Development of High Performance Breath Acetone Sensing Device

Transcription

1 Development of High Performance Breath Acetone Sensing Device Thesis submitted in partial fulfillment of the requirements for the degree of Master of Science (by Research) in Electronics and Communications Engineering by Thati Anand Center fo VLSI and Embedded Systems Technology International Institute of Information Technology Hyderabad , INDIA April 2017

2 Copyright c Thati Anand, 2017 All Rights Reserved

3 International Institute of Information Technology Hyderabad, India CERTIFICATE It is certified that the work contained in this thesis, titled Development of High Performance Breath Acetone Sensing Device by Thati Anand, has been carried out under my supervision and is not submitted elsewhere for a degree. Date Adviser: Dr. Shubhajit Roy Chowdhury Co-adviser: Dr. Tapan Kumar Sau

4 To my Family & Friends

5 Acknowledgments My journey in IIIT-Hyderabad has been a journey worth it. It could not have been so without the support of many people. As I submit my MS thesis, I want to offer my gratitude to all those people who helped me in successfully completing this journey. First of all, I want to thank my guide Dr. Shubhajit Roy Chowdhury, for accepting me as a student and constantly guiding me. His guidance has helped me improve not only as a researcher but also as a person. I can never thank him enough for providing me support at my difficult times and helping me move forward. I would also like to thank my co-adviser Dr. Tapan Kumar Sau for his continuous support and guidance during the course of my research. And special thanks to Arunangshu for his contributions which aided my research work. I am thankful to everyone in CVEST and CCNSB for providing such a positive work environment. Many thanks to Harsh, Arpit, Vasu, Swathi, Ramakrishna and Harinath for all the help, discussions, motivation and support. Thanks to Neeraj, Gopi, anwar, Vidya, Divya, Vigneswaran, Gugan, Bhuvanan and Prateek for remaining by my side rock steady throughout this journey and for being my friends through my ups and downs. Among my friends, Arpit deserves a special mention for being by my side while I was going through some rough patches. Thanks to Raghuram, Narendra,,Amitanshu, mohan, parijat and karthik for their company, advice, help during courses, research and fun-filled moments in IIIT. Thanks to Mr. Sathish Kumar, CVEST staff, and shireesha, CCNSB staff, for their help and support. I could not have accomplished it without the support and understanding of my parents. I wish to thank my Mom, Dad, brothers Ravi and Nagesh for being my constant support and motivation. Last, but not the least, thanks to IIIT community for giving me such a beautiful campus and environment to grow. v

6 Abstract Researchers have demonstrated that breath acetone is an effective biomarker of Type 2 diabetes which is the most common form of diabetes. Diabetes treatment requires intermittent monitoring of the blood glucose levels in the patients. Conventional way of detecting glucose levels is through invasive technique which involves pricking the finger and collecting blood sample. This is not only painful and blood consuming but also time consuming and expensive. Therefore, there has been a great demand for the non-invasive techniques of blood glucose determinations in the commercial market. Researchers have been attempting to develop a number of non-invasive techniques where the glucose can be measured by different methods outside the body, without puncturing the skin or without taking the blood sample. However, out of various non-invasive techniques developed, many of them are not suitable for real-time, point-of-care, and routine uses because they involve high cost and require time-consuming and complicated sample pretreatment, large space and infrastructure for operations. Detection of breath acetone can be a rapid, noninvasive, and patient compliant viable alternative to the conventional methods of blood glucose determination. Acetone in the breath appears due to increased lipolysis. The breath acetone levels are found to be less than 0.9 ppm in healthy people and more than 1.8 ppm in diabetes patients. This has led researchers to develop alternative acetone sensors, especially semiconducting metal oxides (SMOs)-based chemoresistive sensors due to their several favorable attributes. However, achieving the target selectivity, sensitivity, fast response and recovery, and low detection limit have often been very challenging for the SMO-based sensors, especially for the sample like exhaled breath that consists of plethora of gases including a large quantity of water vapor and carbon dioxide. The thesis work involves designing an embedded system to measure blood glucose levels from breath acetone. First, it involves testing with the existing SMO-based acetone sensor followed by reducing the effects of other parameters by using Artificial neural network model. We observed that we still need improvements in various fronts of sensor design. Sensors for breath acetone monitoring must show high sensitivity and selectivity and very small response and recovery times, in addition to a good reversibility and stability. Exhalation is a very fast process ( 3 s) and a small quantity of acetone ( ppm) is present in the exhaled breath. Therefore, detection of acetone below 10 ppm is of great significance in breath acetone sensing. Towards this goal, we developed highly efficient SMO-based acetone sensors made of Pd nanoparticle-loaded nanostructured SnO 2 particles (Pd@SnO 2 ). The sensors were prepared by using simple chemical reduction and sol-gel synthesis methods with no requirements for dispersing vi

7 vii agents or complicated synthesis and fabrication steps. Our 2 sensors exhibited very high sensor response with small response and recovery time at relatively low operating temperature. Further, the sensors exhibit excellent reversibility, selectivity over ethanol and noninterference from water vapor and CO 2, two major constituents of the exhaled breath. These parameters satisfy the requirements of a real-time breath acetone sensor.

8 Contents Chapter Page 1 Introduction Importance of Breath Acetone Sensing Origin of Breath Acetone History of Acetone monitoring Glucose Monitoring Methods Traditional methods to diagnose Diabetes Invasive detection of blood glucose levels: Noninvasive detection of blood glucose levels Problem statement System Design and Testing with TGS822 Acetone sensor Breath Chamber: TGS822 gas sensor DHT11 Digital Humidity and Temperature sensor BMP180 Digital barometric pressure sensor Controller Artificial Neural Network model Acetone characteristics of TGS822 Sensor Results and Discussion Conclusions High Performance Acetone Sensor Development Introduction Sensor Fabrications and Characterizations Results and Discussion Sensitivity of the sensors Humidity and CO 2 gas testing: Response and Recovery time of sensors Conclusions Conclusion Bibliography viii

9 List of Figures Figure Page 1.1 Dextrostix, Ames Reflectance Meter, used to test blood glucose levels based on color change. courtesy: The pursuit of non invasive blood glucose The Eyetone blood glucose reflectance colorimeter. courtesy: The pursuit of non invasive blood glucose The improved version of kyoto Dai-ichi Dextrostix glucose monitoring meter with digital readout. courtesy: The pursuit of non invasive blood glucose Block Diagram of the Noon-invasive blood glucose detection system through breath acetone TGS822 sensor characteristics. Showing high sensitivity towards Acetone gas TGS822 gas sensor circuit Artificial Neural Network Block Diagram Artificial Neural Neuron Model Artificial Neural Network with 30 hidden neurons Acetone detection a). Under normal humidity b). Humidity=90% inside the chamber (a) Network with R= (b) Trained network regression R=1 (c) Network under validation R=1 (d) Network under test R= Variations in various parameters during monitoring of blood glucose levels over time: (a) actual concentration of blood glucose levels of a patient over time during invasive measurement; (b) voltage levels, (c) resistance, (d) pressure, (e) humidity and (f) temperature during non-invasive monitoring of blood glucose Electric circuit diagram for the signal measurement of the sensor XRD Images of as synthesized particles Typical SEM image showing the surface morphologies of the Pd@SnO 2 particles A typical elemental composition mapping of the l-pd@sno 2 sample determined by EDS: Sn (in green), O (in red) and Pd (in yellow) Sensitivity plot for all sensors SnO 2, h-pd@sno 2 and f-pd@sno 2 under various acetone and ethanol concentrations A Typical plot showing the sensitivity of acetone sensing Table showing the sensitivity (Slopes) of sensors prepared Typical response of the Half-coated Pd@SnO 2 sensor exposed to CO 2 gas. The sensor showed no response to CO 2 gas ix

