The Problem of pH ISFET and the Investigation of this

As mentioned in the first section, the environment of pH detection is always in a wet condition. That is, the devices used for the pH measurement must be immersed in the water solution, and then the sensing materials react with the solution and exhibit the pH responses. For ISFETs, the sensing materials used for pH detections are the gate oxides, such as SiO2, Si3N4, Ta2O5, Al2O3, etc., however, these materials are not completely waterproof in nature. Some solution will permeate into the oxide surfaces and that may lead to the hydration effect. The hydration effect is a reaction correlated with time. The period of the hydration effect for Si3N4 to reach a steady state was about 50-60 hours in pH 7 solution [2], and for Al2O3 was about 7 hours [2].

Hydration effect is a critical problem in practical applications, and hence it limits

the commercialization speed and lowers the reliability of ISFET. Hydration can lead to a drift phenomenon of the gate voltage during the measurement, especially in the first two to three hours. When measuring the pH responses, the stability and accuracy are seriously influenced by the drift voltage.

In order to have a further understanding of the drift phenomenon, SiO2 grown by plasma-enhanced chemical vapour deposition (PECVD) was used for the sensing layer of ISFET. PECVD SiO2 is relatively poor in structural integrity and is expected to have an obvious hydration effect which can help us to observe the drift phenomenon more clearly. Detailed experiment steps will be presented in chapter 3, and the results will be discussed in chapter 4.

Chapter 2

Theories for the Investigation of Drift Characteristics

2.0 Introduction

In this chapter, the theories of metal oxide semiconductor field effect transistor (MOSFET) which are relevant to ISFET will firstly be presented. The fundamental principles of ISFET will be developed from these MOSFET theories. The pH response at the oxide-electrolyte interface will also be characterized in the first section.

Subsequently, the drift phenomenon which is caused by the hydration effect and the ions transport in the insulator will be discussed. In the final section, a physical model for drift developed by Jamasb [12] will be presented. This model can help us to have a further understanding of the mechanism responsible for the instability of ISFET under long-term operating.

2.1 Fundamental Principles of ISFET

Since the first report of the ISFET by Bergveld, research on new material and fabrication process to improve the sensitivity and stability has been continuously proposed [3-5]. At the same time, the mechanism of the pH response of pH ISFET has also been studied extensively [4-10]. The followings are the theoretical foundations which are mostly adopted to characterize the ISFET.

2.1.1 From MOSFET to ISFET

Seen by the history of the development of ISFET, it is not difficult to find out the similarities between ISFET and MOSFET. The most obvious characteristic is the similarity between their structures. Therefore, the best way to comprehend the ISFET is to understand the operating principle of a MOSFET first. When MOSFET is operated in the so-called ohmic or non-saturated region, the drain current ID is given by: where COX is the gate insulator capacitance per unit area, μ the electron mobility in the channel and W/L the width-to-length ratio of the channel.

The threshold voltage VT of Eq. (2-1) is described by the following expression:

F where VFB is the flat-band voltage, QB the depletion charge in the substrate, and ^B ψF

the potential difference between the Fermi levels of doped and intrinsic silicon. For a MOSFET with charge present in the oxide and at the oxide-semiconductor interface, the flat-band voltage can be given by:

OX

where ΦM is the workfunction of the gate metal, ΦSi the workfunction of silicon, QOX

the charge in the oxide and QSS the surface state density at the oxide-silicon interface.

Substitution of Eq. (2-3) in Eq. (2-2), the general form of the threshold voltage of a MOSFET becomes: the metal gate is no longer present, so that the term ΦM/q must be revised. Figure 2-1 illustrates the similarities and differences between these two devices. It can be seen

that the reference electrode, the aqueous solution and the phenomena occurring at the oxide-solution interface must be accounted for instead of ΦM/q. Hence the threshold voltage of the ISFET becomes: where Eref represents the constant potential of the reference electrode, χ^sol is the surface dipole potential of the solution which also has a constant value. The term Ψ0

representing the surface potential at the oxide-electrolyte interface is the key element that makes ISFET pH-sensitive.

