Choosing a sensing membrane is central to improving ISFET sensitivity: the material at the electrolyte interface affects how pH-driven surface charge changes the transistor’s electrical response. A comparative Silvaco TCAD study assessed several candidate membranes against conventional silicon dioxide (SiO2), focusing on sensitivity and stability. Its results offer engineers a way to compare materials in simulation before committing to device fabrication, while keeping in mind that simulated performance still requires experimental validation.
How the sensing membrane shapes ISFET response
An ion-sensitive field-effect transistor (ISFET) detects changes in the chemical environment through changes in the electrical potential at the electrolyte–insulator interface. For pH sensing, hydrogen-ion activity influences the charge at the membrane surface. That surface potential couples to the transistor and shifts its threshold voltage, changing the measured drain-to-source current under a given bias.
The interface behaviour can be described using the Gouy–Chapman–Stern framework and a site-binding model. In practical terms, the membrane’s surface sites interact with ions in the electrolyte. As pH changes, the balance of surface charge changes too. The resulting potential shift influences the gate of the ISFET, so the membrane’s chemical and electrical properties contribute to the overall sensor response.
SiO2 has been used as a conventional gate dielectric, but the study describes it as offering comparatively low sensitivity and weaker stability. Candidate alternatives included TiO2, Ta2O5, ZrO2, Al2O3, HfO2 and Si3N4. These materials differ in dielectric properties and surface-site parameters, which affect their modelled response to pH.
Using TCAD to compare high-k ISFET materials
A key challenge in device simulation is representing the electrolyte. Commercial TCAD tools are designed primarily for semiconductor devices and do not inherently include all the material models and electrochemical processes needed to simulate an ISFET interface. The study addressed this limitation by defining the electrolyte as a user-defined material in Silvaco TCAD.
The approach represented electrolyte behaviour through semiconductor-equivalent parameters, including energy-band properties, permittivity and density of states. The model linked pH variation to changes in the represented carrier concentrations. It then used interface traps to introduce pH-dependent surface charge into the electrostatic simulation. This allowed the researchers to examine how the interface charge affected transistor characteristics within a device simulation environment.
The simulated device was evaluated through transfer characteristics: drain-to-source current was examined against reference gate voltage at different pH values. The analysis also considered threshold-voltage shifts across pH and compared the average sensitivity of the candidate membranes with SiO2. This approach connects the chemistry-informed surface model to device-level electrical behaviour, rather than treating material response as a standalone numerical calculation.
What the material comparison showed
Across the modelled materials, the high-k candidates generally showed greater sensitivity and better stability than SiO2. The comparison placed TiO2 at the top for sensitivity, followed by HfO2 and Ta2O5, then ZrO2 and Al2O3. Si3N4 was a notable exception among the alternatives: its sensitivity was below the other high-k materials, but remained higher than that of SiO2.
The study reported the following approximate sensitivities for several membranes:
- TiO2: 59.065 mV/pH, the highest value in the comparison.
- HfO2 and Ta2O5: approximately 59.03 mV/pH.
- Al2O3 and ZrO2: approximately 59.02 mV/pH and 59.007 mV/pH, respectively.
These values are close to the Nernstian sensitivity limit discussed in the study. The reported ordering is more useful as a comparative result within the model than as a guarantee of performance in a fabricated sensor. Actual device behaviour depends on the accuracy of the interface model and on real material and process characteristics, which are not established by simulation alone.
The simulated transfer characteristics also showed that response was not uniform across the full pH range. Sensitivity was lower at the range’s lowest and highest pH values than in the pH 5–9 region. The study additionally reported stability reaching 99.99% for the high-k materials. Because this figure comes from the simulation analysis, it should be interpreted in that context rather than as a measured field-performance result.
Why the simulation approach matters
For device designers and system integrators, the value of TCAD is not simply the ability to produce a ranking. A model that incorporates electrolyte-related surface charge can help link membrane selection to threshold voltage and current response. It can also support investigation of device and circuit behaviour before a fabrication run, when design changes may be less costly to evaluate.
The comparison suggests TiO2 is a promising candidate when modelled pH sensitivity is the primary consideration. HfO2 and Ta2O5 also performed strongly, while ZrO2 and Al2O3 showed similar, slightly lower values. Si3N4 may remain relevant where its distinct surface chemistry is of interest, although its sensitivity ranked below the other alternatives evaluated.
Practical implications for ISFET development
High-k sensing membranes can improve the modelled sensitivity and stability of an ISFET relative to SiO2, but material selection should be treated as one part of a broader design decision. The study’s Silvaco TCAD method provides a framework for comparing candidate membranes and examining their impact on electrical characteristics before fabrication. Its results point to TiO2 as the strongest simulated option for sensitivity, with HfO2 and Ta2O5 close behind. The next step is to validate and refine these predictions through further modelling, optimization and, ultimately, physical measurement.