10 x LIST OF FIGURES 3.9 Typical response of the Half-coated 2 sensor exposed to Humidity. The plot shows humidity sensor response and our sensor response. Humidity sensor showed good response while our sensor showed almost zero response to humidity response when exposed to 5ppm acetone gas. The sensor is exposed to gas and air alternately SnO 2 response when exposed to 5ppm acetone gas. The sensor is exposed to gas and air alternately Response times of SnO 2, Half-coated Pd@SnO 2 and Full-coated Pd@SnO 2 sensors when exposed to Acetone gas under different concentrations Recovery times of SnO 2, Half-coated Pd@SnO 2 and Full-coated Pd@SnO 2 sensors when exposed to Acetone gas under different concentrations Response times of SnO 2, Half-coated Pd@SnO 2 and Full-coated Pd@SnO 2 sensors when exposed to Ethanol gas under different concentrations Recovery times of SnO 2, Half-coated Pd@SnO 2 and Full-coated Pd@SnO 2 sensors when exposed to Ethanol gas under different concentrations SnO 2 sensor exposed to acetone gas concentrations ranging from 1ppm to 100ppm. Shows increased sensitivity with respect to increase in concentration Response reversibility of Pd@SnO 2 sensor under 10ppm acetone gas SnO 2 sensor exposed to ethanol gas concentrations ranging from 1ppm to 100ppm. Shows increased sensitivity with respect to increase in concentration Response reversibility of Pd@SnO 2 sensor under 10ppm ethanol gas SnO 2 sensor response of ethanol and acetone gas concentrations ranging from 1ppm to 100ppm Half-coated Pd@SnO 2 sensor response acetone gas concentrations ranging from 1ppm to 100ppm Half-coated Pd@SnO 2 sensor response of ethanol and acetone gas concentrations ranging from 1ppm to 100ppm Half-coated Pd@SnO 2 sensor response of ethanol gas concentrations ranging from 1ppm to 100ppm Full-coated Pd@SnO 2 sensor response of acetone gas concentrations ranging from 1ppm to 100ppm Full-coated Pd@SnO 2 sensor response of ethanol gas concentrations ranging from 1ppm to 100ppm Full-coated Pd@SnO 2 sensor response of ethanol and acetone gas concentrations ranging from 1ppm to 100ppm

11 List of Tables Table Page xi

12 Chapter 1 Introduction 1.1 Importance of Breath Acetone Sensing There are many volatile organic compounds (VOCs) present in human breath. Over 1000 VOCs have been detected to date in the ppmv (parts per million by volume) to pptv (parts per trillion by volume) concentrations [1]. Acetone gas is one of many such VOCs and is present in the sub-ppm range. Breath acetone shows a correlation with the blood glucose levels present in any human body. It suggests that diabetes in which the blood glucose level increases can be diagnosed from the breath acetone measurement. Breath acetone concentration increases in diabetic patients whereas concentration levels are low for normal people [2]. Conventional diabetes monitoring is not only painful but also needs blood sample and often the instruments are expensive. As the technology is moving forward in all the directions it is the time to reinvent the diabetes monitoring instruments to make it more advanced, pain-free, and user friendly. Monitoring acetone in the breath appears to be a great step in the diabetes diagnosis. 1.2 Origin of Breath Acetone Liver produces three ketone bodies: Acetone (C 3 H 6 O), acetoacetate (AcAc, C 4 H 6 O 3 and 3-βhydroxybutyrate (3BH). Acetone is produced from these sources in the body: one is from the decarboxylation of acetoacetate and the other is from dehydrogenation of isopropanol [3]. The acetone conversion process through acetoacetate is: CH 3 COCH 2 COO+H + CH 3 COCH 2 COOH CH 3 COCH 3 +CO 2 In fasting subject s acetoacetate is observed to be 37% and in diabetic patients it is 52% [4, 5]. Acetone enters to lungs and it passes through exhaled breath, it also eliminated through urine from lungs [6]. Reichard et al pointed out that breath and urinary excretion of acetone accounts for a 2-30% of its endogenous production rate, and the in vivo metabolism accounts for the rest [4]. 1

13 1.3 History of Acetone monitoring Before semiconductor sensors came into picture acetone was measured by using conventional chemical experiments to a number of sophisticated techniques. For example, flame ionization selected ion flow tube mass spectrometry (SIFT-MS) is the one such technique that was used for the accurate determination of the acetone concentration in breath. For this test it is required to collect breath sample in a bag and send the bag to laboratory for analysis [7]. There are other methods also to detect acetone from breath: gas chromatography [8],ion mobility spectrometry [9], cavity ringdown spectroscopy [10]. All these methods requires a bag to collect the breath and send it to laboratory for analysis. Further, often they involve high cost, time-consuming complicated sample preparations, etc. Therefore, they are not suitable for real-time, point-of-care routine uses. 1.4 Glucose Monitoring Methods Diabetes mellitus is a major health problem worldwide [11]. This health condition arises from many complex metabolic disorders leading to abnormal glucose levels in a person [7]. Diabetes is a condition in which the body s natural control of blood sugar has been lost. Whether it is termed type 1 (previously known as juvenile-onset ), type 2 ( adult-onset ), or the gestational diabetes that is a complication of pregnancy, the end result is the same - glucose may be present in the blood in dangerously low, hypoglycemia,or high, hyperglycemia,amounts. Without a means of measuring glucose, treatment is a dangerous guessing game of taking pills, injecting insulin, or deciding how much and what kind of food to eat [12 14]. Insulin is the principal hormone that regulates the uptake of glucose from the blood into most cells of the body, especially liver, muscle, and adipose tissue. Therefore, deficiency of insulin or the insensitivity of its receptors plays a central role in all forms of diabetes mellitus [15]. The classical signal/symptoms of diabetes are weight loss, increased hunger (polyphegia), increased urination (polyuria) and increased thirst (polydipsia) [16]. These symptoms may develop rapidly in type-1 diabetes whereas in type-2 it is very slow. In addition it may cause kidney failure, cardiovascular diseases, blurry vision, headache, fatigue and slow healing of wounds. Prolonged high glucose levels can cause glucose absorption in the lens of eyes, which leads to blurry vision due to the change in lens shape. The symptoms of hypoglycemia, low blood sugar, are sweating, trembling, feeling uneasy and serious issues such as confusion, changes in behavior and (rarely) permanent brain damage or death in severe cases. Mild to moderate cases are self treated by taking extra sugars. Severe cases can lead to unconsciousness which should be treated with injecting glucagon. If patients can measure their glucose levels, they can treat the diabetes easily. Diabetes patients can lead a relatively normal life if they can maintain a proper diet, exercise, meditation and frequent glucose measurements. In both - hyperglycemia and hypoglycemia - cases, it is required to measure patients 2

