NEW THERMAL CONDUCTIVITY MICROSENSOR TO MEASURE THE METHANE NUMBER OF NATURAL GAS

Transcription

1 23 rd World Gas Conference, Amsterdam 2006 NEW THERMAL CONDUCTIVITY MICROSENSOR TO MEASURE THE METHANE NUMBER OF NATURAL GAS Main Author A.M. Gutierrez Spain

2 ABSTRACT As a result of the liberalization of the gas market, more important and frequent fluctuations of the natural gas composition are expected. Owing to this fact, the on-site control of the gas properties will become necessary for end-users. Methane Number is the parameter used to quantify knocking tendency of a gas, parameter especially relevant when natural gas is used as engines fuel. The present paper describes a measurement method to determine the Methane Number of natural gas as well as the microdevices developed to carry out these measurements. The method is based on the measurement of the gas Thermal Conductivity and the existing correlation between this parameter and the Methane Number of natural gas. The developed microsensors are ascribed to Microsystems Technology. They integrate two platinum thermoresistors: a reference one to measure the gas temperature and a sensing one to measure its thermal conductivity. In order to reduce thermal losses to the silicon substrate two different approaches are proposed: a) microbridges to locate the thermoresistors and b) oxidized porous-silicon isolation. In the range of natural gas compositions considered (60-100MN), applying a 50mA constant current to the sensing thermoresistor, obtained sensitivity is 0.95mV/1MN. Thermal stability and mechanical behaviour of both approaches are compared. Furthermore, the microdevices are mounted in a gas line in order to test them in field. Additional tests are still required but it can be concluded that the developed microsensors are a valid alternative to measure in situ the Methane Number of natural gas. Keywords: Gas Distribution, Gas Quality, Methane number, Microsystems Acknowledgements This Project has been partially funded by the Industry Department of the Autonomous Basque Government. Four Spanish companies have been involved in the project: NaturCorp Redes SAU (Naturgas Energia Group), Wärtsilä Ibérica, SA, Kosorkuntza, AIE and CEIT Research Centre.

3 TABLE OF CONTENTS 0. Abstract 1. Introduction 2. Experimental 2.1. Measurement principle 2.2. Fabrication process 2.3. Characterization infrastructure 3. Results and discussion 3.1. Microsensor characteristics 3.2. Platinum thin film TCR 3.3. Microsensor electrical output 3.4. Tests in field 4. Conclusions 5. References 6. List of figures

