Temperature effects on the ISFET behaviour: simulations and measurements

Sensors and Actuators B (1998) 60 – 68

Temperature effects on the ISFET behaviour: simulations and measurements Sergio Martinoia a,*, Leandro Lorenzelli a, Giuseppe Massobrio a, Paolo Conci b, Alberto Lui b a

Department of Biophysical and Electronic Engineering (DIBE), Uni6ersity of Geno6a, Via Opera Pia 11a, I-16145 Geno6a, Italy b Microsensor and System Integration Di6ision, IRST/ITC-Via Sommari6a, Pante` di Po6o, 38050 Trento, Italy Received 28 November 1997; received in revised form 2 April 1998; accepted 6 April 1998

Abstract Temperature effects on ion-sensitive field-effect transistor (ISFETs) are investigated both from theoretical and experimental point of view. An ISFET model has been implemented into a modified version of SPICE and the effects of temperature on the device behaviour over a user-defined range of pH and temperature have been simulated. The simulated and measured results are then compared and discussed. © 1998 Elsevier Science S.A. All rights reserved. Keywords: ISFET; pH sensor; Temperature dependence; SPICE program

1. Introduction Nowadays ion-sensitive field-effect transistor (ISFET)-based pH meters and ISFET devices, are commercially available from several companies (e.g. ORION, Orion Research, Boston, MA; CORNING, New York, NY; SENTRON Integrated Sensor Technology, Roden, The Netherlands). Nevertheless, further improvements and investigations of ISFET performances are needed to promote their use in microsystems for specific applications. Difficulties arise not only from technological issues but also from the lack of considering first-order mechanisms such as the effects of temperature. This limiting operating condition is particularly relevant in respect of possible applications of ISFET-based microsystems in industrial control and monitoring of processes (e.g. biofermenters for biotechnological applications) or where temperature cannot be a controlled parameter. Previous works on the effects of temperature on the ISFET characteristics have been based on the validity

* Corresponding author. Tel.: +39 10 3532251; fax: + 39 10 3532133; e-mail: [email protected] 0925-4005/98/$19.00 © 1998 Elsevier Science S.A. All rights reserved. PII S0925-4005(98)00157-9

of the Nernstian equation [1,2], or have been devoted to propose configurations to obtain temperature compensation [3,4], or to describe the manner in which temperature affects specific operating parameters of the device [5,6]. However, for applications such as those mentioned above, where the device has to operate over wide temperature ranges, the effects of temperature, including possible non-Nernstian behaviour, must be taken into account. In this paper, we present a first-order theoretical model that allows us to fully describe the influence of temperature on the ISFET characteristics. The model has been implemented into an ad hoc modified version of the circuit simulation program SPICE (called BIOSPICE). Then we discuss the effects of temperature on the ISFET behaviour by comparing experimental results with the results obtained from the simulations with BIOSPICE. The use of BIOSPICE has allowed us to focus attention only on the temperature-dependent physico-chemical parameters of the ISFET structure, neglecting those parameters of the MOSFET (modified to form the basis of the ISFET) whose temperature dependence is already modeled in the original version of SPICE.

S. Martinoia et al. / Sensors and Actuators B 50 (1998) 60–68

61

Table 1 .MODEL statement used in the simulations with BIOSPICE

2. ISFET thermal model in BIOSPICE An ISFET model has been developed and implemented into SPICE version 2G (called BIOSPICE). ISFET theory shows the threshold voltage of the device depends on terms common to the MOSFET theory, as well as on terms that are electrochemical in nature. This statement is mathematically expressed by writing the threshold voltage VTH in the form [7]:



VTH(ISFET)= [ERef +8lj]− [8eo(pH) −xeo] −

Qss +Qsc f −28f + s Cox q

n

(1)

In Eq. (1), 8f is the Fermi potential of the semiconductor, Qss is the fixed surface-state charge density at the insulator-semiconductor interface, Qsc is the semiconductor surface depletion region charge density, Cox is the insulator capacitance per unit area, Eref =(Erel + Eabs) is the potential of the reference electrode (Erel is ZEREL in Eq. (2)), 8lj is the potential difference between the reference solution and the electrolyte solution (ZPHILJ in Eq. (3)), 8eo is the potential of the electrolyte solution-insulator interface which determines the ISFET sensitivity to pH, xeo is the electrolyte solutioninsulator surface dipole potential (ZCHIEO in Eq. (4)) and fs is the semiconductor work function. The most important of these parameters is the potential difference across the insulator- electrolyte solution interface which results from the charged insulator surface of the device. The charging of the insulator surface can be explained by the site-binding model which we have used, together with MOSFET physics, to model the response of the ISFET to pH. As a result, a set of equations has been derived which, implemented into BIOSPICE, fully determine the threshold voltage of the ISFET. Typical insulator layers such as Al2O3, Si3N4,

