Chromatography Forum

Recently I came across a method which gave an absorbance reading of over 3 AU. This proved to be in a non-linear region. Personally I would like to keep absorbances in UV detector to 1 or less,prefererably 0.5AU. Any views?

Zero to 0.5AU is the preferred region. Absorbance up to 1.0 is acceptable for me if the calibration remains linear.

UV spectrophotometer (including LC detectors) linearity at high absorbance is typically limited by stray light. Remember that spectrophotometers do not measure absorbance, they compute absorbance. What they measure is transmittance (the fraction of light that gets through). Absorbance is the negative log of that. If you have 1% stray light getting into the photodetector, you will never see a transmittance less than 0.01, absorbance values will asymptotically approach 2.0, and you will see significant non-linearity between 1.0 and 2.0 AU.

In practice, it's best to stay below 1.0 AU

In addition to the previous posts: the higher the concentration of the analyte the greater deviation from the linearity between concentration and absorbance (according to the Beer's law)

Just to clarify, there's a deviation due to practical limitations not Beer's law itself. I wouldn't want to mislead new chromatographers into thinking they should not increase a sample concentration when faced with a small detector response.

(honestly, mislead isn't my word du jour )

Following a recent post querying the limit of a 996 detector I was again procrastinating over the rule of thumb to stay below 1 AU. Until I read Tom's words I never thought 1% of stray light could get to the photodetector -- why can't we direct more than 99% of a focussed light beam through a cell? Is this 1% of stray light stuff which gets around the cell and hits the photodetector without passing through the sample, or is it that which is refracted away from it's intended (straight?) path?

Until I read Tom's words I never thought 1% of stray light could get to the photodetector -- why can't we direct more than 99% of a focussed light beam through a cell? Is this 1% of stray light stuff which gets around the cell and hits the photodetector without passing through the sample, or is it that which is refracted away from it's intended (straight?) path?

There are many issues associated with stray light in detectors, and good stray light performance is necessary to achieve linearity up to 2 Abs, as claimed by HPLC detector manufacturers. Typically, it's measured by using solutions or filters that prevent light transmission at nominated wavelengths, usually between 220 and 390 nm. Usually comparing them to water, for the 100% value.

For good quality UV spectrophotometers, the measured wavements are often at 340 or 370 nm, with stray light limits of <0.0005% transmission ( or better, some top instruments specify x10 lower transmission ).

The significance usually increases as the lamp energy decreases at the target wavelength, hence it's often first visible as apparent absorbance on highly absorbing samples below about 205nm, or when light sources age and decrease in intensity.

The sources of stray light be combinations of light leaking in from outside, internal reflections within the detector, and within the optical path, failure to filter other wavelengths from the light path, etc. etc. Effectively it's any detectable radiation landing on the detector when the sample blocks all the set wavelength energy.

HPLC detectors might use lower wavelengths. but I doubt any would accept 1% at typical operating wavelengths, but might exceed that at low wavelengths ( eg 200nm ).

Stray light is reduced during detector design, including double
monochromators, low reflection good quality gratings and mirrors, and light sealing optical boxes and cell holders, as well as using optical baffles.

There's bound to be good discussions on Stray Light in the UV-Vis spectrophotometer sections on manufacturers' WWW pages.

Bruce Hamilton

Bruce and JA, you have a good point. The 1% number was "pulled out of my hat".

However, you got my curiosity up. A quick Google search turned up a smattering of stray light specifications for spectrophotometers. For single-monochromator units (which would be equivalent to HPLC detector optics), they seem to range from "<0.05%) to < 1%. (so my 1% example is definitely a "worst-case", but not out of the question). Double-monochromator instruments get down to and below the 0.0005% range Bruce suggested.

For HPLC detectors, the only two I could find were for an older Kratos ("<0.1%") and a new Cecil ("<0.02%"). My impression has always been that optical performance takes a back seat to things like low dead volume in HPLC detectors.

Just for reference, I ran a quick-and-dirty spreadsheet computing stray light errors for various levels:


The stray light is just what the name implies: light that gets to the sample-side photodetector from everywhere and anywhere except the designed optical path. Any seams in the monochromator housing, the joints where the inlet tubing enter and exit the enclosure, etc., etc.

All that said, I still prefer to stay below 1AU, (just because I'm paranoid doesn't mean they aren't out to get me! )

Cool, thanks for that Tom!.

I couldn't find the stray light specification for my Agilent 1100 DAD, so used a syringe to push through some 1% KI in Milli-Q. May not be accurate, lamp in middle of life, spectral bandwidth 8 nm ( I think ), but data is interesting....

Anyway, the data was...
200 nm ~2600 mAU, 0.2% Stray Light
220 nm ~3000 mAU, 0.1% Stray Light
240 nm ~3500 mAu, 0.03% Stray Light

Was going to do the VWD, as that is expected to be better, may still try that sometime... Agilent claim both detectors are linear up to 2AU, but don't test linearity above 0.8 AU in OQ.

Main lessons learnt were that detector does seem to slightly adjust lamp energy after periods of high absorbance ( baseline moves ), and Cerity sucks when trying to simply read the detector output...

Please keep having fun,

Bruce Hamilton

In the example which caused me to start the thread the standard deviated less than the sample( tested by injecting volumes of 10 -60 ul) I have no idea why this should be.
Height measurements are much more sensitive to non linearity than area. Having examined the method I was given I suggested a max absorbance of around 0.5AU as even the poorest detector should be good at this absorbance.