silicon wafer and KOH etching solution for N-type silicon wafer, and the content of metal ions in the etching solution is ≤0.1 ppm;

[0014] S12: Controlling the etching temperature to be 85-95℃ and the etching time to be 3-5 minutes, and ensuring that the etching removal amount deviation of each batch of silicon wafer is ≤0.5 μm;

[0015] S13: Real-time monitoring the concentration of the etching solution during the etching process, and supplementing fresh etching solution when the concentration is reduced by more than 10% compared with the initial value to maintain the etching uniformity.

[0016] Preferably, the single-side grinding process in S2 comprises:

[0017] S21: Detecting the initial TTV value of the silicon wafer after S1 treatment by using a laser thickness gauge, and dividing the initial TTV into two to three μm and three to four μm two ranges;

[0018] S22: Setting the grinding speed of the single-side grinding machine to be 3000-3500 rpm and the feed speed to be 8-10 mm / min, and matching the corresponding grinding removal amount according to different TTV ranges;

[0019] S23: Real-time tracking the removal amount by the thickness monitoring module during the grinding process, and automatically adjusting the feed speed of the grinding machine when the actual removal amount deviates from the set value by more than 0.5 μm.

[0020] Preferably, the dynamic adjustment of the polishing machine area pressure in S3 comprises:

[0021] S31: Arranging a millionth table on the polishing machine workbench, collecting the thickness data of the outer side, inner side and center area of the silicon wafer every 10 seconds, and calculating the difference Δt between each area and the target thickness;

[0022] S32: Establishing a polishing pressure adjustment model based on the thickness difference Δt, and the adjustment formula is:

[0023] ΔP = K·Δt

[0024] Wherein, ΔP is the pressure adjustment amount of the corresponding area of the polishing machine, K is the pressure adjustment coefficient and K = 10 KPa / μm, and Δt is the area thickness difference;

[0025] S33: When the thickness difference between the outer side and the inner side of the silicon wafer is 2 μm, the medium pressure needs to be reduced by 20 KPa according to the formula; when the thickness difference is 1.2 μm, the medium pressure needs to be reduced by 12 KPa.

[0026] Preferably, the constant-temperature water control disc topography in S3 comprises:

[0027] S301: Control the constant-temperature water temperature at 25±2℃, and monitor the temperature distribution of the polishing disc in real time through a water temperature sensor;

[0028] S302: When the temperature difference between the edge and the center area of the disc is more than 1℃, adjust the constant-temperature water circulation rate to ensure that the disc thermal deformation amount is less than or equal to 0.05 mm, and achieve the control target of negative longitudinal taper of the silicon wafer.

[0029] Preferably, the etching liquid concentration supplementing rule in S13 comprises:

[0030] When the etching liquid concentration detection value C is lower than the initial concentration C0, the volume V of the fresh etching liquid to be supplemented is determined as follows:

[0031]

[0032] V = (C0-C)×V0 / C0 总 is the total volume of the etching liquid in the etching tank, to ensure that the etching liquid concentration after supplementing is restored to more than 95% of the initial concentration.

[0033] Preferably, the lapping removal amount grading formula in S22 comprises:

[0034] According to the initial TTV value of the silicon wafer, set the lapping removal amount Q, and the grading formula is:

[0035] When 2 μm≤TTV 初始 ≤3 μm, Q=5+2·(TTV 初始 -2);

[0036] When 3 μm<TTV 初始 ≤4 μm, Q=8+2·(TTV 初始 -3);

[0037] The lapping removal amount is accurately matched with the initial TTV, and the TTV of the silicon wafer after lapping is ensured to be less than or equal to 1.5 μm.

[0038] Preferably, the data feedback model in S4 comprises:

[0039] The etching time t 腐蚀 of S1, the lapping removal amount Q 研削 of S2, the polishing medium pressure P 抛光 of S3, and the constant-temperature water temperature T 水温For input variables, the final TTV prediction model of the silicon wafer is established:

[0040] TTV = 0.1 t 腐蚀 + 0.05 Q 研削 - 0.03 P 抛光 + 0.02 T 水温 - 0.5

[0041] Wherein, the model fitting goodness R 2 > 0.9, the TTV value of the new batch of silicon wafers can be predicted by the model, and each process parameter is corrected in reverse.

[0042] Compared with the prior art, the present application has the following beneficial effects:

[0043] 1. The method matches the etching solution according to the type of the silicon wafer, strictly controls the metal ion to be less than or equal to 0.1 ppm, ensures that the etching removal amount deviation is less than or equal to 0.5 microns through the temperature of 85-95 DEG C, the time of 3-5 minutes and the concentration supplement formula, simultaneously divides the initial TTV of the silicon wafer after S1 treatment into two grades of 2-3 microns and 3-4 microns, sets the lapping removal amount according to the corresponding formula, monitors and adjusts the feed speed in real time, and guarantees that the TTV after lapping is less than or equal to 1.5 microns; the method collects the thickness difference of the silicon wafer area every 10 seconds through the microammeter, dynamically adjusts the polishing pressure according to the formula of ΔP = 10KPa / micron*Δt, and realizes the longitudinal taper negative by cooperating with the polishing fixed disc controlled by the constant temperature water of 25+2 DEG C, thereby breaking through the TTV precision bottleneck of the silicon wafer as a whole and meeting the demand of advanced semiconductor process of 7nm and below.

