Fe-doped Ga₂O₃ Photoconductive Switch

• The paper demonstrates that Fe-doped Ga₂O₃ enables sub-bandgap photoconductive switching by activating deep-level traps for high-current avalanche behavior.
• Device architectures—vertical embedded-electrode and grid-anode—are optimized through contact spacing and anode pitch to balance electric field distribution and optical fill factor.
• Wavelength tuning, particularly at 272 nm, yields optimal carrier generation and low on-resistance, highlighting FG-PCSS's potential for high-voltage, high-frequency switching applications.

Searching arXiv for the specified FG-PCSS papers and closely related work. Fe-doped Ga$_2$O$_3$ photoconductive semiconductor switches (FG-PCSSs) are photoconductive switching devices based on Fe-doped $\beta$-Ga$_2$O$_3$, an ultra-wide-bandgap semiconductor in which Fe-related deep levels enable sub-bandgap triggering and, under suitable bias and geometry, high-current photoconductive operation. Recent work has treated FG-PCSSs in two complementary regimes: a simulated high-gain vertical device with embedded electrodes and 450 nm triggering, and a fabricated vertical grid-anode device optimized over anode pitch and excitation wavelength from 235 nm to 500 nm (Karpourazar et al., 6 Aug 2025, Jangir et al., 16 Dec 2025).

1. Material platform and defect energetics

The FG-PCSS material system is centered on $\beta$-Ga$_2$O$_3$, for which the intrinsic bandgap is reported as $$E_g(\beta\text{-Ga}_2\text{O}_3) \simeq 4.8\ \text{eV}$$, corresponding to $\lambda_{\text{edge}} \simeq 258\ \text{nm}$ (Jangir et al., 16 Dec 2025). In the simulated vertical device, the substrate is a 400 $_3$0m-thick $_3$1-Ga$_3$2O$_3$3 wafer with a nominal intrinsic background donor-trap concentration $_3$4 at $_3$5, while Fe compensation produces an acceptor-trap concentration $_3$6 at $_3$7 (Karpourazar et al., 6 Aug 2025). In the fabricated device, the substrate is semi-insulating $_3$8-Ga$_3$9O$\beta$0, orientation (010), thickness 450 $\beta$1m, with Fe doping concentration $\beta$2 in the bulk (Jangir et al., 16 Dec 2025).

The defect physics is central to the operation of FG-PCSSs. The fabricated-device study identifies Fe-related deep levels from DLTS and optical spectroscopy, including $\beta$3 as the major trap associated with Fe doping, together with $\beta$4 and $\beta$5, as well as other levels at $\beta$6, $\beta$7, $\beta$8, and $\beta$9 (Jangir et al., 16 Dec 2025). The simulated vertical study likewise treats Fe-related mid-gap levels as the absorption centers for 450 nm sub-bandgap excitation: 450 nm photons are absorbed via Fe-related mid-gap levels with $_2$0, exciting electrons from Fe$_2$1 to the conduction band and leaving holes in the valence band (Karpourazar et al., 6 Aug 2025).

A useful consequence of these defect levels is that the excitation regime can be tuned away from purely above-bandgap absorption. For $_2$2, the reported absorption coefficient is $_2$3, implying a penetration depth $_2$4; for $_2$5, $_2$6–$_2$7, implying $_2$8–$_2$9; and for $_3$0, $_3$1–$_3$2, implying $_3$3–$_3$4 (Jangir et al., 16 Dec 2025). This wavelength dependence underlies the distinction between surface-confined and volumetric photogeneration.

2. Device architectures and fabrication

Two device configurations are described. The first is a vertical FG-PCSS with embedded electrodes. It uses two embedded ohmic electrodes of Ti/Au, each 1 $_3$5m thick, separated by a contact spacing $_3$6 varied from 10 $_3$7m to 30 $_3$8m, with the anode biased and the cathode grounded. Uniform illumination is applied through the 400 $_3$9m Ga$\beta$0O$\beta$1 thickness at 450 nm. The Ga$\beta$2O$\beta$3 is grown by edge-defined film-fed growth, and the contacts are formed by e-beam evaporation and rapid thermal anneal to achieve ohmic behavior (Karpourazar et al., 6 Aug 2025).

