朝向可詮釋電子舌之磷酸鹽與酚類化合物電化學感測研究

Abstract

電化學感測方法可以直接將化學訊號轉化爲便於處理的電訊號，且具有量測快速、低成本等優勢。隨著人工智慧被發揚光大，愈來愈多研究者將電化學感測陣列結合多變數分析方法進行樣本數據的判讀。然而，電化學感測數據透由這些演算方法所抓取到的特徵不一定與實質的化學成分差異相關，使得這些訓練模型出現可泛化性的問題。其原因包含該陣列缺乏特定的專一性電化學感測器，或是研究者對於非專一性電化學感測器的交錯響應性理解不足。 本研究由電極響應行為之解析出發，開發代表性分析物的專一性感測器並探討對於不同電極材料的電化學交錯響應性與其訊號的科學性詮釋，並進一步延伸至具可詮釋性的電子舌系統。 第二章探索電鍍鈷電極作為磷酸根離子的電位式感測電極的響應特性。磷酸根離子的監測對水耕作物生長十分重要。電鍍在網印碳平面電極上的鈷膜對於磷酸根離子在10−5 M到10−3 M的濃度區間具有super-Nernstian的線性響應，且響應靈敏度隨著pH值改變。這些現象與混合電位機制有關，牽涉到鈷和磷酸根離子以及溶氧的交互作用，以及鈷電極表面性質的改變。在實際的水耕養液環境下，此電極可以在四天內保持誤差小於20%。本研究驗證其於實際樣本的感測表現，預期在水耕離子感測陣列中整合此感測電極，可提升植物營養攝取數據的可詮釋性。 第三章則將待測物拓展到有機分子，探討PEDOT電極作為綠原酸的伏安式感測特性。不同的溶劑 (水和乙腈) 中電鍍所得的PEDOT電極在表現出截然不同的表面形貌以及電化學響應特性。與乙腈中電鍍的PEDOT相比，水中電鍍的PEDOT表面形貌較為平整，雖然靈敏度較低，但在目標區間呈現線性響應且不易吸附，具可重複量測的優勢。接著本研究探討PEDOT對多種有機分子的交錯響應性，發現PEDOT電極是對於綠原酸上的鄰苯二酚有電催化活性。儘管咖啡酸和綠原酸因共同的鄰苯二酚結構而造成圖譜高度重疊，本研究利用其吸附特性的差異，成功在實際咖啡樣本中量測綠原酸。 第四章探討亞鐵氰化銅 (CuHCF) 對於香草精與癒創木酚的伏安式雙重感測。香草精與癒創木酚是食物飲品中常見的兩種分子，結構只差一個醛基，但是味道迥異。醛基帶來的誘導效應造成了亞鐵氰化銅對這兩種分子的氧化電位與圖譜差異，使得本研究得以利用重疊原理，使用峰值電流處的電位進行線性組合來對應分子訊號。基於上述方法，本研究也在真實咖啡和冰淇淋樣本中成功檢測到滴入的香草精與癒創木酚。這些成果顯示了普魯士藍類似物這類具有廣泛催化性的材料，能依據分子結構差異來區分不同成分的訊號，使得電化學圖譜的詮釋有更多應用可能。 第五章應用了PEDOT:ClO4、PEDOT:PSS、普魯士藍 (PB)、CuHCF四種電極作為電子舌感測陣列應用於茶湯量測。雖然此電子舌對於茶湯的發酵程度分辨效果欠佳，但是其峰值電流與茶湯的抗氧化性呈現線性相關。本研究也量測並比較了這些電極對於茶湯以及咖啡中的代表性分子的伏安響應圖譜。除了探討分子官能基與訊號的對應關係之外，也發現以金屬為活性中心的PB和CuHCF對於不同成分比較容易產生區分；以π-π交互作用對待測分子產生響應的PEDOT則在電子舌陣列中扮演輔助性的角色。本研究最後將此電子舌用於區分冷泡茶與熱泡茶。儘管茶湯樣本中的成分複雜，PB和CuHCF的伏安量測圖譜反映了兩種茶湯的差異，且與其中的化學成分有所對應，為電化學指紋帶來了可詮釋性。 本研究由特定電極與待測物交互作用出發，逐步建立電化學感測策略並探討在目標環境中能專一性感測的電鍍鈷電極和PEDOT電極。接著，本研究更進一步利用CuHCF電極的交錯響應性達成雙重感測。最後，本研究整合PEDOT和普魯士藍類似物電極成為電子舌系統，並在彼此疊加的響應下追溯訊號差異至分子來源，做出訊號的詮釋。未來研究可進一步整合交錯響應性電極陣列與多變數分析方法，以提升複雜樣本中化學資訊的擷取與判別能力，並推動電化學感測於智慧農業、食品分析與即時監測系統中的實際應用。