[0044] 2. The method collects the core process parameters such as S1 etching time, S2 lapping removal amount, S3 polishing pressure and constant temperature water temperature in full dimension, synchronously records the equipment running state and the silicon wafer TTV detection data, provides comprehensive and accurate data support for model establishment, constructs the final TTV prediction model of the silicon wafer, can accurately predict the TTV value of the new batch of silicon wafers, and reversely corrects each process parameter according to the deviation between the prediction result and the actual detection value, thereby forming the closed loop of "parameter input-prediction-verification-correction". The closed loop effectively reduces the batch difference of the traditional process, reduces the fluctuation range of the silicon wafer TTV, improves the production stability and yield, and provides guarantee for the high-precision and large-scale production of semiconductor silicon wafers. BRIEF DESCRIPTION OF DRAWINGS

[0045] Figure 1 It is the overall method flow diagram of the present application;

[0046] Figure 2 It is the S1 flow diagram of the present application. DETAILED DESCRIPTION

[0047] In order to make the technical means, creative features, purposes and effects of the present application easy to understand, the present application is further described below in conjunction with specific embodiments.

[0048] As shown in the figure, a silicon wafer TTV control method based on process optimization comprises:

[0049] S1: using a metal-pollution-free alkali etching process, by controlling the type of etching solution, etching temperature and time, a uniform etching removal amount of 3-5 μm is achieved, and the initial flatness of the silicon wafer is improved;

[0050] S2: introducing a single-side lapping process, according to the initial TTV value of the silicon wafer after S1 treatment, a lapping removal amount of 5-10 μm is set according to the initial TTV value of the silicon wafer after S1 treatment, and the thickness distribution of the silicon wafer is precisely adjusted by a single-side lapping machine;

[0051] S3: real-time detect the regional thickness difference of the silicon wafer during the polishing process by using a millionth meter, dynamically adjust the corresponding area pressure of the polishing machine, and combine with the constant temperature water control polishing disc morphology to realize the negative taper of the silicon wafer in the longitudinal direction;

[0052] S4: collect the process parameters, equipment states and silicon wafer TTV detection data of S1-S3, establish a data feedback model, continuously optimize the process parameters, and improve the TTV control stability.

[0053] The implementation method of the present scheme is:

[0054] Firstly, the S1 metal-pollution-free alkali etching process is performed, the type of the silicon wafer is identified first, for example, the resistivity of the silicon wafer is detected by a four-probe tester, if the resistivity is >10 Ω·cm, it is determined as a P-type silicon wafer, and TMAH etching solution is selected, if the resistivity is <1 Ω·cm, it is determined as an N-type silicon wafer, and KOH etching solution is selected, then the metal ion content in the etching solution is detected by an inductively coupled plasma mass spectrometer, 10 ml of etching solution sample is detected, and the content of Fe, Cu, Ni and other metal ions is ensured to be ≤0.1 ppm, such as 0.08 ppm of Fe ion and 0.05 ppm of Cu ion, which meets the requirements, then the etching temperature is set, if the target removal amount is 3 μm, the temperature is controlled at 85℃ for 3 minutes, if the target removal amount is 5 μm, the temperature is controlled at 95℃ for 5 minutes, 50 silicon wafers are taken in each batch, the thickness is measured at the center and the edge of each silicon wafer by using a screw micrometer, the removal amount is calculated, such as the thickness before etching is 500 μm, the thickness after etching is 496 μm, and the removal amount is 4 μm, and the removal amount deviation of each batch is ensured to be ≤0.5 μm;

[0055] Then, the single surface grinding process S2 is performed, the initial TTV value of the silicon wafer after S1 treatment is detected by a laser thickness gauge (laser power 5 mW, measurement distance 10 cm), the center and edge 4-point thickness of each silicon wafer is measured, and the maximum value minus the minimum value is the initial TTV value, for example, the thickness of a certain silicon wafer is 496 μm (center), 495 μm, 493 μm, 494 μm, 495 μm, the initial TTV value is 3 μm, and it is divided into 2-3 μm range, if the thickness is 497 μm, 494 μm, 493 μm, 495 μm, 496 μm, the initial TTV value is 4 μm, and it is divided into 3-4 μm range, and then the single surface grinding machine speed is set to 3000-3500 rpm, and the feed speed is set to 8-10 mm / min, the removal amount is calculated according to Q = 5 + 2 × (TTV initial-2) for 2-3 μm range, for example, the initial TTV value is 2.5 μm, Q = 6 μm, the removal amount is calculated according to Q = 8 + 2 × (TTV initial-3) for 3-4 μm range, for example, the initial TTV value is 3.5 μm, Q = 9 μm, and the thickness monitoring module measures the thickness every second during grinding, and if the actual removal amount deviates from the set value by more than 0.5 μm, the feed speed is automatically adjusted;

[0056] Then, the polishing control S3 is performed, one millionth table (accuracy 0.001 mm, probe pressure 0.5 N) is installed outside (5 mm from the edge), inside (20 mm from the edge), and center of the polishing machine workbench, thickness data is collected every 10 seconds, and the difference Δt between each area and the target thickness is calculated, for example, the target thickness is 490 μm, the outside is measured to be 492 μm, Δt = 2 μm, the pressure is adjusted based on ΔP = K·Δt (K = 10 KPa / μm), the outside ΔP = 20 KPa, that is, the medium pressure is reduced by 20 KPa, and the temperature of the constant temperature water is controlled at 25±2℃, the edge and center temperatures of the disc are measured by a water temperature sensor, if the temperature difference exceeds 1℃, the constant temperature water circulation rate is adjusted to ensure that the disc thermal deformation amount is ≤0.05 mm;

[0057] Finally, the data feedback S4 is performed, the corrosion time t corrosion of S1, the grinding removal amount Q grinding of S2, the polishing medium pressure P polishing of S3, and the constant temperature water temperature T water temperature of S3 are collected, for example, t corrosion = 4 minutes, Q grinding = 6 μm, P polishing = 160 KPa, and T water temperature = 25℃, which are substituted into the TTV prediction model TTV = 0.1 × t corrosion + 0.05 × Q grinding-0.03 × P polishing + 0.02 × T water temperature-0.5 to calculate, the new batch TTV value is predicted by using the model, and the process parameters are corrected in reverse to improve the TTV control stability.

