The deconvolution of FTIR-ATR spectra to five Gaussians for detection of small changes in plant–water clusters

Water is the environment where all the biochemical reactions and intracellular processes occur. All the processes of life are related to water.^1–3 At first glance, the molecule of water appears to be simple, as it consists of one atom of oxygen and two atoms of hydrogen, but due to the specific nature of its physicochemical properties, it has constantly been of interest for research studies. One of the specific features of water molecules is the ability to bind with each other with the help of hydrogen bonds and thereby create three-dimensional structures.^4–8 Scientists share the opinion that different water structures may have a significantly different impact on the internal processes and dynamics of biological objects.^9–11 In order to study the impact of water on a biological system, it is essential to find a reliable and simple method for the determination of differences in the structure of water. Spectral absorptions of fundamental molecular vibrations (transitions 0 → 1) take place in the mid-IR region. Absorption bands in the near infrared (NIR) region are due to overtones and combinations of the fundamental molecular vibrations. In the NIR region, the overtones of v_s bands appear with much weaker intensities than those of fundamental v_s bands. The interpretation of the spectra of overtones and combinations of the fundamental molecular vibrations is more complicated and, consequently, less quantitative than the spectra of 0 → 1 transitions.¹² Infrared spectroscopy in the mid-IR region is considered to be one of the most promising methods for studying hydrogen-bonding between the OH groups of water.^13–15 On the other hand, this method also involves several problematic aspects. For instance, there is a problem in the measurement of the IR transmission spectra in the most intensive region of the OH group absorption of 2500 cm⁻¹ to 4000 cm⁻¹ of how to guarantee the reproducible and stable thickness of ∼2 µm layer of the liquid.^16–19 Another difficult task is to eliminate the impact of water vapor of the environment in the case of IR transmission spectroscopy.^17,20 In the study of aqueous solutions where the quantity of additives that have an impact on the formation of water clusters accounts for over 1%, the above-mentioned method has produced reproducible measurement results. In the event of lower concentrations of additives, when the composition of water clusters changes a little, it is difficult to identify reproducibly the changes in the spectra. When the infrared spectra are measured by Fourier transform infrared (FTIR)-attenuated total reflection (ATR) method, the sample thickness of the measured layer is practically always stable and identical since the depth of IR beam penetration depends only on the refractive index of the sample. An FTIR-ATR system with the temperature control minimizes the influence of the temperature changes occurring due to changes in the surrounding environment and enables good reproducibility. With regard to the above-outlined factors, we introduce how FTIR-ATR spectroscopy with the temperature control enables us to reproducibly measure the spectra of water. We outline how the deconvolution of the measured spectra to five Gaussians enables us to identify the changes in the structure of the water cluster at different temperatures and in the case of different concentrations of additives in water. To describe a bond between the water molecules of different structures, earlier research work^21–24 has applied the method of deconvolution of the infrared spectra in the range of 2500 cm⁻¹ to 4000 cm⁻¹. The spectra have been deconvoluted into three and four Gaussians. The infrared spectrum of water deconvoluted into three Gaussians identifies them as follows: ∼3585 cm⁻¹ “multimer,” ∼3465 cm⁻¹ “intermediary”, and ∼3320 cm⁻¹ “network” water areas, based on the location of the Gaussian.^21,22 The spectrum deconvoluted into four Gaussians identified in accordance with the location of the center—OH vibration of strongly hydrogen-bonded structured water molecules ∼3250 cm⁻¹, OH vibration of weakly hydrogen-bonded structured water molecules ∼3420 cm⁻¹, symmetric OH vibration of non-hydrogen-bonded water molecules ∼3550 cm⁻¹, and OH vibration of asymmetric non-hydrogen-bonded water molecules ∼3620 cm⁻¹.^23,24 Unfortunately, the deconvolution of the water infrared spectrum into three or four Gaussians appears to be insufficient to identify all the changes that occur in the structure of water. The fifth peak also has a physical meaning that should not be left out of consideration. Calculations related to the existence of the fifth peak have already been made in one of the first in-depth research on the structure of water by means of infrared spectroscopy²³ and in the latest research.²⁵ However, it has not been commonly applied in the identification of changes in the state of the water structure in different matrices by means of the deconvoluted spectra. The reason for outlining the fifth peak is the in-plane bending of water at the wavelength 1650 cm⁻¹ and the overtone of which is in Fermi resonance with the OH vibration of the strongly hydrogen-bonded water molecules at 3215 cm⁻¹ to 3220 cm⁻¹. As a result, the fifth (Fermi) maximum becomes identifiable at 3090 cm⁻¹.²³ This research outlines for the first time, how the deconvolution of the water infrared spectrum into five Gaussian maxima increases the amount of information acquired from the spectrum and enables us to identify changes resulting from very small influencing factors.