The second is a fabricated vertical device with a continuous back cathode and a grid-type top anode. The back contact is a Ti/Au bilayer of 50 nm Ti and 100 nm Au. The top anode is a Ti/Au grid of 50 nm/100 nm, patterned by image-reversal lithography and lift-off. The grid pitch $\beta$4 is varied over $\beta$5, the metal-finger width is chosen so that the optical window is approximately equal to $\beta$6, and rapid thermal anneal is carried out at 480 $\beta$7C for 1 min in N$\beta$8 for contact formation. The active area is approximately $\beta$9, and the electric field is predominantly vertical under the grid, with lateral fringing at the finger edges (Jangir et al., 16 Dec 2025).

These two implementations emphasize different control parameters. In the embedded-electrode structure, the dominant geometric variable is the short contact spacing $_2$0, which directly sets the internal field through $_2$1. In the grid-anode structure, the dominant geometric variable is the anode grid pitch, which controls the trade-off between fringing-field overlap and optical window. The reported trade-off is explicit: small $_2$2 gives high field but low optical window, while large $_2$3 gives high window but low field (Jangir et al., 16 Dec 2025).

3. Transport, photogeneration, and switching physics

Carrier transport in the simulated FG-PCSS is governed by the drift-diffusion equations

$_2$4

$_2$5

Under a high bias $_2$6, the internal field is approximated as $_2$7, so carriers are swept to the contacts by drift, with diffusion contributing to transport as well (Karpourazar et al., 6 Aug 2025).

Recombination is treated through SRH recombination with two trap levels, donor at 0.56 eV and acceptor at 0.76 eV, modified to include both levels, together with radiative and Auger terms as given by Eqs. (5)–(6) in the paper. At high electric field, impact ionization is modeled by Selberherr’s form, with ionization rates $_2$8 and $_2$9 producing an additional generation term

$_3$0

The field dependence is written as $_3$1, and similarly for holes (Karpourazar et al., 6 Aug 2025).

For wavelength-dependent excitation in the fabricated device, the carrier generation profile is written as

$_3$2

where $_3$3. This yields surface-confined generation for $_3$4 and volumetric generation for $_3$5. The reported spectral behavior is strongly non-monotonic: the photocurrent peaks at $_3$6, corresponding to $_3$7, which is approximately 12–17 nm longer than the intrinsic edge and is associated with deep-level photoionization (Jangir et al., 16 Dec 2025).

A common misconception is that above-bandgap excitation should necessarily maximize switch performance. The reported data indicate otherwise. Under 235–255 nm excitation, absorption occurs in the top $_3$8, leading to lateral diffusion and surface recombination and yielding $_3$9. By contrast, 272 nm excitation provides an optimal penetration depth of approximately 0.3–0.8 $$E_g(\beta\text{-Ga}_2\text{O}_3) \simeq 4.8\ \text{eV}$$0m, activates Fe levels, and still maintains high absorption, giving bulk generation and $$E_g(\beta\text{-Ga}_2\text{O}_3) \simeq 4.8\ \text{eV}$$1 up to 4.14 A. At 300 nm the trend remains bulk-dominated but with a lower peak than at 272 nm; at 450–500 nm, volumetric excitation of deep Fe/Ir levels yields moderate photocurrent, approximately 1–1.7 A at 450 nm and less than 1 A at 500 nm (Jangir et al., 16 Dec 2025).

