Cutting and polishing combined machining treatment method suitable for polycrystalline silicon

Abstract

The invention discloses a cutting and polishing combined machining treatment method suitable for polycrystalline silicon. The method comprises the following steps: step 1, silicon ingot pretreatment: loading historical process data of a silicon ingot in combination with a block chain technology, generating a machining parameter packet, and issuing the machining parameter packet to cutting and polishing equipment to form a machining quantization instruction set; a plasma channel is formed through combination of femtosecond laser focusing and high-frequency electromagnetic field constraint for cutting, annular laser is synchronously used for deep micro-area fusion polishing, and the surface roughness of the cut silicon wafer is reduced by means of argon protection and the surface tension self-leveling effect; 3, polishing: immersing the anode of the silicon wafer into TEMPO / [EMIM] [BF] electrolyte, and regulating and controlling by using pulse voltage; 4, cleaning and drying, wherein megasonic assisted nanobubble ultrapure water is adopted for flushing, and Malangori drying is carried out; and 5, quality control: verifying the surface appearance and the subsurface defect density, and carrying out secondary processing flow on the abnormal silicon wafer.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of polycrystalline silicon processing, in particular to a cutting and polishing combined processing method suitable for polycrystalline silicon. BACKGROUND

[0002] The core purpose of the cutting and polishing combined processing method suitable for polycrystalline silicon is to realize the efficient, low-cost and high-quality conversion of polycrystalline silicon material from bulk raw material to high-precision silicon wafer by integrating the technical advantages of the two key processes of cutting and polishing, which not only meets the stringent requirements of integrated circuit manufacturing on the flatness, surface roughness and lattice integrity of silicon wafer, but also adapts to the large-scale production needs of large-size, ultra-thin and low-loss silicon wafer for photovoltaic cell, and significantly improves the production efficiency and reduces the energy consumption by reducing the processing steps, improving the material utilization rate and reducing the thickness of the surface damage layer, which has far-reaching significance for promoting the technology upgrading of semiconductor industry chain, reducing the cost of photovoltaic power generation and promoting the popularization of renewable energy.

[0003] In the prior art, diamond wire saw cutting leaves periodic undulations on the surface of the silicon wafer, which increases the material removal amount of subsequent polishing / etching, causes stress concentration in the local silicon wafer, and increases the fragment rate, wherein the line mark area may become a defect source for battery texturing, reducing the minority carrier lifetime. Therefore, a cutting and polishing combined processing method suitable for polycrystalline silicon is proposed. SUMMARY

[0004] The purpose of the present application is to solve the problems existing in the prior art, and a cutting and polishing combined processing method suitable for polycrystalline silicon is proposed.

[0005] In order to achieve the above-mentioned purpose, the technical scheme adopted by the present application is as follows: A cutting and polishing combined processing method suitable for polycrystalline silicon, comprising the following steps: Step 1: Silicon ingot pretreatment: laser scanning is used to construct a three-dimensional crystal orientation map and measure the resistivity / minority carrier lifetime distribution, generate silicon ingot digital information, load silicon ingot historical process data combined with blockchain technology, fuse material characteristics and process database, generate processing parameter package, and the processing parameter package is transmitted to the cutting and polishing equipment in real time through industrial Ethernet to form a set of quantitative processing instructions; Step 2: Cutting: cutting is performed by focusing femtosecond laser combined with high-frequency electromagnetic field to form a plasma channel, and synchronous deep micro-area melting polishing is performed by using annular laser, and argon protection and surface tension self-leveling effect are used to reduce the surface roughness of the silicon wafer after cutting; Step three: polishing: the silicon wafer anode is immersed in TEMPO / [EMIM][BF4] electrolyte, the silicon wafer cathode is spaced 2.95mm-3.05mm, the pulse voltage is 1.2V, the TEMPO catalytic oxidation damages the silicon atoms in the damaged area, the pulse voltage is used for regulation, when the phase angle of the electrochemical impedance increases by 15°, it is determined that the defect layer is completely removed, the polishing depth data is fed back to the digital twin model, and the processing parameters of the subsequent batch are corrected; Step four: cleaning and drying: megasonic wave assisted nanobubble ultrapure water is used for rinsing, an electrochemical deionization module is used for circulating treatment of the cleaning liquid, and a Malan Goni drying machine is used for drying; the drying room temperature and humidity data are synchronized to the MES system; Step five: quality control: the surface topography and subsurface defect density are verified by multi-dimensional detection means, the polishing parameters are corrected using a cutting damage prediction model, abnormal silicon wafers trigger adaptive re-polishing protocols for secondary processing, and a digital quality certificate containing the full parameter hash value and timestamp is generated by combining blockchain storage technology.

