Sensors and Actuators B, 4 (1991) 125-133 125
Future Applications of ISFETs*
P. BERGVELD University of Twente, Box 217, 7OOAE Enschede (The Netherlands)
Abstract seen as directly of interest for the electrophys- iological experiments. The non-selectivity of The application of ISFETs has been fo- silicon dioxide has recently been confirmed cused up to now mainly on static measure- by independent measurements [ 31. ments, comparable with the conventional way With respect to the measurement of ion of applying ion-selective electrodes. The spe- fluxes, it should be noted that the related cific and unique properties of ISFETs, their current densities are in the order of some small dimensions and very fast response, are rnA/cm’, resulting in ionic transports in the not exploited explicitly in the case of static order of picomoles/cm2. However, by catch- measurements. These features will, however, ing the ionic flux in a very small cavity, which be explored more and more in the future, for is achieved by pressing the ISFET with its example in flow-through and FIA systems as surrounding ring of encapsulant firmly well as for ISFETs with an integrated coulo- against the nerve membrane, the ionic metric actuator. In the latter case the ISFET changes may rise to 0.1 millimol/litre, easily is part of an active microchemical reactor. detectable by an ISFET. The recorded Similar systems may in the future contain monophasic pulses [4], as shown in Fig. 1, bioactive materials, whose affinity to a bio- are therefore most probably indeed the result logical analyte can be measured by means of of very local ionic concentration changes, stimulus-response experiments. although this has never been proved in any later paper. In the context of the present paper, it is IntroduetIon important to notice that in the electrophysio- logical ISFET application as mentioned Relevant future applications of ISFETs above, the fast response of an ISFET is ex- cannot be discussed without considering the plicitly exploited, combined with its small original research, development and applica- dimensions. tion aims and the corresponding results as In another, rather isolated paper by the achieved up to now. present author, a chemically insensitive The author’s original intention in develop- device has been described, related to the ear- ing the ISFET was to measure the ionic lier ISFETs [5]. These devices were intended sodium influx and potassium efflux at nerve to be ISFETs, but, probably due to contami- and muscle membranes [ 1,2]. The silicon nation of the gate oxide by a very thin epoxy dioxide of a MOSFET, from which the gate layer, arising from the encapsulation of the metal was etched away, appeared to be sensi- devices, no sensitivity to ions was observed, tive for sodium and potassium ions, and thus at least not for an initial period of contact well suited for the intended application. Sen- with the extracellular liquid. The device oper- sitivity to pH was also observed, but was not ated as a purely capacitive electrode, similar to a MOSFET. Because the only difference from the MOSFET was the omission of the *Invited paper. gate metal, this device was called an OSFET.
09254005/91/$3.50 0 Elsevier Sequoia/Printed in The Netherlands 126
was not directed towards the electrophysio- logical application of ISFETs and related devices, but to the measurement of ionic con- stituents of various body liquids, mostly blood. This was due to the lack of any small conventional type of ion-selective electrode suitable for in vim application. This is still Fig. 1. Original registration of monophasic pulses with the reason that many ISFET projects at Uni- an ISFET pressed to the Locusta flexor tibialis. versity as well as industrial laboratories are devoted to extra- or intracorporeal measure- Nowadays such a device would be called a ments of blood chemical constituents. REFET. However, the original OSFET Application in analytical chemistry is less differed from present REFETs in that the gate pursued, because of the apparent adequacy was surrounded by a relatively large parasitic of conventional ion-selective electrodes and gate. The result was that extracellular action other analytical methods. This may, however, potentials could be measured from a muscle change in the future as soon as the specific fibre which contacted the actual gate as well as properties of ISFETs are more clearly recog- the parasitic gate. In fact the potentials as nized, such as their extremely short response caused by extracellular ionic currents were time, which is of great advantage in the case measured by means of a capacitive division of flow injection analysis (FIA) and other between the actual gate capacitance and the microchemical investigations, for example. parasitic gate capacitance. The recordings ex- Although the electrophysiological experi- celled in a very good signal-to-noise ratio as ments, as mentioned in the Introduction, al- compared to the usual recordings with metal ready showed from the beginning that ion microelectrodes and also the broad frequency fluxes occurring within milliseconds could range (0
