Low frequency noise and drift in Ion Sensitive Field Effect Transistors

Transcription

1 Ž. Sensors and Actuators B Low frequency noise and drift in Ion Sensitive Field Effect Transistors C.G. Jakobson a,), M. Feinsod b, Y. Nemirovsky c a Department of Biomedical Engineering, Technion Israel Institute of Technology, Haifa 3000, Israel b Department of Medicine, Technion Israel Institute of Technology, Haifa 3000, Israel c Department of Electrical Engineering, Technion Israel Institute of Technology, Haifa 3000, Israel Abstract Ion Sensitive Field Effect Transistors Ž ISFETs. are currently produced commercially and promise to become the platform sensors for important biomedical applications. The drift in ISFETs is still an important inherent problem that prevents its application to accurate in vivo measurements. The present paper presents measurements of the drift and the drain current power spectral density Ž PSD. of ph ISFETs in the very low frequency range from 5 mhz to 10 khz. The measurements have been performed in buffer solutions with ph 4, 7 and 10, at room temperature. Above a corner frequency, the measured spectra correspond to 1rf noise introduced by fluctuations at the channel current. Below this corner frequency that depends on the magnitude of the drift, the measured spectra correspond to 1rf. The observed corner frequency is ;1 Hz for a drift of mvrh and shifts to frequencies below 0.01 Hz for a drift of 0.1 mvrh. The measured drift is correlated to leakage currents as well as temperature fluctuations and the inherent behaviour of the ISFET. A method for quality evaluation based on frequency behaviour is introduced. q 000 Elsevier Science S.A. All rights reserved. Keywords: ISFET; Noise; 1rf noise; Drift 1. Introduction Since originally introduced by Bergveld wx 1, Ion Sensitive Field Effect Transistors Ž ISFETs. have been under extensive study ŽRefs. w,3 x, and references therein.. In spite of the considerable progress that has already led to commercial devices, ISFETs still exhibit drift. Noise and drift are limiting issues for the application of the device as a platform for chemical and biomedical sensors. In a recent work wx 4, we have presented extensive measurements of low frequency noise in ISFETs considering: Ž. i various bias conditions corresponding to the gate voltage changing from subthreshold to saturation; Ž ii. several ph values obtained with buffer solutions ranging from phs4 tophs10; Ž iii. a frequency range between 1 Hz and 100 khz. 1rf noise behaviour was observed down to 1 Hz that followed the same bias dependency of a standard MOSFETs, indicating that the noise is dominated by the Si SiO interface. ) Corresponding author. Tel.: q ; fax: q Ž. address: claudio@tx.technion.ac.il C.G. Jakobson. In this paper, the drain current power spectral density Ž PSD. is measured down to 5 mhz. At low drift conditions, which will be shown to correspond to low leakage current, the 1rf is the only noise source observed in the entire frequency range. A 1rf spectral behaviour is observed below a corner frequency, which depends on the magnitude of the drift and varies from ;1 Hz to frequencies below 0.01 Hz. This corner frequency provides a method for quick evaluation of the drift quality of the device. The measured drift varies from to 0.1 mvrh. The drift at the output of an ISFET amplifier may be due to one or more of the following effects: Ž. i changes in the insulator electrolyte interface potential at constant ph or charging of the insulator Ž inherent drift.; Ž ii. changes in the reference electrode Ž RE. potential due to leakage currents, which polarise the RE; Ž iii. polarisation potentials at additional leakage points, which may be present at the set-up; Ž iv. temperature variations. This paper discusses the above effects and relates them to the measurement of low frequency noise. It is observed that for stable measurement set-ups the RE current due to leakage must be kept below 100 pa. Otherwise the RE polarisation, measured by the ISFET as drift, turns to be the dominant drift of the ISFET amplifier. The reduction r00r$ - see front matter q 000 Elsevier Science S.A. All rights reserved. Ž. PII: S