[0058] Further, the alkali corrosion process in S1 comprises:

[0059] S11: Select the corrosion liquid according to the type of silicon wafer, select TMAH corrosion liquid for P-type silicon wafer, and select KOH corrosion liquid for N-type silicon wafer, and the metal ion content in the corrosion liquid is ≤0.1 ppm;

[0060] S12: control the etching temperature to be 85-95℃ and the etching time to be 3-5 minutes, and ensure that the etching removal deviation of each batch of silicon wafer is ≤0.5μm;

[0061] S13: monitor the etching liquid concentration in real time during the etching process, and when the concentration is reduced by more than 10% compared with the initial value, supplement fresh etching liquid to maintain the etching uniformity.

[0062] In the above, S11 first distinguishes the type by detecting the silicon wafer resistivity through the four-probe tester, and tests the center area of the silicon wafer. If the test resistivity is 15Ω·cm, it is determined that the silicon wafer is P-type, and the TMAH etching liquid is selected. If the test resistivity is 0.8Ω·cm, it is determined that the silicon wafer is N-type, and the KOH etching liquid is selected. Then, 10ml of the prepared etching liquid sample is taken, and the metal ion content is detected by the inductively coupled plasma mass spectrometer, with the detection accuracy set to 0.01ppm. If the detected Fe ion is 0.07ppm, the detected Cu ion is 0.04ppm, and the detected Ni ion is 0.02ppm, all meet the requirement of ≤0.1ppm. If the detected content of a certain metal ion is 0.12ppm, the etching liquid needs to be filtered through a filter membrane with a pore size of 0.22μm until the detected metal ion content is ≤0.1ppm again;

[0063] S12 connects the heating device of the etching tank to the temperature controller, sets the etching temperature, and if the target removal amount is 3μm, the temperature is set to 85℃, and if the target removal amount is 5μm, the temperature is set to 95℃. After the temperature is stabilized within the set value±1℃ range, 50 silicon wafers are put in and the timing starts. The removal amount of 3μm corresponds to 3 minutes of timing, and the removal amount of 5μm corresponds to 5 minutes of timing. After the etching is completed, the silicon wafers are taken out, the thickness of each silicon wafer is measured at the center, the east, south, west and north edges of each silicon wafer by using a screw micrometer (accuracy 0.001mm), and the removal amount of each silicon wafer (thickness before etching minus thickness after etching) is calculated. For example, if the thickness of a batch of silicon wafers before etching is all 500μm, and the removal amounts of 10 silicon wafers after etching are measured to be 3.1μm, 3.0μm, 2.9μm, 3.2μm, 2.8μm, 3.0μm, 3.1μm, 2.9μm, 3.0μm and 2.8μm respectively, the average removal amount of this batch is calculated to be 2.98μm, and the maximum deviation is 3.2-2.98=0.22μm, which is ≤0.5μm, meeting the requirement;

[0064] S13 In the etching process, the concentration of the etching solution is detected every 2 minutes with a concentration meter, and the initial concentration C0 is recorded, such as the initial configuration of the TMAH etching solution concentration C0=25%, when the concentration C=22% is detected at a certain time, C0-C=3% is calculated, 3%>2.5% (10%xC0=2.5%), fresh etching solution needs to be supplemented, if the total volume of the etching solution in the etching tank Vtotal=100L, according to the formula of the supplement volume V=100x(25-22) / 25=12L, 12L of fresh TMAH etching solution with a concentration of 25% is prepared, and is slowly added to the etching tank at a rate of 50ml / s, and is stirred with a stirrer (speed 50r / min) while adding, and is left for 2 minutes after adding, and the concentration is detected again with a concentration meter, if the detected concentration is 24.7%, 24.7%>25%*95%=23.75%, the supplement is completed, if the detected concentration is 23.5%, the supplement volume V=100x(25-23.5) / 25=6L is recalculated according to the above formula, and 6L of fresh etching solution is supplemented and detected again until the concentration is greater than or equal to 23.75%.

[0065] Further, the single-sided grinding process in S2 includes:

[0066] S21: The initial TTV value of the silicon wafer after S1 treatment is detected by a laser thickness gauge, and the initial TTV is divided into two to three microns and three to four microns.

[0067] S22: The grinding speed of the single-sided grinding machine is set to 3000-3500 rpm, the feed speed is 8-10 mm / min, and the corresponding grinding removal amount is matched according to different TTV grades.

[0068] S23: The removal amount is tracked in real time during the grinding process through a thickness monitoring module, and when the actual removal amount deviates from the set value by more than 0.5 microns, the feed speed of the grinding machine is automatically adjusted.