4 1. INTRODUCTION Natural gas is a gaseous mixture formed of methane, smaller fractions of higher molecular weight hydrocarbons (C 2 H 6, C 3 H 8, C 4 H 10 ) and inert gases (mainly N 2 and CO 2 ), where the different components ratio depends on the point of origin of the natural gas and on its manipulation during the distribution to end users. The natural gas composition, for its part, determines the chemical and physical properties and, consequently, the quality of the gas. Recent liberalization of natural gas market will give rise to a wider number of natural gas distribution companies making the composition fluctuations observed by end-users to be more frequent and larger. This circumstance will make more and more necessary to measure and control in situ the physical and chemical properties of the natural gas. Concretely, composition fluctuations have direct repercussions into two problems associated to the natural gas combustion: efficiency and cogeneration engines knocking. Knocking is a serious problem in engines where natural gas is used as fuel. Gas mixture self-ignition caused by its high pressure and temperature provokes in turn high pressure waves that impact against piston walls. In addition to decrease engine performance, knocking is extremely damaging for engine lifetime because of the high thermal and mechanical stresses that the latter must withstand under these conditions. Most of the aspects that promote knocking are conditional on engine design and operation conditions, but natural gas composition is also a critical aspect that must be controlled [1]. Methane Number, MN, defined as methane volume percentage of a mixture with hydrogen that provokes the same knocking intensity than the considered gas, is the parameter used to quantify knocking tendency of a gas mixture. Arbitrarily, it has been assigned a value of 100 for pure methane and 0 for pure hydrogen MN. Minimum MN required to keep an engine in good working order depends on the engine characteristics. It is usually admitted that knocking problems are avoided for installations with a MN higher than For cogeneration applications, engine characteristics are specified for gases with a MN higher than When MN is between 55 and 65, taking measures to prevent engine knocking is recommended. For a Methane Number lower than 55, leaving engine out of service is the best option. Alternatives to determine natural gas composition and/or quality now in the market, such as gas chromatography, optical measurements or calorimetric measurements, are sophisticated, expensive and, big size and weighty alternatives: unacceptable for natural gas end-users. Traditionally, gas chromatography has been the most usual alternative to analyze natural gas but its use has been restricted to laboratory or very specific gas lines nodes because of its high price and its complex operation. Nowadays, portable chromatographs, so-called microchromatographs have been developed. The traditional injection systems, columns and filament thermal conductivity detectors (TCD) have been substituted by devices based on MEMS technology. These analyzers [2-6] are available for portable, on line and in situ applications. Their price ( ) has been reduced to half of that of a conventional chromatograph. New measurement alternatives based on correlative methods have also been developed [7-9]. They allow the in situ analysis of natural gas composition and properties. Nevertheless they still are too expensive (1/3 of a conventional chromatograph price) for the sole determination of the Methane Number. In this context, a method to determine on line the Methane Number of natural gas is proposed. The adopted approach, ascribed to the µtas (Micro Total Analysis Systems) field, consists in measuring the gas Thermal Conductivity and using the existing correlation between this parameter and the Methane Number of natural gas. Microdevice fabrication process based on Microsystems Technology provides all the advantages related to microelectronics technology, allowing the on-line monitoring of the gas MN at low cost. The present paper describes the measurement principle and the design of the two proposed configurations for the thermal conductivity microsensor. It also includes their fabrication process and the characterization and field tests results.

5 2. EXPERIMENTAL 2.1. Measurement principle Microsystem configuration integrates two sensing elements or thermoresistors. One of the thermoresistors, R R, low current supplied, with quasi-null self-heating and therefore, in thermal short circuit with the gas, is used to determine natural gas temperature. The second thermoresistor, R S, always self-heated to a temperature higher than that of the gas and exposed to its dissipative influence, allows the natural gas thermal conductivity, and consequently, its Methane Number to be determined. Measuring principle is based on the analysis of the thermal equilibrium of the system: 2 V T TG T P = I R = = = = P R θ θ where, P is supplied power, I, current, R, resistance, V, voltage, T, thermoresistor temperature, T G, natural gas temperature, θ G, thermal resistance in the interface thermoresistor/gas and P D, dissipated power. In reference thermoresistor, R R, P=P D 0, so T T G. Depending on the conditions imposed to the sensing thermoresistor, R S, (I constant, V constant, P constant or T constant), thermal conductivity changes caused by gas composition fluctuations will be reflected by changes in the thermoresistor voltage and/or current and/or temperature. As it was previously mentioned, existing correlation between Thermal Conductivity and Methane Number will allow the latter to be determined. Known a priori the natural gas compositions used to calibrate the microsensors, Thermal Conductivity and Methane Number are respectively calculated in agreement with NBS1039 and ISO TC 193 Standards Microdevice development Thermal conductivity microsensor is defined on a silicon substrate. Thermoresistors are patterned in a platinum thin film deposited on the silicon wafer. In order to reduce thermal losses through the substrate, two different approaches are proposed. In both of them, the techniques used to microdevice fabrication are ascribed to Microsystems Technology. They include thermal oxidation, sputtering and CVD depositions, conventional photolithography or lift-off and, porous silicon formation. 2 G G D Pt thermoresistor Pt thermoresistor n-si / SiO microbridge 2 Silicon SiN / SiO x 2 Porous silicon Silicon a) b) Figure 1: Configurations of the thermal conductivity microsensors: a) approach A and b) approach B