High-quality ultra-pure water with a resistance of 18.2 MΩ cm and a TOC (total organic carbon) of 3 µg l⁻¹ at 25 °C was obtained by the Milli-Q® Advantage A10 Water Purification system (Merck Millipore). Raspberry Rubus idaeus (cultivar Novokitaivska) juice was pressed from defrosted berries of the 2015 year harvest using belt press Voran EBP 500. Apple Malus domestica Borkh (cultivar Tiina) juice was pressed from crushed fresh apples of the 2016 year harvest using a home-made hydraulic juice press. The “apple water” and “raspberry water” were obtained as by products in the lyophilization process of apple slices and raspberry juice using VirTis AdVantage 2.0 Benchtop Freeze Dryer. The temperature changes in the lyophilization process are shown in Figs. S1 and S2 in the supplementary material. The vacuum in the lyophilization chamber during the freezing process was 1000 Pa; during the heating process, it was 20 Pa, and when shelf and product temperature gradients converge, it was 6 Pa. The content of soluble solids in the juices was measured by a pocket refractometer “ATAGO” PAL-α at 22 °C. The measured soluble solid content was 12° ± 2° Brix for apple juice and 9° ± 2° Brix for raspberry juice. All the water and juice samples were preserved in sterile pyrogen-free polypropylene tubes closed with high-density polyethylene caps (Sarstedt).

For the detection of changes in the composition of clustered water, the infrared spectra were measured at controlled temperature with the Bruker ALPHA ATR Platinum system (Bruker Optics GmbH, Germany) equipped with a diamond crystal and a DTGS detector. Each measurement consisted of 128 scans in the range 2500 cm⁻¹ to 4000 cm⁻¹ at a resolution of 4 cm⁻¹. The background spectra were measured under the same conditions from the air. The spectra were recorded by OPUS software version 7.5 (Bruker Optics GmbH, Germany). The spectra were analyzed using the OPUS software wine analysis wizard and calibration data. The total sugar content of juices was determined from the infrared spectra under the same conditions that were used for water analysis, but in the range 375 cm⁻¹ to 4000 cm⁻¹. The background spectra were measured under the same conditions from water.

For the deconvolution of the spectra, OriginPro 2017 b9.4.0.220 (OriginLab Corporation), a software peak analyzer tool (goal: Fit Peaks Pro; baseline mode: constant Y = 0; peak finding method: second derivative; peak filtering: number of peaks 5; fitting function: Gaussian; any parameters was not fixed) was used on all the spectra with adj. R-square >9.99 × 10⁻¹.

The separation and quantification of the contaminants in the water samples (obtained by lyophilization from apple and raspberry juices) was performed by ultra high performance liquid chromatography (UHPLC) on the column ACE Excel 3C18-PFP, using the linear binary gradient of water (A) and methanol (B) with the addition of 1% formic acid in both. The chromatographic system was composed of the UHPLC-mass spectrometer (MS) Shimadzu Nexera X2 system coupled with a photo diode array (PDA) detector SPD-M20A and a triple-quadrupole mass spectrometer LCMS-8040 with an ESI ion source (Shimadzu Scientific Instruments, Kyoto, Japan). To analyze the quantity of contaminants in the samples, UV-Vis chromatograms of the blank and water samples (obtained by lyophilization from apples and raspberry juice) at 280 nm were compared and the quantity of the additives was calculated using the calibration equation of a flavonoid standard cyanidin-3-glucoside, obtained at the same wavelength.

