Abstract
Liver cancer is one of the top ten causes of death in Taiwan. To efficiently increase the survival rate of liver cancer, the development of the detection technique is strongly desired. Some researches indicated that the ammonia concentration in exhaled breath for patients with liver disease was greater than 0.7 ppm, while healthy people have less than 0.3 ppm. We have developed a gas sensor based on ultrathin indium nitride (InN) to detect ammonia at sub-ppm level for diagnosing liver disease. Nevertheless, the InN gas sensor responds to not only ammonia but also other interfering gases (ex: acetone, carbon dioxide, and oxygen) existing in human breath. The interfering gases severely influence the signal captured from the InN sensor, causing difficulty in distinguishing the concentration of ammonia. To reduce the noise from the interfering gases, two methods using single sensor and sensor arrays are proposed in this study to enhance the selectivity of ammonia. For single sensor, bare InN films have the current variation ratio of 1:0.95 upon exposure to 8 ppm ammonia and 8 ppm acetone, respectively. When the InN surface was decorated with pentacene on the top, the current variation ratio remarkably reduced to 1:0.4, indicating pentacene can suppress acetone signal. However, pentacene cannot be effectively suppressed signals caused by high concentration of carbon dioxide existed in breath. To differentiate sub-ppm ammonia under the variation of high concentrations of carbon dioxide (3-5 %) and oxygen (16-18 %), the sensor arrays composed of Bare-InN and Pt-InN are proposed. The array can differentiate 0.8 ppm and 0.2 ppm ammonia using Support Vector Machine (SVM) in plane formed by different responses of the two sensors. The number of classified data correctly represents 98% of total data points captured at a response time of 500 s.