[0069] In the above, S21 places the silicon wafer (e.g. 50 pieces) processed by S1 on the stage of the laser thickness gauge one by one, adjusts the laser emission power of the thickness gauge to 5 mW and the measurement distance to 10 cm, selects 5 measurement points for each silicon wafer, which are the center of the silicon wafer (center), the east, south, west and north points 1 cm away from the edge of the silicon wafer, starts the thickness gauge to measure the thickness of each point, for example, the measurement data of a silicon wafer is center 496 μm, east 495 μm, south 493 μm, west 494 μm and north 495 μm, the initial TTV is calculated as 496 μm - 493 μm = 3 μm, which is determined as the 2-3 μm grade, the measurement data of another silicon wafer is center 497 μm, east 494 μm, south 493 μm, west 495 μm and north 496 μm, the initial TTV is 497 μm - 493 μm = 4 μm, which is determined as the 3-4 μm grade, if the initial TTV of a silicon wafer is 2.4 μm, it is classified as the 2-3 μm grade, and if the initial TTV is 3.3 μm, it is classified as the 3-4 μm grade;

[0070] S22 sets the spindle speed of the single-side grinding machine to 3000 rpm (if the diameter of the silicon wafer is large, e.g. 8-inch silicon wafer, set to 3500 rpm), sets the feed speed of the feed motor to 8 mm / min (if the grinding removal amount is large, set to 10 mm / min), for the silicon wafer of the 2-3 μm grade, the grinding removal amount Q is calculated by the formula Q = 5 + 2 × (TTV initial - 2), for example, TTV initial = 2.2 μm, Q = 5 + 2 × (2.2 - 2) = 5.4 μm, TTV initial = 2.9 μm, Q = 5 + 2 × (2.9 - 2) = 6.8 μm, for the silicon wafer of the 3-4 μm grade, Q is calculated by the formula Q = 8 + 2 × (TTV initial - 3), for example, TTV initial = 3.1 μm, Q = 8 + 2 × (3.1 - 3) = 8.2 μm, TTV initial = 3.8 μm, Q = 8 + 2 × (3.8 - 3) = 9.6 μm;

[0071] S23 install infrared thickness monitoring module (measurement accuracy 0.1 μm, measurement frequency 1 / second) under the diamond tool of the lapping machine, input the calculated lapping removal amount Q into the lapping machine control system, the control system automatically calculates the target lapping thickness according to the thickness of the silicon wafer S1 after processing, such as the thickness of the silicon wafer S1 after processing is 496 μm, Q = 5.4 μm, target thickness = 496-5.4 = 490.6 μm, after lapping starts, the infrared monitoring module measures the thickness of the silicon wafer in real time and transmits it to the control system, when the thickness of the silicon wafer is measured to be 491.2 μm, the actual removal amount = 496-491.2 = 4.8 μm, the deviation from the set value 5.4 μm = 5.4-4.8 = 0.6 μm > 0.5 μm, the control system sends a signal to the feed motor, increases the feed speed from 8 mm / min to 8.5 mm / min, continues to monitor, when the thickness is measured to be 490.8 μm, the actual removal amount = 5.2 μm, the deviation = 0.2 μm ≤ 0.5 μm, maintain the feed speed, when the thickness is measured to be 490.6 μm, the actual removal amount = 5.4 μm, the deviation = 0, the control system controls the lapping machine to stop lapping.

[0072] Further, the dynamic adjustment of the polishing machine area pressure in S3 includes:

[0073] S31: Place a micro-ohmmeter on the polishing machine workbench, collect the thickness data of the silicon wafer outside, inside and center area every 10 seconds, and calculate the difference Δt between each area and the target thickness;

[0074] S32: Establish a polishing pressure adjustment model based on the thickness difference Δt, the adjustment formula is:

[0075] ΔP = K·Δt

[0076] Where ΔP is the medium pressure adjustment amount of the corresponding area of the polishing machine, K is the pressure adjustment coefficient and K = 10 KPa / μm, Δt is the area thickness difference;

[0077] S33: When the thickness difference between the outside and the inside of the silicon wafer is 2 μm, calculate according to the formula that the medium pressure needs to be reduced by 20 KPa; when the thickness difference is 1.2 μm, the medium pressure needs to be reduced by 12 KPa.

[0078] In the above, S31 installs the ohmmeter at the designated position of the polishing machine workbench, the outer ohmmeter is installed at a distance of 5 mm from the edge of the silicon wafer, the inner ohmmeter is installed at a distance of 20 mm from the edge of the silicon wafer, and the center ohmmeter is installed directly above the center of the silicon wafer. The ohmmeter with an accuracy of 0.001 mm is selected, the contact pressure of the ohmmeter probe and the surface of the silicon wafer is adjusted to 0.5 N to avoid damage to the surface of the silicon wafer due to excessive probe pressure, the ohmmeter is connected with the polishing machine control system, and the data collection interval is set to 10 seconds. Every 10 seconds, the control system automatically reads the thickness data of the three ohmmeters. If the target polishing thickness of the silicon wafer is set to 490 μm, the first collection (at 10 seconds) is 491.8 μm for the outer side, 490.2 μm for the inner side, and 489.7 μm for the center. The second collection (at 20 seconds) is 492 μm for the outer side, 490.1 μm for the inner side, and 489.6 μm for the center. The third collection (at 30 seconds) is 491.5 μm for the outer side, 490 μm for the inner side, and 489.8 μm for the center. After each collection, the control system calculates the difference Δt between the thickness of each region and the target thickness. The first time, Δt for the outer side is 491.8-490=1.8 μm, Δt for the inner side is 490.2-490=0.2 μm, and Δt for the center is 489.7-490=-0.3 μm.