[0006] The above further comprises: Further, the generation of the silicon ingot digital information comprises the following steps: Laser scanning to construct a 3D crystal orientation map: a femtosecond laser is used to perform spiral scanning on the surface of the silicon ingot to obtain crystal lattice orientation data and then generate a 3D map, which is represented as wherein θ is the angle between the crystal face normal and the laser incidence direction, and are the parallel and perpendicular polarization backscattering light intensities, respectively; Resistivity / minor carrier lifetime distribution measurement: a four-probe method is used to apply a 1mA current to measure the resistivity distribution, which is represented as wherein t is the thickness of the silicon ingot, d is the probe spacing, s is the sample diameter, and F is the geometric correction factor; a microwave photoconductivity decay instrument is used to fit the minor carrier lifetime decay curve, and the fitting expression is wherein α is the recombination coefficient, is the normalized photoconductivity change; Silicon ingot digital information generation: the 3D map and resistivity / minor carrier lifetime data are fused to generate a multi-dimensional feature vector containing three-dimensional crystal orientation θ(x,y,z), resistivity ρ(x,y,z) and minor carrier lifetime τ(x,y,z); the silicon ingot ID and the multi-dimensional feature vector hash are bound and uploaded to the chain through the SHA-256 algorithm.

[0007] Further, based on a genetic algorithm optimization model, a processing parameter package is generated with the target functions of the cutting seam width, the heat affected zone and the polishing time; the JSON format parameter package is transmitted in real time to the cutting and polishing equipment through the industrial Ethernet by the OPC UA protocol, forming a set of quantitative processing instructions.

[0008] Further, in the cutting, a femtosecond laser with a wavelength of 800 nm and a pulse width of 300 fs is adopted, focused by an objective with a numerical aperture of 0.65 to form a plasma channel with a diameter of 40 μm, wherein the laser power density meets the plasma ionization threshold, represented as wherein c is the speed of light, is the vacuum permittivity, m e is the electron mass, e is the elementary charge, is the ionization probability, an annular electromagnetic field with a frequency f of 5 MHz and a peak current of 500 A is applied to the cutting area to constrain the plasma expansion by Lorentz force, represented as wherein q is the electron charge, is the vacuum permeability, n is the plasma density (10 0 m⁻³), r is the channel radius, is the plasma expansion velocity.

[0009] Further, in the cutting, a ring-shaped spot laser with a wavelength of 1064 nm and a power P of 20 W is used for micro-area remelting at a scanning speed v of 10 m / s, represented as wherein k is the thermal conductivity of silicon, a is the thermal diffusivity, t is the laser action time, and z is the melt depth, a positive pressure gas curtain is formed by argon with a flow rate Q of 15 L / min, the oxygen partial pressure is controlled within 1 ppm, and real-time monitoring is combined with plasma photoluminescence spectrum, represented as wherein K is the reaction rate constant, is the activation energy, k is the Boltzmann constant, the generation of the oxidation layer is inhibited, the actual cutting seam width W and the polishing depth D are measured by a laser interferometer based on a digital twin model, and the parameter package is feedback corrected, represented as wherein , is the PID control parameter.

[0010] Further, the polishing adopts self-terminating electrochemical precision polishing, including the following steps: Electrolyte circulation and electric field construction: the silicon wafer anode is immersed in 1-ethyl-3-methyl imidazole tetrafluoroborate ionic liquid electrolyte containing 0.1 M TEMPO catalyst, circulated at a flow rate of 2 L / min to maintain the convective diffusion coefficient wherein Q is the flow rate, A is the electrode area, and the anode-cathode distance is controlled at 2.95-3.05 mm; Pulse voltage regulation selects corrosion: a pulse voltage of 1.2 V is applied, with an average current density of 20 mA / cm², and TEMPO catalytic oxidation preferentially attacks the grain boundary defect area; Dynamic termination decision and feedback correction: real-time electrochemical impedance spectroscopy is collected, when the defect layer is completely removed, the surface roughness Ra decreases, leading to a sudden increase of 15° in phase angle θ, the polishing depth is fed back to the digital twin model through Faraday's law, and the model automatically corrects the amplitude of the next batch of pulse voltage based on the PID control algorithm.

[0011] Further, the specific steps of the cleaning and drying are: Megasonic wave assisted nanobubble flushing: megasonic waves are used to excite ultrapure water to produce cavitation effect, combined with nanobubble generator to inject bubble groups, through the impact force of cavitation bubble collapse to strip the particles on the surface of the silicon wafer, and the micro-jet generated by the rupture of nanobubbles can remove pollutants deep in the subsurface micropores; EDI cycle deionization treatment: the cleaning liquid is circulated through the electrodeionization module, under the driving of direct current electric field, the ion migration rate through the anion / cation exchange membrane is , where μ is the ion mobility, is the direct current electric field, combined with resin bed adsorption and bipolar membrane water electrolysis regeneration, the concentration of metal ions is reduced, and the mass balance equation describes the decay dynamics of ions, so that the purity of the cleaning liquid meets the ISO 14644-1 Class 1 standard; Marangoni effect drying: by controlling the isopropanol vapor concentration through gradient, the liquid film is retracted by using surface tension gradient driving, and the drying time is optimized by the diffusion equation , where L is the characteristic length, D is the diffusion coefficient, is the saturation concentration, and the inclination angle of 15° maximizes the droplet detachment speed; Temperature and humidity data closed-loop feedback: the temperature and humidity sensors in the drying chamber collect data in real time, the PID controller adjusts the power of the heater to maintain the temperature at 25±0.5℃ and the humidity at 30±2%RH, the collected data is synchronized to the MES system through OPC UA protocol, and the humidity drift trend is predicted by combining machine learning model to provide early warning for equipment maintenance requirements.

