Measuring nitrate concentration in surface waters with a microfluidic device facilitated by a miniaturized capacitive deionization cell

Nitrate is one of the primary anthropogenic pollutants in surface waters. The increase of nitrate ions in surface waters in many places of the world is seasonal and tightly related to the use of nitrogen fertilizer for agricultural purposes (Yu et al. 2020). Most of the time, the concentration of nitrate in the surface water and groundwater supplies remains at a low level. For instance, in the US, it is normally below 4 mg/L. However, a peak concentration of over 1,500 mg/L has been reported in the agricultural area in India (Cotruvo et al. 2011). Nitrate is the key macronutrient that causes eutrophication in natural water (Burt et al. 2010) and it also poses a threat to the health of human beings. As natural waters are important drinking water sources, the changes of the nitrate concentration in surface water, e.g. sewage effluent discharge, reflect the exposure of higher nitrate levels in the drinking water. Previous studies have shown that high nitrate concentrations in the drinking water play a significant role in the occurrence of methemoglobinemia in infants exclusively under the age of 3 months (Cotruvo et al. 2011). Therefore, monitoring nitrate concentration in natural waters is important for the prediction of pollution accidents, e.g. algal blooms and for safeguarding our drinking water sources.

Traditionally, there are many ways of measuring nitrate in surface waters, including chemistry (Moorcroft et al. 2001), electrochemistry (Kaniansky et al. 1994; Davis et al. 2000), chemiluminescence (Aoki et al. 1997), as well as chromatography. The most adopted is the ‘wet chemical method’, which uses Cr or VCl₃ to reduce all nitrate into nitrite and subsequently reacts with the Griess solution to form a pink solution (Gal et al. 2004; Cogan et al. 2013). The absorbance of the pink solution at 525 nm light is normally utilized to quantify the strength of the color, which has a linear relationship vs. nitrate concentration over a wide concentration range. Previous researchers have successfully miniaturized this concept using state-of-the-art microfluidics technology (Beaton et al. 2012; Nightingale et al. 2019). However, the ‘wet chemical’ method uses toxic chemicals that might cause secondary pollution to the environment without careful handling and post-treatment. Nitrate and nitrite have very strong responses to ultraviolet (UV) light with a wavelength of 220 nm and also in natural water nitrite concentration is quite low compared to nitrate, the alternative is to measure nitrate with the UV absorption method (National Environmental Protection Bureau (NEPB) 2007; Baird et al. 2017). This UV220 method requires the compensation of the UV absorption of the natural organic matter (NOM) at wavelength of 275 nm. This is because the NOM has responses at both 220 and 275 nm, while at 275 nm no response of nitrate is found. This method is much more direct and speedy than the ‘wet chemical’ method. Therefore, it is adopted by many countries for the fast analysis of nitrate ions in aqueous solutions. However, if we want to transplant this method to the framework of microfluidics, at the present, the size of the luminous gas source is the greatest hindrance.

Fortunately, nitrate does not only respond to UV light at 220 nm, specifically. Its responding region of UV light extends up to 240 nm, although the response decays along with the increase of UV wavelength (Supplementary information). To be specific, the strength of the responding signal measured by a spectrophotometer of a fixed nitrate concentration at a wavelength of 235 nm is only about one-tenth the signal measured at a wavelength of 220 nm. Because of the much weaker signal level compared to wavelength of 220 nm, it is then very important to keep the UV absorption signal from the disturbance of the organic matter. As in the surface waters, the majority of the NOMs are humic acid and fulvic acid (Rodrigues et al. 2009; Costa et al. 2011). Previous studies have shown that the molecular weights of these organic compounds are larger than 100 g/mol (MacFarlane 1978; Thurman et al. 1982; Klučáková 2018). Taking advantage of these properties, we have fabricated a miniaturized capacitive deionization (CDI) cell with monovalent exchange membranes with the cutoff molecular weight of 100 g/mol in order to separate the monovalent ions, including nitrate from the surface water samples, which contain NOMs, in other words, the chemical oxygen demand (COD). The miniaturized membrane capacitive deionization (MCDI) module was then coupled to a microfluidic device incorporating an UV-235 nm LED in order to measure the surface water samples without the interference of the NOMs.