14 blood glucose levels before taking medication in order to avoid further complications. This can be possible if painless, inexpensive, portable and reliable testing instruments are available to them Traditional methods to diagnose Diabetes The disease has been known since ancient times. Since high blood glucose levels will cause the kidneys to deposit glucose into urine, in early days Chinese started diagnosing diabetes by seeing if the ants are attracted to the sugars in patient s urine [17]. The process of diagnosing diabetes this way was continued for a century and later chemical techniques were introduced. In 1941, the Ames Division of Miles Laboratories, in Elkhart, Indiana, introduced Clinitest tablets. The tablet based on a standard test for certain sugars involving copper sulfate, called Benedict s solution. These Clinitest tablets can be added to a few drops of urine. This will change the color from bright blue to orange and based on the instruction sheet approximate glucose level can be estimated from urine sample [18]. However, diagnosing diabetes from urine test has some serious issues. When a person first develops diabetes, the level of glucose in urine is a reasonable indication of excessive amounts in the blood; however, because both normal and low blood glucose levels result in no glucose in urine, it is never possible to assess those blood levels by using urine tests. As the disease progresses over time, it becomes much less reliable as a marker of high blood glucose. Even early on, its never an accurate measure, and even though improved testing devices (dipsticks) have been developed over the years, its never been more than a semi-quantitative test [19]. To get accurate values, its necessary to measure the amount of glucose in the blood itself, and this has long been done in doctors offices and laboratories. However, for people with diabetes to maintain healthy levels of glucose, there has always been a need for simple, accurate tests they could perform at home [19] Invasive detection of blood glucose levels: In 1964, after developing many dipstick tests for urine, Ernest Adams of Ames developed a practical test strip for measuring glucose in blood named Dextrostix, after dextrose, another name for glucose [20,21]. Instead of using a chemical reaction to measure glucose, as Clinitest had done, Dextrostix used a biochemical reaction with an enzyme called glucose oxidase, which reacted with glucose to produce hydrogen peroxide. The hydrogen peroxide produced a color from another chemical called o-tolidine, and the amount of color on the strip after exposing it to a drop of blood was a good measure of the amount of glucose present [22]. At first, the strength of color developed was simply compared to a series of printed colors on the label, and the glucose concentration was estimated by the color comparison. The procedure was not trivial but could be mastered by people with reasonable dexterity for home use. 3

15 Figure 1.1: Dextrostix, Ames Reflectance Meter, used to test blood glucose levels based on color change. courtesy: The pursuit of non invasive blood glucose The major limitation to this approach, aside from the timing and manipulation involved, is that visual acuity and the ability to perceive color accurately decrease with age [23].The Dextrostix as shown in the Figure 1.1is the first, developed at Ames by Anton Clemens, was called the Ames Reflectance Meter, or A.R.M. According to interviews with Clements, he was ordered to drop the project several times but somehow managed to bring it to the market, and the first electronic blood glucose device could be purchased in about 1970 for about $400. Unfortunately, it had some reliability problems, mostly from its rechargeable lead-acid batteries, and its use didnt become widespread [24]. Figure 1.2: The Eyetone blood glucose reflectance colorimeter. courtesy: The pursuit of non invasive blood glucose 4

16 The next electronic strip reader to appear was in about 1972, called the Eyetone,shown in the Figure 1.2, and was manufactured by a Japanese company, Kyoto Dai-ichi (which later changed the company name to Ark-Ray). It also read Dextrostix, but used a plug-in AC adapter for power instead of batteries. Albeit the increase in performance over Dextrostix, it has a higher standard deviation so that a baby with hyperglycemia might be missed by Eyetone glucose measurement [25]. Figure 1.3 shows the improved version of Dextrostix meter.in about 1979, Kyoto Dai-ichi introduced an improved Dextrostix meter with a digital readout, called the Dextrometer. Figure 1.3: The improved version of kyoto Dai-ichi Dextrostix glucose monitoring meter with digital readout. courtesy: The pursuit of non invasive blood glucose After that there came many instruments to test although they failed at later stages. People came up with the strips and digital meter to monitor the glucose levels. While glucose monitors and strips were once a very profitable business, the market contraction since about 2005 has resulted in less research, reduced sales forces and more intense price competition among the established companies. While it is possible that this trend will also reduce the emphasis on a noninvasive monitoring solution, there is no indication of a slowdown among inventors trying to provide novel approaches to solve the problem, even if the big players appear to be less receptive to ideas presented by these inventors Noninvasive detection of blood glucose levels As discussed in the invasive method, glucose level is determined from a small volume of blood sample collected by finger pricking. Though the test may not pose any risk to a healthy adult who goes for the diabetes checkup in every 2 to 3 months, but it is very painful to the diabetic patients because 5

17 every time they have to prick the finger. The current invasive method is based on the enzymatic catalysis principle where a thin needle is used to prick the finger of the patient to minimize the discomfort [26]. To avoid such painful diagnosis, extensive research has been devoted towards developing non-invasive techniques that measure blood glucose levels without taking the blood sample [27 32]. Non-invasive diagnosis technique is becoming more prominent in diagnosing diseases due to their pain free and simple monitoring methods. Non-invasive detection of blood hemoglobin was already reported by our group in the earlier work [33]. Lieschnegg et al. have developed a sensor to detect failures and material imperfections in total joint prosthesis based on acceleration measurement non-invasively [34]. Diabetes can also be detected using non-invasive methods. Frequent testing and accurate determination of glucose levels is essential for diagnosis, effective management and treatment of diabetes mellitus. Therefore, there have been constant efforts to develop efficient and sensitive techniques for the determination of blood glucose levels. As mentioned above, a number of invasive enzymatic and non-enzymatic methods and systems have been reported for the detection of glucose [35 38]. Luaibi et al. used nuclear magnetic resonance technique to measure the blood glucose levels noninvasively [27]. Other non-invasive techniques used are electrical impedance, NIR spectroscopy, breath analysis, ultrasound and thermal spectroscopy [36 38]. However, none of these methods seems to achieve the desired accuracy due to varying environmental conditions and physical movements and therefore none of them led to any accurate and safe commercial device. Further, compared to the breath analyzer other techniques appear to be expensive due to the sensor components involved. This has led researchers to develop alternative acetone sensors, especially a number of semiconducting metal oxides (SMOs)-based sensors. 1.5 Problem statement The thesis work reports non-invasive systems for the determination of blood glucose levels from the detection of the breath acetone. We have seen earlier that acetone is one of the volatile organic compounds (VOCs) present in the exhaled breath [39, 40]. The acetone present in the exhaled breath is a metabolic product of the body-fat burning [39]. The breakdown of excess acetyl-coa from fatty acid metabolism in diabetic patients leads to increase in the levels of acetone in the blood and finally in the exhaled breath. Therefore, the breath acetone levels could be a measure of the blood glucose levels of a person [41]. The breath acetone concentration ranging from 1.7 ppm to 3.7 ppm can be detected in diabetic patients [2, 42], whereas it varies between 0.3 and 0.9 ppm for healthy humans [43]. Optical fiber made of nanostructured films has been used for the detection of VOCs [44]. A number of metal oxide nanomaterials have shown excellent gas sensing properties [45, 46] because large surface area and nanoscale surface features of such nanomaterials result in increased sensitivity towards gases [47]. With the newly emerging semiconductor technologies it is possible to design and fabricate nanoparticle 6

