Ultraviolet Visible Absorption Spectroscopy

Running UV-Vis Spectra Without Losing Your Mind

Most people treat this technique like it's plug-and-play because the instruments are semi-automated now. You load a cuvette, hit start, and get a graph. The problem is that the graph can lie to you if you haven't bothered to understand what's actually happening inside the spectrophotometer.

The core principle is straightforward enough. Molecules absorb light in the ultraviolet and visible range when photons promote electrons from a ground state to an excited state. Conjugated pi systems absorb at longer wavelengths. Simple alkanes don't absorb in the accessible range at all, which is why UV-Vis is mostly useful for things with chromophores. Beer-Lambert law ties absorbance to concentration: A equals epsilon times l times c. Epsilon is the molar absorptivity coefficient, l is the pathlength, and c is concentration. This is the part every textbook covers and every analyst somehow still messes up in practice. I'll walk through the practical steps because the manual on your instrument probably assumes you already know what you're doing, which is generous to a fault. First, decide your solvent. The solvent cutoff matters a lot. Water is fine down to about 190 nanometers, but acetonitrile starts absorbing heavily below 190. If you're working near 200 nanometers, skip methanol entirely and go with water or a phosphate buffer. I've seen analysts waste three days troubleshooting weird baseline shifts only to realize their solvent was absorbing at the wavelength they were trying to measure. Next, prepare your blank. It should be the exact same solvent with the exact same concentration of any buffers or additives as your sample. Not close. Exact. The difference between your sample cuvette and blank cuvette should ideally be nothing more than the analyte. Use matched cuvettes if you have them. Single-use quartz cuvettes are more consistent than trying to clean and reuse the same ones, especially if you're running samples across multiple days.

For the instrument settings, scan from 800 down to 190 if you need the full range. Narrower windows work faster. A scan from 350 to 600 takes about 30 seconds. A full 190 to 800 scan with a 1 nm step size can take two or three minutes. Choose your slit width based on how much resolution you actually need. Wider slits give you more light and a better signal-to-noise ratio, which matters when your sample is dilute. Narrower slits resolve overlapping peaks but kill your throughput if your sample isn't concentrated enough. When you run the blank, let the instrument stabilize for at least ten minutes after turning it on. Older instruments with deuterium lamps need even longer. Skip this and your baseline will drift throughout the run, and you'll blame your samples instead of the lamp instability.

What Actually Goes Wrong

I spent a week dealing with absorbance values that didn't scale linearly with concentration for a simple dye standard. The curve bent downward at higher concentrations. My first instinct was contamination. I remade the stock solution three times. Same result. The problem turned out to be stray light. At high absorbances, the detector gets swamped by stray light coming through the monochromator, and the reading flattens out. Anything above an absorbance of 2.0 becomes increasingly unreliable. Dilute your samples so your readings sit between 0.1 and 1.0. This is such a basic thing that nobody mentions it until you've already ruined an hour of data. Another issue I run into occasionally is particulate scattering. If your sample has any suspended matter, even microscopically, you'll get a turbidity effect that raises the baseline across all wavelengths. It looks like absorption but it isn't. Centrifuge or filter your samples before loading. I use 0.22 micrometer PTFE syringe filters for aqueous samples and nylon for organic solvents. Never use cellulose acetate with organic solvents unless you've verified compatibility. The filter will dissolve into your sample and then you'll have two problems instead of one. There's also the matter of chemical equilibrium shifting during your scan. If your analyte is pH-sensitive, the absorbance spectrum changes depending on the protonation state. A dye that absorbs at 520 nanometers at pH 3 might shift to 610 nanometers at pH 9. Buffer your samples properly and verify the pH doesn't change when you dilute. I once had an analyst report inconsistent results for a phenolphthalein derivative and couldn't figure out why. The sample was being prepared in unbuffered water, and the CO2 from the air was slowly acidifying the solution over the course of the day. It took a week to figure that out.

Data Processing Without Making It Worse

Baseline correction is necessary but it's also where people introduce artifacts. Don't just draw a line between the first and last data point and call it a day. If your blank scan has its own curvature, apply that as your baseline. Most instruments have a baseline correction function that does this correctly. If you're processing data in software like Origin or even Excel, make sure you're subtracting the blank properly, not just fitting a linear baseline to your sample spectrum. Peak finding algorithms in modern software tend to overcall peaks. Small noise features get labeled as real absorptions. Before you assign a peak to a specific electronic transition, verify it persists across multiple scans at different concentrations. If it disappears when you dilute, it's noise or an artifact. Real peaks scale with concentration. Fake ones don't. Quantitation requires a proper calibration curve. Don't skip the standards. Running a single known concentration and calculating the rest from that is how you get systematic errors. Build a curve with at least five points spanning your expected concentration range. R-squared values above 0.99 are standard for well-behaved systems. If your curve is below 0.98, something is wrong. Check your dilutions, check your cuvettes, check for degradation.

When UV-Vis Is the Wrong Tool

This technique has hard limitations. If your sample doesn't have a chromophore, you're out of luck unless you derivatize it, which introduces its own set of problems. UV-Vis also lacks specificity. Two completely different compounds can absorb at the same wavelength with similar epsilon values. If you have a mixture and both components absorb in the same region, you need either chromatographic separation beforehand or some form of multivariate analysis to deconvolute the spectrum. HPLC coupled with a UV-Vis detector handles this routinely, but standalone UV-Vis can't untangle it reliably. The pathlength constraint is another limitation. Standard cuvettes are 1 centimeter. If your sample has an extremely high molar absorptivity and you can't dilute it further, you'll need a shorter pathlength cell, typically 0.1 centimeter or even 0.01 centimeter. These are specialized and expensive. If your sample has very low absorptivity and you can't concentrate it, you're better off with fluorescence spectroscopy, which is orders of magnitude more sensitive for detecting low concentrations. For structural elucidation, UV-Vis gives you almost nothing on its own. It tells you something about conjugation and electronic transitions but not the actual molecular structure. That's what NMR and mass spectrometry are for. UV-Vis is best used as a quantitation tool or as a supplementary characterization method, not as a primary identification method.

The real value of Ultraviolet Visible Absorption Spectroscopy comes from understanding its constraints and working within them rather than fighting against them. Get your blanks right, keep absorbances in range, and verify your results with independent methods when the stakes are high.