MCDI is the modified form of CDI, which has been emerging as a desalination technology in recent years (Porada et al. 2013). A typical MCDI cell consists of two porous electrodes, which are normally fabricated by carbon materials, e.g. activated carbon, graphene, carbon nanotubes, and recently there are lots of attempts to achieve greater salt adsorption by using novel materials (Noonan et al. 2018; Divyapriya et al. 2019; Ma et al. 2020; Peng et al. 2020; Xu et al. 2020; Li et al. 2021). On the surface of each electrode, an ion-exchange membrane is attached (anion exchange membrane on the anode; cation exchange membrane on the cathode). A spacer channel is between the two membranes, which allows the water to flow through. When an electrical voltage is applied to the cell, the ions in the spacer channel will be absorbed into their counter electrodes, thereby the water is deionized. Recently, there have been many attempts to use CDI or MCDI to selectively remove a certain ion from a mixture of ions, either with the pore size, the electrode's affinity to certain ions or the ion's diffusion coefficient (Mubita et al. 2019; Cerón et al. 2020; Hong et al. 2020). In this work, we employed the monovalent anion membranes, which were capable of retaining most organic matter and divalent anions in the spacer channel, while absorbing nearly all nitrate ions.

All chemicals are of analytical reagent grade and used without further purification. Polyvinylidene fluoride (PVDF) was supplied by Shanghai New Materials Co., Ltd (China). The graphite plates purchased from Baofeng Graphite Ltd (Shangdong, China). Activated carbon powders are made from coconut shell (YP-50, Kuraray, Japan). N-methyl-2-pyrrolidone (NMP) and sodium chloride (NaCl) were purchased from Sinopharm Chemical Reagent Co., Ltd. The monovalent anion exchange membrane (ACS, 0.13 mm) and the monovalent cation exchange membrane (CIMS, 0.15 mm) were supplied by Astom Corporation (Tokyo, Japan). The cutoff molecular weights of the two membranes are 100 Da, according to the technicians from the supplier. Before use, the membranes were soaked in a 0.1 M NaCl for 24 h. All surface water samples in this work were prefiltered with a 0.45 mm filter and the installation by the river has a pretreatment of a microfiltration membrane with a pore size of 0.1 mm. Diluted NaCl solution (10 mM) was used to clean the system and was used as the electrolyte during the discharging step.

The MCDI module has a symmetrical structure. It has two titanium current collectors embracing two graphite plates. On the inner side of each graphite plate's surface, there is a layer of coated activated carbon electrode that was used for ion adsorption. To make the electrode, PVDF (previously dissolved in NMP, 1:30) and carbon black powder were mixed at a ratio of 1:9 to obtain a homogeneous slurry, and then it was coated onto the graphite plates (oval shaped for better hydrodynamics; area = 1,321 mm²). The graphite plates were then dried at 80 °C for 10 h. After drying, the final thickness of the obtained electrode is 250 μm and the activated carbon electrode has a pore volume of 0.6 cm³/g, and a surface area of 1,156 m²/g Brunauer Emmett Teller (BET) area, measured by the nitrogen adsorption method).

In the beginning of the whole process, the fluid is firstly pumped into the detection module. A spindle-like optical cell is designed in order to prevent the formation of the dead volume and the disturbance of the air bubbles. The optical cell has a light path of 10 mm, which is specially designed for measuring nitrate in surface water and its volume is 800 mL. The UV-235 nm LED (OP235-nX-SM, Asahi Kasei, Japan) and photodiode (GUVC-T21GH, Genicom, Korea) were fixed on the chip face-to-face by a dismountable bracket, sandwiching the optical cell. After the first absorbance is measured, the sample then flows into the spacer channel of the MCDI module. At the same time, charging starts, which leads to the fast adsorption of the monovalent inorganic ions. After the adsorption process ends, the stream will be transported to the detection module again to measure the absorbance of the 235 nm LED. After the measurement is over, the liquid residual in the detection module will be discharged. NaCl solution will be used to clean the detection chamber and afterwards will be transferred to the MCDI cell, where the voltage is now set to zero and the adsorbed ions will be released, which is the regeneration step. Afterwards, the solution will be discharged and the system will be regenerated.

Since all the water samples entering our device are prefiltered, the disturbance of the turbidity of the surface water sample is no longer a concern. In our method, the water samples have to be measured twice, before and after the MCDI adsorption. Therefore, the calculation does not need to involve the compensation of a second UV light (275 nm), which is simply the difference of the two measurements (⁠⁠). This value will then be taken into the calibration curve to find out the eventual concentration of nitrate.

Changes in the conductivity may be a challenge for the MCDI module to overcome since the porous carbon electrodes we employed here have a limited capacity. Due to previous reports, the capacity of the electrode made by the given recipe is around 9 mg NaCl/g electrode weight (Porada et al. n.d.). Because the monovalent membranes used in this work can prevent most divalent ions from entering the electrode region as well as the NOMs, the activated carbon electrodes can be used to adsorb as many monovalent ions as possible.