[0012] Further, the specific steps of the quality control are: Multi-dimensional detection and defect quantification: white light interferometer is used to obtain surface topography in vertical scanning interference mode, and surface roughness parameters are extracted by power spectral density analysis , is the power spectral density, which is used to describe the distribution of signal power in frequency domain, is the surface roughness parameter, which represents the arithmetic mean deviation of the surface topography from the average plane, A is the scanning area, and f is the spatial frequency, is the average height of the surface topography, μ-PL spectral scanning system is used synchronously to quantify the subsurface damage density through defect state luminescence peak intensity, and the defect state luminescence peak intensity is represented as , when a crack exists is determined; Cutting damage prediction model driving parameter correction: based on a finite element-machine learning hybrid model, inputting the LIPC cutting parameters, calculating the theoretical damage depth through the damage depth prediction formula, the damage depth prediction formula is expressed as wherein, is the damage depth, P is the laser power, B is the electromagnetic field strength, is the activation energy, is the material correlation coefficient α, when the measured value deviates from the predicted value by more than 10%, the STEP polishing parameter correction is triggered, and the correction formula is expressed as wherein, is the polishing time, is the correction coefficient, is the polishing rate; Adaptive re-polishing protocol execution: when the μ-PL detects that the defect density is greater than 10 / cm2, the re-polishing process is started: the pulse voltage amplitude is dynamically adjusted; the polishing current efficiency is monitored in real time through electrochemical impedance spectroscopy, and the re-polishing is terminated when the polishing current efficiency is greater than 95%; Blockchain storage and digital certificate generation: the whole process parameters are generated into a hash value through the SHA-3-512 algorithm, and are packaged together with the UTC timestamp as a transaction, uploaded to the Hyperledger Fabric consortium chain, and a verifiable digital quality certificate is generated, including the Merkle proof path, and the user verifies the authenticity of the data through the certificate hash in the blockchain browser.

[0013] The present application has the following beneficial effects: 8、In the present application, the laser-induced plasma cutting-melt polishing cutting technology forms a plasma channel through femtosecond laser focusing combined with high-frequency electromagnetic field constraint for cutting, avoiding the periodic surface line marks generated by diamond wire saw cutting. At the same time, the high-frequency electromagnetic field constraint effectively reduces the generation of subsurface micro-cracks, improves the surface quality of the silicon wafer, optimizes the cutting and polishing parameters, and adopts self-terminating electrochemical precision polishing technology, reduces the excessive consumption of materials and polishing time, thereby reducing the cutting and polishing cost. In addition, the high cost of diamond wire consumption is avoided.

[0014] 9、In the present application, the self-terminating electrochemical precision polishing precisely controls the polishing process through an electrochemical method, realizes more uniform surface polishing, reduces local corrosion pits and crystal direction sensitivity problems. At the same time, the risk of metal contamination that may be introduced in the wet etching is avoided, the activation degree of the silicon wafer surface is reduced through precise control of the polishing process and subsequent cleaning steps, thereby reducing the risk of generating a native oxide layer. BRIEF DESCRIPTION OF DRAWINGS

[0015] Fig. 1 A step diagram of a cutting and polishing combined processing method suitable for polycrystalline silicon is proposed for the present application; Fig. 2 A system block diagram used in the cutting and polishing combined processing method suitable for polycrystalline silicon is proposed for the present application. DETAILED DESCRIPTION

[0016] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative labor fall within the scope of protection of the present application.