[0079] S32 presets the pressure adjustment coefficient K=10 KPa / μm in the polishing machine control system. This coefficient is determined through preliminary experiments. Ten silicon wafers of the same specification are selected, and different Δt values are set under the same polishing conditions. The corresponding pressure adjustment amounts are recorded. For example, when Δt=1 μm, the adjustment pressure is increased by 10 KPa, and the next collection Δt is reduced to 0.1 μm. The rationality of K=10 KPa / μm is verified. The pressure adjustment relationship ΔP=K·Δt is established. ΔP is the medium pressure adjustment amount of the corresponding region of the polishing machine. If Δt is positive and ΔP is positive, it indicates that the medium pressure is increased. If ΔP is negative, it indicates that the medium pressure is decreased. If Δt is negative and ΔP is positive, it indicates that the medium pressure is increased to increase the removal amount. If ΔP is negative, it indicates that the medium pressure is decreased. For example, for a region with Δt=0.7 μm, ΔP=10×0.7=7 KPa, i.e., the medium pressure of the region is increased by 7 KPa. For a region with Δt=-0.4 μm, ΔP=10×(-0.4)=-4 KPa, i.e., the medium pressure of the region is decreased by 4 KPa.

[0080] S33 When the polishing machine control system detects that the outer side thickness of the silicon wafer is 492 μm, the inner side thickness is 490 μm, the target thickness is 490 μm, the thickness difference between the outer side and the inner side is calculated as 492-490=2 μm, the thickness difference is taken as the Δt of the outer side area and substituted into the formula, ΔP=10×2=20 KPa, since the outer side thickness is higher than the target thickness, the middle pressure in the outer side area needs to be reduced, if the current middle pressure in the outer side area is 180 KPa, the adjusted middle pressure is 180-20=160 KPa, after the adjustment, the thickness is continuously monitored, when it is detected that the outer side thickness is reduced to 490.5 μm, the inner side thickness is 490 μm, the thickness difference is 0.5 μm, ΔP=5 KPa, the middle pressure is further reduced by 5 KPa to 155 KPa; when it is detected that the outer side thickness of the silicon wafer is 491.2 μm, the inner side thickness is 490 μm, the thickness difference is 1.2 μm, ΔP=10×1.2=12 KPa, if the current outer side middle pressure is 170 KPa, the adjusted middle pressure is 170-12=158 KPa, and it is ensured that the thickness of each area approaches the target thickness through pressure adjustment.

[0081] Further, the constant temperature water control disc topography in S3 includes:

[0082] S301: control the constant temperature water temperature at 25±2℃, and monitor the temperature distribution of the polishing disc in real time through a water temperature sensor;

[0083] S302: when the temperature difference between the edge and the center area of the disc is more than 1℃, adjust the constant temperature water circulation rate to ensure that the disc thermal deformation amount is less than or equal to 0.05 mm, and achieve the control target of negative silicon wafer longitudinal taper.

[0084] In the above, S301 installs an intelligent temperature controller at the inlet of the cooling water path of the polishing disc, sets the constant temperature water temperature target value as 25℃, and the temperature fluctuation allowable range is ±2℃, that is, the temperature control interval is 23℃-27℃, installs one platinum resistance water temperature sensor at the edge position (10 mm away from the disc edge) and the center position (the disc center) of the disc respectively, the sensor accuracy is 0.1℃, connects the signal output end of the sensor with the signal input end of the temperature controller, and the temperature controller receives the temperature data transmitted by the sensor in real time, for example, at a certain moment, the edge sensor detects that the temperature is 24.5℃, and the center sensor detects that the temperature is 25.2℃, both are within the control interval of 23℃-27℃, the temperature controller maintains the current heating or cooling state, when the edge sensor detects that the temperature is 22.8℃, which is lower than the lower limit of 23℃, the temperature controller starts the heating device to heat the constant temperature water to 23℃, when the edge sensor detects that the temperature is 27.3℃, which is higher than the upper limit of 27℃, the temperature controller starts the cooling device to cool the constant temperature water to 27℃;

[0085] S302 sets the temperature difference threshold of the edge and the center area of the platen to 1℃, and the temperature controller calculates the temperature difference detected by the edge and center sensors in real time. When the calculated temperature difference = edge temperature - center temperature, if the temperature difference ≤1℃, the temperature controller maintains the current constant temperature water circulation rate (such as 5L / min), if the temperature difference >1℃, the temperature controller sends an adjustment signal to the constant temperature water circulating pump. For example, if the edge temperature is detected to be 26.5℃ and the center temperature is 25.3℃, the temperature difference =1.2℃>1℃, the temperature controller controls the circulating pump to increase the circulation rate from 5L / min to 6L / min to improve the cooling efficiency. The temperature is continuously monitored, and after 5 minutes, the edge temperature is detected to be 25.8℃ and the center temperature is 25.2℃, the temperature difference =0.6℃≤1℃, the temperature controller controls the circulating pump to restore to the circulation rate of 5L / min;

[0086] At the same time, a laser displacement sensor (measurement accuracy 0.001mm) is installed 10cm above the platen to measure the deformation of the platen surface in real time. The sensor collects displacement data of the edge and center of the platen every 2 minutes, calculates the deformation (difference between displacement data and initial state), and adjusts the circulation rate. For example, after adjusting the circulation rate, the edge deformation of the platen is measured to be 0.03mm and the center deformation is 0.02mm, both ≤0.05mm, which meets the requirements. If the edge deformation is measured to be 0.06mm and the center is 0.04mm, the temperature controller further adjusts the circulation rate to 6.5L / min until the deformation ≤0.05mm.