6 Approach A. In this first approach, thermoresistors are located onto two microbridges defined in silicon by surface micromachining. Porous silicon is used as sacrificial layer [10-13]. Figure 1 shows the design of both approaches. The fabrication process of the approach A microsensor consists of three main steps: a) definition of microbridges, b) definition of thermoresistors and c) porous silicon formation and removal. Both-side polished epitaxial wafers (p substrate/n epitaxy) are used as substrate. Microbridges are constituted by a double layer n-type Si/SiO 2.Thermoresistors are patterned in Cr/Pt and Cr/Ti/Pt thin films. HF-H 2 O-ethanol (2:3:5) electrolyte and a 50mA/cm 2 current density with a porosification rate of 80 µm/h have been established as optimal parameters to define properly the micromachined bridges. Diluted KOH is used to remove the porous silicon. Process is described in detail in previous works [14-15]. Approach B. In this second approach, thermoresistors are positioned on a Si 3 N 4 /SiO 2 bilayer. This structure is thermally isolated from the silicon substrate by an oxidized porous-silicon layer (Figure 1). As in the previous approach, fabrication process consists of three main steps: a) formation of porous silicon, b) definition of Si 3 N 4 /SiO 2 structure and c) definition of thermoresistors. In this case, p-type silicon wafers are used as substrate. PECVD silicon nitride is used as mask in the formation of porous silicon. HF-ethanol (4:1) and 35 ma/cm 2 are established as optimal process parameters. Si 3 N 4 /SiO 2 double layer is deposited by PECVD. Its structure is patterned by conventional photolithography and RIE etching. Porous silicon is also oxidized during this step. Finally, Cr/Ti and Cr/Ti/Pt thermoresistors are defined by lift-off in a sputtered thin film Characterization infrastructure The developed microdevices are located into an analysis chamber for testing. Two electrovalves are placed at chamber input and output channels to control microchamber atmosphere and natural gas flow rate. Electrical connections are required to supply both thermoresistors and to extract microsensor output signals. Both constant voltage and constant current supplying are considered for microdevice operating but finally, constant current is preferred because of easier supply control and device output (voltage) processing. For the reference thermoresistor, which measures the natural gas temperature, 1mA current is selected. For the sensing thermoresistor, which measures the thermal conductivity, 50mA current is chosen. Microsystem characterization tests are carried out considering six natural gas compositions that cover the whole range in which the microdevice is supposed to work (MN from 50 to 100). The microsensor output voltage is correlated to the known Thermal Conductivity and Methane Number of these gas mixtures. Tests are carried out at flow zero. In a first step, analysis chamber gas is renewed. This implies input and output electrovalves opening, chamber gas purge and electrovalves closing. Then, both current supplies are switched on and data are acquired once the microsystem is thermally stabilized. Finally, current supplies are switched off. Microsystem must recover room temperature before a new measuring cycle. Tests in field are carried out in addition to laboratory characterization. Microdevice is mounted on a gas line rig that supplies natural gas to a cogeneration engines test-bed. Microdevice voltage data are acquired in parallel to measurements taken by a gas chromatograph. Microdevice and chromatograph both acquire data on line. Microdevice is working 24 hours per day; chromatograph works from 8 a.m. to 12 p.m. Tests are gone on for 6 months.

