Photodegradation Kinetics and Deep Learning-Based Intelligent Colorimetric Method for Bioavailability-Based Dissolved Iron Speciation

Abstract

Via the photodegradation of dissolved iron (dFe) complexes in the euphotic zone, released free Fe(III) is the most important source of bioavailable iron for eukaryotic phytoplankton. There is an urgent need to establish bioavailability-based dissolved iron speciation (BDIS) methods. Herein, an intelligent system with dFe pretreatment and a colorimetric sensor is developed for real-time monitoring of newly generated Fe(III) ions. According to the photodegradation kinetics of dFe, including kinetic constant and photogenerated time of free Fe(III) ions, 3 sources, 6 kinds, and 12 species of dFe are determined by our photocatalytic-assisted colorimetric sensor and deep learning model within 20.0 min. The algal dFe-uptake for 4 days can be predicted by BDIS with correlation coefficient 0.85, which could be explained by the hard and soft acids and bases theory (HSAB) and density functional theory (DFT). These results successfully demonstrate the proof-of-concept for photodegradation kinetics-based speciation and bioavailability assessments of dissolved metals.

Few-Shot Learning-Based, Long-Term Stable, Sensitive Chemosensor for On-Site Colorimetric Detection of Cr(VI)

“…At present, the detection limits of DPC-based dipstick tests for Cr­(VI) without additional enrichment are reported in the range of 3.000–10.00 mg/L, which is difficult to meet the requirement of WHO and USEPA. Therefore, gray value-based signal processing technology for the sample images should be replaced by the deep learning method.…”

“…DCNN, as one-click image analysis with super-resolution, has already revolutionized colorimetric determination. , But it is severely limited by its data-hungry nature, which is overcome by combining the GAN with DCNN, i.e., FSL is proposed as a new data tool.…”

“…The influence of coexisting ions (29 kinds 22−25 ) on the color and gray value change (ΔG) on DPC/PAN/SG after the reaction with Cr(VI) (10 mg/L), i.e., colorimetric determination, is not obvious (shown in Figure 4). At present, the detection limits of DPC-based dipstick tests for Cr(VI) without additional enrichment are reported in the range of 3.000−10.00 mg/L, 4 which is difficult to meet the requirement of WHO and USEPA. Therefore, gray value-based signal processing technology for the sample images should be replaced by the deep learning method.…”

“…The deep convolutional neural network (DCNN), artificial neural network, support vector machine, and generative adversarial network (GAN) are some examples of DL algorithms frequently employed in chemical analysis. Currently, DCNN has been successfully applied to improve the analytical performance of APP-based colorimetric determination, including denoising, recognition, and summary of every small characteristic change from each image. , However, data-driven DCNN is data-hungry. A large number of samples are typically required for proper learning, which has been the major challenge of its application in analytical chemistry for its time-consuming and resource-greedy part, i.e., “time and money” cost. , Often, this is a serious problem.…”

“…Currently, DCNN has been successfully applied to improve the analytical performance of APP-based colorimetric determination, including denoising, recognition, and summary of every small characteristic change from each image. 4,5 However, data-driven DCNN is data-hungry. A large number of samples are typically required for proper learning, which has been the major challenge of its application in analytical chemistry for its timeconsuming and resource-greedy part, i.e., "time and money" cost.…”