Metal work function engineering can help control the electrical behavior of a metal-oxide-semiconductor (MOS) device by changing the gate material. In germanium (Ge), however, interface states can pin the semiconductor’s Fermi level and limit how much influence a gate metal has. Measurements on epitaxial n-type Ge MOS capacitors show that aluminum and platinum gates can produce substantial flat-band voltage shifts on both (100)Ge and (110)Ge surfaces, indicating meaningful Fermi-level modulation in these structures.
The findings are relevant to engineers evaluating Ge as a channel material for future field-effect transistors (FETs). They also highlight an important design consideration: gate-metal selection cannot be considered separately from the oxide-semiconductor interface and the crystal orientation of the Ge layer.
Why gate work function matters
In a MOSFET, the gate influences the electrical potential of the semiconductor channel. The difference between the gate metal’s work function and the semiconductor’s work function contributes to the flat-band voltage, or VFB. This voltage is the point at which there is no band bending in the semiconductor beneath the gate. Changes in VFB can, in turn, affect the transistor threshold voltage, VTH.
Using different gate metals is one way to adjust this relationship. But at a semiconductor surface, electrically active states and defects can form an interface dipole. This changes the alignment of energy bands and can cause Fermi-level pinning (FLP), reducing the effect of changing the metal. The Schottky pinning parameter, S, describes this response: a value near zero corresponds to strong pinning, while a value closer to one indicates weaker pinning and greater sensitivity to the metal work function.
For two MOS capacitors with the same oxide and semiconductor structure but different gate metals, the difference in flat-band voltage, ΔVFB, can be used to estimate S. A larger shift indicates that more of the difference between the metals’ work functions is expressed at the device interface.
Metal work function engineering on epitaxial germanium
The devices used n-type epitaxial Ge layers grown on GaAs substrates, with an AlAs intermediate buffer. The study compared (100)Ge and (110)Ge structures. Each MOS capacitor included a thermally formed GeOx interfacial layer and an aluminum oxide (Al2O3) gate dielectric. Aluminum and platinum-based gate stacks were selected because their work functions are substantially different, allowing the resulting voltage shift to be evaluated.
Capacitance-voltage and conductance-voltage measurements were performed across a range of frequencies and temperatures. These measurements provide information about the device’s electrostatic response, interface defects, and flat-band voltage. The reported results were:
- (100)Ge: A flat-band voltage difference of about 701 mV and an estimated S value of 0.512.
- (110)Ge: A flat-band voltage difference of about 816 mV and an estimated S value of 0.596.
- Both orientations: Interface trap density (Dit) of approximately 6 × 1011 cm−2 eV−1 in the energy range 0.05 to 0.3 eV below the conduction band.
The measured voltage shifts, greater than 700 mV, demonstrate that the metal choice had a substantial effect in these epitaxial Ge MOS capacitors. The estimated S values indicate partial, rather than complete, Fermi-level pinning. The reported platinum work function can vary depending on the gate stack and processing conditions, so the authors describe the extracted pinning parameter as a lower-bound estimate.
What crystal orientation changes
The two orientations did not show identical electrical behavior. The (110)Ge devices displayed an inversion-like response at low gate overdrive across the measured frequencies. Cooling to 78 K reduced, but did not eliminate, this response. The (100)Ge capacitors did not show the same behavior.
The measurements do not establish a single cause for the difference. Possible contributors discussed include minority-carrier generation, shallow defect states that can respond to rapid voltage changes, and gate leakage or charge injection. The researchers also point to differences in the dielectric properties of the oxide stacks on the two orientations. In their measurements, the electronic dielectric constant was approximately 3.5 for the (100)Ge stack and 3.3 for the (110)Ge stack.
This distinction matters because Fermi-level pinning depends not only on the density of interface states, but also on how the dielectric stack screens their influence. Even a modest difference in measured dielectric properties can accompany differences in interface quality and extracted pinning behavior. Orientation-specific characterization is therefore important when comparing candidate Ge device structures.
Interface quality remains a key design variable
The measured interface trap densities were comparable to values reported for bulk Ge MOS capacitors. This supports the quality of the fabricated epitaxial layers and the MOS-capacitor process, but it does not mean interface defects are irrelevant. Lowering Dit could enable further Fermi-level unpinning and increase the influence of gate-metal work function.
For device development, the practical lesson is to assess the complete gate stack rather than treating the gate metal as an isolated tuning element. The Ge surface, interfacial oxide, high-k dielectric, crystal orientation, and processing history all contribute to the electrostatics. Capacitance-voltage measurements with more than one gate metal can help distinguish work-function effects from interface-related limitations, while conductance-based analysis can provide information about interface traps.
Implications for Ge transistor development
These results show that aluminum and platinum gate metals can shift the flat-band voltage substantially in epitaxial (100)Ge and (110)Ge MOS capacitors. The demonstrated S values above 0.5 indicate that the gate-metal work function is not fully screened by Fermi-level pinning in these structures. At the same time, the orientation-dependent response and the role of interface defects show why material and process details must be evaluated together.
For teams assessing Ge-based FETs, the work provides evidence that epitaxial Ge integrated on GaAs with an AlAs buffer can support MOS-capacitor structures suitable for studying gate work-function control. Continued optimization of interface quality and orientation-specific dielectric behavior will be important for translating that control into device designs.