The stretching vibration of water OH-groups is influenced by temperature due to the thermodynamically relatively labile intermolecular hydrogen bonds. In the comparison of areas of measured ultra-pure water OH stretching vibration spectra absorption maxima at different temperatures, we may notice a correlation where a rise in the temperature reduces the area of absorption maximum in a linear manner. The systematic reduction of such absorption is in correlation with the linear reduction of the density of water when the temperature rises in the range of 30 °C–80 °C.²⁶ In the comparison of the spectra of ultra-pure water in the range of 30 °C–80 °C at a 10° interval, a rise in the temperature results in the progressive shift of the whole 2500 cm⁻¹ to 4000 cm⁻¹ spectra maximum toward higher wavenumber. There is an isosbestic point at 3470 cm⁻¹, where the measured spectral lines intersect. Starting from this point, the intensity of the spectral line that has been measured at a higher temperature of the same wavenumber is higher than the intensity of the spectral line measured at a lower temperature (Fig. 1). The existence of such an isosbestic point that occurs due to the differences in temperature and divides the maxima of OH stretching vibration in 2500 cm⁻¹ to 4000 cm⁻¹ into two areas, where the absorption intensity is moving in different directions, suggests the existence of two different types of water structures.²⁷ 

Such a point has been interpreted as a sign of the existence of hydrogen-bonded and non-hydrogen-bonded water molecules.^28,29 A more precise description of such an isosbestic point says that it is a point in the spectrum reflecting the equilibrium between structurally bonded water molecules and water molecules existing outside this structure.^30,31

Figure 2 shows that when the temperature rises, the maximum of the OH stretching vibration of ultra-pure water shifts progressively toward the higher wavenumbers. The five Gaussian centers of the deconvoluted spectrum are located at the wavenumbers as follows: (A1) 3090 cm⁻¹; (A2) 3220 cm⁻¹; (A3) 3393 cm⁻¹; (A4) 3540 cm⁻¹; and (A5) 3625 cm⁻¹. The division of the spectrum in such a manner is in correlation with the model described by Walrafen.²³ 

According to Walrafen, the peak (A1) 3090 cm⁻¹ is the result of the Fermi resonance of the overtone of OH-in-plane bending (1650 cm⁻¹) with the OH-vibration of the strongly hydrogen-bonded structured water. The other Fermi maximum is at (A2) 3220 cm⁻¹. In the reduction of absorption on the wavenumber (A2) 3220 cm⁻¹, the Gaussian (A1) 3090 cm⁻¹ (Fig. 3) resulting from the Fermi resonance also falls. This may be explained by the fact that bigger water structures will decompose with an increase in the temperature.³² The rise of Gaussian (A3) 3393 cm⁻¹ in line with the rise in the temperature shows a rise in the content of weakly hydrogen-bonded water molecules. The rise in the area of Gaussians (A4) 3540 cm⁻¹ and (A5) 3625 cm⁻¹ (symmetric- and asymmetric-OH stretching vibration) reveals that with the rise in temperatures more non-hydrogen-bonded water molecules are released from big structures. The results of the measurement of ultra-pure water at different temperatures and their deconvolution to five Gaussians are in accordance with earlier understandings concerning the structure of water.³² 

Comparing the Gaussian areas of the FT-IR spectra of ultra-pure water, raspberry juice, and apple juice at a constant temperature (40 °C, what was technically convenient to conduct this experiment), it may be clearly noticed that there is a rise in the areas of Gaussians of juice spectra at (A1) 3090 cm⁻¹ and (A2) 3220 cm⁻¹, as compared to ultra-pure water (Figs. 4 and S3 in the supplementary material).

An increase in Gaussians (A1) and (A2) shows a rise in the proportion of strongly hydrogen-bonded water molecules that indicates the existence of larger water clusters in juices than in ultra-pure water. The absorption of juices at (A3) 3393 cm⁻¹ of the deconvoluted spectra is smaller than that in ultra-pure water, which shows a smaller content of structural water molecules with the weak hydrogen bonds, and at (A5) 3625 cm⁻¹ shows reduction in the content of water molecules with no hydrogen bonds according to symmetric OH vibration. An increase in absorption at (A4) 3540 cm⁻¹ shows a rise in the content of non-hydrogen-bonded water molecules in juices according to asymmetric OH vibration. The concentrations of sugars measured with the FTIR method in raspberry juice and apple juice were 45.9 ± 0.5 g/l and 118.9 ± 0.8 g/l, respectively. Sugars can also be one of the main influencers in the formation of the water structure contained in juices.

Figure 5 provides a comparison of five Gaussian areas received by the deconvolution of the FT-IR spectra of ultra-pure water, apple water, and raspberry water (Fig. S4 in the supplementary material).

