Nanochips® cartridges (Fig 1) are disposable panels with 400 micro arrays (Fig 2) that can be independently used as test sites for various assays like genetic marker. Each site (Fig 3) is a permeation layer coated electrode with 80 μ diameter and 120 μ pitch. Permeation or “perm” layer is a thin film of a proprietary hydrogel material over the proprietary chip design. Perm layers were deposited on the electrodes using spin coating process with a proprietary solution. Performance of the cartridge depended on the quality and thickness of the permeation layer over these 400 electrodes. In this process one millimeter deep tear shaped well was constructed from the ceramic base to which 20×20 array of silicon chip were attached. These wells filled with 180 μl proprietary solution and spun at fixed 1200 RPM for 20 seconds. Process was repeated three times at room temperature in a clean room. Post process treatment included 30 minutes in dry incubator, wash in Milli Q water, and finally, dry at room temperature. Quality of the 1,500 nm thick permeation layer was so demanding that more than half of the cartridges were rejected due to poor quality of the perm layer. Major causes of rejection were bubbles, high SD (Standard Deviation), and thickness (too thick or too thin) of the layer. To understand the problem focus was given to both solution making and spin coating process. Basic hypothesis was that the film thickness was based on viscosity of solution and subsequent evaporation process. Figs 4–6 showed the details of spin coating process for developing a heuristic model. Since viscosity depended strongly on the temperature and time during a chemical reaction, a viscosity profile was developed for the solution during the reaction. From the viscosity curve (Fig. 7) it was established that 75 minute at 50° C was not enough to complete the reaction as was initially thought. The time-Viscosity curve reached the plateau after 150 minutes! So, it was necessary to continue the reaction for 75 more minutes to complete the reaction. This explained a major reason (bubble forming) in the present process. Viscosity of the solution depended on a number of other factors like dispensed volume, temperature of the spinneret, time to dispense and/or in the spinneret, etc. A systematic study on these variables led to an empirical equation of the following form: Hf=Hf0+A*(τe/Vd)*ηf*(Tc−T0)n Where Hf = Predicted Average Thickness in nm. Hf0 is minimum average thickness in nm. T, V, and τ are various temperature, dispensed volume, and time parameters. A and n are curve fitting constants for experimental set-ups. The best of part of this modeling was its ability not only to predict the thickness of the film, rather its ability to control the thickness of the film in real time for a given solution. Above equation allowed appropriate dispensing volume and time to be kept in the well of the chip before spinning for a given solution with a specific viscosity. A tabular form was given to the operator who matched the information to get a specific film thickness. This model helped dropped the rejection rate to less than 10% from more than 50%. Operators were able to control the thickness of the film within 1500 nm +/− 300 nm, as demanded by the Spec. The model further allowed to target, rather control, the film thickness. We were also able to make films as thin as 800 nm or as thick as 1,500 nm with +/− 200 nm variation from the Table developed from the empirical equation. Using the assumption that viscosity plays the most important role in our spin coating process and constructing a corresponding evaporative model, we were able to identify a major shortcoming of an existing process to develop submicron thick permeation layer. Existing process was resulting in more than 50% rejection of an expensive critical component. An empirical model of spin coating process was developed to predict the film thickness within hundreds of nanometers. This dramatically improved the yield to more than 90% from less than 50%. The model allowed to correct the process in real time and allowed targeting the film thickness anywhere around one micron with few hundred nanometer accuracy.
Multivariable Real-Time Control of Viscosity Curve for a Continuous Production Process of a Non-Newtonian Fluid