2 of the leakage current is of particular importance in measurement environments that have relatively low impedance to the ground, as is encountered in in vivo applications. ( ) C.G. Jakobson et al.rsensors and Actuators B Experimental set-up The experimental set-up for measuring the low frequency drain current noise spectrum of the ISFET is shown in Fig. 1. The influence of the different noise sources in this set-up has been analysed in Ref. wx 5. The ISFET operates at constant gate voltage. It is connected in a common source configuration and is DC coupled to an SR570 current amplifier. The sensitivity of this amplifier and the drain bias of the ISFET are controlled by the Personal Computer Ž PC.. The output voltage of the current amplifier is connected to an HP35670A dynamic signal analyser and to an AnalogrDigital converter, providing measurements in the time and the frequency domains. An auxiliary RE connected to a Keithley 6517 electrometer measures directly the voltage drift at the RE. A PC controls the set-up and can work stand-alone for hours as required for these measurements. The ISFETs under study are n-channel commercial ph ISFETs with a relatively matured fabrication and packagwx 6. The ISFET gate insulator is Al O 3, the ing technology channel width W is 600 mm, the channel length L is 0 mm, and the insulator capacitance per gate unit area Ci is 3P10 y8 Frcm. The RE is AgrAgCl and measurements are performed in buffer solutions with ph values of 4, 7 and Results and discussion Fig. shows the drain current vs. gate voltage static characteristics for the ph ISFET, at different phs. The transistor is biased from the subthreshold to the saturation region. The characteristics measured at different phs are parallel both in the subthreshold and the saturation region. Fig. 1. Experimental set-up. Fig.. Drain current characteristics as a function of gate voltage from subthreshold to saturation. This result confirms that the flat-band voltage is the only parameter of the ISFET associated to the ph variations. The inverse logarithmic slope measured in the subthreshold region for any of the curves in Fig. is 98.8 mvrdecade. In a MOSFET, the slope s of the current gate voltage characteristics at subthreshold is given by wx 7 : kt CDqCit ssln10 1q Ž 1. q ž C ox / where k, T and q have their usual meanings, Cox is the silicon oxide capacitance, CD is the silicon depletion capacitance, and Cit is the silicon silicon oxide interface traps capacitance, all capacitances are expressed per unit area. For MOSFETs with high quality interfaces and C ox 4CDqC it, an ideal slope of ;60 mvrdecade is predicted. State-of-the-art MOSFETs exhibit a slope ranging from 80 to 105 mv wx 8. Thus, the observed values of ; 100 mvrdecade for the ISFETs under study correspond to MOSFETs values and indicate the quality of the interfaces. Fig. 3 shows the ISFET gate transconductance characteristics in the saturation region. The threshold voltage is calculated by extrapolating to the gate voltage axis the straight gate transconductance characteristics in the saturation region, as shown by the dashed line in Fig. 3. The threshold voltages values obtained are 0, 380, and 50 mv for phs4, 7, 10, respectively. The average shift of the threshold voltage that expresses the ph sensitivity is 50 mv per unit ph and is roughly linear. Fig. 4 shows the drain current PSD obtained at low to medium frequencies from 1 Hz to 10 khz for two buffer solutions having phs7 and phs10, and at an average drain current of 10 ma. In this range of frequencies, the spectra observed follow a clear 1rf slope. Moreover, the PSD for different phs at the same drain current are completely coincident indicating that in this frequency range the channel contributes the noise independently of the chemical environment.

