Comparing the light response of D-glucose in polyacrylamide hydrogel and water in NIR spectral region by using an LED based portable device

The spectral properties of the molecules depend on the matrix in which the interactions with other molecules inside the matrix affect the vibrational and rotational modes of the molecule. In this study, an absorption-based system was designed to show how the absorbance properties of the glucose change in polyacrylamide (PAAm) hydrogel when compared with water. The measurements were performed at different wavelengths; 960 nm, 1450 nm, 1550 nm, and 1950 nm and it was observed that the system is sensitive to glucose at the wavelengths of 1450 and 1950 nm in PAAm hydrogel, whereas it is only sensitive at 1450 nm in water which is due to the high absorbance of water at 1950 nm. In PAAm hydrogel, water molecules mostly gather around the polymer chains via electrostatic interactions and the absorbance of water decreases which results in an increasing absorbance of glucose. According to the results, the responsivity of the system at 960 and 1550 nm, which are the wavelengths commonly used in LED-based systems for measuring glucose in literature, is not high enough for reliable glucose measurements when compared with 1450 and 1950 nm.


Introduction
Light response of the molecules directly depends on the properties of the matrix (environment) in which the molecules are present. The rigidity, viscosity, polarity, temperature are some of the important parameters which define the matrix. For example, the number of possible vibrational states of a molecule increases with the increasing temperature or decreas-ing viscosity (Lakowicz 2006).Therefore, by controlling the parameters of the matrix, the light response of the molecules can be altered.
In the absorption experiments of organic materials like glucose in aqueous solution, the IR region is very responsive to the O-H vibration mode, therefore measurements with aqueous solutions containing O-H bonds become difficult. Whereas in the NIR, the measurements of these aqueous solutions are enabled because the intensity of the O-H overtone vibration mode substantially decreases. Since the overtones of a sample's fundamental vibrations in the IR region correspond to the NIR absorptions, samples with functional groups such as O-H, C-H, and N-H are sensitive to the NIR. With all organic materials consisting of hydrogen bonds, NIR spectroscopy has been used widely in the medical fields for analysis (Rondonuwu et al. 2019). Thus, studies such as Yang's are concerned within the NIR, and chosen to work with (900-1700) nm range to provide stable data (Yang et al. 2018). Other works such as Goodarzi et al. also focused on NIR, while investigating the shortwave band (900-1450 nm), the first overtone (1450-1700 nm) and the combination bands (2000-2300 nm). They indicated that a mixture of the first overtone band between (1600-1700) nm and the combination band between (2100-2300) nm give the high performance (Goodarzi and Saeys 2016). In addition, Heise et al. also concluded that the ranges (1473-1831) nm and (2111-2374) nm were optimal for glucose measurement (Heise et al. 1994). There are many other studies focusing on the optimal wavelength for glucose measurements, among them Abd Rahim et al. also specified that the highest significant peaks of blood glucose detected in the range  nm similar to the previous works (Abd Rahim et al. 2016). The wavelengths 1450 and 1950 nm found efficient in our study are in accordance with the literature given above. On the other hand, no study was found in literature about the absorption properties of the glucose in polymer matrix which is the important novelty of our work.
Blood measuring devices work based on two different methods which are photometric and electrochemical. In the photometric method, sugar in the blood and a particular reagent interact, and the color change is analyzed optically by the device (Demitri and Zoubir 2017). In the electrochemical method, glucose interacts with the test strip and generates an electric current, this current is measured, and the sugar value is calculated (Salacinski et al. 2014).
Blood measurements have played an essential role in the treatment of diabetic patients until now, but continuous measurements make this method painful and uncomfortable for patients considering they need blood samples for measurements (Dalvi 2013;Ahmad et al. 2015;Javid et al. 2018). Measurements from the blood became a bigger problem in children with diabetes. Thus, modern measurement methods have become necessary for diabetic patients, enabling comfortable and frequent tests (Fei et al. 2004;Formosa 2013). In light of the same idea, optical systems that work without taking blood samples from the patient have begun to be studied. Systems that measure optical absorbance, reflectance, and measurements from intercellular fluid (reverse iontophoresis) can be given as examples (Leboulanger et al. 2004;Poddar et al. 2008;Zhang et al. 2016). Among these systems, measurement by optical absorbance method was preferred in this paper due to its low cost and easy application (Jung and Hwang 2013;Xue et al. 2014;Ionescu and Doctorala 2019). Because it is optically based, it provides excellent ease of use. The low production cost allows better access to people in need (Haxha and Jhoja 2016). With the increased awareness, the value and the number of optical-based systems are rising every day (Gonzales et al. 2019).
It is aimed in this paper to test and compare the absorption properties of glucose in different matrices, which are aqueous and PAAm gel, and present it to the literature to enable further advancement in the area of non-invasive glucose measurements. Therefore, the wavelength at which the absorbance is highest for the glucose molecule will be determined in the artificially designed system that may simulate the tissue. The experiment consists of three stages; (i) preparation of the samples in different concentrations, (ii) design and installation of the experimental setup, and (iii) performing the measurements. After the experiment structure was planned, a circuit was designed accordingly. Then a device was designed to hold the electronic circuit and sample together and was printed using a 3D printer. The wavelengths selected in this study were 960 nm, 1450 nm, 1550 nm, and 1950 nm according to the spectrum of the aqueous solution of the glucose. It was observed that the system is sensitive to glucose at 1450 and 1950 nm wavelengths in PAAm gel, whereas it is only sensitive at 1450 nm for water which is due to high absorbance values of free water. According to the results, the responsivity of the system at 960 and 1550 nm, which are the wavelengths commonly used in led-based systems for measuring glucose in literature, is not high enough for reliable glucose measurements when compared with 1450 and 1950 nm.

