For engineers developing III–V semiconductor devices, surface preparation is not simply a cleaning step: it can shape the electrical behavior of the dielectric interface. In InSb MOS capacitors, hydrochloric acid (HCl) treatment can leave an indium chloride (InCl3) surface layer, depending on the rinse chemistry. That layer is associated with a substantial flatband voltage shift when Al2O3 is deposited at 200 °C, while a higher deposition temperature removes it and changes other interface characteristics.
The results highlight a useful process consideration for InSb MOS structures: cleaning chemistry and thermal budget should be evaluated together. In particular, an interfacial layer may provide a way to adjust voltage behavior, but higher-temperature processing can affect interface-trap density and hysteresis.
How the HCl rinse changes the InSb surface
X-ray photoelectron spectroscopy (XPS) was used to compare InSb samples cleaned with HCl diluted in isopropanol (IPA) and rinsed with IPA against samples cleaned and rinsed with deionized water. Both surfaces showed signals associated with InSb and indium oxide. The IPA-treated surface also showed an In–Cl component, with the measured indium-to-chlorine ratio supporting its identification as InCl3. The water-treated surface showed negligible chlorine.
This difference matters because the acid treatment alone does not fully describe the resulting surface. The rinse influences whether the chloride remains: InCl3 is more soluble in water than in IPA, consistent with its absence after the HCl-water process. The study also found that brief exposure to ambient air during sample transfer did not prevent the InCl3 layer from being detected on the IPA-treated surface.
Heating progressively reduced the chloride signal. The reduction accelerated above 150 °C, and the InCl3 signal was no longer detected after annealing at 300 °C. In a separate temperature-ramp test, the chlorine signal fell below the instrument’s detection limit within five minutes after the sample reached 250 °C. These observations place the layer’s removal within the temperature range relevant to Al2O3 atomic layer deposition (ALD).
InSb MOS capacitors: linking interface chemistry to voltage
To examine the electrical consequences, MOS capacitors were fabricated with Al2O3 deposited at either 200 °C or 250 °C. The comparison included the two HCl cleaning and rinse methods, along with an untreated control. Capacitance–voltage measurements were used to assess flatband voltage (VFB), interface-trap density (Dit) and hysteresis voltage (VH).
At 200 °C, the HCl-IPA process produced a positive VFB shift of 0.79 V relative to the control. The HCl-water process did not produce a significant shift. This aligns with the surface analysis: the InCl3 layer was present after the IPA rinse but not after the water rinse. The findings associate the retained chloride layer with the flatband voltage change.
At 250 °C, the chloride layer was substantially desorbed. The voltage shift reversed, with both cleaning processes showing negative shifts relative to the control; there was no statistically significant difference between the two cleaning methods at that temperature. The authors attributed changes observed at 250 °C not only to chloride removal but also as likely related to the temperature sensitivity and degradation of the InSb substrate.
Voltage adjustment comes with process trade-offs
The 200 °C HCl-IPA result is notable because the positive flatband voltage shift was not accompanied by a significant change in Dit or hysteresis compared with the corresponding HCl-water process. This suggests that, under the tested conditions, the retained InCl3 layer can shift the capacitor’s voltage characteristic without a measured penalty in those other figures of merit.
The higher-temperature process presents a different balance. At 250 °C, Dit increased significantly for both cleaning methods. The increase was larger for the HCl-IPA process, reaching an average minimum Dit of 4.34 × 1012 cm-2 eV-1, compared with 2.25 × 1012 cm-2 eV-1 for HCl-water. The average minimum Dit for the control was 8.71 × 1011 cm-2 eV-1. At 200 °C, neither cleaning process showed a statistically significant Dit improvement over the control.
Hysteresis also warrants attention. All cleaning processes reduced hysteresis relative to the untreated control, with the lower deposition temperature giving the best results. However, the study did not find a statistically significant difference in hysteresis between the two cleaning methods. Together, these results show why process optimization should consider multiple electrical metrics rather than treating flatband voltage as the only outcome.
Implications for III–V device process integration
Flatband voltage is relevant to the voltage behavior of MOS transistors and can influence threshold voltage. The observed shift suggests a potential route to producing enhancement-mode and depletion-mode devices using otherwise similar fabrication processes. That possibility is promising for integration planning, but the demonstrated result is from MOS capacitors; transistor-level performance and process robustness require their own validation.
For process engineers and system integrators, the practical lessons are:
- Specify the full wet-clean sequence. The rinse chemistry can determine whether a chloride-containing interfacial layer remains.
- Manage the ALD thermal budget. Processing at 250 °C removes much of the InCl3 observed after HCl-IPA treatment, while the higher temperature is also associated with increased Dit.
- Evaluate electrical metrics together. Flatband voltage, interface traps and hysteresis respond differently to the cleaning and deposition conditions.
- Validate at the device level. A MOSCAP voltage shift indicates a possible threshold-voltage control mechanism, not a complete demonstration of transistor behavior.
Conclusion
HCl cleaning can create distinct InSb surface conditions depending on whether IPA or water is used for dilution and rinsing. With HCl-IPA treatment and 200 °C Al2O3 ALD, the associated InCl3 layer coincided with a +0.79 V flatband voltage shift, without a statistically significant difference in Dit or hysteresis relative to the HCl-water process. Raising deposition temperature to 250 °C desorbed the chloride and reversed the voltage shift, but significantly increased interface-trap density. The central process implication is to treat surface chemistry and ALD temperature as linked design choices: they may help tune voltage behavior, but must be balanced against interface quality.