Search

CN-121997106-A - Electrolytic process condition establishment method and system based on combination optimization

CN121997106ACN 121997106 ACN121997106 ACN 121997106ACN-121997106-A

Abstract

The invention discloses a method and a system for establishing electrolytic process conditions based on combination optimization, and belongs to the technical field of semiconductor element electrolytic plating. The method comprises the steps of collecting electrolytic plating process condition data, device operation data and product data to construct a training sample set, constructing a neural network prediction model to predict the thickness of plated metal film, extracting predicted process conditions to conduct actual plating verification, obtaining actual measured film thickness and judging whether a preset threshold is met, conducting multiple performance tests and judging on products meeting the threshold, feeding back the actual measured data meeting single judgment standards to update the training set, retraining the model, repeating the steps until process conditions enabling all performance tests to meet excellent standards are screened, and establishing the process conditions as optimized process conditions. According to the invention, model self iteration is realized through a closed loop mechanism of prediction, verification and feedback, and an electrolysis process window capable of stably producing a full-optimal product can be rapidly and accurately established by combining a multi-dimensional performance full-optimal evaluation system.

Inventors

  • Door pine Pearl
  • ZHOU ZHIHAN
  • ZHOU AIHE

Assignees

  • 昆山一鼎工业科技有限公司

Dates

Publication Date
20260508
Application Date
20260410

Claims (9)

  1. 1. An electrolytic process condition establishment method based on combination optimization is characterized by comprising the following steps: S1, collecting electrolytic plating process condition data, electrolytic plating device morphology and operation data, semiconductor element morphology and plating product data, and constructing a training sample set; s2, constructing a neural network prediction model, and predicting the thickness of the electrolytic plating metal film by taking key factors influencing the electrolytic plating process conditions as input variables; S3, extracting at least two groups of predicted electrolytic plating process conditions from the neural network prediction model, performing actual plating verification through an electrolytic plating device, obtaining the actually measured plating metal film thickness, and judging whether the actually measured plating metal film thickness meets a preset threshold range or not; S4, performing multiple performance tests on the actually measured semiconductor element plating products meeting the threshold range, and judging according to preset single performance judgment standards; S5, the electrolytic plating process condition data and the relevant data thereof corresponding to the measured data which are judged to meet the preset single performance judgment standard in the step S4 are used for updating the training sample set and retraining the neural network prediction model to obtain an upgraded and optimized neural network model; S6, repeatedly executing S3 to S5 until screening to enable the judging results of the performance tests in S4 to meet the preset excellent standard electrolytic plating process conditions, and establishing the optimal electrolytic plating process conditions.
  2. 2. The method according to claim 1, wherein the key factors in step S2 include current intensity, current density, plating area surface area, plating time, and electrolyte pumping flow rate for characterizing cathode current efficiency, which are determined based on faraday' S law.
  3. 3. The method for establishing electrolytic process conditions based on combinatorial optimization according to claim 1, wherein the constructing a neural network prediction model in step S2 specifically comprises: s21, randomly selecting a group of network structure parameters to construct a neural network, wherein the network structure parameters comprise the number of neurons of a first layer and the number of neurons of a second layer; s22, training the constructed neural network by using a training sample to obtain a trained neural network; S23, inputting test set data into a trained neural network, and calculating an average percentage error of an output value and a true value; and S24, optimizing network structure parameters through loop iteration until the average percentage error meets preset conditions, outputting optimal network structure parameters, and constructing and training based on the optimal network structure parameters to obtain an optimal neural network prediction model.
  4. 4. The method according to claim 1, wherein the electrolytic plating process condition data in the step S1 includes electrolytic time, current density, electrolytic solution temperature, electrolytic solution density, pumping flow rate for delivering electrolytic solution, semiconductor element morphology and plating product data include semiconductor element dimensions, electrolytic plating area surface area, specification standard of plating metal film thickness, and morphology appearance, binding force, gas corrosion rate, resistance change rate, and frictional wear rate of the plating layer, and the electrolytic plating device morphology and operation data include real-time records of electrolytic die plating area morphology, electrolytic anode-to-semiconductor element spacing, electrolytic power supply specification, and electrolytic time, solution temperature, current, pumping flow rate, solution component concentration.
  5. 5. The method according to claim 1, wherein the threshold value range preset in the step S3 is a film thickness qualification standard, and the method comprises the steps of determining that the percentage error value of the measured plating metal film thickness and the standard specification value is less than 15.0% and is qualified and less than 5.0% is excellent.
  6. 6. The method for establishing electrolytic process conditions based on combination optimization according to claim 1, wherein the multiple performance tests and the preset single performance judgment standards in the step S4 at least comprise a plating binding force judgment standard, a gas corrosion rate judgment standard, a resistance change rate judgment standard before and after a plug test and a quality difference rate judgment standard before and after a friction wear test, wherein the plating binding force judgment standard is that the ratio of a falling area to a total area of a plated layer after bending is less than 1.0% and is excellent, the ratio of the gas corrosion rate judgment standard is that the ratio of the corrosion area to the total area is less than 0.6% and is excellent, the ratio of the gas corrosion rate judgment standard is that the change rate is less than or equal to 3.0% and is excellent, the resistance change rate judgment standard before and after the plug test is less than or equal to 0.5%, the quality difference rate judgment standard before and after the friction wear test is that the difference rate is less than or equal to 3.0% and is excellent, and the ratio is less than or equal to 0.5% and is excellent.
  7. 7. The method according to claim 1, wherein the predetermined criteria in step S6 are that the determination results of the percentage error value of the thickness of the plated metal film, the surface area error value of the plated area, the binding force of the plated layer, the appearance of the plated area, the gas corrosion rate, the resistance change rate before and after the gas corrosion, the resistance change rate in the plugging test, and the frictional wear quality difference rate are all excellent.
  8. 8. A combinatorial-optimization-based electrolytic process condition establishment system for implementing the method of any one of claims 1 to 7, comprising: an electrolytic plating device for actually plating the semiconductor element according to the inputted electrolytic process conditions; The data processing system of the electrolytic plating process conditions is embedded with a neural network prediction model and is used for executing an electrolytic process condition establishment flow based on combination optimization so as to establish optimized electrolytic process conditions and outputting the optimized electrolytic process conditions to the electrolytic plating device for production debugging.
  9. 9. The combinatorial optimization-based electrolytic process condition establishment system according to claim 8, wherein the electrolytic plating apparatus further comprises an electrolytic plating solution composition automatic analysis-automatic replenishment system for maintaining the stability of the concentrations of the respective components of the electrolytic solution.