Electrochemical sensing directly converts chemical information to readily processable electrical signals, featuring rapid measurement and low cost. With the development of artificial intelligence, researchers are combining electrochemical sensor arrays with multivariate analysis methods to discriminate samples. However, the features captured by these methods are not necessarily related to differences in the chemical components of the samples, which may limit the models' generalizability. This study begins with an analysis of electrode response behavior, develops selective sensors for representative analytes, investigates the electrochemical cross-sensitivity of different electrode materials and the scientific interpretation of their signals, and further extends to an interpretable electronic tongue system. Chapter 2 investigates the potentiometric response of electrodeposited cobalt electrodes to phosphate ions. Phosphate monitoring is crucial for hydroponic plant growth. The cobalt film electrodeposited on a screen-printed carbon electrode exhibited a super-Nernstian linear response to phosphate ions over the range 10−5 to 10−3 M, and the sensitivity varies with pH value. Such phenomena are related to the mixed potential mechanism, involving interactions among cobalt, phosphate ions, and dissolved oxygen, and the change of cobalt surface property. The fabricated electrode showed an error <20% in a real hydroponic solution over 4 days. This work verified its performance in a real sample. It is envisaged that integrating this cobalt electrode into the hydroponic monitoring system can improve the interpretability of nutrient intake data. In Chapter 3, the analyte was extended to organic molecules, where PEDOT's voltammetric sensing behavior of chlorogenic acid was investigated. The PEDOT electrodes electrodeposited in different solvents (water and acetonitrile) showed distinct surface morphology and electrochemical responses. Compared with the PEDOT deposited in acetonitrile, the PEDOT deposited in water had a flatter surface. Despite its lower sensitivity, it shows a linear response and less residual after sensing, implying that it can be used for repetitive measurements. This work also investigated the cross-sensitivity of this PEDOT electrode to several molecules, finding that the PEDOT exhibits electrocatalytic activity toward the catechol moiety of CGA. Although the voltammetric pattern of caffeic acid and chlorogenic acid are highly overlapped due to their common catechol moiety, chlorogenic acid's concentration was successfully measured in real coffee samples by exploiting differences in their adsorption characteristics. Chapter 4 investigates the dual sensing of vanillin and guaiacol using copper hexacyanoferrate (CuHCF). Vanillin and guaiacol are common in foods and beverages. Although their structures differ by only an aldehyde group, they taste distinctly different. The inductive effect of the aldehyde group is responsible for their voltammetric pattern and the peak potential measured by CuHCF, thereby enabling a correlation between molecular concentration and a linear combination of the current signals. By this method, added vanillin and guaiacol could be detected in real coffee and ice cream samples. These results demonstrate that Prussian blue analogues, despite their broad catalytic activity, can generate distinguishable signals for different components based on molecular structural differences, thereby expanding the potential applications of electrochemical signal interpretation. Chapter 5 uses PEDOT:ClO4, PEDOT:PSS, Prussian blue (PB), and CuHCF as the electronic tongue sensor array for tea measurements. Although this electronic tongue showed limited performance in distinguishing the fermentation degree of tea, its peak current exhibited a linear correlation with the antioxidant activity of the teas. The voltammetric responses of these electrodes to representative molecules in tea and coffee were also measured and compared. In addition to investigating the correspondence between molecular functional groups and electrochemical signals, this study also found that PB and CuHCF, with metal-centered active sites, more readily differentiated among different components; PEDOT, which responds to analytes mainly through π–π interactions, played a more auxiliary role in the electronic tongue array. Lastly, the electronic tongue was used to distinguish cold-brew and hot-brew teas. Despite the complex matrix of tea samples, the voltammetric responses from PB and CuHCF can reflect the chemical compositional differences between cold-brew and hot-brew teas, thereby enhancing the interpretability of electrochemical fingerprints. This work begins with the interaction between the chosen electrodes and analytes, establishing the electrochemical sensing strategy and investigating the electrodeposited cobalt and the PEDOT electrode as selective sensors in the target samples. Then, the cross-sensitivity of the CuHCF electrode was utilized for dual sensing. Lastly, the PEDOT and CuHCF electrode were integrated as an electronic tongue. Despite the superimposed responses, the signal difference can be traced back to the underlying molecular difference and interpreted accordingly. Future research can further integrate cross-sensitive electrode arrays with multivariate analysis methods to enhance the ability to capture and distinguish chemical information in complex samples and to motivate the real-world application of electrochemical sensing in smart agriculture, food analysis, and real-time monitoring.