The sample solutions used in the experiment were CaCl₂ and MgCl₂ solutions, the concentration of Ca²⁺ was 160 mg/L and Mg²⁺ was 48 mg/L. CEM CMI-7000 S (Membrane International Inc., American) was used as an adsorbent with a thickness of 0.42mm, its exchange group was SO₃-Na, exchange capacity was 1.6 meq/g, and it was applicable in solution with pH 1–14. Ion exchange is the process of substituting ions between an insoluble solid with exchangeable ions and ions of the same charge in a solution. In this way, CEM can extract other cations from the solution and convert the target ions from liquid to solid state. Therefore, CEM was used as both an adsorbent and spectral-detecting matrix. To ensure the effectiveness of the CEM, it was saturated in 1 mol/L HCl for 24 h and rinsed with deionized water until pH = 7 before it was used. Considering the upper limit of the exchange capacity of CEM, we used 3 cm² CEM each time to prevent it from reaching the limit during the exchange of Ca²⁺ and Mg²⁺.

An automatic device was designed for sample handling and detecting. As shown in Figure 1, the device consisted of an automatic sampling component, an element ion extraction component, a sample cleaning and drying component, and a LIBS spectroscopy measuring system. The workflow of the device is shown in Figure 2. First, the automatic sampling module injected the sample solution directly into the container after the CEM was installed with fixture and started the device. A magnetic stirring device at the bottom of the container ensured that Ca²⁺ and Mg²⁺ were uniformly and fully exchanged on the surface of the CEM. Next, the two-dimensional mobile component held the CEM with a fixture in the sample container. After stirring the sample solution for the set time of the program, the two-dimensional mobile component automatically moved the CEM to the cleaning and drying component successively so that the residuals on the CEM surface can be removed and the moisture interference can be minimized. Deionized water was used for CEM cleaning. The target object Ca²⁺, Mg²⁺ were absorbed on the CEM surface through covalent bonding which was strong enough that it could not be broken under the water cleaning condition. Therefore, there was no analyte loss during the water cleaning process. At last, the CEM was taken to the detecting position which was precisely the focal plane of the laser. At this point, the LIBS system measured the Ca²⁺ and Mg²⁺ on the CEM and the results were displayed on the screen of the control module. The CEM exposed area is round, and the position of the CEM surface coincides precisely with the laser focal plane. The spectral intensity of each sample was obtained by averaging five spectra collected at different positions on the CEM surface. The position switching was realized by the stepper motor to control the vertical movement of CEM in the laser focal plane position.

The LIBS measuring system, as shown in Figure 1, mainly included a laser, spectrometer and control board. The laser source was a diode pumped solid state laser (LD201B, Lemeng Company, China) with wavelength 1,064 nm, maximum output energy 50 mJ, pulse width 8 ns, and repetition rate 1 Hz. The spectrometer was a USB2000 + fiber optic spectrometer (Ocean Optics, America), which had a spectral response range of 200–1,100 nm, spectral resolution of 0.2 nm, and signal to noise ratio 250:1. In order to obtain the best spectral, the laser energy was set to 50 mJ, the spectrometer CCD integration gate width was set to 1 ms and the detector sampling delay was set to 1 us. The LIBS system worked in paraxial mode, the laser focused on the CEM surface through a focusing lens, and the optic lens collected laser plasma emission spectra from the 45 °direction. The focal length of the laser focusing lens is 30 mm. For the collection of laser plasma, we installed an optical collimator at one end of the optical fiber, and the other end of the optical fiber was connected to the spectrometer. The spectral intensity of each sample was obtained by averaging five spectra collected at different positions on the CEM surface. The position switching was realized by a stepper motor to control the CEM vertical movement in the laser focal plane position.

First, we used the device to detect Ca²⁺ and Mg²⁺ in solution and the spectrums are shown in Figure 3. According to the NIST spectrum library, Ca II 393.3 nm and Mg II 279.5 nm were selected for the characteristic peak of Ca and Mg for further analysis. Figure 3 showed that no spectral signal of Ca²⁺ and Mg²⁺ elements could be observed on blank CEM. However, when CEM was used for Ca²⁺ and Mg²⁺ elements detection, there was no spectral interference near the characteristic peak of Ca and Mg elements. Consequently, it was feasible to use CEM combined with LIBS for Ca and Mg ions determination in solution.