[0087] Further, the corrosion liquid concentration replenishment rule in S13 includes:

[0088] When the corrosion liquid concentration detection value C decreases from the initial concentration C0, the volume V of fresh corrosion liquid to be replenished is determined as follows:

[0089]

[0090] Wherein, V 总 is the total volume of the corrosion liquid in the corrosion tank, to ensure that the concentration of the corrosion liquid after replenishment is restored to more than 95% of the initial concentration.

[0091] In the above, the corrosion liquid concentration replenishment rule in S13 is that when the corrosion liquid concentration detection value C decreases from the initial concentration C0, the volume V of fresh corrosion liquid to be replenished is determined according to the formula. First, measure the total volume Vtotal of the corrosion liquid in the corrosion tank. The corrosion tank is a cuboid structure with a length of 120cm, a width of 60cm and a height of 50cm. The actual liquid filling height is 35cm. According to the cuboid volume formula Vtotal = length x width x height, the Vtotal = 120 x 60 x 35 = 252000cm3 = 252L is calculated;

[0092] After the etching liquid is just prepared and stirred uniformly, the initial concentration C0 is measured by the concentration meter, for example, the C0 of the TMAH etching liquid is measured as 28%; during the etching process of the silicon wafer, 5 ml of the etching liquid sample is extracted from the middle position of the etching tank every 3 minutes by the concentration meter, the current concentration C is measured, the first measurement C is 27.5%, the calculation C0-C is 0.5%, 0.5% < 2.8% (10% x C0 = 2.8%), and no fresh etching liquid needs to be supplemented, the second measurement C is 24.5%, C0-C is 3.5% > 2.8%, and the fresh etching liquid needs to be supplemented;

[0093] According to the supplement volume formula V = Vtotal x (C0-C) / C0, Vtotal = 252 L, C0 = 28%, and C = 24.5%, V = 252 x (28-24.5) / 28 = 252 x 3.5 / 28 = 31.5 L is calculated; 31.5 L of fresh TMAH etching liquid with a concentration of 28% is prepared, the fresh etching liquid is poured into a corrosion-resistant storage tank, the fresh etching liquid is slowly injected into the etching tank through a peristaltic pump (flow rate 1 L / min), the stirrer (speed 60 r / min) in the etching tank is started during the injection process to ensure uniform mixing of the etching liquid, and after the injection is completed, the etching liquid is left to stand for 3 minutes. The concentration of the etching liquid is measured again by the concentration meter. If the measured concentration is 27.2%, 27.2% ≥ 28% x 95% = 26.6%, it meets the supplement requirement, and if the measured concentration is 26.2%, it is lower than 26.6%, the supplement volume is recalculated, C0 = 28%, the current C = 26.2%, and V = 252 x (28-26.2) / 28 = 252 x 1.8 / 28 = 16.2 L. After 16.2 L of fresh etching liquid is supplemented, the concentration is measured again until the concentration ≥ 26.6%.

[0094] Further, the lapping removal amount grading formula in S22 includes:

[0095] According to the initial TTV value of the silicon wafer, the lapping removal amount Q is set, and the grading formula is:

[0096] When 2 μm ≤ TTV 初始 ≤ 3 μm, Q = 5 + 2·(TTV 初始 - 2);

[0097] When 3 μm < TTV 初始 ≤ 4 μm, Q = 8 + 2·(TTV 初始 - 3);

[0098] The lapping removal amount and the initial TTV are accurately matched to ensure that the TTV of the silicon wafer after lapping is ≤ 1.5 μm.

[0099] In the above, the lapping removal amount grading formula in S22 is set according to the initial TTV value of the silicon wafer, and is calculated in two grades. First, the initial TTV value of the silicon wafer after S1 processing is measured by using a laser thickness gauge (measurement accuracy 0.1 μm), and five measurement points are selected for each silicon wafer, which are one center point and four evenly distributed edge points (east, south, west, and north). After measuring the thickness of each point, the initial TTV value TTVinitial is obtained by subtracting the minimum value from the maximum value.

[0100] For the first grade, when 2 μm≤TTVinitial≤3 μm, the lapping removal amount Q is calculated by the formula Q=5+2×(TTVinitial-2). For example, if the thickness of each point of a certain silicon wafer is 495 μm (center), 494.5 μm (east), 493.2 μm (south), 493.8 μm (west), and 494.2 μm (north), the maximum value is 495 μm, the minimum value is 493.2 μm, and TTVinitial=1.8 μm, which does not conform to the grade. If the thickness of each point of a certain silicon wafer is 496 μm (center), 495.5 μm (east), 494.1 μm (south), 494.7 μm (west), and 495.1 μm (north), TTVinitial=1.9 μm, which also does not conform. If the thickness of each point of a certain silicon wafer is 496.2 μm (center), 495.8 μm (east), 494.2 μm (south), 494.9 μm (west), and 495.3 μm (north), TTVinitial=2.0 μm, and the formula Q=5+2×(2.0-2)=5 μm is substituted, TTVinitial=2.6 μm of a certain silicon wafer, Q=5+2×(2.6-2)=6.2 μm, and TTVinitial=3.0 μm of a certain silicon wafer, Q=5+2×(3.0-2)=7 μm;

[0101] For the second grade, when 3 μm≤TTVinitial≤4 μm, the lapping removal amount Q is calculated by the formula Q=8+2×(TTVinitial-3). For example, if the thickness of each point of a certain silicon wafer is 497.1 μm (center), 496.5 μm (east), 493.8 μm (south), 494.5 μm (west), and 495.2 μm (north), TTVinitial=3.3 μm, and the formula Q=8+2×(3.3-3)=8.6 μm is substituted, TTVinitial=3.7 μm of a certain silicon wafer, Q=8+2×(3.7-3)=9.4 μm, and TTVinitial=4.0 μm of a certain silicon wafer, Q=8+2×(4.0-3)=10 μm;