Differential Measurements Application of ISFETs in, Flow-through Cells To solve the reference electrode problem, a The fast response of an ISFET has already differential measurement set-up has been pro- been exploited in flow-through cells, where a posed [ 1 l] in which case the ISFET response sample is injected into a carrier stream, as is measured with respect to a REFET [ 121, schematically drawn in Fig. 2. while the analyte potential is defined by a The first system based on this approach platinum wire, as shown schematically in was published in 1980 [ 81, while Sibbald in- Fig. 3. troduced it for on-line clinical analysis of The unstable electrolyte/platinum elec- blood ionic composition, using a flow- trode potential is a common signal to the through cell with four integrated membrane- differential system and will thus not interfere covered ISFETs [9]. Especially adapted to the with the final output signal. Therefore the flow-through cell configuration, backside platinum electrode is usually called a pseudo- contacted ISFETs can be used with great reference electrode. advantage for this type of measurement, as However, for static measurements such recently shown by Alegret [lo]. systems require that the REFET is at least as In all the examples mentioned above, the stable as the ISFET, while furthermore the reference electrode consisted of a conven- electrical characteristics of the ISFET and
Fig. 2. Schematic representation of single ISFET flow- Fig. 3. ISFET/REFET differential measurement with through cell arrangement. platinum pseudo-reference electrode. 128
REFET should be identical. The development such that one of the devices should become in of REFETs which can meet these require- equilibrium with the sample solution and that ments is still in its infancy, especially for cases during this period the other should continu- where the analyte contains an unknown com- ously be in contact with the carrier solution. position of many types of ions. For dynamic This requirement will generally result in rela- measurements the REFET can, however, be tively long distances between the two separate replaced by an ISFET which is temporarily in chips. contact with a constant pH; this condition can This type of flow-through cell is explored be fulfilled in a flow-through cell. by Gumbrecht et al. for the measurement of blood samples in a bedside unit, where blood samples and calibration solution are carried to Flow-throughCell with Two IdenticalISFETs the cell by means of a double lumen needle [131. Differential measurements as proposed in A similar system for the measurement of the previous Section can be performed with a urea is applied by Chandler et al., where flow-through system as schematically shown between the two identical ISFETs an enzyme in Fig. 4. column is mounted, in this particular case If only the carrier solution is pumped loaded with urease [ 141. Urea injection results along the two ISFETs, the differential signal in measured peaks (differential), whose ampli- should be zero. If not, the difference can be tude is proportional to the concentration of set to zero. After injection of the sample urea. solution, ISFET, will first sense the sample, while ISFETz is still in contact with the car- rier solution, thus acting at this moment as a ISFET with RetardedResponse pseudo-REFET. The differential signal thus as Pseudo-REFET reflects the sample concentration with respect to the carrier concentration. As soon as the As already mentioned, a maximum signal sample contacts both ISFETs, the differential in the flow-through cell, as described in the signal will again be zero. After the sample previous Section, can only be achieved if the leaves ISFET, , this ISFET will be the distance between the two ISFETs is relatively pseudo-REFET and ISFET2 will measure the large. This distance can be decreased in the sample with respect to the carrier solution. case that the response time of the second The signal will then be reversed in polarity, ISFET has been artificially increased by but will show the same amplitude as the first adding a membrane which limits the diffusion signal. of ions. This system is shown in Fig. 5. A positive point of this type of flow- In contrast to the system shown in Fig. 4, through cell is that two identical sensors can both ISFETs can now be in contact with the be used, but a negative point is that they sample solution at the same time. The second cannot be integrated in one chip. The dis- ISFET should be slow enough to achieve a tance between the two ISFETs should be maximum signal with the first ISFET before
1-t ISmm. Pi Fig. 5. Flow-through cell with one fast ISFET and one Fig. 4. Flow-through cell with two identical ISFETs slow ISFET (retarded response) and pseudo-reference and pseudo-reference electrode. electrode. 129 the second ISFET starts to respond. A draw- back of this approach may be that the inter- sample time now also increases, because the membrane of the second ISFET should be washed after each sample, limiting the num- ber of samples per unit of time. The system is described in the literature by van den Vlekkert et al., using a polyHEMA layer of 200 pm for the pseudo-REFET, re- Fig. 6. Schematic representation of flow injection volu- sulting in a response delay of 200 s [ 151. This metric titration system. is sufficiently low as compared to the time constant ( 10 s) of the membrane-covered ISFET ion sensors (K+ and Ca+ +) in the be proportional to the concentration of the flow-through cell. sample. Although van den Vlekkert et al. use sepa- Comparing this FIA titration method with rate chips for the ion sensors and the pseudo- the different flow-through methods described REFET, it is technologically not difficult to in the previous Sections, it should be noted integrate the devices into one chip. Using that now not the peak height but the peak micromachining techniques it might be possi- width of the response curves is the actual ble that in the future a sample inlet valve will signal which provides the analytical data. also be integrated. This reduces the requirements of stability and reproducibility of the applied ISFETs. It will be clear that for the detection of the Flow Injection Titration Systems occurring pH changes, in the order of 10 to 30 s, an ISFET is an attractive type of sensor. Volumetric acid/base titration is a well- This is even more the case with a coulometric known technique to