2.1.2 The pH Response at Oxide-Electrolyte Interface

The surface of any metal oxide always contains hydroxyl groups, in the case of silicon dioxide SiOH groups. These groups can be protonated and deprotonated, and thus, when the gate oxide contacts an aqueous solution, a change of pH will change the SiO2 surface potential. These reactions can be expressed by

+

The pH response, which is generally called the sensitivity, is defined as the surface potential change over a pH unit change. This response is given by

q α

where pHB is the pH value in the solution bulk, k is the Boltzmann constant, T is the absolute temperature, C

B

dif is the differential capacitance, and βint is the intrinsic buffer capacity. α is a dimensionless sensitivity parameter. The value of α varies between 0 and 1 depending on the intrinsic buffer capacity and the differential capacitance. If α equals 1, the theoretical maximum sensitivity of -59.2mV/pH at room temperature can be obtained.

The potential between electrolyte solution and insulator surface causes a proton concentration difference between bulk and surface that is according to Boltzmann:

kT bulk, respectively. According to the definition of pH, Eq. (2-10) can be expressed by

kT

Drift phenomenon can be considered by two aspects of view, the hydration of the insulator surface after immersing it in pH buffer solution, and the trap of hydrogen-bearing species by the binding sites when they transport through the insulator. The former’s influence on drift is generally smaller than the latter. This result can be found in the previous work [2, 11-13] and also in the measurement data of this research. The following models, which are classified according to the location where the mechanism of pH-sensitivity is presumed to occur, will help us to have a further understanding of the transport of mobile ions [14]:

(1) Models based on the reactivity of the insulator surface. The surface sites on

the insulator react with ions in the solution and build up a surface potential. This will lead to the formation of the electrical double layer in the electrolyte at the interface with the insulator. This model is generally regarded as the site-binding model.

(2) Models based on the presence of mobile ions in the insulating layer. This implies the existence of a transport mechanism to establish the required thermodynamic equilibrium, and leads directly to a Nernst equation. This model is generally regarded as the gel model.

(3) Models based on the modification of the Si/SiO2 interface through a pH-controlled change in the surface state density via transport of a hydrogen-bearing species.

The above discussions are only the characteristics of ions transport in the insulator, while the physical model for the gate voltage drift is going to be presented in the next section.

2.3 Physical Model for Drift

The physical model for drift was firstly proposed by Jamasb in 1997 [11]. The key point of this model was that employing the dispersive transport theory to express the gate voltage drift which is caused by the hydration effect at the insulator-electrolyte interface.

2.3.1 Dispersive Transport

Dispersive transport was brief reviewed in [12] and it could be characterized by a power-law time decay of the mobility or diffusivity of the form t^β-1, 0<β<1. This time dependence is based on a model that interprets transport in terms of a “random walk”.

The origin of random walk is either a) hopping motion through localized states giving rise to hopping transport, or b) multiple trapping from a band of extended states or localized states leading to multiple-trap transport. The multiple-trap transport is generally associated with the motion of electrons or holes in disordered materials.

Regardless of the specific dispersive mechanism involved, dispersive transport leads to a characteristic power-law time decay of diffusivity which can be described by where D00 is a temperature-dependent diffusion coefficient which obeys an Arrhenius relationship, ω0 is the hopping attempt frequency, and β is the dispersion parameter satisfying 0<β<1. Dispersive transport leads to a decay in the density of sites/traps occupied by the species undergoing transport. This decay is described by the stretched-exponential time dependence given by where ΔNS/T(t) is the area density (units of cm^-2) of sites/traps occupied, τ is the time constant associated with structural relaxation, and β is the dispersion parameter.

2.3.2 Expression for Drift

Since hydration leads to a change of the chemical composition of the sensing oxide surface, it is reasonable to assume that the dielectric constant of the hydrated surface layer differs from that of the sensing oxide bulk. The overall insulator capacitance, which is determined by the series combination of the surface hydration layer and the underlying oxide, will exhibit a slow, temporal change. When drift phenomenon occurs at the surface of an actively-biased ISFET, the gate voltage will

simultaneously exhibit a change to keep a constant drain current. The change of the gate voltage can be written as

) Since the voltage drop inside of the semiconductor is kept constant, ΔVG(t) becomes

)] where VFB is the flatband voltage and Vins is the voltage drop across the insulator. VFB

and Vins are given by

where Qinv is the inversion charge. If the temperature, pH, and the ionic strength of the solution are held constant, Eref, χ^sol, Ψ0, and ΦSi can be neglected, so the drift can be