Ta2O5 have been used in the simulations. A detailed analysis of the model can be found in [7,8]. The model [7,8] has been upgraded by implementing the temperature-dependence of the electro-chemical parameters of the ISFET, in order to simulate the effects of temperature on the characteristics of the device over a user-defined pH and temperature range. This thermal model is based on the standard electrochemical and MOSFET theories and on previously developed theoretical models [5,9]. In this section, we present the equations, implemented into BIOSPICE and affecting the temperaturedependence of the ISFET parameters, but we will not consider those parameters of the MOSFET (modified to form the basis of the ISFET) whose temperature dependence has already been modeled in SPICE version 2G [10]. The parameter notation used in the equations follows the one specified in the .MODEL statement (which contains the parameters of the device model) of BIOSPICE.

2.1. Equations for the temperature dependence of the reference electrode [5,11] Potential of Ag/AgCl reference electrode relative to a hydrogen electrode: ZEREL(T)=ZEREL(Tbio)+ 1.4 · 10 − 4(T−Tbio)

(2)

where Tbio = (25+ 273.15) in degrees Kelvin. In Eq. (2), the temperature coefficient value is that which holds when the standard hydrogen electrode, to which the Ag/AgCl reference electrode is relative, is kept at 25°C. Liquid junction potential for the Ag/AgCl electrode: ZPHILJ(T)= ZPHILJ(Tbio)+ 10 − 5(T−Tbio)

(3)

where the temperature coefficient value applies for a typical liquid junction potential of 3 mV.

S. Martinoia et al. / Sensors and Actuators B 50 (1998) 60–68

62

Fig. 1. Sketch of the measurement system.

2.2. Equations for the temperature dependence of the electrolyte solution [5,9]

where X stands for A or B for the dissociation constants of the silanol binding sites and N for the dissociation constant of the ammine binding sites (see Table 1). Electrolyte surface dipole potential: In Eqs. (2)–(6), Tbio points out that the values of each involved parameter are assumed to have been measured 0.86 0.4 · 10 − 3 · ln(ZI) · 1− ZCHIEO(T)= ZCHIEO(Tbio) · 1 −exp (T− Tbio) 2.303 ZCHIEO(Tbio) (4) where ZI is the ionic strength and the temperature (input in the .MODEL statement) at 25°C, as reported coefficient is that for pure water. in [5]. As all the other parameter values specified in the Solution permittivities: .MODEL statement are assumed to have been measured





n 

n

EPSX(T)= EPSX(Tbio) · o0[1 −4.6 · 10 − 3(T − Tbio)+ 8.8 · 10 − 6(T−Tbio)2] (5) where X stands for EL for the bulk of the electrolyte at T = TNOM=27°C (default value for TNOM in solution, IHP for the portion of electrolyte solution in BIOSPICE), we have initialized and updated at TNOM the inner Helmholtz plane and OHP for the portion of the ‘bio’ parameter values in the MODCHK subroutine electrolyte solution in the outer Helmholtz plane. As of BIOSPICE before the temperature analysis loop is already considered [8], the water permittivity was divided executed. into three compartments and each of them has been simulated with the same temperature dependence.

2.3. Equations for the temperature dependence of the dissociation constants [5,7]

3. Experimental methods

Dissociation constants: ZXKAP(T)=[ZXKAP(Tbio)]Tbio/T

(6)

A fully automated experimental system has been developed to carry out the measurements on ISFETs at different operating temperatures.

S. Martinoia et al. / Sensors and Actuators B 50 (1998) 60–68

63

Fig. 2. Thermal dependence of the pH buffers, as given by the producer.