Three different surface water samples were examined before and after the adsorption process in order to select the optimal running condition for our device. For each sample, fixed voltages (0.5, 0.8, and 1.1 V) were consecutively applied and the adsorption duration for each water sample was varied for each voltage (300, 600, and 900 s). Before and after the adsorption procedure, the optical absorbance values at 235 and 275 nm were measured by a spectrophotometer (UVmini-1285, Shimadzu, Japan), which represent concentrations of nitrate and NOMs, respectively.

Figure 3(a) shows that in the beginning the nitrate can sustain a 100% adsorption ratio, but when we continue to increase the background chloride concentrations, finally the adsorption ratio starts to drop at the point that the chloride concentration reaches 240 mg/L, and the respective conductivity is over 1,000 μS/cm. Whereas for the magnesium sulphate as the background ion, the nitrate adsorption is not affected by the increase of the magnesium sulphate concentration. The monovalent ion-exchange membrane plays an important role here, which prevents most of the divalent ions from entering the membranes (Figure 3(b)). Therefore, the more important factor that ensures the full adsorption of the nitrate ions is increasing the salt adsorption capacity of the porous carbon materials, either by increasing the thickness of the electrode region (Porada et al. 2012), by employing porous electrode materials with higher capacity (Cohen et al. 2011; Porada et al. 2013; Yeh et al. 2015; Xu et al. 2016; Zhang et al. 2019; Gao et al. 2016), or by using an electrode material, which has a selective feature of nitrate ions over chloride ions (Oyarzun et al. 2019). In our case, using the current setup, given the experimental results, we are confident to apply our device to surface waters, which have a conductivity of less than 1,000 μS/cm.

A common and inevitable drawback of LEDs is temperature shift, which means changing the ambient operational temperature may have an effect on the wavelength and the light strength of the LED leading to a higher error. Therefore, we tested the device at three different temperatures (10, 25, and 40 °C) in an environmental test chamber (JHY-H-408L, Moon Mountain, Xiamen, China). For each temperature, experiments were done at concentrations of 5, 10, 15, and 20 mg/L (Figure 7(a)) and the final result was calculated only using the calibration curve made at 25 °C. It is shown that most normalized values have an error within 10%, except for one data point at 10 °C, 5 mg/L (Figure 7(b)). It is suggested to operate the device under a relative stable temperature. In our next sections, all experiments were tested in the laboratory with a stable room temperature of around 25 °C and for the long-term experiment by the river, the device was installed in an apparatus cabinet with an air conditioner always set at 25 °C.

After having the device calibrated in the laboratory, we installed it in a river on downstream of a wastewater treatment plant discharge point deliberately in order to witness more variation in the nitrate concentration. The location of this installation is not permitted to be unveiled. The device was deployed from the end of August to the beginning of November 2021. This sensor was installed together with a pretreatment unit (Haoli Tech, Shanghai, China). This unit consists of a ceramic microfiltration membrane module and a water reservoir. The ceramic microfiltration membrane has an average pore size of 0.16 micrometer, which filters the water every 1 h. The filtered water was then pumped into the reservoir, where, the measurement starts, the sample was taken by the MFD. After the measurement, the back-wash process will start, and pressurized air is used to clean the membrane. Every day at 9 O'clock in the morning, one sample was taken from the water reservoir to be measured in the laboratory by ion chromatography.

In this work, we investigated a novel method for the detection of nitrate concentrations in surface water using a miniaturized MCDI cell in combination with a microfluidic device. The MCDI cell was used for the separation of natural organic compounds and the nitrate ions prior to the determination of nitrate concentration by a 235-nm LED. It is shown that during the separation step, the optimal parameter setting (600 s charging time and 0.5 V cell potential) can lead to thorough adsorption of the nitrate ions, while most natural organic matters are left in the spacer channel. As a result, this new method can measure the nitrate concentration more accurately and directly using the 235-nm LED compared to the method adopted by the Chinese national standard. The measuring range of nitrate ions of our equipment is set at 0–20 mg/L, which can meet the detection requirements of nitrate in surface water in most cases. The long-term study shows that this online equipment can be used to monitor the trend of the nitrate concentration variation in surface water with conductivity lower than 1,000 μS/cm in the world and the data show an excellent agreement with the lab-analyzed values.

This work was supported by the horizontal project between Shanghai Bozhongguanche Intelligent Technology Co., Ltd (Haoli Technology^®) and East China Normal University (project number: 2020KFR0257).

All relevant data are included in the paper or its Supplementary Information.

The authors declare there is no conflict.