In this study, the gate oxide of the fabricated ISFET was composed of two layers, a lower layer of thermally-grown SiO2 of thickness, xL, and an upper layer of PECVD SiO2 of thickness, xU.. CI(0) is the effective insulator capacitance given by the series combination of the thermally-grown SiO2 capacitance, εL/xL, and the PECVD SiO2

capacitance, εU/xU. Ci(t) is analogous to CI(0), but an additional hydrated layer of capacitance, εHL/xHL, at the oxide-electrolyte interface must be took into consideration, and the PECVD SiO2 capacitance is now given by εU/[xU－xHL]. The series combinations of the capacitances are illustrated in Figure 2-2. Therefore, the drift is given by

From this equation, we observed that drift is directly proportional to the thickness of the hydrated layer. By applying dispersive transport theory, an expression for xHL(t) is given by [12]

where AD represents the cross-sectional area, and Nhydr is the average density of the hydrating species per unit volume of hydration layer. Thus, the overall expression for the gate voltage drift is

From this equation, we can expect that if the time of gate oxide immersing in the test-solution is long enough (determined by the constant τ ), the gate voltage drift will approach a constant value which is greatly dependent on the hydration depth, xHL(∞).

Chapter 3

Experiment and Measurement

3.0 Introduction

In this chapter, the advantage of making differential measurements between an ISFET and a reference FET (REFET) will be interpreted in the first section. The importance of a stable reference electrode (RE) in the miniaturized device will also be discussed. The second section is the fabrication process flow of the ISFET and REFET devices which are used for investigating the drift characteristics. Finally, the measurement setup, and the detection principles of pH and drift will be presented.

3.1 Differential Sensing

How to detect a correct and consistent pH value is always the direction of research. Besides adopting materials that have good linearity, sensitivity, and stability, there are two important subjects in measurement that provide alternative ways to obtain a reliable pH value. One is the design of a stable reference electrode, the other is the introduction of a REFET.

3.1.1 Reference Electrode

An ideal reference electrode for use as the ISFET gate terminal should provide [15]

a) an electrical contact to the solution from which to define the solution potential;

b) an electrode/solution potential difference (Eref) that does not vary with solution composition.

The conventional silver chloride or calomel electrode provides both of these functions by maintaining an electrochemical equilibrium with the solution. Novel techniques are to fabricate the reference electrodes in miniaturized dimensions [16,17]. The on-chip fabrication of a reference electrode with IC-compatible techniques would make ISFETs suitable for biomedical sensing because of the low cost, small size and rigidity.

3.1.2 Reference FET

An alternative technique to achieve consistent pH detections is through the co-fabricating of an ISFET and a REFET. An ideal REFET is a FET that insensitive to ions in pH measurement [18], but identical to the ISFET in terms of transconductance, thermal response, etc. The differential measurement between an ISFET and a REFET thus eliminates the variations of the environment, such as temperature, light, instable reference electrode/solution contact potential, etc.

In this study, an alternative approach to the implementation of a REFET is introduced. Ta2O5 is a material that exhibits good linearity, sensitivity, and stability in the pH measurement, so we are trying to take this material as the sensing layer material instead of an ion-insensitive one.

3.2 Fabrication Process

As mentioned in the previous section, Ta2O5 was taken as the sensing layer material of REFET, while SiO2 was the sensing material of ISFET. Ta2O5 was

deposited by sputtering, and SiO2 was deposited using plasma-enhanced chemical vapour deposition (PECVD). Because of the relatively low deposition temperature, PECVD SiO2 is poor in structural integrity in comparison with other preparation methods, such as LPCVD and thermally grown silica [19]. The poor integrity of the gate oxide structure is expected to have an obvious hydration effect which can help us to observe the drift phenomenon more clearly during the pH measurement. The following is the procedure for fabricating the ISFET and REFET devices, and the process is illustrated in Figure 3-1:

a) RCA clean

i) Al evaporation, 5000Å (mask 5)

Al sintering, 400°C, 30min

3.3 Measurement Principle

The setup of the measurement system is illustrated in Figure 3-2. ISFET and REFET were designed to share the same source terminal, and the drain-to-source voltage were biased at the same condition (VDS1=VDS2). The drain currents of ISFET and REFET were set at a constant magnitude, therefore, if the pH of the solution varies, the gate voltage must adjust its magnitude to maintain the constant current.

Consequently, the variation of the gate voltage exhibits the pH sensitivity of the sensing oxide. Figure 3-3 illustrates the detection principle of pH. For the drift measurement, the detection principle is in a similar manner to that of the pH measurement and is shown in Figure 3-4.

