Detection Method Of Heavy Metals Of Chromium And Cadmium In Soil Using LIBS

The soil plays a fundamental role in ensuring national food security.In recent years,with the rapid development of industry and the widespread use of agricultural chemicals,the issue of soil heavy metal pollution has intensified significantly,leading to the degradation of cultivated soil and posing a significant threat to food quality and safety.Soil heavy metals enter the food chain through the roots of crops,such as"cadmium rice",which is a serious threat to human health.In December 2021,China’s Ministry of Ecology and Environment proposed the"14th Five-Year Plan for the Protection of Soil,Groundwater,and Rural Ecological Environment",which addressed the key issues of restoring cultivated land soil.Traditional soil heavy metal detection techniques,such as inductively coupled plasma mass spectrometry,are time-consuming,laborious,and require highly skilled operators,which cannot meet the demands of rapid soil heavy metal detection.The LIBS（Laser-induced breakdown spectroscopy）technology has the advantages of fast detection speed,simple sample pretreatment,and multi-element simultaneous detection,which is expected to achieve rapid and accurate detection of soil heavy metals.However,due to the complex soil matrix and the presence of many interfering elements,the sensitivity and universality of soil heavy metal detection by LIBS are reduced.Aiming to address the issue of low detection sensitivity,the enhancement mechanism of LIBS signals for soil heavy metal was studied,and a highly sensitive detection network with strong interpretability for heavy mentals was established.Aiming at the problem that matrix effect of heavy metal detection by LIBS in mixed soil types seriously leads to poor universality,the physicochemical causes of the matrix effect were investgated,and a framework for removing the matrix effect in heavy metal detection by LIBS was developed.The highly sensitive fusion detection of heavy metals in different soil types has been achieved,which is of great significance for the control and prevention of soil heavy metal pollution.The main research contents and results are as follows:（1）Aiming to address the problems of low sensitivity in the LIBS spectral quantitative detection of the soil heavy metal chromium（Cr）and the serious matrix effect by LIBS,the response rules of LIBS spectra of heavy metal Cr in soil were excavated.The linear weighted network（LWNet）detection network model was created,achieving highly sensitive and accurate detection of the total Cr in different soil types.Moreover,the physicochemical reasons for the matrix effect in the quantitative detection of total Cr in mixed soil types were revealed.The matrix effect removal framework of adaptive weighting normalization-LWNet（AWN-LWNet）was developed,achieving highly sensitive and accurate fusion detection of the total Cr in mixed soil types,providing model support for the rapid detection of soil heavy metals using LIBS.The results showed that:（1）The surface of yellow-brown soil particles appeared smooth with clear texture,while the surface of lateritic red soil particles exhibited rough and uneven characteristics without obvious texture.The ablative volume and maximum depth of the two soil tablets differed significantly,and the corresponding average LIBS spectra were also different,which led to the matrix effect in the detection of Cr by LIBS between the two soil types;（2）For the quantitative detection of total Cr in different soil types,the self-designed highly sensitive detection network LWNet was superior to the traditional spectral quantitative methods,and the average relative error in prediction set（AREP）of total Cr quantitative detection in yellow-brown soil and lateritic red soil reached 2.08%and 3.03%,respectively,indicating the effectiveness of LWNet in extracting Cr characteristic peaks.（3）For the quantitative detection of the total Cr in mixed soil types,based on the characteristic peaks of Cr,the self-designed AWN-LWNet framework successfully reduced the matrix effect and enhanced the quantitative accuracy of Cr（ARE=4.12%）.By extracting the Cr spectral lines after applying AWN in the AWN-LWNet framework,a calibration curve between the intensity of the Cr spectral line and the concentration was established.It was proposed that a closer slope of the calibration curve for heavy metals in different matrixes（two soil types）indicated a better matrix effect removal effect.The spectral analysis of matrix effect removal was realized.（2）Aiming to address the problems of the detection of cadmium（Cd）in soil disturbed seriously by soil matrix,the mechanism of enhancing Cd spectral signal by Graphite（C）doping was investigated.The LWNet detection network method was independently designed to realize the enhanced detection of total Cd in different soil types.A spectral fusion method from a monochromator and middle step was proposed,which effectively fused soil Cd information and soil matrix information.Combined with offset coefficient（O）,AWN-O-LWNet matrix effect removal framework was built,achieving the fusion accurate detection of Cd in different soil types,providing method and model support for the enhanced detection of total Cd in different soil types and the accurate fusion detection of total Cd in mixed soil types.The results showed that:（1）The reason why C doping could improve the intensity of Cd lines was that the ablation surface area was reduced,while the ablation depth was increased,which caused the laser energy more concentrated.（2）For the quantitative detection of total Cd in a single soil type,C doping combined with LWNet achieved the best results.The AREP of quantitative detection of Cd in yellowbrown soil and lateritic red soil were 5.10%and 9.71%,respectively.Moreover,LWNet could effectively mine Cd characteristic peaks.Therefore,the proposed LWNet model was a highly adaptable and interpretive LIBS quantitative model.