Scholarlink

<p>This work is about predicting the etching time to remove the saw damage and achieve the desired thickness of the monocrystalline silicon wafer using a polynomial regression machine learning model. More than 100 wafers were sliced using a diamond wire saw. All the wafers were etched in KOH solution. The initial and final thicknesses were logged to train the ML model. The equations were formulated to attain the final thickness precisely. The number of features was reduced using mathematical calculations. The reduced features and outputs were fed into a polynomial regression algorithm to get a trained model. The trained model was exported into a pickle file. The calculations were made to validate the quantity of KOH. Using the trained model and the calculations, the user-friendly interfacing software was developed with tkinter, an own graphical user interface of Python. The graphical user interface gets 7 parameters as inputs. They are Length, breadth, thickness, desired thickness of the wafer, temperature, quantity of KOH, and water. Using all the inputs, it can precisely calculate the etching time of the reaction. The primary beneficiaries of the calculator are laboratories seeking to achieve precise and specific target wafer thickness. The tool is particularly advantageous for industries looking for new wafer dimensions and for those aiming to remove saw damage.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A Generalized Precise Approach for KOH Etching in Monocrystalline Silicon Wafer with an Orientation of < 100 > using Machine Learning

  • 期刊论文

  • 发表日期：

    2025年5月30日

查看全文
  • Lawrence Daniel Stephen Tamil,
  • Jeffrey Joseph John Jeya Kamaraj,
  • Senthil Pandian Muthu,
  • Ramasamy Perumalsamy

摘要

This work is about predicting the etching time to remove the saw damage and achieve the desired thickness of the monocrystalline silicon wafer using a polynomial regression machine learning model. More than 100 wafers were sliced using a diamond wire saw. All the wafers were etched in KOH solution. The initial and final thicknesses were logged to train the ML model. The equations were formulated to attain the final thickness precisely. The number of features was reduced using mathematical calculations. The reduced features and outputs were fed into a polynomial regression algorithm to get a trained model. The trained model was exported into a pickle file. The calculations were made to validate the quantity of KOH. Using the trained model and the calculations, the user-friendly interfacing software was developed with tkinter, an own graphical user interface of Python. The graphical user interface gets 7 parameters as inputs. They are Length, breadth, thickness, desired thickness of the wafer, temperature, quantity of KOH, and water. Using all the inputs, it can precisely calculate the etching time of the reaction. The primary beneficiaries of the calculator are laboratories seeking to achieve precise and specific target wafer thickness. The tool is particularly advantageous for industries looking for new wafer dimensions and for those aiming to remove saw damage.