Description

Electrolytic process condition establishment method and system based on combination optimization Technical Field The invention belongs to the technical field of semiconductor element electrolytic plating, and particularly relates to a method and a system for establishing electrolytic process conditions based on combination optimization. Background In the field of semiconductor packaging and manufacturing, as electronic components continue to develop toward high density, miniaturization and high reliability, the performance requirements for the surface functional plating layer thereof are becoming more stringent. The electrolytic plating technique is one of the core processes for depositing nanoscale metal films (e.g., au, ni, pd, pt and alloys thereof), and the establishment of process conditions directly determines the quality of the coating and the final properties of the product. Particularly, in key components such as connectors, lead frames and the like, the plating layer is required to have excellent compactness, thickness uniformity and wear resistance, and also is required to exhibit excellent and long-acting corrosion resistance under severe environments such as high temperature, high humidity, salt mist, industrial corrosive gas and the like so as to ensure high fidelity and long-term stability of electronic signal transmission. At present, the industry generally adopts severe test results such as a nitric acid vapor corrosion test, a mixed gas corrosion test and the like as important technical indexes for measuring the comprehensive performance and the process of the plating layer. Currently, establishing electrolytic plating process conditions (e.g., current density, plating time, solution temperature, flow, electrode spacing, etc.) is largely dependent on two conventional methods: (1) Experience trial and error, i.e. the engineer sets the initial process parameters based on faraday's law and past experience, and optimizes by "test-strip-test-adjust" cycles. The method is highly dependent on personal experience, long in research and development period and high in cost, and a global optimal solution is difficult to find in a complex system of multi-parameter coupling. (2) Single factor test method in which, when the influence of a certain parameter (such as current density) is studied, other parameters are fixed and the influence trend and preferred range are determined by test. This approach ignores interactions between process parameters, such as the optimum current density value, which tend to change with changes in solution flow or temperature. Therefore, the "optimization" condition obtained by the method is often locally optimal, and it is difficult to ensure that the plating layer obtains a fully excellent judgment result in a complex and severe comprehensive performance test (such as high corrosion resistance, low contact resistance and strong binding force are satisfied). In recent years, although studies have been made to predict plating properties or process parameters using artificial intelligence models such as neural networks, these methods have been mostly in the off-line modeling and simulation prediction stages. The accuracy of model prediction is highly dependent on the comprehensiveness and quality of training data, and the prediction result usually lacks closed-loop verification and feedback correction of a complete physical system of 'device-solution-product' under actual production conditions. Therefore, the existing method cannot form an optimization system capable of self-iteration and continuous evolution, so that predicted process conditions often fail due to dynamic factors such as device states, solution aging and the like in practical application, and the final product still has difficulty in stably passing through the severe detection of all performance indexes. Therefore, how to provide an electrolytic process condition establishment method capable of comprehensively considering multi-parameter coupling, fusing theoretical models and measured data and realizing self-optimization through closed-loop feedback, so as to rapidly and accurately screen out an optimal process window capable of stably meeting all key performance indexes (especially the full-optimal standard) of a semiconductor element, and the method becomes a technical problem to be solved currently urgently. Disclosure of Invention The invention aims to provide an electrolytic process condition establishment method capable of integrating a theoretical model and measured data to carry out closed-loop feedback and self-iterative optimization so as to rapidly and accurately screen out an optimal process window capable of stably meeting the comprehensive performance full-excellent standard of a semiconductor element. In order to solve the problems, the invention provides a method and a system for establishing electrolytic process conditions based on combination optimization, which are applied to a double-side