An Improved Optimization Model to Predict the Deposition Rate and Smoothness of Ni Pulse-Reverse Electroplating Based on ANN and Experimental Results

Abstract

The metallic layers are an essential part of MEMS (micro electromechanical system) devices, and their deposition process must be accurately controlled; this may lead to difficulties as there are many input parameters for such a process. This research focuses on the input parameters’ effects on the Ni pulse-reverse electroplating. A neural network was constructed to characterize the pulse-reverse nickel electroforming process parameters. The sample training has accurately established the mapping relationship between input and output parameters. The nickel layer thickness and surface roughness prediction in the pulse-reverse electroplating process was realized and verified by experimental tests with a test error of 3.3%. Then, the effect of direct and reverse current density, deposition time, structure width, and stirring speed as input parameters on the thickness and surface roughness are investigated. Finally, a novel 4D diagram has been developed to derive the optimal values of direct and reverse current density relative to thickness, surface roughness, and deposition time. This diagram can help researchers and industries find suitable parameters to achieve the desired deposited Ni layer’s properties.

Keywords

• metal MEMS
• pulse-reverse electroplating
• neural network
• surface roughness
• optimization