3 136 ( ) C.G. Jakobson et al.rsensors and Actuators B Fig. 3. Gate transconductance characteristics in saturation with V V, VDS s V,pHs4, 7, 10. BS s0 The drift at the output of an ISFET measurement set-up may be due to drift of the RE potential, polarisation potentials due to leakage currents, temperature variations, as well as changes in the sensing insulator or the insulator electrolyte interface. In this study, two REs are used to characterise the drift contributed by the RE. The auxiliary RE, which is incorporated to the measurement set-up, provides direct measurement of the RE voltage drift. In our first attempt to measure the voltage between both REs, a voltage amplifier having an input impedance to ground of 100 MV was used instead of the electrometer shown in Fig. 1. A voltage of ;1 V was applied to the main RE that controlled the solution potential, and generated a current of ;10 na. The measured ISFET drain current and the voltage between REs under this high leakage current are presented in Fig. 5a and b, for buffer solutions with phs4 and phs10. To provide comparable scales, the right axis shows the RE voltage values Fig. 5. Ž. a Drain current and reference electrode drift over 15 h on a buffer solution of phs4 with reference electrode current of ;10 na. Ž. b Drain current and reference electrode drift over 3 h on a buffer solution of phs10 with reference electrode current of ;10 na. Ž. c Drain current and reference electrode drift over 3 h on a buffer solution of phs7 with reference electrode current of ;1 na. Fig. 4. Drain current power spectral density for two different buffer solutions having phs7 and phs10. The average drain current is 10 ma. obtained after dividing the drain current by the ISFET transconductance Ž Fig. 3., and the voltage between REs is shifted to the right axis scale by adding an arbitrary constant. The drain current drift and the RE drift in these figures show a clear correlation. It is evident that the high leakage currents induce polarisation of the RE, generating a drift in the RE voltage that is measured by the ISFET. It should be noted that such high leakage currents are often present in practical applications, such as in-vivo measurements. Fig. 5c shows a measurement performed with the

4 ( ) C.G. Jakobson et al.rsensors and Actuators B high input impedance electrometer at phs7, which reduced the leakage current. However, a residual leakage between the measurement beaker and the shielding box induced an RE current of the order of 1 na in the set-up. The same correlation between the RE leakage current and the ISFET drain current drift is observed indicating that even the 1-nA leakage current induces polarisation of the RE. However, the low leakage current results in low drift. Thus, for the 10 na RE leakage current of Fig. 5a and b, the drift observed is ; 5 mv over the first 3 h of operation, while for the lower 1 na RE leakage current of Fig. 5c, it is reduced to ;1.5 mv over the same time period. Fig. 6 compares between the ISFET amplifier drift observed on the same device for different RE leakage currents: a significant leakage of 10 na and a small leakage below 100 pa. For the largest leakage current of 10 na, the measured drift is of the order of 0.8 ma, and for the lower leakage current below 100 pa, the measured drift is of the order of 0. ma. It was observed that for the leakage current below 100 pa, the ISFET amplifier drift does not correlate to the RE drift since the voltage variation between the two REs in this case was lower than 0.1 mv. Thus, the measurement set-up is effective to distinguish between RE drift measured by the ISFET and other drift sources. The drift that is still observed exhibits two phases: Ž. a The first phase drift observed during the initial ; h of operation is the inherent ISFET drift and shows a slope of 0.5 mvrh. Ž. b The second phase drift is due to temperature fluctuations between day and night and shows a slope of 0.1 mvrh. Fig. 7 shows the drain current PSD for the very low frequency range from 5 mhz to Hz, for a buffer solution with ph s 7 and measured at different times after the beginning of the operation of the device. The spectra are measured after 1, and 10 h of operation. The first two spectra measured exhibit the same slope corresponding approximately to 1rf, and similar PSD. After 10 h of Fig. 7. Drain current power spectral density with a buffer solution of phs7 at an average drain current of 100 ma. The spectra are measured at different time intervals after 1, and 10 h of operation. operation, the spectrum follows 1rf behaviour for frequencies above 0.1 Hz, and the higher slope is observed only below this frequency. Fig. 8 shows spectra corresponding to the low leakage current Ž pa. presented in Fig. 6. As the leakage current is reduced, the 1rf characteristic shifts to lower frequencies. Moreover, for the second phase drift that is due to temperature variations and is measured after more than ;3 h of operation, the 1rf noise is the only characteristic observed down to 0.01 Hz. A linear dependency in time provides a 1rf dependency in frequency. This is easily understood remembering Ž. < Ž.< that the Fourier transform of f t s t is F v s 1rv wx 9. Hence, it is concluded that the 1rf slope is related to the initial higher drift, and dominates the spectrum below 1 Hz during the first hours of operation. As the drift is reduced, the 1rf slope corresponding to the noise at the wx Si SiO interface is observed 5. A 1rf dependency has been previously reported w10 1x and was considered to be related to a adsorptionrdesorption process with a single Ž time constant t providing a tr 1qv t. Lorentzian dependency. The results presented in this work show that a Fig. 6. Drain current and reference electrode drift over 15 h on a buffer solution of phs10 at an average drain current of 10 ma and different reference electrode currents: I RE ;10 na and IRE -100 pa. Fig. 8. Drain current power spectral density corresponding to the I - RE 100 pa time trace in Fig. 6 after 1 and 10 h of operation.