7 3. RESULTS AND DISCUSSION 3.1. Microsensor characteristics Microsensor global size is 6x6 mm 2, obtaining 60 samples from a 3 wafer. Each sample integrates two thermoresistors. Figure 2 shows a processed silicon wafer corresponding to approach A. Several configurations with different microbridge and thermoresistor width and length and, symmetric or in-a-line pads distribution can be observed. Figure 2: Processed silicon wafer corresponding to approach A In approach A samples, the thickness of microbridges bilayer is 7µm (5µm n-type silicon epitaxy + 2µm thermal silicon oxide). This circumstance and the low internal stresses of the double layer confer good mechanical resistance to the microstructure. On the other hand, the porous silicon sacrificial layer must be thick enough to allow the microbridge release. So, it must be at least, half of the bridge width. Thicknesses up to 150µm are obtained. Resistance to HF of n-type silicon used as masking material in the anodization process is strong enough to produce the required thick porous silicon layers. In approach B samples, Si 3 N 4 /SiO 2 double layer offers lower thickness (250nm Si 3 N 4 /190nm SiO 2 ) because of the high internal stress of the PECVD-grown films and the consequent adherence failures. The oxidized porous silicon layer has therefore a double function: to provide mechanical resistance to the structure and to thermally isolate the sensing element from the substrate [16-17]. In this case, resistance to HF of silicon nitride used as mask is not long enough to obtain thick porous silicon layers. Thicknesses up to 30 µm are obtained.

8 3.2. Platinum thin film TCR Platinum thin film TCR, Thermal Coefficient of Resistance, is determined since this parameter is essential to relate the measured voltage in R R thermoresistor to its temperature and therefore to the natural gas temperature (R R thermoresistor and natural gas are in thermal short-circuit). Two configurations are considered: Cr/Pt and Cr/Ti/Pt multilayers. At room temperature, 75Ω resistances have been obtained for thermoresistors. Both alternatives show a linear TCR with correlation index and for Cr/Pt and Cr/Ti/Pt respectively. The former presents a value of 1390 ppm/ºc; the latter, 1820 ppm/ºc, far both from bulk platinum TCR (3920 ppm/ºc). As it could be presumed, trilayer structure shows a higher TCR since titanium thin layer acts as barrier to chromium diffusion towards thermoresistor surface. Additionally, it provides a better temporal stability to the structure Microsensor electrical output Microsensor output voltage is measured when immersed in the six different natural gases considered. Results corresponding to approach A were presented in previous work [14-15]. No influence of the thermoresistors dimensions is observed for the developed microsensors. In the range of the natural gas compositions considered, microsensor output span is 38 mv, which implies a sensitivity of 0.95mV/1MN. Output voltage/mn relation shows good linearity with correlation index of However, it takes about 30 minutes stabilizing the output signal. The high thermal conductivity of n- type silicon present in the microbridge is likely the main reason of this long stabilization time. It allows heat conduction towards the silicon substrate, which prevents the microdevice from quickly reaching the thermal equilibrium. Additionally, once the thermal equilibrium is reached, the silicon substrate temperature is 20ºC over room temperature. This circumstance makes that, in order to obtain reproducible results, microsensor must be cool before a new measuring. Once the current supply is switched off, it takes 8 minutes to recover room temperature. 3,04 OUTPUT VOLTAGE (V) 3,03 3,02 3,01 3, TIME (min) Figure 3: Stabilization curve of the output voltage for the approach A

9 In approach B microsensors, the low thermal conductivity of the silicon nitride and the silicon oxide that constitute the double layer where the sensing element is located with the also low thermal conductivity of the oxidized porous silicon [16-17] makes the microsystem to reach the thermal equilibrium in 3 minutes. Additionally, heat conduction towards silicon substrate is not allowed and it keeps at room temperature during the whole measuring cycle. These two circumstances imply a higher data acquisition frequency and at the same time, lower power consumption. 2,95 OUTPUT VOLTAGE (V) 2,90 2,85 2,80 2,75 2, TIME (min) Figure 4: Stabilization curve of the output voltage for the approach B 3.4. Tests in field Tests in field have been carried out for 6 months in a gas line rig that supplies natural gas to a cogeneration engines test-bed. With a few exceptions, no great change of the supplied natural gas composition was observed during this period and, therefore, neither of the Methane Number. Anyway, some interesting results are obtained. From the analyzed data, it can be said that output signal of developed microsensors follows the measurements taken by the chromatograph. For example, Figures 5 and 6 compare natural gas Methane Number measured by the Agilent chromatograph and the output voltage measured by a microsensor corresponding to approach A. The Methane Number drop detected by the chromatograph is immediately reflected in the output signal of the microsensor.