4. Gain, efficiency, and high-field vertical operation

The simulated vertical study defines the device photocurrent under bias and illumination as $$E_g(\beta\text{-Ga}_2\text{O}_3) \simeq 4.8\ \text{eV}$$2, and takes $$E_g(\beta\text{-Ga}_2\text{O}_3) \simeq 4.8\ \text{eV}$$3, the photocurrent for 100% collection with no gain. The current gain is then

$$E_g(\beta\text{-Ga}_2\text{O}_3) \simeq 4.8\ \text{eV}$$4

with

$$E_g(\beta\text{-Ga}_2\text{O}_3) \simeq 4.8\ \text{eV}$$5

and, for uniform $$E_g(\beta\text{-Ga}_2\text{O}_3) \simeq 4.8\ \text{eV}$$6,

$$E_g(\beta\text{-Ga}_2\text{O}_3) \simeq 4.8\ \text{eV}$$7

The quantum efficiency is defined as

$$E_g(\beta\text{-Ga}_2\text{O}_3) \simeq 4.8\ \text{eV}$$8

The on-state resistance is given as

$$E_g(\beta\text{-Ga}_2\text{O}_3) \simeq 4.8\ \text{eV}$$9

where $\lambda_{\text{edge}} \simeq 258\ \text{nm}$0, $\lambda_{\text{edge}} \simeq 258\ \text{nm}$1 is the photon energy, and $\lambda_{\text{edge}} \simeq 258\ \text{nm}$2 is the surface reflection (Karpourazar et al., 6 Aug 2025).

The SILVACO ATLAS simulation uses approximately 31,920 2D mesh points, ohmic contacts on the electrodes, and Neumann insulating boundaries elsewhere. The material parameters include carrier lifetime $\lambda_{\text{edge}} \simeq 258\ \text{nm}$3, obtained from a 30 ps, 355 nm pulse on a coplanar waveguide; $\lambda_{\text{edge}} \simeq 258\ \text{nm}$4; low-field mobilities $\lambda_{\text{edge}} \simeq 258\ \text{nm}$5 and $\lambda_{\text{edge}} \simeq 258\ \text{nm}$6; high-field mobilities from the ATLAS field-dependent model benchmarked to DFT results; and thermal conductivity $\lambda_{\text{edge}} \simeq 258\ \text{nm}$7. The optical excitation is a single-pulse 450 nm laser with energy swept from 1 $\lambda_{\text{edge}} \simeq 258\ \text{nm}$8J to 100 $\lambda_{\text{edge}} \simeq 258\ \text{nm}$9J, while DC bias is swept to 2.5 kV and transients are run until the steady photoswitch event (Karpourazar et al., 6 Aug 2025).

The reported performance depends strongly on contact spacing. For $_3$00 and $_3$01, the photocurrent is approximately 5–10 A up to 2 kV and rises exponentially above 2 kV due to avalanche onset, reaching approximately 100 A at 2.5 kV with $_3$02. For $_3$03, no nonlinear avalanche is observed up to 2.5 kV and $_3$04. At $_3$05, 1.5 kV operation remains linear in optical energy with $_3$06, while 2.5 kV operation is superlinear in the avalanche region. The rise time at 2.5 kV and $_3$07 is approximately 0.18 ns and improves with higher bias and larger $_3$08, giving sub-ns switching (Karpourazar et al., 6 Aug 2025).

The high-gain condition is reported to occur above the impact-ionization threshold $_3$09, corresponding to approximately 2 kV across 10 $_3$10m. Because $_3$11 is small, the on-resistance is low, so even a low-energy, approximately 10 $_3$12J, low-cost 450 nm diode laser suffices. The reported optimal point is $_3$13, $_3$14–2.5 kV, and $_3$15–20 $_3$16J, yielding $_3$17–15, $_3$18 up to hundreds of amps, and sub-ns rise time enabling $_3$19 repetition (Karpourazar et al., 6 Aug 2025).

5. Spectral and geometric optimization in fabricated FG-PCSSs

The fabricated study introduces a responsivity-conductance figure of merit,

$_3$20

where

$_3$21

At $_3$22, $_3$23, and $_3$24, the reported values are $_3$25, $_3$26, $_3$27, incident pulse energy $_3$28 over a 7 ns pulse, and peak incident power $_3$29. The responsivity is reported as $_3$30, while the paper reports $_3$31 using normalized $_3$32 in A·cm/W·kV. The switching times at 272 nm and $_3$33 are $_3$34 and $_3$35–20 ns, with the fall time attributed to long-lived hole traps (Jangir et al., 16 Dec 2025).