[0017] Please refer to Figs. 1-2 The present application is a cutting and polishing combined processing method suitable for polycrystalline silicon, which comprises the following steps: Step one: silicon ingot pretreatment: laser scanning is used to construct a three-dimensional crystal orientation map and measure the resistivity and minority carrier lifetime distribution, generate silicon ingot digital information (digital identity card), combine blockchain technology to load silicon ingot historical process data, fuse material characteristics and process database, generate processing parameter package containing cutting path, laser power and electrolytic parameters, and the processing parameter package is transmitted to the cutting and polishing equipment in real time through industrial Ethernet to form a set of quantitative processing instructions; Step two: cutting (laser-induced plasma cutting-melt polishing (LIPC-SMP)): cutting is performed by focusing femtosecond laser combined with 5MHz high-frequency electromagnetic field to form a 40μm diameter plasma channel, 80μm kerf width and ≤1.5μm heat affected zone are achieved at a cutting speed of 8m / s, 4μm deep micro-area melt polishing is simultaneously performed using 1064nm ring laser, argon gas protection and surface tension self-leveling effect are used to reduce the surface roughness of the cut silicon wafer; Step three: polishing (self-terminating electrochemical precision polishing (STEP)): the silicon wafer anode is immersed in TEMPO / [EMIM][BF4] electrolyte (circulating flow rate 2L / min), the silicon wafer cathode spacing is 2.95mm-3.05mm, the pulse voltage is 1.2V (duty ratio 50%, frequency 5kHz), TEMPO catalytic oxidation of damaged area silicon atoms (reaction priority: grain boundary>dislocation>complete crystal), pulse voltage is used for regulation (control grain boundary corrosion rate is 120 times that of complete area), when the electrochemical impedance phase angle increases by 15°, it is determined that the defect layer is completely removed, the polishing depth data is fed back to the digital twin model, and the processing parameters of the subsequent batch are corrected; Step four: cleaning and drying: megasonic (frequency 950 kHz, power density 0.5 W / cm²) assisted nanobubble ultrapure water rinsing, electrochemical deionization (EDI) module recycling treatment of cleaning liquid, metal ion concentration controlled at ≤1 pp, Malan Goni drying (IPA vapor gradient control), drying room temperature and humidity data synchronized to the MES system; Step five: quality control: surface topography and subsurface defect density verified by white light interferometer, μ-PL scanning and other multi-dimensional detection means, cutting damage prediction model used to correct polishing parameters, abnormal silicon wafer triggers adaptive re-polishing protocol for secondary processing, combined with blockchain storage technology to generate digital quality certificate containing full parameter hash value and timestamp.

[0018] In one embodiment, the generating silicon ingot digital information comprises the following steps: Laser scanning to construct 3D crystal orientation map: through femtosecond laser (wavelength λ = 515 nm, pulse energy E = 50 μJ, repetition frequency f = 1 MHz) spiral scanning (scanning speed v = 2 m / s, spot diameter d = 10 μm) on the surface of silicon ingot, lattice orientation data is obtained to generate 3D map, expressed as Wherein, θ is the angle between the crystal plane normal and the laser incident direction, And are the parallel and perpendicular polarized backscattering light intensity, function: positioning grain boundaries (θ mutation area) and defect dense area (backscattering signal fluctuation > 15%), generating 3D map with spatial resolution of 2 μm, assuming that the scanning of a certain area of the silicon ingot shows that θ changes from 5° to 85°, it is determined as {111} crystal plane boundary, marked as defect high risk area; Resistivity / minority carrier lifetime distribution measurement: combined with four-probe method to apply 1 mA current to measure resistivity distribution, expressed as Wherein, t is the thickness of the silicon ingot, d is the probe spacing, s is the sample diameter, F is the geometric correction factor (1.02), and the microwave photoconductance decay instrument is used to fit the minority carrier lifetime decay curve, the fitting expression is Wherein, α is the recombination coefficient, Is the normalized photoconductance change, function: resistivity abnormal area (ρ > 1.5 Ω·cm) and minority carrier lifetime low value area (τ < 5 μs) are marked as potential performance bottleneck, assuming that the ρ of a certain area is 2.1 Ω·cm (higher than the average value 1.2 Ω·cm), τ = 3.2 μs (lower than the average value 6.5 μs), it is determined as unevenly doped area; Silicon ingot digital information generation: fuse 3D map, resistivity / minority carrier lifetime data to generate multi-dimensional feature vector containing three-dimensional crystal orientation θ (x, y, z), resistivity ρ (x, y, z) and minority carrier lifetime τ (x, y, z), hash binding silicon ingot ID and multi-dimensional feature vector through SHA-256 algorithm Chain, ensure that historical process data cannot be tampered with.

[0019] In one embodiment, based on genetic algorithm optimization model, the processing parameter package is generated with the target function of the seam width, heat affected zone and polishing time, including the avoidance of the grain boundary Dynamic adjustment of laser power , and combined with the minority carrier lifetime reference value Modified electrolytic voltage , through OPC UA protocol, the JSON format parameter package is real-time issued to the cutting and polishing equipment through industrial Ethernet, forming a processing quantitative instruction set containing cutting speed 8 m / s, laser power 57.5 μJ and electrolytic voltage 1.30 V, and realizing closed loop verification of MAE <0.5 μm through digital twin model residual error analysis.

[0020] In one embodiment, in the cutting, femtosecond laser with wavelength λ of 800 nm and pulse width τ of 300 fs is used, focused through an objective with numerical aperture NA of 0.65, forming a plasma channel with diameter d of 40 μm, wherein the laser power density meets the plasma ionization threshold, expressed as , wherein c is the speed of light, is the vacuum permittivity, m e is the electron mass, e is the elementary charge, is the ionization probability (0.01), when the laser power P > πd² / 4, multi-photon ionization is triggered to form plasma, and the plasma channel restricts the cutting path, ensuring that the seam width is stable at 80 ± 2 μm at a high speed of 8 m / s, and the heat affected zone (HAZ) ≤1.5 μm (verified by infrared thermal imager), when P = 50 W, ≈1.2×10¹³W / cm², the actual power density I = 50 / (π×20²) = 3.98×10¹³W / cm² , a plasma channel is successfully constructed, an annular electromagnetic field with frequency f of 5 MHz and peak current 500A is applied to the cutting area, the plasma expansion is restricted by the Lorentz force, expressed as , wherein q is the electron charge, is the vacuum permeability, n is the plasma density (10² 0 m⁻³), r is the channel radius, To limit the plasma lateral diffusion within the 40-µm channel, the heat-affected zone is reduced by 60% compared to the conventional method, and the plasma expansion force is calculated to be ≈1.2×10⁻³ N, which effectively counteracts the plasma expansion force and ensures the kerf straightness <1 µm / 100 mm.

