無機氣體壓電晶體感測器研究與應用

Preparation and Applicaion of Piezoelectric Crystal Inorganic Gas Sensors

郭佳文(Kuo Chia-Wen)

指導教授 ： 施正雄

國立臺灣師範大學/理學院/化學系/碩士(1999年)

摘要

無機氣體壓電晶體感測器研製與應用
摘 要
壓電晶體感測器是一種質量感測器，利用質量變化造成頻率改變，再藉由此頻率變化值了來了解被偵測物的特性。本實驗是利用鋅離子/大環胺醚( Zn2+/Cryptand22 ) 、鈦離子/大環胺醚( Ti4+/
Cryptand22 )做為石英壓電晶體膜感測器之塗佈物質，分別對一氧化碳( CO )及二氧化氮( NO2 )進行感測。在此實驗系統中，是利用自製微電腦介面來讀取振盪頻率及其變化，其中包含了數位計數的ALTERA 及8255卡，並以軟體控制介面讀取數據及繪圖。
在本研究曾利用各種金屬離子/大環胺醚錯合物(如Ti4+/
Cryptand22、Zn2+/Cryptand22、Al3+/Cryptand22、Ru3+/Cryptand22等)做為石英晶體塗佈物，用以偵測 CO 及 NO2 ，結果發現，其中以 Zn2+/Cryptand22 為塗佈物之石英壓電晶體感測器中，對 CO為物理性吸附具有可逆性，其吸附、脫附速率分別為0.571、0.404
(Hz/sec)，且具有不錯的偵測下限( 0.26ppm )及再現性( RSD =
2.15%,n=10 )，其中 NO2 不會造成干擾，所以可以使用以 Zn2+/
Cryptand22 為塗佈物之石英壓電晶體感測器，可成功地分辨 CO及NO2。而在 Ti4+/Cryptand22為塗佈物之石英壓電晶體感測器中，對 NO2 的也為物理性吸附具有可逆性，其吸附、脫附速率分別為0.535、0.386(Hz/sec)，有不錯的偵測下限( 0.16ppm )及再現性( RSD
=2.09%,n=10 )，但 CO 會有干擾產生。
水氣與溫度皆對上述之 CO 及 NO2 石英壓電晶體感測器產生干擾，其中使用五氧化二磷( P2O5 )為乾燥劑的效果最好，可將水氣完全去除且允許90%的待測氣體通過。而在溫度干擾的部分，由於本研究中的吸附現象皆屬於物理性吸附，所以感測器的訊號隨著溫度上升而有下降的趨勢。此外，在本研究中亦將石英壓電感測器與氣相層析儀之熱導偵測器( TCD )做比較，結果發現，當待測氣體通過層析管柱後，不僅可以將有機干擾物分離，也可以由石英壓電感測器可得到較好的感測結果，而 TCD 則只能感測到有機干擾物，對無機氣體CO 及 NO2 則不靈敏。
本研究中亦對感測器之使用壽命做探討，結果發現，使用一個月後，感測器之訊號大約下降40%，仍具有相當程度的靈敏度，只需對訊號稍做修正即可繼續使用，具有實用價值。
前述實驗中 CO 、 NO2 個別偵測或混合偵測後，各30組數據經倒傳遞類神經網路( BPN )分析，不管是網路學習或網路測試皆可100%將CO 、 NO2分辨出來，誤判率為0，此結果表示 BPN 可成功地將其定性化。此外，應用多元迴歸分析( MRA )也可成功地將CO 及 NO2 的濃度定量化，個別偵測後CO 及 NO2 的濃度誤差分別為-7.5 ~10%、-5.0~1.3%，而混合偵測後CO 及 NO2 的濃度誤差分別為-17.5 ~5.8%、-3.3~20.0%。此結果表示，本實驗已成功地發展出 CO、NO2 無機氣體石英壓電晶體感測器，輔助以倒傳遞類神經網路及多元迴歸分析，可將 CO及NO2 定性定量化。