determine the acidity of titration, when the titrant is generated at an a sample by means of adding a certain adjacent electrode by means of coulometry. amount of base or acid to the sample and to Especially with small samples, this technique detect the so-called equivalence point in the results in very fast titrations, as will be shown titration curve. This is the point where the in the next Section. added reagent exactly neutralizes the sample. For high-speed titrations the pH of the mix- ture will change very rapidly and therefore a Coulometric Titrations fast pH sensor, such as the ISFET, is a very attractive detector. Wang et al. showed the Van der Schoot et al. constructed for the usefulness of an ISFET by injecting a 40 ~1 first time a complete coulometric titrator in sample via a gradient chamber in a titrant silicon [ 171. solution, which flows continuously along an The titration ‘vessel’ consisted of a cavity ISFET and a conventional reference elec- of 80 pm depth, etched in a piece of silicon, trode, placed directly beyond the ISFET [ 161. containing also two etched holes, which is The system is schematically shown in Fig. 6. sealed on a planar array of ten ISFETs as The concentration gradient of the sample schematically shown in Fig. 7. passes the ISFET twice, so the equivalence Each of the ISFETs is surrounded by a point between sample solution and titrant is thin-film gold electrode. The application of also encountered twice during one titration a current between two of these electrodes cycle. The time span between the two equiva- will result in the electrolysis of water after lence points is used to determine the sample the cavity is filled via the holes with a sam- concentration and this parameter appeared to ple solution from which the acid or base 130
tion time is proportional to the acid or base concentration of the sample. By providing the system shown in Fig. 7 analyzer chip with an additional hole in the silicon cover, over which a Teflon membrane is deposited, a COZ sensor has been created [ 191. Because the acidity of an internal bicarbonate solution will be a function of the external COZ pres- AU T@? sure, this pressure could be determined by SO, applying continuously forward and back ti- tration cycles. The system could measure, after one initial calibration, for at least 50 Fig. 7. Cross section of coulometric titration system. days without recalibration, resulting in a drift of 10%. This is in terms of pH only 0.001 concentration has to be determined. The elec- units per day, which is a great improvement trolysis of water results in a local decrease of compared with a ‘normal’ ISFET drift of the pH at the anode and the reverse at the 0.02 pH/h. cathode. This local coulometric titration is measured by the ISFET inside the actuator electrode with respect to an ISFET which is Ion-step Experiments with ImmunoFETs not positioned in a titrated area. Note that in this case a normal ISFET acts as pseudo- It has long been debated whether it would REFET, while one of the gold electrodes not be possible to measure the charge redistribu- applied for the pH generation is used as a tion that is characteristic for an antibody/ pseudo-reference electrode. It appeared that antigen interaction with an ISFET, if first the the sample could be titrated in a few seconds, antibodies are immobilized on the surface with the ISFET acting only as an indicator to and then the created IMFET is incubated determine the equivalence point of the coulo- with the corresponding antigens. Nowadays metric titration. It will be obvious that the it is recognized that this is not possible, be- fast response, small dimensions, as well as the cause, although proteins are charged planar performance of the ISFET are in this molecules, the counter charges are in the example in fact the key features of the tech- direct vicinity, preventing them from generat- nique. ing a static external electric field. Neverthe- In the dynamic measurement mentioned less, the existence of fixed charges of proteins, above the ISFET is no longer used in a static immobolized in a layer onto an ISFET sur- way and the d.c. value of the ISFET amplifier face, can be detected by using a dynamic output voltage is no longer of importance. method of measuring. For that, the protein- Instead the elapsed time between the onset of covered ISFET should be exposed to a sud- the actuator current and the detection of the den change in electrolyte concentration [20]. equivalence point is measured. This titration This process is schematically represented in time is proportional to the acid or base con- Fig. 8. centration of the sample. Possible ISFET After a stepwise increase of the electrolyte drift or sensitivity degradation does not influ- salt concentration, usually KCl, the ions will ence the measurement results. diffuse into the protein membrane, but not The same titration experiments are also with the same speed, due to a difference in possible in an open vessel with a dip-stick mobility caused by the presence of the fixed system, provided that the liquid is unstirred charges. This means that the initially constant [ 181. As long as the mass transfer is diffusion output voltage of the ISFET amplifier shows controlled, now the square root of the titra- a temporary disturbance. The amplitude of 131
on the measurement of induced transient membrane potentials, occurring within sec- onds, requiring a sensor with a fast response. Again in this case the ISFET is an excellent candidate, where the sensitivity for electrical potentials, superimposed on a relatively sta- ble baseline, is the only requirement. It is striking that in fact this is the same require- p-3 ment as for the original electrophysiological experiments mentioned in the Introduction.