Chemicals and synthesis of PAAm gels
The monomer (acrylamide, AAm), the initiator (ammonium persulfate, APS), and glucose were supplied by Merck (Darmstadt, Germany). All chemicals were used as received. Distilled water was used in the preparation of the glucose solutions and for polymerization. At five different concentrations, the aqueous glucose solutions, 20, 40, 60, 80, 100 mg/ml, were prepared in the polystyrene spectrophotometer cuvettes. The polymer concentration was adjusted to 4 M and was synthesized via free radical polymerization of AAm. The prepolymer solutions were prepared by including five different glucose concentrations given above, and the polymerization was performed in the presence of glucose in the heat bath. The reaction temperature was adjusted to 60 o C, and the reaction took 30 min. The polymer samples also were prepared in the polystyrene spectrophotometer cuvettes. The SEM image of the PAAm gel was taken in dry-freezed condition by using Zeiss EVO LS 10 model SEM device.

Experimental setup
The block diagram of the measurement setup was given in Fig. 1. The light source and the photodiode compartments are modular, and they can easily be changed depending on the wavelength ( Fig. 2(b)). The collimators were used to obtain an almost parallel light beam through the sample. A transimpedance amplifier with an adjustable gain was used in the data acquisition section, as seen in Fig. 2(a). The photodiode was used in the photovoltaic mode for the low noise response.
Due to changing light intensity of the different LEDs and different photodiode responses, circuit gain has been adjusted to the wavelength in which the measurements were performed. Therefore, to obtain accurate results, the gain, which is -R f , was optimized.
The operational amplifier was selected as TL081CP (ST Microelectronics, General purpose JFET OPAMP), the output voltage was recorded by using a data logger (UNI-T 71 C) with a software, and the system is powered by using a low-noise power source (Rigol-DP831A).
The properties of the LEDs and photodiodes used in this study were tabulated in Table 1.  The LED, PD, and sample holder were designed by SolidWorks 3D design program and printed by FlashForge Inventor model 3D printer. The printed holders were shown in Fig. 2(b) and it is clearly seen that the system is designed for 10 mm spectrometer cuvettes by which all measurements were performed in these cuvettes except the PAAm gel at 1950 nm. At this wavelength, the PAAm gel was sandwiched between two glass slides with a light path of 0.4 mm due to the very high absorption in 10 mm cuvette which makes the system unresponsive

Measuring NIR Spectra
The NIR spectra of the glucose in water and PAAm gel were taken by using Shimadzu UV-3600 Spectrometer to compare the results obtained by the LED-based system. The spectra of the water samples were taken in quartz cuvettes with a 1 mm light path, and the PAAm gel samples were sandwiched between two glass slides with a light path of 0.4 mm.