18 gas sensors that can detect sub-ppm VOC concentration in the breath [48]. However, in addition to the quantity of the gas, natures of both the gas and the metal oxide determine the degree of response of the sensor [49]. Tin oxide (SnO 2 ) based sensor have been widely used in the gas sensing applications due to their unique gas detection properties [50]. Here we have used the SnO 2 particle system as acetone sensors. The resistance of this SMO sensor varies depending on the quantity of acetone present and can be detected by potential divider circuit [51]. Several other parameters such as the temperature, humidity, and pressure have been taken into consideration in the estimation of acetone because these parameters affect the quantity of acetone sensed by the sensor. We have used the artificial neural network to calculate the glucose levels from breath acetone measurements. We have fabricated new Pd loaded SnO 2 particle based systems as high performance acetone sensors.the thesis is organized as follows: First, we discuss the use of the commercial TGS SnO 2 particle system as the acetone sensor. Second, we explored the acetone sensing performance of three SMO-based chemiresistive acetone sensor materials, namely pristine SnO 2 and two Pd nanoparticle-loaded nanostructured SnO 2 particles (Pd@SnO 2 ) prepared in our laboratory. Finally, we compared the performances of the commercial and laboratory made sensors and conclude the work. 7

19 Chapter 2 System Design and Testing with TGS822 Acetone sensor Tin oxide (SnO 2 ) based sensors have been widely used in the gas sensing applications due to their unique gas detection properties. First, we discuss the use of the commercial TGS SnO 2 particle system as the acetone sensor. The resistance of this sensor varies depending on the quantity of acetone present and can be detected by potential divider circuit. Several other parameters such as the breath chamber temperature, humidity and pressure have been taken into consideration in the estimation of acetone because these parameters affect the quantity of acetone sensed by the sensor. Further, the acetone concentration in the breath chamber will not be the same for the same person every time due to different flow rates, because all humans cant blow at the same rate into the mouth piece/breath chamber. This can result in wrong diagnosis. Therefore, we specified the time duration for breathing into the mouth piece. Finally, we have used the artificial neural network to calculate the glucose levels. As mentioned earlier, non-invasive detection of blood glucose levels from breath involves monitoring of several other parameters other than acetone itself. The working prototype of the system is designed to monitor, record and compute the parameters. The Figure 2.1 represents the block diagram of the Non-invasive glucose detection system through breath acetone. The system is divided into three parts: 1. Breath Chamber 2. Controller 3. Artificial Neural Network 2.1 Breath Chamber: A 215 cm 3 mouth piece was designed and three sensors were placed inside the mouth piece for analysis. Due to the presence of sensors inside the mouth piece the volume of this test chamber is taken to be approximately 200 cm 3. For analysis of glucose levels from the breath, we considered a total of five parameters from three different sensors: Voltage and resistance from acetone sensor (Figaro TGS822), pressure from Digital barometric pressure sensor (BMP180) and temperature & humidity from DHT11 sensor. 8

20 Figure 2.1: Block Diagram of the Noon-invasive blood glucose detection system through breath acetone a. TSG822 gas sensor b. DHT11 Digital Humidity and Temperature sensor c. BMP180 Digital barometric pressure sensor TGS822 gas sensor The TGS822 is gas sensor which has a high sensitivity towards acetone gas and minimally sensitive to other gases like ethanol, Benzene, n-hexane, Isobutene, CO and methane. The sensing element of the figaro TGS822 sensor has very less conductivity at clean air and the conductivity increases in the presence of a detectable gas depending on the concentration. The acetone sensing application is used when the ethanol is absent in the breath. The acetone sensing relies on the changes in electrical conductivity due to the change in the sensor surface arising from the reactions between ionosorbed surface oxygen and acetone gas. In the presence of a deoxidizing gas, the surface density of the negatively charged oxygen decreases. This results in the decrease of the barrier height in the grain boundary and hence decrease of sensor resistance. Gas characteristics of the TGS822 sensor: Features of Figaro TGS822 sensor: 1. High sensitive to organic solvent vapors such as acetone 2. High stability and reliability over a long period 3. Long life and simple circuitry 9

21 Figure 2.2: TGS822 sensor characteristics. Showing high sensitivity towards Acetone gas From the Figure 2.2 it is clearly visible that the temperature is effecting the sensitivity of the sensor. And also increasing humidity also effects the sensor s sensitivity towards the gases. Figure 2.3: TGS822 gas sensor circuit V C - Sensor input voltage, V H - Micro heater voltage, R L - Load resistor. V C = 5 V; V H = 5 V; R L = 100 K Ω V RL = V c R L R L + R S (2.1) 10

22 R s = ( V C V RL 1) R L (2.2) Voltage VRL and RS are calculated from the above Figure 2 for analysis. The sensor accuracy, especially at low analyte concentrations, shows nonlinear characteristics of the response and depends on the temperature and relative humidity in addition to the nature of the gas [49]. As mentioned earlier, the test cant be conducted when alcohol vapor is present in the breath. Lee et al [52] have studied the fabrication and characteristics of SnO 2 gas sensor array for many VOCs. A voltage detecting method was used to calculate the sensitivity (S) of the sensor: S = R air R gas R air 100 (2.3) where, R air and R gas are the sensor resistances in normal air and under gas [32] DHT11 Digital Humidity and Temperature sensor Flow rate and volume of breath blow into the mouth piece cant be controlled and are different for each person. To compensate this, the effects of pressure, temperature and humidity levels have been considered for each and every person apart from the actual parameters (voltage and resistance) detected from the acetone sensor. DHT11 is a Digital Temperature and Humidity sensor from micropik. It measures humidity, ranges from 20-90% with a resolution of 1temperature from 0-50 o C with a resolution of 1 at accuracy of 2 o C. The sensor is operated at 5VDC supply. DHT11 sensor sends 40bit data on a single data-line which includes 16 bit Relative Humidity (8bit integer RH data + 8bit decimal RH data), 16 bit Temperature (8bit integer temperature data + 8bit decimal temperature data) and 8bit checksum. The 40bit single data-line is connected to the microcontroller board to read the Relative Humidity and Temperatures BMP180 Digital barometric pressure sensor BMP180 is a Barometric pressure sensor from BOSCH Sensortec. It measures atmospheric pressure ranges from hpa (hectopascal) with a resolution of 0.01 hpa under 0-65 o C temperature conditions. The sensor is operated at 3.3V DC supply. The data is transferred via I 2 C bus. Sensor is connected to the microcontroller board via SCL and SDA lines to read data from I 2 C bus. Microcontroller board reads the data from all the three sensors and sends back to the artificial neural network model via serial port. 2.2 Controller All the three sensors data were given to the controller board (Arduino Uno R3). I2C communication protocol is implemented in the controller to communicate with digital sensors (DHT11 and BMP180) to read data. The read data is converted to standard values and transferred serially to the attached computer 11