The differences in calculated values are tiny, but reproducible. The differences between the deconvoluted spectra of ultra-pure water, apple water, and raspberry water in the range of A1 and A2 indicate a higher concentration of water molecules containing a strong hydrogen bond and in the range of A3 decrease in the proportion of water molecules with a weak hydrogen bond. Greater absorption in the range of A4 (symmetric OH vibration) shows that both types of water are generated as a result of lyophilization. The proportion of water molecules that are structurally not bound will increase. The deconvoluted spectra of apple water in the area of A5 reveals a decrease in the proportion of asymmetric OH vibration of water molecules, but in the raspberry water, it shows increase, as compared to ultra-pure water. The increase in absorption resulting from asymmetric OH vibration in raspberry water compared to ultrapure water shows that contaminants in minute amounts may cause unpredictable effects on the formation of the water structure. Identifying the reasons for this kind of influences can be a challenge for future research. The researchers have been describing water structures many times before in various great ways by using deconvolution,^33,34 but not shown the possibility to detect differences in the water structure when water contains contaminants in minute amounts, as described here. The HPLC-MS analysis of apple water and raspberry water revealed that both water samples contain five substances in microscopic amounts, with the total amount below 20 ppm. Therefore, in the deconvolution of the FT-IR spectra of apple water and raspberry water into five Gaussian peaks, we have registered changes in the structure of water that may have been caused by a minute amount of substance that has been dissolved in water.

The deconvolution of the absorption maxima of OH– stretching vibration at 2500 cm⁻¹ to 4000 cm⁻¹ in the FT-IR spectra into five Gaussians enables us to describe the water structures and the transitional processes between them. The deconvolution of ultra-pure water FT-IR spectra in the range of temperatures 30 °C–80 °C into five Gaussians demonstrates very clearly how the expected temperature dependent changes in the structure of water are reflected in the changes of peak areas. The results of comparison of the deconvoluted spectra of juices and ultra-pure water show an increase in the proportion of bigger water structures in juices, which is quite natural and logical. The OH-groups of sugar molecules contained in juices are suitable partners in the formation of hydrogen bonds with water molecules, and act as a core in the formation of structure in the solution. Polar hydroxyl groups in sugars can easily form hydrogen bonds with highly polar water molecules. The configurations in sugar–water hydrogen bonds also may affect the changes of bond polarity of the hydroxyl groups and could cause the formation of new hydrogen bonds.³⁵ The comparison of deconvoluted spectra of ultra-pure water and apple water and raspberry water provides unexpected results. The 20 ppm concentration of substances contained in apple water and raspberry water is so low that their existence cannot be identified in the ordinary infrared spectrum. A systematic difference of five Gaussian areas, formed as the result of deconvolution of the FT-IR spectra of apple water and raspberry water, from the corresponding areas of ultra-pure water demonstrate that water molecules need a minute amount of a dissolved substance in order to create structures with hydrogen bonds around it, and thereby reveal the existence of additives. The above-mentioned results not only demonstrate the excellent suitability of deconvolution of the FT-IR water spectra measured at a controlled temperature into five Gaussians for the study of water structures. This method is essential to effectively investigate the various effects caused by water clusters and also has a great prospect in the development of quantitative analytical methods in the study of the low concentration of additives in water solutions.

See the supplementary material: Fig. S1. The temperature changes during the apple slices lyophilization process. (A) Temperature of the shelf. (B) Temperature of the apple pulp; Fig. S2. The temperature changes at the time of raspberry juice lyophilization process. (A) Temperature self. (B) Temperature of the raspberry juice; Fig. S3. The FTIR spectra of ultra-pure water, raspberry juice, and apple juice deconvoluted into five Gaussians; Fig. S4. The FTIR spectra of ultra-pure water, raspberry water, and apple water deconvoluted into five Gaussians.

P.L. designed and performed the experiments; acquired, analyzed, and interpreted data; and wrote and edited the manuscript. P.R. performed and interpreted the data of UHPLC-MS analysis. H.K., A.K., and U.M. supervised the research. All authors reviewed the data and the manuscript.

The data that support the findings of this study are available within the article and its supplementary material. The raw data of spectra that support the findings of this study are available from the corresponding author upon reasonable request.

This research was supported by the European Regional Development Fund project Grant Nos. EU50282 and IUT20-17 of the Estonian Ministry of Education and Research.