[0102] The calculated Q value is input into the single-sided grinding machine control system, and the control system determines the target grinding thickness according to the thickness of the S1 processed silicon wafer, such as the thickness of the S1 processed silicon wafer is 496 μm, Q = 6.2 μm, the target thickness = 496-6.2 = 489.8 μm, the thickness monitoring module tracks the removal amount in real time during the grinding process to ensure that the actual removal amount deviates from Q by ≤0.5 μm, and after the grinding is completed, the silicon wafer TTV value is measured again using the laser thickness gauge, such as the thickness of a certain silicon wafer after grinding is 489.8 μm, 489.7 μm, 489.6 μm, 489.5 μm, 489.7 μm, TTV = 0.3 μm ≤1.5 μm, which meets the requirements.

[0103] Further, the data feedback model in S4 includes:

[0104] The etching time t of S1 腐蚀 , the grinding removal amount Q of S2 研削 , the polishing pressure P of S3 抛光 , the constant temperature water temperature T of S3 水温 are input variables, and a final TTV prediction model of the silicon wafer is established:

[0105] TTV = 0.1·t 腐蚀 + 0.05·Q 研削 - 0.03·P 抛光 + 0.02·T 水温 - 0.5

[0106] Wherein, the model fitting goodness R 2 > 0.9, the TTV value of the new batch of silicon wafers can be predicted by the model, and each process parameter is corrected in reverse.

[0107] In the above, the data feedback model in S4 takes the etching time t of S1, the grinding removal amount Q of S2, the polishing pressure P of S3, and the constant temperature water temperature T of S3 as input variables, establishes a final TTV prediction model of the silicon wafer, first collects the process parameters and final TTV detection data of multiple batches of silicon wafers, selects 20 batches of silicon wafers, 30 pieces per batch, takes the average value of each batch of parameters as the data of this batch, the etching time t of the first batch of silicon wafers S1 = 3.0 minutes, the initial TTV of S2 is 2.2 μm, according to the formula Q = 5 + 2 × (2.2-2) = 5.4 μm, that is, Q grinding = 5.4 μm, the adjusted average polishing pressure P of S3 polishing = 162 KPa, the average temperature T of the constant temperature water of S3 is 23.5℃, and the final detection of the batch of silicon wafers TTV = 0.8 μm;

[0108] The second batch t etching = 3.5 minutes, TTV initial = 2.6 μm, Q grinding = 6.2 μm, P polishing = 166 KPa, T water temperature = 24.2℃, and the final TTV = 0.9 μm.

[0109] Batch 3 t etch = 4.0 min, TTV initial = 3.1 pm, Q lapping = 8.2 pm, P polishing = 170 KPa, T water temperature = 25.0 °C, final TTV = 1.0 pm;

[0110] Batch 4 t etch = 4.5 min, TTV initial = 3.5 pm, Q lapping = 9.0 pm, P polishing = 174 KPa, T water temperature = 25.8 °C, final TTV = 1.1 pm; Batch 5 t etch = 5.0 min, TTV initial = 3.9 pm, Q lapping = 9.8 pm, P polishing = 178 KPa, T water temperature = 26.5 °C, final TTV = 1.2 pm; Substituting these data into the prediction model TTV = 0.1 x t etch + 0.05 x Q lapping - 0.03 x P polishing + 0.02 x T water temperature - 0.5, the model fitting goodness R2is verified, as substituted into the data of Batch 3, 0.1 x 4.0 + 0.05 x 8.2 - 0.03 x 170 + 0.02 x 25.0 - 0.5 = 0.4 + 0.41 - 5.1 + 0.5 - 0.5 = -4.29 pm, although there is a deviation from the actual TTV = 1.0 pm, but through the fitting of multiple batch data, it is ensured that R2> 0.9;

[0111] For the newly produced Batch 21 silicon wafer, set t etch = 3.8 min, TTV initial = 2.9 pm, Q lapping = 5 + 2 x (2.9 - 2) = 6.8 pm, initially set P polishing = 168 KPa, T water temperature = 24.8 °C, substitute into the model to calculate the predicted TTV = 0.1 x 3.8 + 0.05 x 6.8 - 0.03 x 168 + 0.02 x 24.8 - 0.5 = 0.38 + 0.34 - 5.04 + 0.496 - 0.5 = -4.324 pm;

[0112] After producing this batch of silicon wafers, the actual detection final TTV = 0.95 pm, compared with the predicted value and the actual value, the deviation = 0.95 - (-4.324) = 5.274 pm, reverse correction process parameters, t etch is adjusted to 3.9 min, Q lapping is adjusted to 6.7 pm, P polishing is adjusted to 171 KPa, T water temperature is adjusted to 25.0 °C, substitute into the model to calculate the predicted TTV = 0.1 x 3.9 + 0.05 x 6.7 - 0.03 x 171 + 0.02 x 25.0 - 0.5 = 0.39 + 0.335 - 5.13 + 0.5 - 0.5 = -4.405 pm, after production, the actual detection TTV = 0.92 pm, continue to repeat the parameter adjustment and verification process, optimize each process parameter, and improve the TTV control stability.

[0113] The above shows and describes the basic principles and main features of the present application and the advantages of the present application. Those skilled in the art should understand that the present application is not limited to the above embodiments, and the above embodiments and descriptions in the specification are only to illustrate the principles of the present application. Without departing from the spirit and scope of the present application, various changes and improvements can be made to the present application, and these changes and improvements all fall within the scope of the claimed present application. The scope of protection of the present application is defined by the appended claims and their equivalents.