Conclusions and Future Outlook I\__ I Flow -through Cells With respect to the different flow-through cells described in the first Sections, it should l&q----y , be mentioned that these type of measure- 0 - kne (s”, ments are often called flow-injection analysis (FIA) experiments, because a small sample is Fig. 8. Schematic representation of immunoFET to injected into a carrier stream. However, all which the ion step procedure is applied. systems described in the literature up to now make use of equilibrium measurements be- the pulsatile membrane voltage appears to be tween the sample solution and the ISFET, a function of the total effective fixed mem- while real FIA systems employ non-equi- brane charge. This could be proved by gradu- librium data [21]. These measurements re- ally changing the pH of the solution and in quire however, that the calibration and the meantime carrying out the ion step proce- sample solutions have the same invariant vis- dure at regular intervals. Because proteins are cosity. Furthermore, the hydrodynamics of amphoteric molecules, the protein charge will the injection valve and the analysis cell, in- change as a function of the changing pH. cluding the applied tubing, as well as the Consequently a point, the so-called isoelectric response time of the sensor should remain point, will be passed where the proteins are constant during the experiment. Only when uncharged. At this point the amplitude of the these requirements are fulfilled can non-equi- transient membrane potential indeed ap- librium data also be used for the chemical peared to be zero. Before or beyond this analysis of the sample, like peak height, peak point the amplitudes were positive or nega- width at selected level and peak area. Maybe tive, due to the corresponding negative or in the future these requirements can be met positive protein charge at that specific pH. In by constructing the complete FIA system in this way the difference in isoelectric points of silicon as an integration of ISFET, valves and antibodies, adsorbed to a porous membrane channels by means of micromachining. which was deposited onto the ISFET surface, and the subsequently formed antibody/anti- Microlitre Titrations gen complexes could be detected. It should be noted that the coulometric Although the method has still to be evalu- microlitre titration experiments are limited to ated and further developed to achieve an the measurement of acid-base concentrations IMFET which can be used practically for the as long as no solid-state actuators exist which direct monitoring of an immunological reac- can generate or consume ions other than tion, it will be clear that the method is based protons, to be measured by corresponding 132 ion sensors. The exploitation of other elec- REFETs. A REFET is in fact a capacitive troactive materials, such as Prussian Blue type of electrode, applied to measure the [22], may change this in the future, although electrical potential of the electrolyte. In the no experience exists at the present moment electrophysiological experiments mentioned with respect to actuator properties of this in the Introduction [5], the capacitive refer- type of materials. ence electrode consisted of a parasitic gate In the microlitre titration vessel [ 171 and surrounding the actual gate of the measuring the dip-stick titrator [ 181, the actuator was FET. In this way voltage pulses in the milli- shaped around the ISFET gate. Generated second range could easily be detected. Espe- protons or hydroxyl ions have therefore to cially in the case of dynamic measurements, diffuse over a certain distance, typically which result in ever shorter pulses, this ap- 10 ,um, before the ISFET will respond. This proach may in the future again be worth limits the speed of titration. As soon as the considering. distance between the ISFET and the actuator is decreased, for example, by applying a porous electrode over the gate of an ISFET, References it can be expected that the speed of titration will be enlarged, in which case the fast 1 P. Bergveld, Development of an ion-sensitive solid response time of an ISFET is again very state device for neurophysiological measurements, IEEE Trans. Biomed. Eng., BME-I 7 (1970) 70-71. favourable. 2 P. Bergveld, Development, operation and applica- By ultimate integration of ISFET and ac- tion of the ion-sensitive field effect transistor as a tuator in the near future, a new area of tool for electrophysiology, IEEE Trans. Biomed. microchemical analysis may come into Eng., BME-19 (1972) 342-351. existence. 