Classification and Quantification of Consumer Drink Products Using Solution Gated Graphene Field Effect Transistors
Restricted (Penn State Only)
- Author:
- Price, Collin
- Area of Honors:
- Engineering Science
- Degree:
- Bachelor of Science
- Document Type:
- Thesis
- Thesis Supervisors:
- Saptarshi Das, Thesis Supervisor
Joseph Paul Cusumano, Thesis Honors Advisor - Keywords:
- Graphene
ISFET
Sensing
Semiconductors
Nanomaterials
Machine Learning
Consumer Products - Abstract:
- Food fraud - economically motivated tampering of food products, is an expensive global issue. Each year it leads to billions in costs to consumers and retailers around the world. Consuming diluted, adulterated, or counterfeited food products can have serious health consequences for many people, especially when it comes to cornerstone nutritional products like milk. One potential solution to this issue is a small, robust, and scalable food identification sensor system based on graphene ion-sensitive field effect transistors (ISFETs). Graphene, a single atom thick sheet of carbon, is a novel nano-material used in many exciting applications, including the manufacture of semiconductor devices. Graphene has many amazing properties, including incredible abilities in chemical and biological sensing challenges. By placing liquid analytes on the surface of graphene ISFETs and applying a changing gate voltage through the liquid, electrical transfer curve data can be obtained. These transfer curves are unique to each liquid analyte. While differences between transfer curves can be minimal to the human eye, transfer curve data can be used to develop physically derived Figures of Merit (FOMs) which show key data on device operation. Analysis using these FOMs allows identification of each unique liquid analyte and allows examination and ranking of the most effective FOMs for creating identification models. Using this information, data analysis and machine learning models such as analysis of variance, K-nearest neighbors, and principle component analysis can be used to classify and quantify common consumer food products with an average accuracy of over 99%.