多壁碳奈米管輔助聚合物氣體感測器陣列於電子鼻系統之應用

Abstract

本研究的目的為開發具八個多壁碳奈米管(MWCNT)輔助之聚合物氣體感測陣列晶 片，並以此提昇電子鼻系統在室溫下感測毒化物氣體的靈敏度以及對於多種氣體之辨識 能力。此氣體感測晶片共分成兩類，分別是直接以MWCNTs+PVP 複合材料感測器 (MWCNTs+PVP polymer composite sensors)及以聚合物溶液與多壁奈米碳管(MWCNTs) 組成之雙層膜堆疊式感測器(polymer / CNTs stacked sensors)。本研究中所使用之測試氣 體有三種化學戰劑模擬劑氣體、六種工業毒化物(toxic industrial compound)氣體，以及四 種不同酒類等十三種氣體，分別為二甲基甲基磷酸酯(DMMP)、二氯甲烷(DCM)、氰甲 烷(ACN)、四氯化碳(CCl4)、氯仿(CHCl3)、四氫呋喃(THF)、甲苯(Toluene)、二甲苯 (Xylene)、甲乙酮(MEK)、、清酒(Japanese sake)、高梁酒(Kinmen sorghum)、藥酒(medicinal liquor)、威士忌(Scotch whisky)，在室溫下針對不同濃度的目標氣體進行測試。 而多壁碳奈米管與PVP 聚合物複材感測器(MWCNTs+PVP polymer composite sensors)及以聚合物與多壁奈米碳管(MWCNTs)形成之雙層膜堆疊式感測器(polymer / CNTs stacked sensors)所使用之多壁奈米碳管(MWCNTs)材料，是在熱化學氣相沉積 (thermal CVD)系統中，以鐵鈷(FeCo)合金濺鍍於矽基板上當觸媒並於其上覆蓋一層氧化 鎂材料輔助成長，再以乙烯(C2H2)及氫氣(H2)為成長氣體所獲得。之後並以掃瞄式電子 顯微鏡(SEM)及拉曼光譜儀(Raman spectroscopy)來檢測其材質特性。另外亦將多壁奈米 碳管(MWCNTs)由試片上刮下以穿透式電子顯微鏡(TEM)進行檢測。而所使用的八種高 分子是根據線性溶合方程式(linear salvation energy relationship (LSER) theory)理論，及其 物理吸附鍵結(physical absorption bonding)性質所挑選。兩個類型的氣體感測器均是以液 滴滴定的方式製作以便簡化氣體感測膜的製程。 此類元件用來進行氣體感測的原理是因為感測膜吸附目標氣體之後其電性會產生 變化，從而了解其對氣體的感測行為並將所獲得數據加以整理呈現。而聚合物與多壁奈 米碳管(MWCNTs)形成之雙層膜堆疊式感測器(polymer / CNTs stacked sensors)對各種氣 體接觸後產生多組不同的電阻改變量，接著歸納反應情形為長條圖及雷達圖，進而建立 各類型化學氣體感測資料庫。之後再使用兩種主要的辨識方法：主成分分析法(principal IV component analysis (PCA))於電腦上運算以及最近鄰演算法(k-nearest neighbor (k-NN))於 電子鼻系統上進行辨識，進而判定所感測的氣體。 與多壁碳奈米管與PVP 聚合物複材感測器(MWCNTs+PVP polymer composite sensors)比較，此種雙層式堆疊的氣體感測元件最大的好處是碳奈米管層可以被聚合物層 包覆及保護，避免其直接與反應氣體接觸以延長其使用時間、元件壽命及其氣體感測靈 敏度。另外實驗結果亦顯示，本研究所使用的十三種氣體，於主成分分析法(PCA)以及 最近鄰演算法(k-NN)上的辨識結果均非常好。此外，針對部份氣體的感測結果亦與其氣 體濃度呈現線性關係，且如加以控制環境及減少外界的干擾，其感測靈敏度應可達到 ppm 等級，是十分值得發展的氣體感測系統。

For room temperature toxic gas sensing and gas specificity improving, two system chips with two different sensor configurations of a multiwalled carbon nanotube (MWCNT) - assisted polymer gas sensor array of eight sensor types were successfully developed and compared their performance. One chip employed MWCNTs + polyvinylpyrrolidone (PVP) polymer composite sensors and another chip polymer/carbon-nanotubes stacked sensors. Gases tested include three simulants of chemical warfare agents, six toxic industrial compound gas and four commercial liquors, i.e. including dichloromethane, acetonitrile, dimethyl- methyl phosphonate, carbon tetrachloride, chloroform, tetrahydrofuran, toluene, xylene, methyl-ethyl ketone, Japanese sake, Kinmen sorghum, medicinal liquor, and Scotch whisky, respectively. The chip with composite sensors used a mixture of MWCNTs + PVP polymer as sensing material. The chip with stacked sensors employed sensing materials of MWCNTs as the base layer and one of the eight polymer types as the top layer materials. MWCNT powders were scratched from Si wafer, which were prepared by thermal chemical vapor deposition on MgO/FeCo/Si substrate with C2H2 + H2 as source gases, where FeCo acts as catalyst. IV Morphology and bonding structure of the as-deposited MWCNTs were characterized by SEM, TEM and Raman spectroscopy to identify their metallic properties. The eight polymer types were selected according to their linear salvation energy relationship, and physical absorption bonding property differences with respect to different gases in the group for greater gas specificity improvement. Both chips were prepared on Si (001) wafer by solution drop casting method to simplify the process. The principle of gas sensing is basically to measure the different degrees of resistivity changes of the eight sensor types upon contact with a target gas. The sensing responses of eight sensor types on a chip were recorded as a function of time to form a so-called “sensor radar plot”. These data were then analyzed by two different mathematical analysis methods, including principal component analysis (PCA) on a personal computer or laptop, and k-nearest neighbor (k-NN) classification algorithm on an electronic nose system. By comparing the chip with the composite sensors, one of the advantages of the chip with the stacked sensors having a polymer overlayer above the MWCNT layer is to protect the MWCNT from direct interaction with the gas to improve sensor life and sensitivity. The results also indicate that the specificity for one of the three analyte groupss can be determined by its specific radar plot pattern at room temperature for each testing run, using the chip with the stacked sensors made of eight different selected polymer types. The pattern analyses of the radar plots can be simplified through PCA and k-NN analyses. By extrapolation and careful process monitoring, the maximum sensitivity of few ppm among the eight different sensor types is likely. The results also show that a linear relationship between the resistance response and analyte concentration is clearly evident for these toxic gases. The gas sensing mechanisms are discussed in the text