23 where MATLAB program is running. MATLAB is used to plot the real-time values detected from all the three sensors for mentioned duration of time. And it also saves all parameters for further analysis. 2.3 Artificial Neural Network model Breath samples have been collected from 30 persons. Each person was advised to blow into the mouth piece for 5 seconds continuously with the same flow rate. Sensor data was recorded on MATLAB tool. In each of five parameters maximum and minimum values were taken in the specified time duration. Invasive test was also performed with ACCUCHECK instrument immediately after taking the breath test. To analyze and get the relation between the recorded parameters and glucose levels Neural Network Tool in MATLAB has been used. Figure 2.4 shows the modeled neural network block diagram. The network is selected to have 30 hidden layer neurons and 1 output layer neuron with 5 inputs and a single output. In hidden layer, the neurons have the sigmoid transfer function at the output, and in the output layer neurons have a linear function at the output. Figure 2.4: Artificial Neural Network Block Diagram A single neuron model is shown in Figure 2.5 with a sigmoid transfer function at the output. Here the transfer function is Output of neuron is given to the sigmoid function G j = Σ i x i W ij + Bias (2.4) Out = Sigm(G) = 2 1 (2.5) 1 + e 2G For a hidden layer neuron, output = G which is the same as the sum of all the inputs to the neuron. Figure 5 shows the complete neural network model with 5 inputs and 30 hidden layer neurons and one output layer neuron. 30 samples of each 5 elements (V, R, P, H and T) have been given as the input matrix to the network with a target output array of 30 Glucose samples, are given below 12

24 Figure 2.5: Artificial Neural Neuron Model V 1 V 2 V 3... V 30 R1 R2 R3... R30 Input = P 1 P 2 P 3... P 30 H1 H2 H3... H30 T 1 T 2 T 3... T 30 Input = [ ] G1 G2 G3... G Acetone characteristics of TGS822 Sensor Acetone concentration detection is done in 200 cm 3 test chamber. To create required concentration inside the chamber we injected liquid acetone into the chamber. Acetone has a molecular weight of MW=50.08 g/mol and density δ= g/cm3. The gas density of acetone is given by [33] δ = P MW R T (2.6) Where, δ = the density of the gas of acetone in g/l, P = the standard Atmospheric Pressure (1 atm) 13

25 Figure 2.6: Artificial Neural Network with 30 hidden neurons MW= Molecular Weight in g/mol R= Universal gas constant in atm/mol.k (equal to atm/mol.k), T= temperature in Kelvin. Thus, one gets 2.36g/L gas density for acetone. 100 ppm acetone stock solution was prepared for testing. For 1 ppm acetone gas concentration inside the 200 cm 3 test chamber required volume is where x = 2 cm 3 V gas = x (2.7) Mass = V gas δ = V liquid ρ (2.8) 14

26 where V gas is the volume occupied by the gas of acetone which is equal to 2 cm 3, δ is the density of the gas of acetone as calculated before, ρ is the constant density of liquid acetone. Therefore, V liquid = V gas δ ρ (2.9) From the eq. (9) one can calculate the liquid acetone required. Thus one gets V liquid = 6 µ L for 2cm 3 acetone gas that gives 1 ppm acetone concentration in the mouth piece. Therefore, for a concentration of n ppm acetone inside the chamber n*6 µl liquid acetone is required Results and Discussion Acetone concentrations have been studied in 1ppm - 10ppm levels with and without humidity effect. First, Figaro TGS822 acetone sensor is tested under normal atmospheric conditions. Humidity effect has been calculated for acetone sensor under 85-90chamber. The test results are shown in Figure 6. It is observed that there is an increase of 0.2V in high humidity sensing compared to the low humidity sensing. Artificial neural network was used to compensate these effects. The network Voltage variation in Volts without humidity with humidity Acetone Concentration in PPM Figure 2.7: Acetone detection a). Under normal humidity b). Humidity=90% inside the chamber 15

27 is trained such that the MSE (Mean Square Error) is very low and Regression reaches 1. The trained network knowledge is stored in terms of weights and biases. The network is fixed with 180 I/O layer weights and 31 neuron biases. The trained network MSE is 2.75 x e 27 and the total regression ratio is as shown in Figure 7. The relation between the acetone concentration and sensor response with humidity effect is already explained and is modeled in ANN tool [14]. From the collected samples it is observed that different values of voltage, resistance, pressure, humidity and temperature have been recorded for every person depending on their glucose levels and flow rate. (a) (b) (c) (d) Figure 2.8: (a) Network with R= (b) Trained network regression R=1 (c) Network under validation R=1 (d) Network under test R=1 Figure 2.8 shows the network trained with regression coefficient R=1. Validation and Test results also show regression coefficient R=1 as shown in Figures 2.8 (c) and 2.8 (d). After training the network 16

28 it has been tested with different breath inputs and the results compare closely with the actual glucose levels. From the results it is observed that the voltage and resistance are playing important roles in measuring the glucose level non-invasively whereas the other parameters affect the measurements very little. The temperature effect is more compared to other two parameters, the relative humidity and pressure. The training data is collected from non-diabetic patients with glucose levels in between 80 mg/dl and 140 mg/dl and a few pre-diabetic patients with glucose levels in between 140 mg/dl and 180 mg/dl. Figure11 (a) shows the variation in concentration of actual blood glucose levels of a patient over time. Figures 11b to 11f show the variation in different parameters during the non-invasive monitoring of blood glucose over time. From the Figures 2.9 (a) to (f) it can be observed that if the glucose levels are high at some point the voltage will also be high and the resistance will go down and the other parameters are moderate. In some cases, albeit the glucose levels are high, voltage levels are low because of considerably low pressure, humidity and temperature levels, which indicate the person did not blow correctly into the mouth piece. These effects are minimized with the help of neural network tool Conclusions In this study, the applicability of the breath acetone sensing method to the determination of glucose in human blood is demonstrated. We used acetone sensor for monitoring acetone levels in the exhaled breath and compared with actual blood glucose levels. We also considered the effects of the pressure, temperature and humidity parameters on the acetone sensing. This test involved studies of non-diabetic and pre-diabetic persons. We have also used Artificial Neural Network model for analyzing the data. The test results show that it is possible to measure the blood glucose levels via breath acetone sensing. The accuracy of the system can be improved with a large set of data. 17

29 (a) (b) Figure 2.9: Variations in various parameters during monitoring of blood glucose levels over time: (a) actual concentration of blood glucose levels of a patient over time during invasive measurement; (b) voltage levels, (c) resistance, (d) pressure, (e) humidity and (f) temperature during non-invasive monitoring of blood glucose. 18