關鍵字

壓電 ； 感測器 ； 一氧化碳 ； 二氧化氮 ； 類神經網路

並列摘要

Preparation and Application of Piezoelectric Crystal Inorganic Gas Sensors
Abstract
Zinc(Ⅱ)/Cryptand22 and Titanium(Ⅳ)/Cryptand22-coated piezoelectric crystal detection system with a home-made computer interface for data acquisition and processing was prepared and applied for CO and NO2. The home-made computer interface includes the digital counting system Altera and Intell-8255 data processing system. Software was written to control the interface and data acquisition.
In this study, many kinds of metal-ion/Cryptand22 complex(e.g. Ti4+/Cryptand22、Zn2+/Cryptand22、Al3+/Cryptand22、Ru3+/Cryptand22 etc.) were used to detect CO and NO2. A Zn2+/Cryptand22-coated piezoelectric crystal detection system is a reversible type. It is a physical adsorption, and its adsorption、desorption rate is 0.571、0.404 Hz/sec. The detection system also showed the good detection limit of 0.26 ppm and good reproducibility with RSD of 2.15%( n=10 ) for CO, but was insensitive to NO2. The sensor can successfully distinguish between CO and NO2. Furthermore, a Ti4+/Cryptand22-coated piezoelectric crystal detection system is also a reversible type. It is a physical adsorption, and its adsorption、desorption rate is 0.535、0.386 Hz/sec. The detection system showed the good detection limit of 0.16 ppm and good reproducibility with RSD of 2.09%( n=10 ) for NO2, but the sensor was interfered by CO.
The interference effect of water vapor and temperature on the piezoelectric detection system was not negligible. P2O5 was the best dry filter because it reacted completely with water vapor and it permitted that a 90% of tested gases reached the sensor. Besides, the frequency response apparently decreased with the increased temperature, indicating that the adsorption decreased with increase in temperature which is a characteristic of physical adsorption. The piezoelectric crystal detector system was successively connected on-line with a gas chromatograph( GC ) and compared with the thermal conductivity detector( TCD ). The organic interference compounds e.g. acid, amine are separated with the Porapak GC column and, subsequently, detected with the piezoelectric crystal detector and TCD. The piezoelectric crystal detector system seems to show the better response and selection.
The detector lifetime was also discussed in this study. The frequency response decayed 40% after it was used for a month. It only do a easy calibration. The detection system is economical.
The frequency signals from the two channel crystal analyzed sensor array were processed by back- propagation artificial neural network( BPN ) and multiple regression analysis( MRA ). The qualitative and quantitative analyses of CO and NO2 in their mixtures has been successfully realized by using the sensor array、BPN and MRA. In the qualitative analysis, the error of network learning and training was zero. In the quantitative analysis, the error of CO and NO2 concentration was -7.5~10% and -5.0~1.3% for each other. In their mixtures, the error of CO and NO2 concentration was -17.5~5.8%、-3.3~20.0% for each other.

並列關鍵字

piezoelectric ； sensor ； CO ； NO2 ； Artificial neural network

參考文獻

4.行政院環境保護署, 空氣品質歷年趨勢分析—空氣污染物濃度趨勢圖

1.C. J. Perderson, Cyclic polyethers and their complexs with metal salts, J. Am. Chem. Soc., 89 (1967) 7017-7036.

2.B. Dietrich, J. M. Lehn, T. P. Sauvage, Les cryptates, Tetrahedron Letters, 34 (1969) 2889-2892.

10.Klaus Steiner, Ulrich Hofer, Gerd Kuhner, Gerd Sulz, Elmar Wagner, Ca- and Pt-catalysed thin-film SnO2 gas sensors for CO and CO2, Sensors and Actuators B, 24-25 (1995) 529-531.

11.Keiichi Ohta, Hisayoshi Terai, Ikuhiko Kimura, Katsuyuki Tanaka, Simultaneous determination of hydrogen, methane, and carbon monoxide in water by gas chromatography with a semiconductor detector, Anal. Chem., 71 (1999) 2697-2699.