The ion extraction of CEM in solution was affected by many factors, including the CEM dosage (size), solution volume, ion concentration, stirring speed and time. In this study, the sample volume was set to 20 mL and CEM size was 3 cm². In order to achieve optimal working condition, we mainly investigated the effects of magnetic stirring speed and adsorption time on spectral signals. As shown in Figure 4, the spectral intensity of Mg²⁺ and Ca²⁺ increased first and then slowly decreased with the stirring speed increased. It indicated that within a certain range of magnetic stirring speed, high speed led to high exchange efficiency between ions on CEM surface and in solution. However, when the magnetic stirring speed exceeded the range, the exchange efficiency reduced. The reason was that excessive speed led to unstable ion exchange reaction and there was not enough time for Ca²⁺ and Mg²⁺ formatting stable chemical bonding with CEM matrix. The inadequate reaction also led to weak spectral intensity. Based on above, we chose the magnetic stirring speed of 600 rpm to get good spectrum.

We further investigated the effect of time on spectral intensity. Figure 5 shows that the spectral intensity of Mg²⁺ and Ca²⁺ increased at first and then stabilized gradually with time. It illustrates that the reaction between the exchange group of CEM and the Mg²⁺ and Ca²⁺ in the solution progressed gradually from a rapid reaction to a dynamic equilibrium state. Moreover, the exchange efficiency of Mg²⁺ in different concentrations kept stable from 1–3 min and 3–6 min. In contrast, the CEM exchange efficiency of Ca²⁺ with different concentrations was high from 1 to 3 min and gradually decreased after 3 min. Within 3–6 min, the Ca²⁺ exchange reaction rate decreased and was in the transition stage from rapid exchange to exchange equilibrium. After 6 min, the ion exchange reaction reached equilibrium and the spectral intensity kept in a certain range. To achieve quantitative analysis, 3 min was selected as the adsorption time for rapid determination. This time point was in the linear range of ion exchange reaction and the CEM could be prevented from reaching adsorption saturation state at the same time.

Aiming to verify the detection ability of the automatic device for Ca and Mg in solution, the quantification model was established using characteristic spectral lines of Mg II at 279.5 nm and Ca II at 393.6 nm with C I 247.8 nm as the internal standard. The spectrums and calibration curves are shown in Figure 6. The spectral intensity increased with the increase of the Mg²⁺ and Ca²⁺ concentrations, but the trend is nonlinear. Inflection points turn out at Mg 5 mg/L and Ca 20 mg/L, and the whole range can be divided into two parts, in which the signal intensity shows linear variety with the concentration gradient change separately. Moreover, the slope in the low concentration part (Mg²⁺ 0–5 mg/L and Ca²⁺ 0–20 mg/L) is obviously high than that in high concentration. The reason might be that the ion exchange speed was high in high concentration difference between target ions and exchange groups in the CEM matrix under certain setup parameters including stirring speed and time. Meanwhile, the other part of the curve also shows linear after the inflection point and does not appear nonlinear caused by the self-adsorption effect.

In the calibration curve, the signal intensity is not zero when the Ca²⁺ and Mg²⁺ analyte concentrations are zero, which is probably due to the interference of other element lines emitted close to the wavelength. Meanwhile, in the real water sample, the emitted lines intensity of Ca and Mg are much higher than that of interferential element on the CEM matrix, so it does not represent to be a problem for real sample quantitative analysis. Moreover, the purpose of the device and method is to meet the detection needs of Ca²⁺ and Mg²⁺ concentrations in the real water sample, and their concentrations are mainly in the linear range of the second part of the calibration curve. Therefore, we choose the second part of the calibration curve for quantitative analysis, and the correlation coefficient R ² for Mg is 0.99 and Ca is 0.98.

We also investigated a new definition for the LOD of a given element in a univariate calibration. We used the linear part of the calibration curve at low concentrations to calculate the LOD by Eqs 1, 2, [42]. The LOD of Mg is 1.63 mg/L and that of Ca is 3.62 mg/L. Then, in the second part of the calibration curve, root-mean-square error (RMSE) was calculated using Eq. 3, and more detailed parameters of the quantitative model are shown in Table.1. Furthermore, we verified the prediction ability of the established mode by calculating the standard recovery rate of samples with and without interfering cations. The results are shown in Table 2. The results demonstrate that the device is applicable for the automatic rapid detection of Mg²⁺ and Ca²⁺ in solution. L O D = 3.3 σ y / x 1 + C ¯ 2 C i − C ¯ 2 b (1) σ y / x = I i − I ^ i 2 N (2)

In Eqs 1, 2, N represents the number of points in the calibration curve, b is the slope of calibration curve. I i are spectral signal, I ⌢ i are the value predicted by the calibration curve. C i are the concentrations of the standards, C ¯ is the average of these concentrations. R M S E C = Y i − Y i p r e 2 n (3)

In Equation 3, n is the number of points in the calibration curve. Y i are the concentrations of the standards, Y i p r e are the predicted value of the concentrations.