Chapter 4

Results and Discussions

4.1 Sensitivities of PECVD SiO

and Sputtered Ta

O

Figure 4-1~4-12 are the measured sensitivities of PECVD SiO2 and sputtered Ta2O5, and the data are sorted in Table 4-1. The samples of SiO2 were prepared in dimensions of W/L=400μm/20μm, 400μm/30μm and 400μm/40μm as well as in thickness of 100Å, 300Å, 500Å and 1000Å. But the Ta2O5-gate REFETs were of the thickness of 300Å only. Followings are the discussions of the measurement results:

(1) The variations in thickness of the SiO2 sensing oxide showed no influence on the sensitivity. This result can be explained by the Nernst equation for sensitivity and also the site-dissociation model [10] that the pH sensitivity is presumed to occur at the surface of the sensing layer.

(2) There is no consistency of sensitivities among the PECVD SiO2-gate ISFETs. For PECVD deposited SiO2 films, lacking of uniformity were innate problems such that the PECVD SiO2-gate ISFET devices, even were fabricated on the same wafer, could exhibit different characteristics.

(3) Devices of different dimensions have no critical influence on the sensitivity.

As what have been discussed in (2) and (3), the sensitivity of ISFET device was determined by the surface characteristics of its specific sensing film, but not by the FET device itself.

(4) Ta2O₅ shows a good linearity of the sensitivity from pH 1 to pH 13, and it also has a high sensitivity of 54.6mV/pH in average.

(5) The sensitivity of SiO2 is relatively higher in the pH range of 1-9 than in

the range of 9-13.

(6) By Taking Ta2O5-ISFET as the REFET, the SiO2-ISFET exhibits a relatively higher sensitivity in the pH range of 9-13. It is contrary to the result of applying solely a SiO2-ISFET that the sensitivity is relatively lower in the pH range of 9-13.

4.2 Drift Characteristics

Drift characteristics of PECVD SiO2-gate ISFETs are shown in Figure 4-13~4-16.

Measurements were carried out in pH 7, and the duration of each measurement was about 6 hours. Average drift rate per hour was calculated by averaging the gate voltage drift in the last 5 hour, the results were 7.4, 12.8, 15.4 and 16.2mV/hour for 100Å, 300Å, 500Å and 1000Å of PECVD SiO2 respectively. These results are listed in table 4-2. Figure 4-18 shows the drift characteristic of Ta2O5. The measured drift rate of Ta2O5 was 1.6mV/hour. Followings are the discussions of the drift characteristics:

(1) The drift rate of 300Å SiO2 has a critical increase from 100Å SiO₂. This critical increase reflects that the 300Å SiO2 captures more H⁺ ions than 100Å SiO2

can do in the same time duration. However, the transport of H⁺ ions is presumed that it takes place in the hydrated layer because the hydration layer has a sufficiently open structure that ionic mobilities are much higher than in the rigid bulk region [14]. That is, 300Å SiO2 has a deeper hydration depth and more binding sites than 100Å SiO2. This suggests that the 100Å SiO2 is fully hydrated, and the 100Å hydration depth of a PECVD-deposited SiO2 film is reasonable in comparison with the LPCVD-grown Si3N4 film that has a hydration depth of about 105Å [12].

(2) The drift rate of 500Å SiO2 also has a critical increase from 300Å SiO2, but

the increment is not as large as that from 100Å to 300Å. The explanation is same as it of (1), and the 300Å SiO2 is suggested fully hydrated. The large hydration depth of the SiO2 film is reasonable to a PECVD-related process that has a lattice structure not as dense as a high-temperature process.

(3) The drift rate of 1000Å SiO2 has only increased by 5% from 500Å SiO₂. It suggests that 500Å SiO2 has almost provided an enough thickness for the maximum hydration depth it may reach after 6 hours of hydrating as well as 1000Å SiO2 can do.

(4) The drift rate of PECVD SiO2 is obviously larger then other sensing oxides proposed in the literature [20] such as Al₂O₃ and Si₃N₄, and it is also larger than the measured drift rate of Ta2O5 of this research. This result is consistent with the expectation of the drift rate of PECVD SiO2 mentioned in chapter 3.

(5) The initial drift rates of the four samples are roughly the same in the initially first hour, and the drift of the gate voltage is obviously larger than the total amount in the last 5 hours. It was regarded that the density of Si-OH sites near the SiO2 surface was large, and the OH sites buried in SiO2 was several orders in density smaller than the surface OH sites [14].

From (1), (2) and (3) it is not hard to figure out that the hydration depth is around

From (1), (2) and (3) it is not hard to figure out that the hydration depth is around