[rev_slider_vc alias=”applications”]

Monitoring Water in Organic Solvents

water bubbles - app note

Solvents are the backbone for many chemical processes, as reagents often require dissolving before use. It is therefore imperative that solvent quality can be trusted, with known levels of impurities. One potential impurity is water, which can be particularly challenging to remove, for example from polar solvents which actively solvate large amounts of water from the atmosphere.

Moreover, certain processes require strict limits on the amount of water present for efficiency, purity and safety reasons. For example, a process containing chemicals that react violently with water will require anhydrous solvents to prevent dangerous side reactions occurring. How can you trust that your solvent is as dry as you’d like?

Challenges with current techniques

There are several recognised moisture quantification techniques, all with their particular challenge:

  • Karl Fischer titration – a common technique for quantifying water levels in liquids. Highly accurate and precise, but requires taking a sample extract, is slow, and expensive.
  • Mass loss calculations – a very simple technique, but affected by all volatile compounds present and can be very inaccurate. It still requires a sample extract and can be very slow.
  • Online spectroscopic techniques – these are real-time measurements with good levels of precision and accuracy and include near infrared (NIR), mid infraed Fourier transform (FTIR) and Raman spectroscopy. Both Raman and NIR instruments can be used in industrial processes, but they lack any real sensitivity for water. Conventional FTIR spectrometers are more sensitive but necessitate the use of fragile fibre probes, which are wholly unsuited for the industrial environment. Until recently they haven’t been truly suitable for online analysis, but the development of Keit’s IRmadillo now enables real-time, online measurements using a spectroscopic technique.

Innovation in process monitoring & control

Keit offers the innovative IRmadillo FTIR spectrometer – an industrial mid-infrared instrument that is sensitive enough to measure water content, developed with a solid-state design that eliminates the need for fragile fibre probes and is wholly suited for the industrial environment for real-time, online process monitoring of water content in a variety of solvents.

Measuring water content in solvents

In this application note, we demonstrate the use of a rugged, online FTIR spectrometer, the IRmadilloTM , to monitor the water content in a variety of solvents. The water content in methanol, IPA, acetone and acetonitrile are monitored, with limits of detection (LoDs) of 100, 101, 65 and 161 ppm respectively.

To properly test the performance of the spectrometer in a range of commonly used solvents both protic (methanol and IPA) and aprotic (acetone and acetonitrile) solvents were used for the trial. The protic solvents contain -OH groups, which behave in a similar way to water molecules, meaning they are potentially challenging mixtures to analyse. Acetone and acetonitrile are both aprotic solvents (as neither have -OH groups), but still have polar features (a C=O bond in acetone and a C­­­≡N bond in acetonitrile), which will strongly interact with water molecules.

It is necessary to use OSC transformations on the data because at very low concentrations of water the spectral response becomes very non-linear. While complicated chemometric transformations are often undesirable, the robust and reliable nature of the IRmadillo makes them a genuine possibility as the model does not drift with time and needs minimal maintenance.


The probe of the IRmadillo was inserted into dried glassware and the system sealed from ambient atmosphere under a N2 purge. The solvent to be studied (anhydrous) was added to the glassware and spectra were acquired for 120 s in batches of six. Five of these spectra were used to build the chemometric models and the remaining one spectrum per batch was used to test the model. Known quantities of water were then added to the solvent and the spectral acquisition repeated.

This was performed for multiple water concentrations for the following solvents and ranges: acetone 50 – 3500 ppm, isopropanol (IPA) 50 – 2500 ppm, methanol 35 -2500 ppm and acetonitrile 5 – 10,000 ppm (10 %wt). The lower limit was dictated by the residual water present in as supplied anhydrous solvents. (Even anhydrous solvents contain trace amounts of water, so these values were provided by the manufacturer).

The spectra were analysed using Partial Least Squared (PLS) or Support Vector Machine Regression (SVMR) models. A spectral range of 900 – 2000 cm-1 was chosen and an orthogonal signal correction (OSC) pretreatment was applied in all cases.

Figure 1: Plot of predicted vs actual values for water measurement in common solvents from PLS and SVMR models


The results from SVMR (in the case of methanol and IPA) and PLS models (acetone) are shown in Figure 1. There is good agreement between the predicted and reference values for the alcohols, with excellent correlation for the acetone samples.

The models were used to predict values for previously unseen samples (shown in red). These were used to calculate the root mean squared error in prediction (RMSEP) which equates to an LoD. The LoDs are therefore:
Methanol – 100 ppm
IPA – 101 ppm
Acetone – 65 ppm

This shows how the spectrometer performs at the low concentration end, but it is possible that a process may include both medium and low concentrations water (for example throughout a drying process). To test this water in acetonitrile was monitored from 5 ppm through to 10,000 ppm.

The results for this PLS model are shown in Figure 2. There is an exceptional correlation between the predicted and reference values, with almost perfect agreement. The LoD is low, despite a very large range being studied:
Acetonitrile – 161 ppm

Figure 2: Plot of predicted vs actual values for water measurement in acetonitrile from PLS model.


These results show that the IRmadillo is capable of monitoring water levels in a range of solvents, over a range of concentrations. The LoDs are all < 200 ppm, enabling real-time online monitoring of both solvent drying processes and solvent quality checks.

Water content can be monitored both in protic solvents such as alcohols (methanol and IPA) and aprotic solvents such as acetone. The measurements are also not limited to low concentrations, as we demonstrate that measurements of 5 through 10,000 ppm water are made in acetonitrile.

While four different solvents have been demonstrated here, the performance of the IRmadillo will be similar for a whole host of different solvents.

Keit IRmadillo25 FTIR Spectrometer

Download the application note

If you are interested in downloading our application note on monitoring water in organic solvents, click here to complete a request form.

Interested in finding out more?

Send us a message to find out more by completing the form below.  Or, see what other types of analysis we offer to the chemical industry, and to other industries.

  • This field is for validation purposes and should be left unchanged.