4H-SiC Detectors: Design & Performance

• 4H-SiC detectors are semiconductor radiation sensors based on the 4H polytype, featuring a 3.26 eV bandgap, high electric field tolerance, and fast carrier transport.
• They employ diverse architectures—Schottky, p–n, PIN, SPAD, LGAD, 3D structures, and graphene-contact designs—to optimize charge collection, timing resolution, and internal gain.
• Advanced studies using TCAD modeling and transient current techniques highlight key trade-offs in field shaping, depletion depth, and radiation hardness for varied applications.

4H-SiC detectors are radiation detectors fabricated on the 4H polytype of silicon carbide and realized in Schottky, p–n, PIN, SPAD, LGAD, interdigitated, graphene-contact, and 3D columnar forms. Their importance derives from a material combination that repeatedly appears across the literature: a wide bandgap around $3.26\ \text{eV}$, critical electric field on the order of $2$–$3\times10^6\ \text{V/cm}$, high carrier saturation velocity, low leakage current, strong thermal conductivity, and high displacement threshold energy, which together enable operation in high-field, high-rate, radiation-hard, and elevated-temperature regimes (Gaggl et al., 2022, Tan et al., 2021). Recent work spans charge-collection mapping in interdigitated photodetectors, direct drift-velocity extraction in thin PIN diodes, ultraviolet single-photon detection, segmented 4H-SiC LGADs, and system-level instruments for ozone DIAL and collider luminosity monitoring (Vittone et al., 2016, Gsponer et al., 15 Sep 2025, Zhao et al., 2024, Kráčmar et al., 16 May 2026, Li et al., 31 Jul 2025).

1. Material platform and signal-formation framework

The 4H polytype is a hexagonal SiC polytype with ABAC stacking, giving a unit cell with four Si–C bilayers along the c-axis (Christanell et al., 2021). Reported material parameters include electron mobility $$\mu_e \simeq 800\ \text{cm}^2/\text{V}\cdot\text{s}$$ perpendicular to the c-axis and $$\mu_e \simeq 900\ \text{cm}^2/\text{V}\cdot\text{s}$$ parallel to it, hole mobility $$\mu_h \simeq 115\ \text{cm}^2/\text{V}\cdot\text{s}$$, saturation velocity $v_{\text{sat}} \simeq 2.2\times10^7\ \text{cm/s}$, and thermal conductivity in the $300$–$490\ \text{W/m}\cdot\text{K}$ range (Christanell et al., 2021). A direct drift-velocity study in thin PIN detectors fixed $$\mu_{\text{low}}=950\ \text{cm}^2/\text{V}\cdot\text{s}$$ for electrons and $2$0 for holes, and extracted $2$1 and $2$2 from current transients (Gsponer et al., 15 Sep 2025).

For junction devices, depletion is commonly described by the one-sided abrupt-junction relation

$2$3

or equivalent forms using the specific layer doping and built-in voltage of the device under study (Gaggl et al., 2022). This scaling is central to the literature because many operating points are set by the bias needed either to deplete the active epilayer or to deplete only the region sampled by the incident radiation.

Signal formation is generally analyzed through Shockley–Ramo theory. In interdigitated devices, the induced current is written as

$2$4

with the weighting potentials $2$5 and $2$6 determining how specific regions couple to the chosen readout electrode (Vittone et al., 2016). In UV-TCT studies of irradiated PiN devices, the current is written in the equivalent form

$2$7

with $2$8 and $2$9 denoting weighting-field factors (Gsponer et al., 2024). Across architectures, the standard detector figure of merit is charge collection efficiency,

$3\times10^6\ \text{V/cm}$0

sometimes normalized to a reference detector or to an injected optical charge (Vittone et al., 2016, Jiang et al., 28 May 2025).

A persistent literature issue is the electron–hole pair creation energy. One dedicated measurement on $3\times10^6\ \text{V/cm}$1 4H-SiC p–n diodes obtained $3\times10^6\ \text{V/cm}$2 from a linear fit of pulse height to deposited $3\times10^6\ \text{V/cm}$3 energy and reported a linewidth of $3\times10^6\ \text{V/cm}$4 FWHM, corresponding to an energy resolution of $3\times10^6\ \text{V/cm}$5 (Gsponer et al., 2023). A high-temperature Schottky study reported $3\times10^6\ \text{V/cm}$6 and found that $3\times10^6\ \text{V/cm}$7 decreases by approximately $3\times10^6\ \text{V/cm}$8 between $3\times10^6\ \text{V/cm}$9 and $$\mu_e \simeq 800\ \text{cm}^2/\text{V}\cdot\text{s}$$0 (Garcia et al., 2013). By contrast, a beam-monitoring study extracted $$\mu_e \simeq 800\ \text{cm}^2/\text{V}\cdot\text{s}$$1 from strip-sensor data (Christanell et al., 2021). This literature spread is explicit rather than incidental: one recent measurement states that published values vary significantly (Gsponer et al., 2023). A plausible implication is that detector calibration and comparison across experiments remain strongly readout- and methodology-dependent.

2. Device architectures and fabrication directions

The device family now includes multiple architectures that target distinct operating regimes, from resolved charge spectroscopy to Geiger-mode photon counting and internal avalanche gain.