The core of the characterization system, shown in Fig. 1, is a thermostatic chamber, designed to hold, at the same time, four different sensor chips connected to an HP 4145B Semiconductor Parameter Analyzer. A switching matrix system allowed us to connect each sensor with the Semiconductor Parameter Analyzer for different measurement configurations. The temperature and pH in the thermostatic bath were continuously monitored in order to ensure ISFET electrochemical characterizations are at a uniform buffer condition; Carlo Erba buffered solutions (pH= 4, 7, 10) were used. The experimental protocol consisted of the following steps. 1. A preliminary current leakage test on each device was carried out in order to verify the packaging integrity. The leakage current was measured through the reference electrode and the sourcedrain-bulk shorted. 2. A set of Ids –Vgs curves was acquired for each pH buffer value at each given temperature. The range of Vgs has been chosen to vary from 0 to 4 V. The curves have been parameterized by the drainsource voltage Vds (from 0.1 to 1.3 V with a voltage step of 0.2 V). 3. Sets of curves were obtained at different temperature values (T= 5, 15, 25, 37°C). After each change of the pH buffer, the ISFET under test has been first rinsed with deionized water and then conditioned with the pH buffer to be used in the subsequent measurement. The ISFETs under test were fabricated by a nonconventional p-well metal-gate process, called CMNOS/ISFET (Complementary Metal Nitride-Oxide

Semiconductor/ISFET) process [12] with added field channel stop and separate threshold adjustment implants. The chemical sensing layer was made of stoichiometric Si3N4 deposited at 775°C in an LPCVD system (ASM DFR210) using ammonia and dichlorosilane in a 4:1 flow ratio.

4. Results and discussion Following the experimental protocol described in Section 3, measurements were carried out on ISFETs fabricated from different wafers. The experimental data were then compared with the results obtained from BIOSPICE simulations. The ISFET parameter values, specified in the .MODEL statement, were extracted from the test structures fabricated on the same wafer containing the sensor. The values of the standard electro-chemical parameters, such as surface dissociation constants, solution permittivities and total concentration of binding sites, are assumed from literature [13,14]. Table 1 shows, as an example, the .MODEL statement for the Si3N4-gate ISFET of wafer 6. The slight pH changes of the buffer solutions due to the temperature variations have been taken into account, in the simulations, by interpolating the data given by the producer for the thermal dependence of the pH buffer (Fig. 2). Fig. 3 shows the simulated and measured input curves Ids –Vgs (with source and bulk grounded) for the Si3N4-gate ISFET under test (wafer 6), at pH= 4 and for the indicated temperature values, at Vds = 0.1 V. The measured data refer to a typical device taken from wafer 6. Similar results (data not shown) were

64

S. Martinoia et al. / Sensors and Actuators B 50 (1998) 60–68

Fig. 3. Simulated and measured input curves (Ids –Vgs) of the Si3N4-gate ISFET under test, at pH=4, Vds =0.1 V, for (a) T =5°C; (b) T= 15°C, (c) T = 25°C, (d) T = 37°C.

obtained with other devices from different wafers and a good agreement was found between measured and simulated curves by changing the BIOSPICE simulation parameters according to the figures obtained from the parameter extraction procedures. Fig. 4 shows the measured and simulated input characteristics at pH= 4 and pH = 10, each at T= 5°C and T=37°C. In this figure, the athermal point, where (Ids/(T is a minimum, i.e. where the drain current is not significantly thermally dependent, can be clearly seen, both in the measured and simulated curves. Moreover, it should be noted that the athermal point, usually affected by the physico-geometric parameters of the MOSFET, appears to be also dependent on the pH value of the electrolyte solution so that, for strong basic solutions (pH\ 10), the athermal point tends to shift to lower drain currents (sub-threshold region). This effect, also reported in literature [15], if correctly predicted, can be usefully used to design integrated circuits containing ISFETs whose operating point is near the athermal point, in order to minimize the effects of temperature variations on pH measurements.

Experimental and simulated results of threshold voltage variations, normalized at T= 5°C, are shown in Fig. 5a and b for different ISFETs obtained from different wafers at pH=4 and pH= 10. For comparison, the figure also shows the analogous curve for the MOSFET fabricated on the same chip. As reported in literature, the threshold voltage and as a consequence the pH sensitivity, are temperature dependent [16] and this dependence is affected by the physicochemical parameters of the ISFET. The measured data refer to several devices taken from three different wafers and each device was simulated by using the appropriate parameters. By carrying out the linear regression on the experimental and simulated curves, comparable average slopes (dVTH/ dT) were obtained for the given pH values. It should be noted that the temperature dependence of the electrochemical component of the ISFET threshold voltage counteracts the temperature dependence of the solid-state structure underlying the sensitive layer. The electrochemical component which contributes to the threshold voltage and which is temperature depen-

S. Martinoia et al. / Sensors and Actuators B 50 (1998) 60–68

65

Fig. 4. Simulated and measured input curves (Ids-Vgs) of the Si3N4-gate ISFET under test, at Vds =0.1 V, pH =4 and pH =10 for T =5 °C and 37 °C. The athermal point is also indicated.