30 Chapter 3 High Performance Acetone Sensor Development 3.1 Introduction Developing efficient sensors for the detection of low concentrations of various gases such as alcohol, acetone, etc. has drawn a lot of attention due to their uses in the fields of biomedicines and public safety. We have mentioned earlier that the detection and measurement of breath acetone can help in the treatment and management of diabetes mellitus and drug-resistant epilepsy in children [3, 53, 54]. Breath acetone is an effective biomarker of Type 2 diabetes [55]. As mentioned earlier, diabetes is a metabolic problem that leads to the rise in blood glucose (sugar) levels in a person. Blood glucose levels are often measured by collecting blood via painful skin pricking. In order to avoid such painful testing, a number of non-invasive methods for the determination of blood sugar level have been reported [56 58]. Breath acetone sensing appears to be one of the promising non-invasive methods of blood sugar level measurements [58, 59].Acetone in the breath appears due to increased lipolysis [3, 60].The breath acetone levels are found to be less than 0.9 ppm in healthy people and more than 1.8 ppm in diabetes patients [61]. A number of research demonstrated that the breath acetone concentration correlated with the human blood sugar levels [3, 53, 58, 62]. Thus, detection of the breath acetone can be a rapid, noninvasive, and viable alternative to the conventional methods of blood glucose determination. Further, breath acetone had also been shown to be a measure of mild to moderate systemic ketosis [54, 63]. Ketogenic diet is prescribed for the treatment for the drug-resistant epilepsy. The treatment of both the diabetes and the drug-resistant epilepsy require intermittent blood sampling which is not only time consuming and expensive but also painful and blood consuming, especially for pediatric patients. Breath acetone-based noninvasive measurements would allow frequent analyses without the requirement of any blood sampling. This has paramount importance in the daily implementation of the diet, search for better diets, etc. A few sophisticated techniques like gas chromatography with flame ionization, selected ion flow tube mass spectrometry, etc. has been used for the accurate determination of the breath acetone. However, in addition to their high cost, this kind of instruments requires time-consuming, complicated sample pretreatment and large space and infrastructure for operations. Therefore, they are not suitable for real- 19

31 time, point-of-care, and routine uses. This has led researchers to develop alternative acetone sensors, especially a number of semiconducting metal oxides (SMOs)-based sensors [64 80]. Semiconducting metal oxides (SMOs)-based sensors have been emerging as one of the most reliable, efcient, and costeffective sensors. A number of SMOs such as WO 3, SnO 2, ZnO, Fe 2 O 3, Fe 3 O 4, Co 3 O 4, etc. have been used for acetone sensing [68,73 82].We have chosen nanostructured SnO 2 for our present studies. SnO 2 is an inexpensive, nontoxic, and highly stable n-type SMO. It can be obtained by various fabrication techniques. SnO 2 has been extensively used to detect a number of gases and organic volatile compounds. SnO 2 has shown excellent sensitivity and response reversibility toward several target gases [83, 84] Despite various favorable attributes of the SMO-based sensors, often additional treatments and improvements are required for target selectivity, enhanced response and recovery, and low detection limit. SMOs often exhibit low sensitivity towards the target gases under ambient temperature conditions. Therefore, SMOs require to be heated to high temperatures ( o C) to show the desired response and recovery properties. The selectivity of the target gases poses another major problem due to the similar redox behavior of many gases toward an SMO. For example, the exhaled breath consists of plethora of gases including a large quantity of water vapor [67, 85]. Water vapor, like acetone, is a reducing gas. The presence of water vapor poses severe challenges toward the breath acetone determination by using SMO-based sensors. SnO 2, an n-type semiconductor metal oxide, show a drop in the resistance upon interacting with reducing gases like acetone, water vapor, etc. Various efforts have been made to overcome such limitations and improve the gas sensing performance of the SMO-based sensors. For example, SMOs have been prepared in the nanostructural and nanoparticulate forms for enhancing the efficiency of their gas sensing properties. Nanoparticles offer larger surface areas, plenty of active sites, and higher reactivity for the gas adsorption [86]. The other strategies have been doping and surface modification by various additives, like metals, metal oxides, etc. [87 90, 90 93]. These additives enhance the SMO surface reactions of the target gases thereby significantly enhancing the response of the SMOs. We would like to point out here that though we have witnessed a great progress, we still need improvements in various fronts of sensor design. Often the sensor preparations involve complicated synthesis and fabrication steps and the sensors require high operating temperatures for a good response. Further, sensors for breath acetone monitoring must show high sensitivity and selectivity and very small response and recovery times, in addition to a good reversibility and stability. Exhalation is a very fast process ( 3 s) and a small quantity of acetone ( ppm) is present in the exhaled breath. Therefore, detection of acetone below 10 ppm is of great significance in breath acetone sensing. Here, we report a highly efficient SMO-based chemiresistive acetone sensor made of Pd nanoparticle-loaded nanostructured SnO 2 particles (Pd@SnO 2 ). The sensors were prepared by using simple chemical reduction and sol-gel synthesis methods with no requirements for dispersing agents or complicated fabrications. Our Pd@SnO 2 sensors exhibit a sensor response of with a < 2s response and 8s recovery time at 2 ppm acetone (approx. onset concentration of diabetic zone) at 200 o C operating temperature. These parameters satisfy the requirements of a real-time breath acetone 20

32 sensor. Further, the sensors exhibit excellent reversibility, selectivity over ethanol and noninterference from water vapor and CO2, the two major constituents of the exhaled breath. 3.2 Sensor Fabrications and Characterizations We have prepared three SMO-based chemiresistive acetone sensor materials, namely pristine SnO 2 and two Pd nanoparticle-loaded nanostructured SnO 2 particles (Pd@SnO 2 ). The SnO 2 particles were prepared by facile sol-gel methods from stannous chloride (SnCl 2.2H 2 O) precursor Pd particles were generated by the sodium borohydride (NaBH 4 ) reduction of potassium tetrachloropalladate (K 2 PdCl 4 ) salt. In order to see the effects of Pd contents on the acetone sensing behavior of our sensors elements, we prepared Pd@SnO 2 sensors with two different Pd concentrations, namely 0.5 mol% and 1.0 mol% (l-pd@sno 2 (half-pd doped) and h-pd@sno 2 (full-pd doped)), respectively. Ma et al. reported that the amount of Pd loading onto SnO 2 particles reached a saturation level at 0.7 mol% precursor concentration for similar sized tin oxide particles [94, 95]. In fact, there was very little change in the Pd loading after 0.5 mol% of precursor concentration. Therefore, we have chosen two Pd concentrations, namely 0.5 mol% and 1.0 mol%, a little below and above the saturation points in order to verify the effects of Pd amount on the sensor properties. The SnO 2 and Pd@SnO 2 particles were characterized by using a scanning electron microscopy (SEM) (Carl Zeiss ultra 55) showed in Figure 3.3 equipped with energy dispersive X-ray spectroscopy (EDS) showed in Figure 3.4, and X-ray diffraction (XRD) showed in Figure 3.2. The sensor was designed on a ceramic tube with preinstalled gold electrodes on it and platinum wires attached to it. We had taken 20 mg of SnO 2 or Pd@SnO 2 particles as required and coated on the ceramic tube. The sensor element construction was done according to ref [96]. The sensor element was kept under room temperature for a while until it is dried. After that the sensor was placed in the circuit for testing. We used nichrome wire as a heating element that generates 200 o C at 5V/0.32A. 200 o C falls at the relatively low temperature end of the optimal operating temperature range used for most of the acetone and ethanol sensing experiments. This heating offers enhanced sensitivity and performance to the sensor elements via inter-particle sintering and enhanced transducing and surface catalysis [97]. Sensor was tested with acetone vapor in the range of ppm concentrations. Ethanol, moisture (upto 90%), and CO 2 testing were also performed to check the interference of the common gases. The relative humidity (RH) was 40-50% while performing experiments. 3.3 Results and Discussion Metal oxide semiconductors have been intensively used as solid-state sensor materials for detecting various hazardous, flammable, and pollutant gases. In addition to the chemical compositions and intrinsic structural features, materials properties such as size, shape, and morphological characteristics of 21

