Raw Data and Noise in Spectrophotometry

Abstract

Background

Spectrophotometers are ubiquitous in chemical and biological science; however, their precision limits are under-appreciated. Rules-of-thumb and IUPAC referenced guidance restricting absorbance ranges to minimize uncertainty are based on historically important instruments which are no longer as widely used. Advances over the last half-century changed most “raw” data from absorbance and transmittance values directly produced in analog electronics to digitised intensities. The latter are rarely provided in favour of digitally transformed derived data. Assessment of spectrophotometer limitations using digitised intensities would be straightforward with mean-variance analysis. However, in their absence, derived data evaluated at scale allows efficient characterisation of modern spectrophotometers. (100)

Results

This study analyses signals when I and I₀ are not available and only absorbance or transmittance are obtained. Current IUPAC referenced guidance indicates that absorbance should be limited between 0.1 and 1.0 a.u. with optimal performance (minimum relative standard deviation (RSD)) at 0.43 a.u. or 0.86 a.u. depending on the type of limiting noise. We characterised noise in UV-Vis spectrophotometers using three methods and report optimality spectra for the first time. We found the instruments were not Poisson optimal and best RSDs were sometimes above 1.0 a.u. We could find no evidence justifying guidance restricting absorbance to between 0.1 and 1.0 a.u. Measured RSD and light intensity are more important than absorbance values for assuring good quality measurements. However, estimating light intensity is a difficult inverse problem when I and I₀ are not available, and the tested commercial instruments did not provide these. (143)

Significance

Based on this work, classical theories are insufficient to describe spectrophotometers accurately. Furthermore, we urge IUPAC to modernise the references in its Gold Book and press instrument makers to improve data transparency. These steps are crucial to use spectrophotometers optimally. (40)