Date of Award

Summer 2019

Document Type

Open Access Dissertation

Department

Chemistry and Biochemistry

First Advisor

Stephen L. Morgan

Abstract

This dissertation focuses on creating a non-destructive method for the identification of degraded magnetic audio tapes using attenuated total reflectance Fourier transform infrared spectroscopy (ATR FT-IR) and chemometrics. The primary recording medium during the second half of the twentieth century was magnetic tapes. These tapes hold some of the world’s modern cultural history. Unfortunately, the majority of these tapes were made with a polyester urethane (PEU) binder, which has shown to degrade through hydrolysis of the ester component, called sticky shed syndrome (SSS). As the name implies, degraded tapes will stick and shed onto player guides and heads during playback. This can irreversibly destroy the data, since the binder layer houses the magnetic particles responsible for the content. Therefore, libraries, institutions, and archives are digitizing their collections. The first half of this manuscript focuses on creating a model that can accurately determine degradation using a non-destructive approach for a mixture of tape manufacturers and models. Eleven different brand/model tapes were used, acquired from three different sources. Principal component analysis (PCA), followed by quadratic discriminant analysis (QDA) was used for classification. This was successfully performed for 95.19% of the calibration set consisting of 154 tapes and 97.79% of the test set, containing 52 tapes.

The second half of this manuscript focuses on applying this model to external tape test sets that were not used in the creation of the model. It was determined that the different binders and additives included during the manufacturing process would affect the model. Therefore, a new calibration model was created from the original 154-tape set. The new model was validated yielding a 94.64% accuracy. It was tested with the original 52-tape test set, correctly identifying the degradation status for 95.58% of those tapes. The model was then externally validated with tapes acquired from the Library of Congress (LC) and University of Maryland (UMD). The model successfully identified 90.79% of the LC tapes and 78.85% of the UMD tapes. The outcome of this work will greatly enhance preservationists’ goal of triaging magnetic audio tape before digitization.

Included in

Chemistry Commons

Share

COinS