[0021] In one embodiment, in the cutting, a 1064-nm-wavelength, 20-W-power annular spot laser (40-µm inner diameter / 80-µm outer diameter) is used for micro-area remelting at a scanning speed v of 10 m / s, which is expressed as where k is the silicon thermal conductivity (150 W / m·K), α is the thermal diffusivity (0.8 cm² / s), t is the laser action time, and z is the melt depth. A 4-µm-deep melt pool is generated, and the Ra is reduced from 0.8 µm after cutting to 0.15 µm through the surface tension self-leveling effect (γ∇²T, γ is the surface tension coefficient). When P=20 W and t=10 µs, T_melt is calculated to be ≈1687 K (the silicon melting point is 1687 K), realizing precise melting polishing. A 99.999% argon gas with a flow rate Q of 15 L / min forms a positive pressure gas curtain, and the oxygen partial pressure is controlled to be within 1 ppm. In combination with real-time monitoring of the plasma photoluminescence spectrum (OES), which is expressed as where K is the reaction rate constant, is the activation energy (2.5 eV), and k is the Boltzmann constant. The generation of the oxide layer is inhibited, and the polished surface SiO2 thickness is ensured to be <0.5 nm (XPS detection). When the oxygen partial pressure is 0.5 ppm, the calculation is ≈0.3×10³ counts / s, which is lower than the oxidation threshold of 0.5×10³ counts / s, realizing non-oxidation polishing. Based on the digital twin model, the actual kerf width W and polishing depth D are measured by a laser interferometer (λ=633 nm), and the parameter package is feedback corrected, which is expressed as where =0.2 W / µm, =0.05 W·s / µm is the PID control parameter.

[0022] In one embodiment, the polishing employs a self-terminating electrochemical precision polishing, which includes the following steps: Electrolyte circulation and electric field construction: the silicon wafer anode is immersed in a 1-ethyl-3-methylimidazolium tetrafluoroborate ([EMIM][BF4]) ionic liquid electrolyte containing 0.1 M TEMPO catalyst, which is circulated at a flow rate of 2 L / min to maintain the convective diffusion coefficient ​Where Q is the flow rate, A is the electrode area, ensuring uniform supply of reactants and rapid removal of products. The distance between the anode and cathode is controlled at 2.95-3.05 mm, and the standard deviation of the current density distribution is calculated to be less than 2% by Ohm's law, ensuring uniformity of the electric field;

[0023] Pulse voltage regulation selective corrosion: apply 1.2V pulse voltage (duty cycle 50%, frequency 5kHz), average current density is 20mA / cm², TEMPO catalytic oxidation preferentially attacks the grain boundary defect area, the reaction rate is 120 times higher than that of the complete lattice area, and the mechanism is explained by the difference in activation energy: The difference in activation energy is ΔEa, and T is the local temperature. The difference in activation energy at the grain boundary is 0.3eV which leads to an exponential increase in reaction rate; Dynamic termination criterion and feedback correction: real-time acquisition of electrochemical impedance spectroscopy (EIS), when the defect layer is completely removed, the surface roughness Ra decreases from 15nm to 0.3nm, resulting in a sudden increase of phase angle θ by 15° (calculated by Z'' is the capacitive reactance, and Z' is the resistance), the polishing depth (M is the molar mass of silicon, n is the number of electron transfers, F is the Faraday constant, and ρ is the density of silicon) is fed back to the digital twin model based on the PID control algorithm (e(t) is the depth error) to automatically correct the amplitude of the next batch of pulse voltage, achieving a polishing depth accuracy of ±0.1μm.

[0024] In this embodiment, when θ increases from -60° to -45°, the surface Ra is calculated to be 0.3nm, and the model determines that the polishing is complete. If the actual depth h=5.2μm exceeds the target 5.0μm, the PID controller calculates ΔV=-0.1V, and the next batch of voltage is adjusted to 1.1V to suppress over-corrosion.

