Date of Award


Degree Type


Degree Name

Master of Science (MS)



Committee Chair

Dor Ben-Amotz

Committee Member 1

Garth J. Simpson

Committee Member 2

Tim Zwier


Water has an identifiable Raman feature between 3000-4000 cm-1 that can act as a probe for its environment. Generally, on addition of hydrophobic solutes into water, the OH stretching mode is perturbed enough to show changes that indicate a disruption in the hydrogen bonding network. That stretching mode is also highly temperature dependent, exhibiting spectral changes at low temperatures, around 3000-3200 cm-1 & high temperatures, around 3600 cm-1. These regions respectively indicate tetrahedral order or disordered state enhancement. The state enhancement is respectively from either an increase or a breakdown of the rigid hydrogen bonding network. Previous methods of water temperature analysis are dependent on a ratio defined by an inflection point that separates these regions traditionally called ’hydrogen bonding’ and ’non-hydrogen bonding’ of the OH stretch, has been able to determine ≈ 1oC accuracy in a few seconds. However, when the signal-to-noise ratio of acquired spectra are low, this method is untenable when quick determination of the temperature is desired. Thus, if one were able to develop a method that doesn’t suffer as much from a low signal-to-noise limitation, then it would be possible to quickly and accurately determine temperature. Here we show, that by using Non-Negative Matrix Factorization (NMF) to generate a linear combination that best represent the low and high temperature components of water, it is possible to attain accuracy ≈ 0.5 oC in 0.1s. We have further demonstrated an accuracy ≈ 10 oC at 10x reduced laser power to simulate faster time scales. Temperature error was not significantly altered at each laser power between CCD modes despite an increase in noise. These methods will allow for a quicker, non-invasive temperature determination.