Theoretical model
In this study, the glucose absorption was evaluated by using Beer-Lambert's Law which is given in Eq. 1.
where I o is the initial light intensity, I is the light intensity after passing the sample as seen in Fig. 3, ε, C and d are molar absorption coefficient, the concentration, and the sample's thickness, respectively. The output voltage, v out , of the amplifier, which was given in Figs. 2 and 3, depending on the photodiode current is given in Eq. 2.
where v out , i pd , and R f are the output potential, the photodiode current, and the resistance of the feedback resistor, respectively. As seen in this equation, the output potential is directly proportional to the photodiode current. The photodiode current, i pd , is proportional to the light intensity [20] and therefore, Eq. 1 can be written in terms of the potential as given in Eq. 3.
where v out (0) and v out (C) are the output potential measured without and with glucose, respectively. By taking the logarithm of both sides of Eq. 3, the absorbance can be obtained as The molar absorption coefficient of the glucose for each wavelength was calculated by fitting the measured data to equation of Beer-Lambert's Law given in Eq. 4.

Results and discussion
The glucose absorption measurements were performed at four different wavelengths, 960 nm, 1450 nm, 1550 nm, and 1900 nm, to carry out the most sensitive wavelength for detecting the glucose in water and in PAAm gel. The wavelengths were selected based on the data given in reference (Hotmartua et al. 2015) and the spectra given in Fig. 4. In fact, glucose has higher absorbance at higher wavelengths, but the photodiodes and light sources operate in this region are costly; therefore, working at high wavelength is not cost-effective for a wearable device. When the NIR spectra given in Fig. 4 were compared, it is clearly seen that the overall absorption in water is much higher than in PAAm gel and it saturates in water above 1900 nm. This is due to the high absorption of water in this wavelength region which makes the system insensitive to glucose. In PAAm gel, the overall absorption is lower and the responsivity of the system against glucose is higher, that is the change in the absorption is higher for changing concentration of the glucose (Fig. 4(b)). In addition, the absorption peak of glucose at 1950 nm is visible for glucose in PAAm gel which is the most efficient wavelength for glucose measuring.
Before starting the absorbance measurements for glucose, the aging of glucose in water was studied first. For this purpose, five glucose samples with different concentrations, 20, 40, 60, 80, and 100 mg/mL were prepared, and the measurements for each sample were performed for 20 days at 23 o C. The slopes of the concentration-output voltage graph were given in Table 2. It is seen that the slope did not change considerably in the first seven days, but it changes considerably at the end of the 20 days. Therefore, it was concluded that the aqueous glucose solution should not be used after seven days.
The background-subtracted potential values, which were measured by using an LEDbased system for glucose in water, at 960 nm, 1450 nm, 1550 nm, and 1950 nm were given in Fig. 5 as a function of glucose concentration. The potential changes linearly with the concentration of the glucose in water and the slopes were calculated as 0.0287 mV/mM, 0.0422 mV/mM, and 0.0071 mV/mM at 960 nm, 1450 nm, and 1550 nm, respectively. The highest slope was obtained for 1450 nm, and it is clearly seen from Fig. 5(a) that the system's response to the glucose is highest at this wavelength. It can be concluded that this  Table 2 The slope of the concentration-output voltage graph with respect to time wavelength is more suitable for designing a system for measuring the glucose for wearable technology. This is also in accordance with the results given in the absorption spectrum of the glucose in water as seen in Fig. 4(a) where the glucose absorbs the light more efficiently at 1450 nm. Additionally, the LED-based system does not respond to the concentration changes of the glucose in water at 1950 nm due to the high absorption of water which is again in accordance with the results given in Fig. 4(a) where the absorption reached saturation above 1900 nm. The absorbances of glucose in water as a function of glucose concentration at 960 nm, 1450 nm, and 1550 nm were given in Fig. 6. The absorbance changes linearly with the concentration and decreases with increasing glucose concentration contrary to the expectation. However, this unexpected behavior was also observed by different groups in literature and it was explained as the waveguide effect of the glucose molecules in water (Javid et al. 2018). That is, when the concentration of the glucose increases in water, they are aligned to construct a waveguide-like structure for light beam and leads to an increase in the transmitted light intensity through the sample.