[0025] In one embodiment, the specific steps of the cleaning and drying are: Megasonic wave assisted nanobubble flushing: megasonic waves with a frequency of 950kHz and a power density of 0.5W / cm² are used to excite ultrapure water (resistivity 18.2MΩ·cm) to produce cavitation effect, combined with nanobubble generator to inject bubble groups with a diameter of 50-200nm, through the collapse impact force of cavitation bubbles , is the gas pressure, is the liquid density, is the maximum radius of the bubble, and ω is the angular frequency) to strip the particles on the surface of the silicon wafer, and the micro-jet generated by the rupture of the nanobubbles (micro-jet γ is the surface tension, and r is the bubble radius) penetrates the subsurface micropores to remove contaminants; ​EDI cycle deionization: cleaning fluid is circulated through an electrodeionization module (EDI), under the driving of a direct current electric field, ions migrate through the cation / anion exchange membrane at a rate of where μ is the ion mobility, is the direct current electric field, combined with resin bed adsorption and bipolar membrane water electrolysis regeneration, the metal ion concentration is reduced from the initial 50 ppb to ≤1 ppb, the mass balance equation describes the ion decay dynamics, so that the purity of the cleaning fluid meets the ISO 14644-1 Class 1 standard; Marangoni effect drying: by controlling the gradient of isopropyl alcohol (IPA) vapor concentration (linearly rising from 0% to 100% IPA saturated vapor pressure), the liquid film is retracted by using the surface tension gradient as the driving force, and the drying time is described by the diffusion equation Optimization, where L is the characteristic length, D is the diffusion coefficient, is the saturation concentration, and the inclination angle is 15° to maximize the droplet detachment speed, achieving water mark-free drying and a contact angle <5°; Temperature and humidity data closed-loop feedback: the drying chamber temperature and humidity sensor (accuracy ±0.1℃ / ±1%RH) collects data in real time, which is adjusted by the PID controller to regulate the heater power, maintaining the temperature at 25±0.5℃ and the humidity at 30±2%RH. The collected data is synchronized to the MES (Manufacturing Execution) system through the OPC UA protocol, and combined with a machine learning model to predict humidity drift trends and provide early warning of equipment maintenance needs.

[0026] In this embodiment, when the IPA vapor concentration gradient Δc=50% is detected, the surface tension difference Δσ=15mN / m is calculated, driving the liquid film to retract at v=0.8mm / s; if the MES system monitors a sudden increase in humidity to 35%RH, the PID controller will increase the heating power by 15% within 10 seconds to restore the set value, ensuring drying stability.

[0027] In one embodiment, the specific steps of the quality control are: Multi-dimensional detection and defect quantification: a white light interferometer (Wyko NT9300) is used to obtain the surface topography in vertical scanning interferometry (VSI) mode, and the surface roughness parameters are extracted by power spectral density (PSD) analysis , is the power spectral density (Power Spectral Density), which describes the distribution of signal power in the frequency domain, is the surface roughness parameter, specifically the arithmetic average roughness, which represents the arithmetic average deviation of the surface topography from the average plane, A is the scanning area, and f is the spatial frequency, For the average height of the surface topography, the μ-PL spectral scanning system (excitation wavelength 325 nm) is used synchronously to quantify the subsurface damage density by the defect state luminescence peak intensity, which is expressed as When , it is determined that there is a crack; Cutting damage prediction model driving parameter correction: based on the finite element-machine learning hybrid model, input LIPC cutting parameters (laser power P, electromagnetic field strength B), calculate the theoretical damage depth through the damage depth prediction formula, which is expressed as Wherein, is the damage depth, the subsurface damage depth generated during the cutting process, which is an important indicator for evaluating the cutting quality, P is the laser power, the output power of the laser during the laser cutting process, which directly affects the cutting ability and speed, B is the electromagnetic field strength, the electromagnetic field strength applied during the cutting process, which is used to constrain the plasma channel and affects the cutting precision and heat affected zone, is the activation energy, the activation energy required for material damage during cutting, which is related to material properties and cutting conditions, is the material-related coefficient α, a coefficient related to material properties, used in the cutting damage prediction model, reflecting the response characteristics of the material to laser and electromagnetic field, when the measured deviates more than 10% from the predicted value, trigger STEP polishing parameter correction, the correction formula is expressed as Wherein, is the polishing time, the polishing time that needs to be increased calculated according to the deviation between the damage depth prediction value and the measured value, is the correction coefficient, a coefficient used to correct the polishing time, determined according to the deviation between the damage depth prediction value and the measured value, is the polishing rate, the rate of material removal during polishing, which affects the polishing time and surface quality; If the predicted damage is 1.2 μm but the measured value is 1.5 μm, the polishing time is extended Δt=0.5×(0.3 / 0.2)=0.75 minutes.

[0028] Adaptive re-polishing protocol execution: when the μ-PL detects a defect density of 1.2×108cm-2, start the re-polishing process: dynamically adjust the pulse voltage amplitude; real-time monitor the polishing current efficiency by electrochemical impedance spectroscopy (EIS), and terminate the re-polishing when the polishing current efficiency is greater than 95%; Blockchain storage and digital certificate generation: the whole process parameters (laser power, polishing voltage, detection data, etc.) are generated by SHA-3-512 algorithm to generate hash value, and UTC timestamp is encapsulated as transaction together, uploaded to Hyperledger Fabric consortium chain, and verifiable digital quality certificate (DQC) is generated, including Merkle proof path, and users verify the authenticity of data through certificate hash in blockchain browser.