5 138 ( ) C.G. Jakobson et al.rsensors and Actuators B rf slope is correlated to the drift and is observed below corner frequencies of ; 1 Hz before stabilisation and below corner frequencies of ; 0.01 Hz after stabilisation. These results indicate the last decade technological improvements leading to reduced drift in ISFETs. The 1rf noise spectra observed at the low drift conditions and their operating conditions dependence indicate that the channel trapping detrapping mechanism is the dominant noise source of n-channel ISFETs wx 4. It is important to notice that while drift characterisation requires measurements lasting several hours, the corner frequency between the 1rf and the 1rf characteristics is measured in several minutes. Hence, this corner provides a fast method for the evaluation of the drift performance of the ISFET and the RE. In addition, it should be noted that practical noise measurements are performed in a limited time window. This measurement procedure applies only to stationary processes. Since we attribute the 1rf dependency to drift, which is non-stationary, this frequency behaviour cannot be considered a noise spectrum in a strict sense and must be understood as an apparent PSD. The correlation between drift and the 1rf behaviour of the measured apparent PSD is evident. 4. Conclusions A considerable drift observed in ISFET amplifiers is introduced by the RE, due to leakage currents of the order of 1 na or larger. This current may be present due to the packaging of the device or parasitic paths in the measurement set-up. The second case is of particular importance for in vivo applications, where the electrical isolation of the measurement environment is usually deficient. Hence, it is advised to work with leakage currents below 100 pa. This often requires that the impedance seen by the RE must be higher than 1 GV. When the leakage currents are adequately reduced, the observed drift is either due to temperature changes or inherent to the ISFET. The drift in the drain current of the ISFET is correlated in the frequency domain with a 1rf behaviour. This spectral dependency is typically observed below 0.1 Hz in state-of-the-art ISFETs, provided the leakage current is properly reduced. ISFET drift measurements based on the time domain require long time traces that are usually difficult to analyse and require measurements lasting several hours. The evaluation in the frequency domain is more efficient for the determination of the quality of the device. The spectra presented in this paper show that this evaluation can be easily obtained from the position of the corner frequency between the 1rf and the 1rf characteristic. At particular conditions presented in this work, 1rf noise is observed down to 0.01 Hz. The noise originates from the MOSFET channel since the spectra are independent of the ph of the solution and the gate voltage dependencies correspond to a typical n-channel MOSFET following the trapping detrapping model. Acknowledgements The Eshkol Scholarship granted to C.G. Jakobson by the Israeli Ministry of Science is gratefully acknowledged. The research support of the Chutik Foundation is gratefully acknowledged. References wx 1 P. Bergveld, Development of an ion-sensitive solid-state device for neurophysiological measurements, IEEE Trans. Biomed. Eng. 17 Ž wx C. Cane, I. Garcia, A. Merlos, Microtechnologies for ph ISFET chemical sensors, Microelectron. J. 8 Ž wx 3 A. Lui et al., Chemical sensors based on ISFET transducers, Joint 4th Int. Conf. on Microelectronics MIEL96, 1996, pp wx 4 C.G. Jakobson, Y. Nemirovsky, 1r f noise in Ion Sensitive Field Effect Transistors from subthreshold to saturation, IEEE Trans. Electron Devices 46 Ž wx 5 C.G. Jakobson, I. Bloom, Y. Nemirovsky, 1r f noise in CMOS transistors for analog applications from subthreshold to saturation, Solid State Electron. 