Fig. 5. (a) Measured results of threshold voltage variations normalized at T =5°C obtained from different ISFETs, at pH = 4 and pH= 10. The analogous curve for the n-MOSFET fabricated on the same chip is also shown for comparison; (b) Simulated results of threshold voltage variations normalized at T = 5°C obtained for a typical device, at pH =4 and pH = 10. The analogous curve for the n-MOSFET underlying the ISFET structure is also shown for comparison with the experimental data.

dent, can be divided into two parts. The first one refers to the voltage drop between the reference electrode and the electrolyte solution (Vref el). As reported in the – literature [11] its temperature dependence is small as also indicated by the simulations obtained by our implemented model. The second one refers to the voltage drop between the Si3N4 surface and the electrolyte solution (Vsol in) and it takes into account the tempera– ture dependence of the voltage across the diffuse layer [5], as well as the dependences of the electrolyte surface dipole potential (Eq. (4)), the solution permittivities

approaching the surface (Eq. (5)) and the dissociation constants (Eq. (6)). These contributions are shown in Fig. 6. Simulation results have been also carried out over a temperature range greater than the one taken into account for experimental measurements. As shown in Fig. 7, the model predicts a behaviour, for the temperature coefficient of the threshold voltage that is almost linear up to an operating temperature of 80°C, suggesting that ISFET sensors (if properly packaged) can also be used at high temperatures.

S. Martinoia et al. / Sensors and Actuators B 50 (1998) 60–68

66

Fig. 6. Contribution of the two electrochemical components (the reference electrode, Vref el and the electrolyte solution-insulator interface, Vsol in) – – and the electronic components (VTH MOSFET) to the ISFET threshold voltage as a function of temperature variations. (Simulated results). –

Fig. 7. Simulated input curves (Ids –Vgs) of the Si3N4-gate ISFET over a wide range of temperatures (T= 5 – 80°C). The inset shows the linear regression of the threshold voltage variations obtained from the input curves.

It should be noted that the slope for the threshold voltage variations obtained from the simulated results presented in the inset of Fig. 7, is in agreement with the experimental data as far as the lower temperature range is considered (5–37°C).

5. Conclusions The results obtained from the simulations with BIOSPICE are in a good agreement with the measured

curves, proving the validity of the proposed ISFET model also under temperature variations. Therefore, the developed model could be conveniently used for designing ISFETs with their integrated controlling electronics, or for optimizing an ISFET fabrication process devoted to specific applications. Two facts affect the goodness of the BIOSPICE simulations with respect to the experimental results. First, the simulation parameters are wafer dependent, because the fabrication technology shows an actual good reproducibility inside the same wafer but not

S. Martinoia et al. / Sensors and Actuators B 50 (1998) 60–68

among different wafers or in the worst case among different fabrication runs. In our simulations the parameters were then calibrated with reference to a single wafer. Second, the procedures for extracting the electrochemical model parameter values have not been optimized yet. In fact, the temperature dependent parameter values of the reference electrode and of the electrochemical parameters of the ISFET have not been extracted from specific test structures, but they are obtained from the bare devices at room temperature or from literature. Moreover, the measurements and simulations show that the behaviour of the ISFET is thermally affected by two opposite contributions due to the electronic component (MOSFET underlying the sensor) and to the electrochemical component. The electronic component effect predominates, in terms of mV °C − 1, on the threshold voltage variations and it could be easily counterbalanced by using an ad hoc configuration of an ISFET and a MOSFET fabricated on the same chip [3].

Acknowledgements The authors are grateful to A. Piccini for his help in the experimental measurements. Work supported by a grant from the University of Genova (‘Fondi di Ateneo’).