3 S. C. Chen, Y-K. Su and J. S. Tzeng, The fabrication and characterisation of ISFETs with silicon dioxide gate, J. Phys. D: Appl. Phys., 19 (1986) 1951- 1956. Ion Step Experiments 4 P. Bergveld, Ion sensitive field effect transistor, The new approach of dynamic measure- Proc. Conf. Biocapt ‘75, Vol. 1, Paris, 1975, pp. ments with ImmunoFETs is at present a first 299-304. example of a stimulus-response type of mea- 5 P. Bergveld, J. Wiersma and H. Meertens, Extracel- lular potential recordings by means of a field effect surement. The ion step induces a membrane transistor without gate metal, called OSFET, IEEE diffusion potential due to the difference of ion Trans. Biomed. Eng., BME-23 (1976) 136- 144. mobilities of cations and anions caused by 6 A. Cambiaso, M. Grattarola, G. Amaldi, S. Marti- the net charge density of the biomolecules noia and G. Massobrio, Detection of cell activity via inside the membrane. The thickness of the ISFET devices; modelling and computer simula- membrane has been chosen relatively arbi- tions, Sensors and Actuators B, I (1990) 373-379. 7 P. Bergveld, Exploiting the dynamic properties of trarily around 5 pm, resulting in pulses with a FET-based chemical sensors, J. Phys. E: Sci. In- duration of seconds. Optimization of the pro- strum., 22 (1989) 678-683. cedure may however result in much thinner 8 A. U. Ramsing, J. Janata, J. Ruzicka and M. Levy, membranes, which are necessary to decrease Miniaturization in analytical chemistry-a combi- the incubation time, for example, leading to nation of flow injection analysis and ion-sensitive much faster responses. The application of an field effect transistors for determination of pH, and potassium and calcium ions, Anal. Chim. Acta, 118 ISFET guarantees in this case that these (1980) 45-52. pulses can also be measured. 9 A. Sibbald, P. D. Whalley and A. K. Covington, A miniature flow-through cell with a four-function in- Reference Electrodes tegrated circuit for simultaneous measurements of In the flow-through cell experiments it was K+, H+, Ca+ +, and Na+ ions, Anal. Chim. Acta, 159 (1984) 47-62. shown that the conventional liquid-filled 10 S. Alegret, J. Alonso, J. Bartroli and J. Domenech, reference electrode can be replaced by Flow-through pH-ISFET detector for flow injection solid-state electrodes like REFETs or pseudo- analysis, Anal. Chim. Acta, 222 (1989) 373-377. 133
11 P. Bergveld, A. van den Berg, P. D. van der Wal, 17 B. H. van der Schoot and P. Bergveld, An ISFET- M. Skowronska-Ptasinka, E. J. R. Sudhiilter and D. based microlitre titrator: integration of a chemical N. Reinhoudt, How electrical and chemical require- sensor-actuator system, Sensors and Actuators, 8 ments for REFETs may coincide, Sensors and Actu- (1985) 11-22. ators, 18 (1989) 309-327. 18 W. Olthuis, B. H. van der Schoot, F. Chavez and P. 12 M. Skowronska-Ptasinska, P. D. van der Wal, A. Bergveld, A dip-stick sensor for coulometric acid- van den Berg, P. Bergveld, E. J. R. Sudhiilter and D. base titrations, Sensors and Actuators, 17 (1989) N. Reinhoudt, Reference field effect transistors 279-284. based on chemically modified ISFETs, Anal. Chim. 19 B. H. van der Schoot and P. Bergveld, Coulometric Acta, 230 (1990) 67-73. sensors, the application of a sensor-actuator system 13 W. Gumbrecht, W. Schelter, B. Montag, M. Rasin- for long-term stability in chemical sensing, Sensors ski and U. Pfeiffer, On-line blood electrolyte moni- and Actuators, 13 (1988) 251-262. toring with a CHEMFET microcell system, Sensors 20 R. B. M. Schasfoort, R. P. H. Kooyman, P. and Actuators B, 1 (1990) 477-480. Bergveld and J. Greve, A new approach to Im- 14 G. K. Chandler, J. R. Dodgson and M. J. Eddowes, munoFET operation, Biosensors Bioelectron., 5 An ISFET-based flow injection analysis system for (1990) 103-124. determination of urea; experiment and theory, Sen- 21 E. H. Hansen, Flow injection analysis-recent devel- sors and Actuators B, I (1990) 433-437. opments and future trends, in G. G. Guibault and M. 15 H. H. van den Vlekkert, N. F. de Rooij, A. van den Mascini (eds.), Analytical Uses of Immobilized Bio- Berg and A. Grisel, Multi-ion sensing system based logical Compounds for Detection, Medical and Indus- on glass-encapsulated pH-ISFETs and a pseudo- trial Uses, Reidel, Dordrecht, 1988, pp. 291-308. REFET, Sensors and Actuators B, l(l990) 395-400. 22 H. Hartman, E. W. Grabner and P. Bergveld, Prus- 16 G. H. Wang, D. Yu and Y. Sun, pH-ISEFT used as sian Blue-coated interdigitated array electrodes for a detector in a flow injection titration system, Sen- possible analytical application, Anal. Chins. Acta., sors and Actuators, 19 (1989) 41-52. submitted for publication.