（3）For the quantitative detection of Cd in mixed soil types,we first established a spectral fusion method for Cd information and matrix information.Compared with using only Cd information,the determination coefficient（R²）of the Cd calibration curve increased from 0.56 to 0.90.Then,by introducing the offset coefficient O,we improved AWN-LWNet and designed the AWN-O-LWNet framework.The fusion accurate detection of Cd in mixed soil types was realized,and the AREP reached 7.17%,which was superior to AWN-LWNet and traditional quantitative methods.（4）By extracting the Cd spectral lines after AWN-O,the calibration curve between the intensity of Cd line and the concentration was established.The results showed that the more similar the slope of the calibration curve of heavy metal for different matrixes,the better the matrix effect removal effect,which was consistent with the spectral analytical conclusion of matrix effect removal for Cr detection.（3）Aiming to address the serious moisture effect on heavy metal detection in soil solution and the decrease in spectral stability caused by the irregular shape and size of direct solution evaporation,the influence of different substrates on the LIBS signals of Cr and Cd elements in soil solution was investigated.The circular hole constraint method and spectral averaging method was proposed to improve the stability of Cr and Cd spectral lines.A quantitative detection method for soil available heavy metals using LIBS with substrate auxiliary was constructed to achieve highly sensitive detection of Cr and Cd,providing method support for the rapid,stable,and sensitive detection of available heavy metals in soil.The results showed that:（1）Cu substrate would stimulate the hybrid peak,which interfered with the Cd fingerprint peak,and Al substrate had serious baseline drift.The signal-to-noise ratio of Cd,Cr by the smooth glass was higher than that of rough glass,so the best substrate was smooth glass.（2）C₆H₁₅NO₃（TEA）in the soil available heavy metal extract had strong viscosity,which could weaken the coffee ring effect,and the circular hole constraint method and spectral average method could also weaken the coffee ring effect to a certain extent,which jointly enhanced the spectral stability.So the stable acquisition of Cr and Cd spectral signals could be ensured without adding additional thickeners,increasing the convenience of the experiment.（3）The feature peaks of Cr selected by the correlation method were 357.89,359.36,360.56,425.47,427.52 nm,and that of Cd were 214.44,226.50,and 228.80 nm,which were consistent with the National institute of standards and technology（NIST）database.（4）Based on four spectral pretreatment methods,all Cr and Cd feature peak calibration curves were established,and the R² of them were all greater than 0.9,indicating the effectiveness and applicability of this method.For yellow-brown soil and peat soil,the optimal R² for Cr detection was 0.97 and 0.99,respectively,and the optimal detection limit was both 0.03 mg/L,and the optimal R² for Cd detection was 0.96 and 0.99,respectively,and the optimal detection limit was also both 0.03 mg/L.The Cr and Cd detection limits of the method meet the requirements of the Cr and Cd limits of rice stipulated in the Food safety national standard（GB 2762-2017）even if the soil available heavy metal were fully absorbed by rice.（4）Aiming to address the low detection accuracy of available heavy metal Cd detection in soil under natural environment by LIBS,the models of LWNet,AWN-LWNet,and AWN-O-LWNet were built,and the fusion detection method of available heavy metals in soil using the substrate auxiliary was built,achieving highly sensitive detection of the available Cd in soil under natural environment by LIBS and fusion accurate detection of available Cd in mixed soil types,which provided effective data and model method support for the practical application of quantitative detection of available heavy metals in soil using LIBS.The results showed that:（1）LWNet was still superior to the traditional machine learning model in the quantitative detection of Cd in different soils.For peat soil,the optimal determination coefficient in prediction set（R_p²）was 0.998,and the root mean standard error in prediction set（RMSEP）was0.099 mg/L.For paddy soil,the optimal R_p² of the prediction set was 0.993,the RMSEP was 0.175mg/L,and the characteristic peaks of Cd could be effectively mined by LWNet.（2）It was found that Ca and Si elements might have some influence on the content of available Cd in soil under natural conditions.（3）In the quantitative detection of Cd in mixed soils,both AWN-LWNet and AWN-O-LWNet were superior to traditional machine learning methods.There was no significant difference in the Cd detection between them.The R_p² in the prediction set was both 0.986 and RMSEP was 0.223 mg/L and 0.222 mg/L,respectively.By extracting the spectra after AWN and AWN-O,a linear fitting model between the intensity of Cd spectral lines and the concentration in mixed soils was established respectively.The higher the R² of the model,the more obvious the removal of the matrix effect.The spectral analysis of matrix effect removal was realized,essentially explaining the superiority of AWN-LWNet and AWN-O-LWNet.Aiming to achieve rapid and highly sensitive detection of heavy metal in soil,the response rules of LIBS spectra of heavy metal were excavated and a highly sensitive network for the LIBS detection of heavy metal was created,achieving the rapid and highly sensitive detection of heavy metal in different soil types by LIBS.The physicochemical causes of the matrix effect in the LIBS detection of heavy metal in mixed soil types were revealed and a framework for removing matrix effect was created,achieving the rapid fusion accurate detection of heavy metal in mixed soil types,which is of great significance for soil heavy metal pollution control and prevention.