4 Ž wx w 6 Sentron Integrated Sensor Technology, Sentron 1000 ph electrode, The Netherlands. wx 7 S.M. Sze, Physics of Semiconductor Devices, nd edn., Wiley-Interscience, New York, wx 8 I. Brouk, Study of CMOS Photodiodes and Low Noise Analog Readout for Visible Photon Detection, MSc Thesis, Technion I.I.T., Haifa, Israel, 000. wx 9 A. Papoulis, Circuits and Systems, a Modern Approach, H.R.W., New York, w10x A. Haemerli, J. Janata, J.J. Brophy, Ion noise in ISFETs, Proc. of the Sixth International Conference on Noise in Physical Systems, 1981, pp w11x A. Haemerli, J. Janata, J.J. Brophy, Equilibrium noise in ion selective field effect transistors, J. Electrochem. Soc. Ž w1x J. Janata, Electrochemistry of chemically sensitive field effect transistors, Sens. Actuators 4 Ž Biographies Claudio G. Jakobson was born in Buenos Aires, Argentina, on April 0, He received his Electronic Engineer degree from the University of Buenos Aires in 199 and his MSc degree in Electrical Engineering from the Technion Israel Institute of Technology in He is currently working toward his DSc degree at the Technion. His DSc research focuses on noise and drift phenomena in ISFETs for in vivo microsystems and CMOS compatible ISFET microsystems for brain monitoring at the cerebro-spinal fluid. His MSc thesis was on low noise CMOS analog channels for X-ray detection, and his design is currently operating in the X-ray detection experiment at the Technion satellite TECHSAT. Other fields of interest are analog electronics in VLSI, readout interfaces for CMOS compatible sensors, noise phenomena in MOSFETs and Micro- Electro-Mechanical Systems Ž MEMS..

6 ( ) C.G. Jakobson et al.rsensors and Actuators B Moshe Feinsod is a MD Professor and Head Department of Neurosurgery at Rambam Ž Maimonides. Medical Center, Faculty of Medicine, The Technion Israel Institute of Technology Ž Haifa, Israel.. He received his MD degree from the Hebrew University in Jerusalem in He got his training in Neurosurgery at Hadassah University Hospital. The training combined with laboratory work in visual evoked potentials was followed by fellowship in Neuro-ophthalmology at University of California medical center in San Francisco. Upon returning to Hadassah, he functioned as senior staff and Deputy Head of the Department of Neurosurgery. He headed the Neuro-ophthalmology Unit and the Laboratory of Applied Sensor Physiology. In 198, he was called to chair the Department of Neurosurgery at Rambam Ž Maimonides. Medical Center and joined the Faculty of Medicine at the Technion. The main academic interests of the department stern from the imposed challenge of military and civilian head injuries. Sensory and motor evoked potentials are employed to asses neural functions in comatose patients, the dynamics of intracranial pressure and cerebral blood flow are monitored and studied for hemometabolic optimization of the injured brain for preservation and salvage of brain tissue. He is now involved in the development of microsensors for multimodality recordings from the human brain, non-invasive detection of cerebral oxygen saturation using near infrared technology and the study of brain ischemia in the animal model. Ž. Yael NemiroÕsky IEEE fellow, IEE fellow received her BSc degree in 1966 and her DSc degree in 1971 from Technion Israel Institute of Technology. Yael Nemirovsky joined in 1980 the Department of Electrical Engineering, Technion Israel Institute of Technology and she has been an associate professor since Twice she was the head of the microelectronics research center of the Department of EE at the Technion. Prior to that, she was a research scientist specializing in microelectronics in Rafael, a national R&D organization. For about 0 years, she has been active in electro-optical devices in II VI compound semiconductors and additional advanced semiconductor materials as well as infrared focal plane arrays. She has been involved in growth, processing, device design and modeling of detectors as well as VLSI circuits. She has been a principal investigator in large funded research programs that ended in prototype infrared detectors and systems that were transferred to industry. Currently, her research focuses on microsensors, CMOS compatible micromachining and microsystems implemented in CMOS technology and silicon devices as well as in II VI semiconductors. She has published close to 100 papers in the open literature and a large number of classified reports. She has supervised over 30 graduate students for MSc and DSc. She is a distinguished lecturer of the electron device society of IEEE and is the chairperson of the Israeli Association for Crystal Growth as well as the chairperson of the microelectronics and photonics section of URSI.