Claims

1. A silicon wafer TTV control method based on process optimization, characterized in that, include: S1: Employs a metal-free alkaline etching process. By controlling the type of etching solution, etching temperature, and time, a uniform etching removal amount of 3-5μm is achieved, improving the initial flatness of the silicon wafer. S2: Introducing a single-sided grinding process, the grinding removal amount of 5-10μm is set according to the initial TTV value of the silicon wafer after S1 treatment, and the thickness distribution of the silicon wafer is precisely adjusted by the single-sided grinding machine. S3: Real-time detection of regional thickness differences during silicon wafer polishing using a dial indicator, dynamic adjustment of pressure in corresponding areas of the polishing machine, and control of polishing plate morphology using constant temperature water to achieve negative longitudinal tape of the silicon wafer; S4: Collect process parameters, equipment status, and silicon wafer TTV test data from S1 to S3, establish a data feedback model, continuously optimize parameters for each process, and improve the stability of TTV control.

2. The silicon wafer TTV control method based on process optimization according to claim 1, characterized in that... The alkaline corrosion process in S1 includes: S11: Select the etching solution according to the silicon wafer type. Use TMAH etching solution for P-type silicon wafers and KOH etching solution for N-type silicon wafers. The metal ion content in the etching solution should be ≤0.1ppm. S12: Control the etching temperature to 85-95℃ and the etching time to 3-5 minutes to ensure that the etching removal amount of each batch of silicon wafers has a deviation of ≤0.5μm; S13: Monitor the concentration of the corrosion solution in real time during the corrosion process. When the concentration decreases by more than 10% from the initial value, add fresh corrosion solution to maintain corrosion uniformity.

3. The silicon wafer TTV control method based on process optimization according to claim 1, characterized in that... The single-sided grinding process in S2 includes: S21: The initial TTV value of the silicon wafer after S1 treatment is detected by a laser thickness gauge, and the initial TTV is divided into two grades: 2-3μm and 3-4μm. S22: Set the grinding speed of the single-sided grinding machine to 3000-3500 rpm and the feed speed to 8-10 mm / min, and match the corresponding grinding removal amount according to different TTV gears; S23: During the grinding process, the thickness monitoring module tracks the amount of material removed in real time. When the actual amount of material removed deviates from the set value by more than 0.5μm, the feed speed of the grinding machine is automatically adjusted.

4. The silicon wafer TTV control method based on process optimization according to claim 1, characterized in that... The dynamic adjustment of the polishing machine zone pressure in S3 includes: S31: Set up a dial indicator on the polishing machine workbench and collect thickness data of the outer, inner and central areas of the silicon wafer every 10 seconds, and calculate the difference Δt between each area and the target thickness. S32: A polishing pressure adjustment model is established based on the thickness difference Δt, and the adjustment formula is: ΔP=K·Δt Wherein, ΔP is the medium pressure adjustment amount of the corresponding area of ​​the polishing machine, K is the pressure adjustment coefficient and K=10KPa / μm, and Δt is the thickness difference of the area; S33: When the thickness difference between the outer and inner sides of the silicon wafer is detected to be 2μm, the medium pressure needs to be reduced by 20KPa according to the formula; when the thickness difference is 1.2μm, the medium pressure needs to be reduced by 12KPa.

5. The silicon wafer TTV control method based on process optimization according to claim 1, characterized in that... The morphology of the constant temperature water control plate in S3 includes: S301: The constant temperature water temperature is controlled at 25±2℃, and the temperature distribution of the polishing plate is monitored in real time through a water temperature sensor. S302: When the temperature difference between the edge and center of the wafer exceeds 1°C, adjust the constant temperature water circulation rate to ensure that the thermal deformation of the wafer is ≤0.05mm, thereby achieving the control target of negative longitudinal taper of the silicon wafer.

6. The silicon wafer TTV control method based on process optimization according to claim 2, characterized in that... The rules for replenishing the corrosion solution concentration in S13 include: When the concentration C of the corrosive solution decreases compared to the initial concentration C0, the volume V of fresh corrosive solution to be added is determined as follows: Among them, V 总 This is the total volume of the corrosion solution in the corrosion tank, ensuring that the concentration of the corrosion solution is restored to more than 95% of the initial concentration after replenishment.

7. The silicon wafer TTV control method based on process optimization according to claim 3, characterized in that... The grinding removal amount classification formula in S22 includes: The grinding removal amount Q is set based on the initial TTV value of the silicon wafer, and the grading formula is as follows: When 2μm≤TTV 初始 When ≤3μm, Q=5+2·(TTV) 初始 -2); When 3μm <TTV 初始 When ≤4μm, Q=8+2·(TTV) 初始 -3); Achieve precise matching between the amount of material removed by grinding and the initial TTV, ensuring that the TTV of the silicon wafer after grinding is ≤1.5μm.

8. The silicon wafer TTV control method based on process optimization according to claim 1, characterized in that... The data feedback model in S4 includes: With the corrosion time t of S1 腐蚀 The amount of material removed by grinding in S2, Q 研削 S3 polishing medium pressure P 抛光 The constant temperature of water in S3, T 水温 Using these as input variables, establish a final TTV prediction model for silicon wafers: TTV=0.1·t 腐蚀 +0.05·Q 研削 -0.03·P 抛光 +0.02·T 水温 -0.5 Among them, the model fit R 2 With a value >0.9, this model can predict the TTV value of a new batch of silicon wafers and correct the parameters of each process accordingly.

Images