The geometry optimization shows a clear optimum at $_3$36 under 272 nm excitation at 250 V. The reported trend is that narrow pitch gives stronger lateral-field overlap and better near-surface collection, whereas wide pitch increases optical fill factor but introduces low-field regions that limit bulk collection (Jangir et al., 16 Dec 2025).

Anode pitch $_3$37	$_3$38	$_3$39
20 $_3$40m	$_3$41 A	$_3$42
40 $_3$43m	4.14 A	10.4 $_3$44
60 $_3$45m	$_3$46 A	$_3$47
80 $_3$48m	$_3$49 A	$_3$50

The same study compares spectral regimes directly. Above-bandgap 245–255 nm excitation gives the fastest rise, below 2 ns, but low responsivity of approximately $_3$51, surface-limited generation, $_3$52, and $_3$53. Sub-bandgap 272 nm excitation gives $_3$54 up to 4.14 A, $_3$55, and $_3$56, with rise time approximately 4.5 ns and fall time approximately 10–20 ns. This is presented as a record simultaneous responsivity-conductance performance in sub-bandgap-triggered Ga$_3$57O$_3$58 PCSS (Jangir et al., 16 Dec 2025).

6. Design rules, application space, and open issues

Several design rules emerge consistently. In the embedded-electrode vertical architecture, reducing the contact spacing to 10 $_3$59m lowers the on-resistance to a few $_3$60, reduces the required optical energy, and enables high-gain avalanche operation at relatively low laser cost. Beam placement is also consequential: for $_3$61 and $_3$62, the best case occurs when the beam is adjacent to the anode, with $_3$63 at 2.5 kV, whereas the worst case is when the beam is centered, with $_3$64. The stated optimum is injection near the anode contact to minimize hole transit loss (Karpourazar et al., 6 Aug 2025).

In the grid-anode architecture, the key rule is spectral and geometric co-optimization. The crucial sub-bandgap regime is centered at 272 nm, where deep-level defect states are effectively activated and bulk carrier transport is promoted. The optimum anode pitch is 40 $_3$65m, at which the fabricated device achieves a high peak photocurrent of 4.14 A, a record-low on-resistance of 10.4 $_3$66, and a record $_3$67 of $_3$68 (Jangir et al., 16 Dec 2025).

The application space described in the two studies includes high-voltage switching, microwave generation applications, and next-generation high-power optoelectronic switching applications. The simulated work emphasizes the combination of very high breakdown field, approximately $_3$69, with mid-gap absorption and sub-bandgap triggering at 450 nm; the fabricated work notes that ultra-wide-bandgap $_3$70-Ga$_3$71O$_3$72 is tolerant to $_3$73, supporting scalability toward kV/kA pulsed power (Karpourazar et al., 6 Aug 2025, Jangir et al., 16 Dec 2025).

The principal open issues are also explicit. One is the distinction between simulated and experimentally demonstrated gain: the vertical embedded-electrode study reports that future work should experimentally validate high-gain operation greater than $_3$74 in fabricated devices. Additional proposed directions are optimization of trap engineering, specifically trap energy and density to tailor $_3$75 and $_3$76, and integration with micro-laser sources for compact, high-repetition FG-PCSS modulators (Karpourazar et al., 6 Aug 2025). Another issue is turn-off behavior in the spectral regime near 280 nm: persistent photocurrent suggests care in turn-off gating but also potential for memory effects (Jangir et al., 16 Dec 2025).

Taken together, these results define FG-PCSSs as a defect-engineered Ga$_3$77O$_3$78 switching platform in which Fe-related deep levels mediate sub-bandgap absorption, geometry controls field distribution and carrier collection, and operating regime determines whether the device behaves primarily as a low-resistance bulk photoconductor or as a high-gain avalanche photoswitch.