[0029] While embodiments of the present application have been shown and described with reference to particular embodiments thereof, it will be understood by those skilled in the art that various changes in form and details can be made therein without departing from the spirit and scope of the application. The scope of the application is defined by the appended claims and their equivalents.

Claims

1. A method for combined cutting and polishing processing of polycrystalline silicon, characterized in that, The method comprises the following steps: Step 1: Silicon ingot pretreatment: Construct a three-dimensional crystal orientation map by laser scanning and measure the resistivity and minority carrier lifetime distribution to generate digital information of the silicon ingot, load the historical process data of the silicon ingot by combining with the blockchain technology, fuse the material characteristics and process database to generate a processing parameter package, and the processing parameter package is transmitted to the cutting and polishing equipment in real time through the industrial Ethernet to form a quantitative instruction set for processing; Step 2: Cutting: Cutting is performed by focusing a femtosecond laser in combination with a high-frequency electromagnetic field to form a plasma channel, and synchronous deep micro-area melting and polishing are performed by using a ring laser, and the surface roughness of the silicon wafer after cutting is reduced by means of argon protection and surface tension self-leveling effect; Step 3: Polishing: The silicon wafer anode is immersed in TEMPO / [EMIM][BF4] electrolyte, the silicon wafer cathode spacing is 2.95mm-3.05mm, the pulse voltage is 1.2V, TEMPO catalyzes and oxidizes the silicon atoms in the damaged area, and the pulse voltage is used for regulation and control, when the phase angle of the electrochemical impedance spectrum increases by 15°, it is determined that the defect layer is completely removed, and the polishing depth data is fed back to the digital twin model, and the processing parameters of the subsequent batch are corrected; Step 4: Cleaning and drying: megasonic wave assisted nanobubble ultrapure water flushing is adopted, an electrochemical deionization module is used for circulating treatment of the cleaning liquid, and a maringoni drying is adopted, and the drying room temperature and humidity data are synchronously transmitted to the MES system; Step 5: Quality control: The surface topography and subsurface defect density are verified by multi-dimensional detection means, the polishing parameters are corrected by using a cutting damage prediction model, the abnormal silicon wafer triggers an adaptive re-polishing protocol for secondary processing, and a digital quality certificate containing a full parameter hash value and a time stamp is generated by combining with a blockchain storage technology.

2. The method according to claim 1, wherein the method is used for cutting and polishing of polysilicon. 2 The generation of the silicon ingot digital information comprises the following steps: Laser scanning to construct 3D crystal orientation map: by means of the femtosecond laser to carry out the spiral scanning to the silicon ingot surface, the crystal lattice orientation data is obtained and then the 3D map is generated, which is expressed as Wherein, θ is the included angle between the crystal face normal and the laser incidence direction, and respectively are the parallel and vertical polarization backscattering light intensity; Resistivity / minority carrier lifetime profiling: Resistivity profiles were measured in conjunction with the four-point probe method applying a 1 mA current, expressed as where t is the silicon ingot thickness, d is the probe spacing, s is the sample diameter, and F is a geometric correction factor. The minority carrier lifetime decay curves were fitted using a microwave photoconductance decay instrument, with the fitting expression given by where α is the recombination coefficient, is the normalized change in photoconductance. Silicon ingot digital information generation: fuse the 3D map, resistivity and minority carrier lifetime data to generate a multi-dimensional feature vector containing three-dimensional crystal orientation θ(x,y,z), resistivity ρ(x,y,z) and minority carrier lifetime τ(x,y,z), and hash bind the silicon ingot ID and the multi-dimensional feature vector by using an SHA-256 algorithm and chain.

3. The method for combined cutting and polishing processing of polycrystalline silicon according to claim 1, characterized in that, Based on a genetic algorithm optimization model, a processing parameter package is generated by taking the kerf width, heat affected zone and polishing time as target functions, the JSON format parameter package is transmitted to the cutting and polishing equipment in real time through the industrial Ethernet by using an OPC UA protocol, and a quantitative instruction set for processing is formed.

4. The method for combined cutting and polishing processing of polycrystalline silicon according to claim 1, characterized in that, In the cutting, femtosecond laser with wavelength of 800 nm and pulse width of 300 fs is adopted, focused by an objective with numerical aperture of 0.65 to form a plasma channel with diameter of 40 μm, wherein the laser power density meets the plasma ionization threshold, represented as wherein c is the speed of light, is the vacuum permittivity, m e is the electron mass, e is the elementary charge, is the ionization probability, an annular electromagnetic field with frequency f of 5 MHz and peak current of 500 A is applied to the cutting area to constrain the plasma expansion by Lorentz force, represented as wherein q is the electron charge, is the vacuum permeability, n is the plasma density (10² 0 m⁻³), and r is the channel radius, is the plasma expansion velocity.