In water, since the absorbance dependence on the glucose concentration is completely different, enhancement factors were calculated from the slopes of Fig. 6 instead of molar absorption coefficients and they were found to be 5.22 × 10 − 3 , 5.49 × 10 − 3, and 1.01 × 10 − 3 mM −1 for 960 nm, 1450 nm, and 1550 nm, respectively. The enhancement factor is higher at1450 nm than the other wavelengths. When the absorption spectrum given in Fig. 4 is examined in detail, it is seen that the absorption is higher at 1450 nm, which Fig. 6 The absorbance of glucose in water as a function of glucose concentration at different wavelengths. The measurements were performed by using an LED-based system designed in this study shows that the system designed in this study gives results that are in accordance with the commercial spectrometer.
The potential changes, which were measured by using an LED-based system, as a function of glucose concentration in PAAm gel were given in Fig. 7 at different wavelengths. The system's responsivity to the glucose was found to be 0.0549 mV/mM, 0.1563 mV/mM, 0.0939, and 0.3242 mV/mM for 960 nm, 1450 nm, 1550 nm, and 1950 nm, respectively. This result clearly shows that the system's responsivity to the glucose is highest at 1950 nm which couldn't be observed in water due to the high absorption.
The absorbance variations as a function of glucose concentration in PAAm gel were given in Fig. 7 at different wavelengths. Here, it is seen that the change at 1950 nm is more significant which is in accordance with the spectra given in Fig. 6. When the absorbance values given in Figs. 6 and 7 were compared, it is clearly seen that the glucose absorbs light more in PAAm gel than in water.
The responsivities and the calculated molar absorption coefficients were tabulated in Table 3. The responsivity of the LED system is higher at all wavelengths in PAAm gel and it is highest at 1950 nm in gel. In addition, the molar absorption coefficients calculated by using the LED based system and the NIR Spectrometer are close to each other, except at 960 nm, which shows that the LED based system designed in this study works correctly and gives reliable results.
The SEM image of the PAAm gel was shown in Fig. 8 and it is clearly seen that it has a very porous structure. The glucose molecules can easily be trapped in the pores of the gel.   It is well known in literature that PAAm gel is highly hydrophilic and the positively charged amine groups on the polymer chains interact with the water electrostatically which gathers the water molecules around the polymer chains (Yilmaz et al. 2009). Possible mechanism which affects the light response of the glucose in water and in PAAm gel can be depicted as seen in Scheme 1. In the left scheme, the distribution of the glucose dissolved in water was shown. Here, the water in the left scheme was named as free water to be able to distinguish from the water in PAAm gel. In the left scheme, water molecules are distributed almost homogeneously and the glucose molecules are mostly surrounded by water. During the light absorption experiments performed in the free water system, while the Scheme 1 The schematic representation to describe the effect of the PAAm gel on the light response of glucose in free water (left) and in PAAm gel (right). Lines in right indicate the polymer chains most of the light is absorbed by water molecules only small amount of light is absorbed by the glucose due to the shielding effect of water. This description is consistent with the results given in Fig. 4(a) where the absorption of water is very high when compared with glucose. In the right scheme, water molecules are mostly gathered around the polymer chains due to the electrostatic interactions between water and positively charged amine groups of the polymer chains. In the pores of the gel, since some part of the water molecules are gathered around the polymer chains, the number density of the glucose molecules increases relatively when compared with water molecules. During the light absorption experiments performed in the PAAm gel, the overall absorption decreases since some part of the water molecules are shielded by the polymer chains and the absorption by the glucose molecules increases due to the decreased number of water molecules which shield the glucose molecules before. This description is also consistent with the results given in Fig. 4(b) where the overall absorption decreased and the glucose absorption became more visible.

Conclusion
It was concluded that the water plays a crucial role in the measurement of the glucose by using an optical system based on absorption/transmission of light. Water absorbs light in NIR region considerably and makes the optical system blind against the glucose as observed in the experiments performed in this study. To be able to increase the responsivity of the optical system against glucose, water molecules should be passivated. In this study, we showed that the PAAm gel can be used for passivating the water molecules where the water molecules are gathered around the polymer chains due to electrostatic interactions and polymer chains prevent the water molecules to absorb light.
According to the results, it was concluded that the 1950 nm wavelength is the most efficient wavelength. However, this wavelength cannot be used in water due to the very high absorption of water at this wavelength and the light at 1450 nm wavelength should be used for the system used for water.
It can be said that the system designed in this study is appropriate for detecting glucose. Therefore, the results of this study will provide a basis for future works about wearable noninvasive glucose measuring devices.