10 84 METHANE NUMBER TIME (hours) Figure 5: Approach A micro sensor. MN measured by the chromatograph 2,944 OUTPUT VOLTAGE (V) 2,94 2,936 2,932 2,928 2, TIME (hours) b) Figure 6: Approach A micro sensor. Output voltage of the microdevice

11 4. CONCLUSIONS A thermal conductivity microsensor ascribed to Microfabrication Technologies is proposed to measure on line the Methane Number of natural gas. Microdevice consists of two thermoresistors: reference and sensing elements. The reference element is used to determine the natural gas temperature. The sensing element measures the thermal conductivity and consequently, the Methane Number of the gas. The results obtained for both temperature and Methane Number show good linearity. The Cr/Ti/Pt thin film TCR is 1820 ppm/ºc. The 38 mv output voltage span of the sensing element implies a sensitivity of 0.95mV/1MN. Two approaches have been considered. In approach A microsensors, both thermoresistors are patterned on two n-si/sio 2 microbridges defined by surface micromachining of silicon substrate. Porous silicon is used as sacrificial layer. N-type silicon is chosen as masking material for selective porosification of silicon. In approach B microsensors, thermoresistors are patterned on a Si 3 N 4 /SiO 2 double layer defined in turn on an oxidized porous silicon layer. Approach A microsensors offer good mechanical behaviour but the high thermal conductivity of the n-type silicon slows down the data acquisition rate since it takes 30 minutes to reach the thermal equilibrium of the microsystem before a measure and 8 minutes to cool it from measure to measure. In approach B microsensors, the thickness of the Si 3 N 4 /SiO 2 double-layer is limited by the high internal stresses of the PECVD-grown thin films. Oxidized porous silicon provides mechanical resistance to the microdevice as well as thermal isolation to the sensing element. This last fact allows the microsystem to increment data acquisition frequency. In addition to laboratory characterization, microsensors are tested in field. The analyses carried out simultaneously by a gas chromatograph corroborate the output voltage measurements offered by the developed microdevices. 5. REFERENCES 1. A. Julià. Effects of natural gas composition on alternative engines. Parameters involved in knocking. Introduction to Methane Number. Natural gas and its effects on alternative motors Workshop. Barcelona, December W. Lang, P. Steiner. Porous silicon for thermal sensors. Sensors and Materials vol. 8, no. 6 (1996) T. Bischoff, G. Müller, W. Welser, F. Koch. Frontside micromachining using porous silicon sacrificial layer technologies. Sensors & Actuators A60 (1997) C. Dücsö, E. Vázsonyi, M. Ádám, I. Szabó, I. Bársony, J.G.E. Gardeniers, A. van den Berg. Porous silicon bulk micromachining for thermally isolated membrane formation. Sensors & Actuators A60 (1997) F. Hedrich, S. Billat, W. Lang. Structuring of membrane sensors using sacrificial porous silicon. Sensors & Actuators A84 (2000) A.M. Gutierrez et al. Development of a novel and low cost thermal conductivity microsensor to measure the methane number of natural gas. IGRC 2004 Proceedings. Vancouver, November D. Puente, F.J. Gracia, I. Ayerdi. Thermal conductivity microsensor for determining the methane number of natural gas. Sensors and actuators B 110 (2005)

12 6. LIST OF FIGURES Figure 1: Configurations of the thermal conductivity microsensors: a) approach A and b) approach B Figure 2: Processed silicon wafer corresponding to approach A Figure 3: Stabilization curve of output voltage for the approach A Figure 4: Stabilization curve of output voltage for the approach B Figure 5: Approach A micro sensor. MN measured by the chromatograph Figure 6: Approach A micro sensor. Output voltage of the microdevice