5. The method of claim 4, wherein the method is applied to a polycrystalline silicon. In cutting, a laser with a wavelength of 1064 nm and a power P of 20 W is used to perform micro-re-melting at a scanning speed v of 10 m / s, represented as wherein k is the thermal conductivity of silicon, a is the thermal diffusivity, t is the laser action time, and z is the melting depth, a positive pressure gas curtain with a flow rate Q of 15 L / min is formed, the oxygen partial pressure is controlled within 1 ppm, and real-time monitoring is performed by combining plasma photoluminescence spectrum, represented as wherein K is the reaction rate constant, is the activation energy, and k is the Boltzmann constant, the generation of the oxidation layer is inhibited, the actual kerf width W and the polishing depth D are measured by a laser interferometer based on a digital twin model, and the parameter package is feedback corrected, represented as wherein , is the PID control parameter.

6. The method of claim 1, wherein the method is used for cutting and polishing of a polysilicon wafer. The polishing adopts self-terminating electrochemical precision polishing, which comprises the following steps: Electrolyte circulation and electric field construction: The silicon wafer anode was immersed in 1-ethyl-3-methylimidazolium tetrafluoroborate ionic liquid electrolyte containing 0.1 M TEMPO catalyst, and circulated at a flow rate of 2 L / min to maintain the convective diffusion coefficient wherein Q is the flow rate, A is the electrode area, and the anode-cathode spacing is controlled at 2.95-3.05 mm; Pulse voltage regulation and selective corrosion: 1.2V pulse voltage is applied, the average current density is 20mA / cm², and TEMPO catalyzes and oxidizes the grain boundary defect area preferentially; Dynamic termination judgment and feedback correction: real-time acquisition of electrochemical impedance spectrum, when the defect layer is completely removed, the surface roughness Ra is reduced, resulting in a sudden increase of 15° in the phase angle θ, the polishing depth is fed back to the digital twin model by Faraday's law, and the model automatically corrects the pulse voltage amplitude of the next batch based on a PID control algorithm.

7. The method of claim 1, wherein the method is used for cutting and polishing of a polysilicon. The specific steps of the cleaning and drying are: ​ Millions of sound wave assisted nano bubble flushing: megasonic wave is used to excite ultrapure water to produce cavitation effect, combined with nano bubble generator to inject bubble group, through the collapse of cavitation bubble impact force to strip the particles on the surface of silicon wafer, at the same time, the micro jet generated by the rupture of nano bubble can remove the pollutants in the subsurface micropore; EDI cycle deionization treatment: the cleaning solution is circulated through the electric deionization module, and under the driving of the direct current electric field, the ion migration rate through the anion / cation exchange membrane is where μ is the ion migration rate, is the direct current electric field, combined with resin bed adsorption and bipolar membrane water electrolysis regeneration, the metal ion concentration is reduced, and the mass balance equation describes the ion decay dynamics, so that the purity of the cleaning solution meets the ISO 14644-1 Class 1 standard; Marangoni effect drying: drying time is governed by diffusion equation through gradient control of isopropanol vapor concentration, using surface tension gradient to drive liquid film retraction Optimization, where L is characteristic length, D is diffusion coefficient, is the saturation concentration, and the drop-off velocity is maximized with a tilt angle of 15°; Temperature and humidity data closed loop feedback: the temperature and humidity sensor in the drying chamber collects data in real time, the PID controller adjusts the power of the heater to maintain the temperature at 25±0.5℃ and the humidity at 30±2%RH, the collected data is synchronized to the MES system through OPC UA protocol, and the humidity drift trend is predicted by combining with the machine learning model to provide early warning for equipment maintenance requirements.

8. The method for combined cutting and polishing processing of polycrystalline silicon according to claim 1, characterized in that, The specific steps of the quality control are: Multi-dimensional detection and defect quantification: white light interferometer was used to acquire surface topography in vertical scanning interferometry mode, and surface roughness parameters were extracted by power spectral density analysis , is power spectral density, which is used to describe the distribution of signal power in frequency domain, is surface roughness parameter, which represents the arithmetic mean deviation of surface topography from the average plane, A is the scanning area, and f is the spatial frequency, is the average height of surface topography, and a μ-PL spectral scanning system was used synchronously to quantify the subsurface damage density by the intensity of defect state luminescence peak, which is represented as , when , it is determined that there is a crack; Cutting damage prediction model driving parameter correction: based on the finite element-machine learning hybrid model, input LIPC cutting parameters, calculate the theoretical damage depth through the damage depth prediction formula, the damage depth prediction formula is expressed as wherein, is the damage depth, P is the laser power, B is the electromagnetic field intensity, is the activation energy, is the material correlation coefficient α, when the measured and the predicted value deviation is greater than 10%, trigger STEP polishing parameter correction, the correction formula is expressed as wherein, is the polishing time, is the correction coefficient, is the polishing rate; Adaptive re-polish protocol execution: when the μ-PL detects the defect density is higher than 95%. Blockchain storage and digital certificate generation: the full-process parameters are converted into hash values through SHA-3-512 algorithm, and are packaged into transactions together with UTC timestamp, uploaded to Hyperledger Fabric consortium chain, and verified digital quality certificate is generated, including Merkle proof path, and users can verify the authenticity of data through certificate hash in blockchain browser.

Images