33 Figure 3.1: Electric circuit diagram for the signal measurement of the sensor. the MOS play important roles in determining their gas sensing properties. Therefore, various synthesis strategies such as sol-gel, solvothermal, hydrothermal, spray pyrolysis, etc. have been employed for the preparation of the SMOs. We employed the facile sol-gel method to prepare nanostructured SnO 2 particles. Materials in the nanoparticulate forms can be a smart choice as chemiresistive sensors due to their large surface area to volume ratios, plenty of active sites, and high surface reactivity [98]. In the next step, Pd nanoparticles were directly deposited onto the nanostructured SnO 2 particles by the reduction of PdCl 4 2 ions in the dispersion of SnO 2 particles. PdCl 4 2 complexes were supposed to bind with surface hydroxyls of SnO 2 particles in solution which were then reduced by the borohydride ions forming Pd nanoparticles on the nanostructured SnO 2 particles. This protocol is suitable for the large-scale, rapid preparation of nanoparticle-loaded SnO 2 particles. There are a few metastable and stable tin oxides formed during the chemical synthesis. We carried out XRD studies in order to know the identity of the tin oxides formed by our chemical synthesis methods. The XRD pattern shown in Figure 3.2 indicates that the particles are mainly made of SnO 2 particles along with some other tin oxides. Measurements based on the full width at half maximum of the diffraction peaks in the Debye-Scherrer equation gave an average crsytallite size of 10 nm which is in agreement with the SEM data given below. Figure 3.3 shows the SEM images of pristine SnO 2 and Pd@SnO 2 particles. One should note that Pd particles are hardly distinguishable from the background of nanostructured SnO 2. The lack in contrast between Pd and SnO 2 phases is expected because there is a very little difference in their atomic numbers ( 46 Pd and 50 Sn) [99]. However, clear nanostructured surface roughness was visible in all the samples. The roughness appears to result from the aggregation of several smaller nanoparticles created by the aggregation/deposition of the nanoparticles. The particle size range from 10 nm to 25 nm. Though the lack in phase contrast in SEM images could not confirm the presence of Pd particles on the SnO 2 background, the electron energy dispersive x-ray spectroscopy (EDS) studies clearly showed it. Sn, O, and Pd elemental composition analysis and mapping of the samples were done by EDS studies. 22

34 Figure 3.2: XRD Images of as synthesized particles Figure 3.3: Typical SEM image showing the surface morphologies of the 2 particles 23

35 Figure 3.4: A typical elemental composition mapping of the l-pd@sno 2 sample determined by EDS: Sn (in green), O (in red) and Pd (in yellow). The elemental mapping Figure 3.4 shows that Pd nanoparticles were successfully and almost uniformly distributed on the surfaces of the SnO 2 particles. Rough surface nanostructures and evenly distributed Pd nanoparticles multiply the accessible active sites and interactive surface area of the sensor materials for the adsorption and diffusion of the target gas molecules; thereby enhancing the sensitivity and response rate of the sensors. It is interesting to compare the atom% compositions of the elements obtained from the EDS analyses. For all the Pd@SnO 2 samples, Pd mol% determined from EDS studies were always higher than that expected from the calculations based on the precursor amounts. The analyses give O:Sn atom% =2.56 for pristine SnO 2 particles. Further, the O:Sn atom% ratio of the l-pd@sno 2 >> h-pd@sno 2 pristine SnO 2 particles. The high oxygen quantity might arise due to adsorptions of oxygen species from ambience. In the following, we shall explore the main performance parameters, viz., the sensitivity, rate of response, selectivity, and signal-to-noise ratio limit of detection (LOD), of the sensors elements. The sensitivity was determined by comparing the test signals of the sensor element before and after exposure to the target gas. SnO 2 is a wide band-gap (3.6 ev) n-type semiconductor metal oxide (SMO). Typically n-type semiconductor metal oxides show a drop in the resistance upon interacting with reducing gases like acetone, water vapor, etc. When an n-type SMO particle is exposed to air, atmospheric oxygen adsorbs onto the surface of the SMO particles and captures electrons from the conduction band of the SMO producing oxygen ions, O (ads), O 2 (ads), etc. Such surface reactions lead to a decrease in the carrier concentration (and hence increase in electrical resistance) due to the formation of depletion layers on the surface of the SMO particles. The reducing target gases like acetone, ethanol, moisture, etc., upon exposure to the SMO particles react with the surface-adsorbed oxygen species. Acetone and ethanol are oxidized to CO 2 and H 2 O upon reaction with the surface-adsorbed oxygen species. These reactions release the electrons captured by the oxygen species to the SMO thereby decreasing the resistance. Since the value of the resistance decreased upon exposure to the reducing gases, the sensitivity, S, was determined from the relation S (%) = (R a - R g )/R g x 100, 24

36 where R a is the sensor resistance in air (original signal or base resistance) and R g is the resistance in the mixture of the target gas and air. Figure 3.5 shows the dependence of the sensitivities of the pristine SnO 2 and two Pd@SnO 2 sensor elements/ on the quantities of acetone and ethanol vapors in the range of 1 to 100 ppm. As discussed above, ethanol can be a serious interfering agent to acetone sensing. Therefore, we have also examined its sensor response along with acetone. It is clear from the figure that the sensitivity of all the sensor elements increased with the increase in the concentrations of both acetone and ethanol vapors. In the case of pristine SnO 2, sensitivities of both acetone and ethanol sensing were almost similar. The sensitivity increased dramatically upon loading of Pd nanoparticles onto the SnO 2 particles. However, the increase in the sensitivity was observed to be prominently higher in the case of acetone than that of ethanol. An increase in the Pd nanoparticle loading showed an increase in the sensitivity Sensitivity of the sensors Our sensor elements exhibited an almost linear, remarkably sharp increase in the sensitivity up to 10 ppm and then the increase was slow which appeared to be almost marginal in the case of ethanol. It is worth noting that even at concentration as low as 1 ppm level of acetone, our sensor elements showed a sensitivity of magnitude which appears to be very promising from the application point of view. Further, the sensitivity increases up to 65 when acetone concentration changes from 1 to 5 ppm only as compared to 15 to 30 in the case of ethanol. 1 to 10 ppm is an important concentration range in the acetone detection because a healthy human and a diabetic patient show respectively <0.9 ppm and >1.8 ppm acetone in the exhaled breath. Further, it is reported that the concentration range of acetone in exhaled breath lies in ppm. Therefore, our Pd@SnO 2 sensors can easily distinguish a healthy person from a diabetic one. We know that the resistance and hence the sensitivity change on exposure to the reducing gases is a result of a number of surface-controlled phenomena such as the quantity of chemisorbed oxygen on the surface of the SMO, the reactions between the chemisorbed oxygen and the target gases, the nature of contacts among particles, etc. The receptor and transducer properties of the SMOs are determined by these phenomena [100]. Researchers reported earlier that the loading SMOs with promoters like Pd, Ag, etc. change both the receptor and transducer properties of the SMOs by changing their surface electronic structures [97,100]. For example, it is reported that Pd can activate the dissociation of molecular oxygen via the formation of stable oxide, PdO, in air. P-type PdO causes a withdrawal of electrons from the n-type SnO 2 particle underneath [101]. Thus, Pd loading produces a wider electron-depleted layer on the SnO 2 particle surface and hence a higher electrical resistance in air than that produced by the oxygen chemisorption only. The space-charge layer disappears on exposure to reducing gas like acetone or alcohol, which reduces PdO to Pd thereby releasing the electrons to the SnO 2 surface. Thus, introduction of Pd nanoparticle gives higher gas sensitivity than the pristine SnO 2. Pd nanoparticles being good oxidation catalysts can increase the catalytic dissociation of the molecular oxygen and can decrease the activation energy barrier 25

