Analysis of different parameters in water

nirjammin's picture

Hi everyone,
I would like to know your opinion about the technical fesibility of measuring different compounds in water by means of NIR (sucrose, starch ...)
The first measurements seem that only the absorption due to water is observed, despite the fact that some of the compounds have C-H bonds which should be observed in spectral regions between O-H absoprtion bands.
The measurements are performed with a transflectance probe with optical path between 2-12 mm. Is there any possible strategy to follow to measure these compounds?
Thanks in advance,

dwhopkins's picture

Hi John,
I have made some very succesful measurements of glucose and malic acid mixtures in water.  It all depends on what concentration ranges you wish to measure, and what other compounds are present.  Measurements in water can be tricky, and with compounds such as sugars and starches with OH groups, hydrogen bonding with water and water-water can complicate the spectra, so that temperature compensation or control will probably be an issue.  There are good wavelength ranges above and below the major water band at ~1940 nm that are well suited to meaure the CH overtones, and the region  you choose will be influenced by the concentration range you wish to measure. It may seem that only the water is observed in the spectra, but CH absorptions are definitely present and can be enhanced by observing derivatives and using derivatives as pretreatment for chemometric models to determine the compounds.
You did not mention whether your samples will be highly light scattering or relatively clear, or whether these are natural or samples you can weigh out.  My advice is, set up a nice experimental design and see whether you can determine the compounds of interest at the concentrations of interest.
Best wishes,
Dave Hopkins

hlmark's picture

John - I pretty much agree with Dave. I have been able to observe the -CH band in glucose, but it was very difficult. The problem is that the -CH bands in sugars are most tertiary -CH, which have very low absorbtivities. In order to see the glucose -CH band I had to use such high concentrations that it probably distorted the spectrum, almost certainly the peak absorbtivity and most likely the peak location. And even then the band was still very weak. This was a number of years ago and maybe there's been some progress made since then, but I would not want to base an important analysis on results such as what I had. The object back then was to detect glucose in blood for diabetes diagnosis. Huge efforts were made but to this day there is no method for measuring glucose in blood, so that can serve as a caution.
As for starch and possibly other unnamed constituents you implied, what is the nature of the starch? Is it raw solid starch or has it been cooked? Raw starch that is a solid powder has a very nice spectrum that can be measured by reflection methods. In water you will probably have to deal with the interference of water in addition. If the starch has been cooked then you will probably have similar problems that you do with sugars.

gabiruth's picture

Hi John

I think you have already good advice although I was able to measure low ranges (few %) of sugar in water while it was clear. However the point I want to make is that don't try to measure the individual sugars in mixtures of sucrose glucose and fructose because once they dissolve in water their spectra are 100% identical and you can't get any reasonable predictions even though you can get nice regressions for each. So, you might be able to measure the total, but not the individual ones. In my early days in NIR more than 20 years ago I thought I will be smart and didn't have anyone to consut with so I learned the hard way


Be well


Gabi Levin, Ph.D


ianm's picture

We have successfully measured 5 components of liquid manure which has contained up to 99% water. The reproducibility of your reference testing for whatever you want to measure, including water or dry matter, is very important.

(posted on behalf of Phil Williams)

dardenne's picture

We have worked a lot with high moisture products (silage, fruits, milk, ..). NIR can predict the main constituents but the models must be verified on a dry matter basis. Sometimes the dilution effect is such that all the parameters are highly correlated (negatively) with water. So, all the parameters seem working well. When the predicted values are obtained on a “as is” base, calculate the values on the DM base and compare with the corresponding reference values. You could be surprised of the decrease of the R2.