References [1] P. Bergveld, The operation of an ISFET as an electronic device, Sensors and Actuators 1 (1981) 17–29. [2] O. Leistiko, The selectivity and temperature characteristics of ion sensitive field effect transistors, Physica Scripta 18 (1978) 445 – 450. [3] A. Sibbald, A chemical-sensitive integrated-circuit: the operational transducer, Sensors and Actuators 7 (1985) 23–38. [4] P. Bergveld, Design considerations for an ISFET multiplexer and amplifier, Sensors and Actuators 5 (1984) 13–20. [5] P.R. Barabash, R.S.C. Cobbold, W.B. Wlodarski, Analysis of the threshold voltage and its temperature dependence in electrolyte-insulator-semiconductor field-effect transistor (EISFET’s), IEEE Trans. Electron Devices 34 (6) (1987) 1271 – 1282. [6] H.-H. van der Vlekkert, N.F. de Rooij, Design, fabrication and characterization of pH Sensitive ISFETs, Analysis 16 (2) (1988) 110 – 119. [7] M. Grattarola, G. Massobrio, S. Martinoia, Modelling H + sensitive FETs with SPICE, IEEE Trans. Electron Devices 39 (4) (1992) 813 – 819. [8] G. Massobrio, S. Martinoia, M. Grattarola, Use of SPICE for modeling silicon-based chemical sensors, Sensors and Materials 6 (2) (1994) 101 – 123. [9] G. Massobrio, S. Martinoia, Modelling the ISFET behaviour under temperature variations using BIOSPICE, Electron. Lett. 32 (10) (1996) 936 – 937. [10] G. Massobrio, P. Antognetti, Semiconductor Device Modeling with SPICE, 2nd ed., McGraw-Hill, New York, 1993.

67

[11] H. Gelsfer, pH Measurements: Fundamentals, Methods, Applications, Instrumentation, VCH, Weinheim, 1991. [12] A. Lavarian, A. Lui, B. Margesin, G. Soncini, V. Zanini, An ISFET Compatible CMNOS Technology, Proceedings of MIELSD 93 Conference, Bled, Slovenia, 1993. [13] C. Fung, P. Cheung, W. Ko, A generalized theory of an electrolyte-insulator-semiconductor field-effect-transistor, IEEE Trans. Electron Devices 33 (1) (1986) 8 – 18. [14] D. Harame, L. Bousse, J. Shott, J. Meindl, Ion sensing devices with silicon nitride and borosilicate glass insulator, IEEE Trans. Electron Devices 34 (8) (1987) 1700 – 1706. [15] A. Sibbald, Chemical-sensitive field-effect transistors, IEE Proc. 130 (5) (1983) 233 – 244. [16] H.-H. van den Vlekkert, L. Bousse, N. de Rooij, The temperature dependence of the surface potential at the Al2O3/electrolyte interface, J. Colloid Interface Sci. 122 (2) (1988) 336 – 345.

Biographies Sergio Martinoia was born in Sanremo, Italy, in 1964. He received the Laurea degree in Electronic Engineering from the University of Genova in 1989 and the Ph.D. degree in Electronic Bioengineering in 1993. At present, he is a researcher in the Department of Biophysical and Electronic Engineering (DIBE). His main research interests are in the area of solid-state sensor modeling, solid-state sensor applications and multichannel microelectrodes for signal recording from networks of cells. Leandro Lorenzelli received the Laurea degree in Electronic Engineering from the University of Genova in 1994 and the Ph.D. degree in Materials, Components and Technologies for Electronics, in 1998. His research interests have focused on the development of solid state sensor based microsystems. He is currently working in the Department of Biophysical and Electronic Engineering (DIBE) of the University of Genova. Giuseppe Massobrio received the Laurea degree in Electronic Engineering from the University of Genova, Italy, in 1976. He is Research Associate in the Department of Biophysical and Electronic Engineering (DIBE) at the University of Genova. Since 1976 he has worked on semiconductor power device modeling and circuit design and simulation. Since 1987 he has been working on modeling semiconductor-based biosensors and neuronal structures. His extensive background in microelectronic device modeling includes teaching and research activities. He has contributed to several papers in the fields of bioelectronics and of semiconductor device modeling. He is co-author of the books ‘Semiconductor Device Modeling with SPICE’ published by McGraw-Hill and ‘Bioelectronics Handbook: Mosfets, Biosensors and Neurons’ published by McGraw-Hill.

68

S. Martinoia et al. / Sensors and Actuators B 50 (1998) 60–68

Paolo Conci was born in Trento in 1967. He received the Laurea Degree in Materials Engineering from the University of Trento in 1995. Since 1996 he has worked in the Microfabrication Group of IRST. His main research interest is in the area of solid-state sensor applications.

.

Alberto Lui was born in Quistello (MN) in 1960. He received the Laurea degree in Physics from the University of Bologna in 1986. In 1989 he joined the Microfabrication Groupe of IRST for characterization and failure analysis task.