37 of the reaction between the chemisorbed oxygen and the reducing gases. Further, the nanoparticulate states of the 2 give enhanced adsorption and diffusion of the reacting species. Therefore, Pd loading can increase not only the sensitivity but also the rate of response of the SMO sensors, which is corroborated by our dynamic response experiments as given later. Figure 3.5: Sensitivity plot for all sensors SnO 2, 2 and 2 under various acetone and ethanol concentrations The increase in the sensitivity with the increasing concentration of the target gas can be attributed to the higher surface coverage of gas molecules. The slower increase in the sensitivity above certain gas concentrations could be result of the availability of decreasing number of active sites with increasing gas concentrations. Yamazoe reported that out of the possible three factors, viz., mean particle diameter, the number of Pd particles per unit surface area of SnO 2, and the total surface area of loaded Pd particles, the population density of Pd particles played a major role in their electronic and chemical sensitizations [97]. In addition to the sensitivity, selectivity is another important parameter of the sensor performance. We find that pristine SnO 2 hardly show any selectivity to acetone over ethanol. On the other hand, Pd@SnO 2 sensors exhibit certain selectivity for acetone over ethanol. It is not only the responses to ethanol was less compared to acetone, a clear picture is also obtained from the plots of sensitivity vs. 26

38 log C, where C is the concentration of target gas as shown in Figure 3.6. A good linear relationship is observed with clearly distinct slopes as shown in Table 3.7. Figure 3.6: A Typical plot showing the sensitivity of acetone sensing Figure 3.7: Table showing the sensitivity (Slopes) of sensors prepared Humidity and CO 2 gas testing: Figures 3.8 and 3.9 show the effects of moisture and CO 2 (major constituents present in the exhaled breath) on the sensor elements. It is obvious that the conductivities of the sensors are hardly not affected 27

39 by the moisture and CO 2. This could be attributed to the presence of Pd on the SnO 2 particles and a different sensing mechanism. Figure 3.8: Typical response of the Half-coated Pd@SnO 2 sensor exposed to CO 2 gas. The sensor showed no response to CO 2 gas Figure 3.9: Typical response of the Half-coated Pd@SnO 2 sensor exposed to Humidity. The plot shows humidity sensor response and our sensor response. Humidity sensor showed good response while our sensor showed almost zero response to humidity Earlier researchers demonstrated that, in the presence of water vapor, pristine SnO 2 and Pd@SnO 2 particles have different major surface-adsorbed oxygen species, namely O and O 2, respectively [94]. The authors reported that Pd loading could suppress the water vapor adsorption onto the SnO 2 parti- 28

40 cles due to the presence of O 2. The presence of different oxygen species could also be ascribed to the change in sensor performance between the pristine SnO 2 and Pd@SnO 2 particles. However, the origin of the difference in the acetone and ethanol gases toward a given sensor is not clear at present. It could arise from the higher dipole moment of acetone than that of ethanol or some other favorable acetone- Pd@SnO 2 interactions [81]. It requires further studies. However, this demonstrates that the Pd@SnO 2 sensors are suitable for the low concentration acetone detection in the presence of ethanol, CO 2, and moisture Response and Recovery time of sensors Pd nanoparticle loading onto SnO 2 particles showed spectacular performance in terms of the speed of response. Figures 3.10 and 3.11 shows typical plots of the time-dependent signal changes of Pd@SnO 2 and pristine SnO 2 sensor elements exposed alternately to the target gas and air. The response and recovery times were as usual taken as the time required for the signal to undergo a change of 90% in the case of adsorption and desorption of the target gas, respectively. The test signal in the cases of Pd@SnO 2 sensors shows almost instantaneous rise on exposure to the target gas. The signal remains almost stable under constant concentration condition and drops abruptly as soon as the test gas is removed. It evident Figure 3.10: Pd@SnO 2 response when exposed to 5ppm acetone gas. The sensor is exposed to gas and air alternately. that the range of the response and recovery times was very narrow, mostly in the single-digit numbers. The deposition of Pd nanoparticles onto the SnO 2 particle surface decreased the response and recovery times significantly in the case of acetone. The faster response and recovery processes can be ascribed to the Pd nanoparticle catalyzed faster surface oxygen dissociation and reactions as well as faster diffusion and transport of the gas molecules and their reaction products. 29

41 Figure 3.11: SnO 2 response when exposed to 5ppm acetone gas. The sensor is exposed to gas and air alternately. We achieved a response times as small as 1 s and 4 s and recovery times 7 s and 4 s for 2 ppm and 100 ppm acetones, respectively, for h-pd@sno 2. It demonstrates that our Pd@SnO 2 sensors have extremely fast response and recovery rate for acetone. Such fast response and recovery to a target gas are very useful for practical applications, especially for a reliable determination of acetone concentration in the exhaled air during the fast exhalation process ( 3 s). However, it was observed that the Pd loading could not improve the recovery process in the case of ethanol, though the response speed was enhanced considerably shown in Figures 3.12 to Figure 3.12: Response times of SnO 2, Half-coated Pd@SnO 2 and Full-coated Pd@SnO 2 sensors when exposed to Acetone gas under different concentrations 30

42 Figure 3.13: Recovery times of SnO 2, Half-coated 2 and Full-coated 2 sensors when exposed to Acetone gas under different concentrations Figure 3.14: Response times of SnO 2, Half-coated 2 and Full-coated 2 sensors when exposed to Ethanol gas under different concentrations We have tested the response reversibility of the sensors for all concentrations of both acetone and ethanol. 2 sensors showed good response reversibility. Figures 3.16 and 3.17 shows typical response and recovery cycles of the sensor elements toward 10ppm acetone gas and 10ppm ethanol gas. It is clear from the figures that there was neither sign of any decay of response amplitude nor aging performance over the repeating test cycles. When the sensor was exposed to air, the resistance returned nearly to the baseline level. On the other hand, the response and recovery curves for the pristine SnO 2 particles were irregular in shape which can be attributed to the restriction of the gas reaction and diffusion. 31

43 Figure 3.15: Recovery times of SnO 2, Half-coated 2 and Full-coated 2 sensors when exposed to Ethanol gas under different concentrations Figure 3.18 shows the sensitivity of the SnO 2 sensor under different gas concentrations ranging from 1ppm to 100ppm. From the plot it is clearly observed that the sensitivity is increasing linearly with increase in acetone gas concentration. At low concentrations the sensor showed low response and recovery times as compared to the high concentrations. The sensitivity difference between 1ppm acetone and 2ppm acetone concentrations is about 2% which can be identified easily. After exposing to each gas concentration it was exposed to air to come back to recovery state. Figure 3.18: SnO 2 sensor exposed to acetone gas concentrations ranging from 1ppm to 100ppm. Shows increased sensitivity with respect to increase in concentration 32