Annealed PtSe2 Films Reach 60 GHz Telecom Photodetection

PtSe2 is attractive for infrared optoelectronics because multilayer films are semimetallic, absorb broadly, and remain stable in air. The practical bottleneck is less about the intrinsic material and more about process control: films grown on insulating, low-cost substrates need enough crystalline order and sheet conductance to support RF devices. This paper addresses that gap by growing multilayer PtSe2 by molecular beam epitaxy on sapphire, then tying Raman spectra, X-ray diffraction, STEM, electrical transport, and device measurements into one process window.

The central result is that post-growth annealing changes both the structural order and the electrical usefulness of the films. The authors report sheet conductance up to 1.6 mS, air-stable conductance after 1.5 years, and devices operating at the 1.55 μm telecom wavelength with 60 GHz photodetector bandwidth and above-30 GHz optoelectronic mixer bandwidth.

Core Contribution

The useful idea is the use of Raman linewidths as actionable growth metrics rather than post hoc characterization. The paper argues that the in-plane $E_g$ mode and out-of-plane $A_{1g}$ mode carry different information: narrower $E_g$ FWHM tracks better in-plane crystalline quality, while narrower $A_{1g}$ FWHM tracks better out-of-plane order. Films with both narrower peaks show higher sheet conductance.

That link matters because high-conductance PtSe2 requires more than the right stoichiometry or thickness. The structural analysis separates two film morphologies: mainly vertically single crystalline domains, which give higher conductance, and superimposed twisted domains, which give lower conductance. This is a process-control paper as much as a device paper.

Figure 3 captures the Raman comparison between as-grown and annealed films, which is the measurement the authors turn into a quality metric for the rest of the study.

Technical Approach

The films are grown on sapphire by MBE and optimized across growth temperature, selenium flux, and annealing conditions. The main device material is a multilayer semimetallic film around 12–15 monolayers, synthesized at 520 °C and then annealed at 690 °C for 30 minutes under selenium. Thickness is calibrated through EDX using Pt/Al ratios from the film and sapphire substrate; the supplement shows that this ratio is approximately linear with thickness for the 0–9 nm regime used in the study.

The structural evidence is deliberately redundant. Raman and EDX screen the films; GIXRD measures in-plane lattice parameters, domain size, and microstrain using a Williamson-Hall analysis; HRXRD checks out-of-plane order; STEM resolves domain arrangements. The X-ray data show that annealing reduces non-uniform in-plane lattice distortions by about a factor of two: microstrain decreases from 0.38% in the as-grown film to 0.22% after annealing, while the in-plane domain size remains broadly similar, 724 ± 54 Å versus 630 ± 107 Å. HRXRD also shows the annealed out-of-plane lattice parameter moving from 5.300 ± 0.10 Å to 5.165 ± 0.02 Å, closer to the 5.07 Å bulk value.

Results and Analysis

The electrical gain is large enough to justify the process complexity. In the conductance plots, annealed 12–15 monolayer films reach 1.1–1.6 mS, while as-grown samples sit around 0.5–0.6 mS. The best reported value, 1.6 mS, is comparable to the highest literature value shown in the paper, with the caveat that the literature point was measured at 10 K whereas these MBE films are discussed in the room-temperature device context.

Air stability is also checked directly rather than assumed. After 1.5 years in air, Raman spectra and EDX Se/Pt ratio remain similar, and sheet conductance changes only from 1.49 mS to 1.46 mS. That is a useful result for devices, because many 2D-material demonstrations fail on process durability rather than peak lab performance.

The device section is the strongest application-level evidence. The authors fabricate arrays of DC and RF devices on a 2-inch sapphire substrate. Across 378 four-point DC devices, they report a median sheet resistance of 0.95 kΩ per square, corresponding to 1.05 mS sheet conductance, with 0.20 kΩ per square standard deviation. The contact resistance is below the measurement precision, less than 100 Ω·μm.

For 1.55 μm photodetection, a 3.5 × 3 μm² PtSe2 channel integrated in a coplanar waveguide is biased at 4 V and illuminated by a modulated laser from 2 to 67 GHz. The response is flat enough to give a 3 dB cutoff at 60 GHz, with responsivity around 0.2 mA/W, or about 8.5 μA at 41.5 mW. The trade-off is clear: semimetallic PtSe2 is fast, but low responsivity and high dark current make it less competitive with conventional photodiodes as a detector.

The same semimetallic behavior is better matched to optoelectronic mixing. With optical modulation fixed at 30 GHz and RF input swept from 0.01 to 29.99 GHz, the PtSe2 mixer maintains operation over an intermediate-frequency range up to 29.99 GHz and reaches -73 dB conversion efficiency. That is 10 dB better than the CVD graphene CPW comparison cited by the authors, though still below the -66 dB LT-GaAs reference in their table.