Architecture	Representative structure	Representative reported result
Interdigitated Schottky detector	$$\mu_e \simeq 800\ \text{cm}^2/\text{V}\cdot\text{s}$$2 n-type epitaxial layer; finger width $$\mu_e \simeq 800\ \text{cm}^2/\text{V}\cdot\text{s}$$3, length $$\mu_e \simeq 800\ \text{cm}^2/\text{V}\cdot\text{s}$$4, gap $$\mu_e \simeq 800\ \text{cm}^2/\text{V}\cdot\text{s}$$5	Peak CCE $$\mu_e \simeq 800\ \text{cm}^2/\text{V}\cdot\text{s}$$6 for $$\mu_e \simeq 800\ \text{cm}^2/\text{V}\cdot\text{s}$$7 protons at $$\mu_e \simeq 800\ \text{cm}^2/\text{V}\cdot\text{s}$$8 (Vittone et al., 2016)
Planar p-in-n / PIN diode	Typical active thickness $$\mu_e \simeq 800\ \text{cm}^2/\text{V}\cdot\text{s}$$9; $$\mu_e \simeq 900\ \text{cm}^2/\text{V}\cdot\text{s}$$0 pads or $$\mu_e \simeq 900\ \text{cm}^2/\text{V}\cdot\text{s}$$1 circular diodes	Full depletion reported around $$\mu_e \simeq 900\ \text{cm}^2/\text{V}\cdot\text{s}$$2–$$\mu_e \simeq 900\ \text{cm}^2/\text{V}\cdot\text{s}$$3 depending on doping and geometry (Gaggl et al., 2022, Gsponer et al., 15 Sep 2025)
UV SPAD	Beveled mesa; active diameter $$\mu_e \simeq 900\ \text{cm}^2/\text{V}\cdot\text{s}$$4 or $$\mu_e \simeq 900\ \text{cm}^2/\text{V}\cdot\text{s}$$5	PDE $$\mu_e \simeq 900\ \text{cm}^2/\text{V}\cdot\text{s}$$6 or $$\mu_e \simeq 900\ \text{cm}^2/\text{V}\cdot\text{s}$$7 at $$\mu_e \simeq 900\ \text{cm}^2/\text{V}\cdot\text{s}$$8 (Yu et al., 2023, Zhao et al., 2024)
4H-SiC LGAD	Shallow gain layer approximately $$\mu_e \simeq 900\ \text{cm}^2/\text{V}\cdot\text{s}$$9 below the surface; strips at $$\mu_h \simeq 115\ \text{cm}^2/\text{V}\cdot\text{s}$$0 pitch and pixels at $$\mu_h \simeq 115\ \text{cm}^2/\text{V}\cdot\text{s}$$1 or $$\mu_h \simeq 115\ \text{cm}^2/\text{V}\cdot\text{s}$$2 pitch	Internal gain $$\mu_h \simeq 115\ \text{cm}^2/\text{V}\cdot\text{s}$$3–$$\mu_h \simeq 115\ \text{cm}^2/\text{V}\cdot\text{s}$$4 in segmented devices; up to $$\mu_h \simeq 115\ \text{cm}^2/\text{V}\cdot\text{s}$$5 at $$\mu_h \simeq 115\ \text{cm}^2/\text{V}\cdot\text{s}$$6 in first-generation pads (Kráčmar et al., 16 May 2026, Švihra et al., 12 Apr 2025)
Graphene-optimized PIN	Monolayer graphene ring electrode on $$\mu_h \simeq 115\ \text{cm}^2/\text{V}\cdot\text{s}$$7 epi	CCE $$\mu_h \simeq 115\ \text{cm}^2/\text{V}\cdot\text{s}$$8 at $$\mu_h \simeq 115\ \text{cm}^2/\text{V}\cdot\text{s}$$9 and a $v_{\text{sat}} \simeq 2.2\times10^7\ \text{cm/s}$0 rise-time reduction relative to a ring-electrode detector (Jiang et al., 28 May 2025)
3D 4H-SiC detector	Column radius $v_{\text{sat}} \simeq 2.2\times10^7\ \text{cm/s}$1; optimized $v_{\text{sat}} \simeq 2.2\times10^7\ \text{cm/s}$2 cell with $v_{\text{sat}} \simeq 2.2\times10^7\ \text{cm/s}$3 pitch in a $v_{\text{sat}} \simeq 2.2\times10^7\ \text{cm/s}$4 substrate	RASER predicts $v_{\text{sat}} \simeq 2.2\times10^7\ \text{cm/s}$5 at $v_{\text{sat}} \simeq 2.2\times10^7\ \text{cm/s}$6 pitch (Tan et al., 2021)

Planar p–n and PIN devices remain the baseline for spectroscopy, timing, and irradiation studies. Representative examples include IMB-CNM $v_{\text{sat}} \simeq 2.2\times10^7\ \text{cm/s}$7 p-in-n pads with a $v_{\text{sat}} \simeq 2.2\times10^7\ \text{cm/s}$8 n-doped epilayer on n$v_{\text{sat}} \simeq 2.2\times10^7\ \text{cm/s}$9 substrate (Gaggl et al., 2022), thin circular PiN diodes of diameter $300$0 and thickness $300$1 for high-frequency transient analysis (Gsponer et al., 15 Sep 2025), and the SICAR PIN detector with a $300$2 depleted region and ring geometry for $300$3 detection from room temperature to $300$4 (Li et al., 18 Jul 2025).

Avalanche architectures divide into SPADs and LGADs. SPADs use thin absorption–multiplication regions and beveled mesas to sustain Geiger mode at ultraviolet wavelengths (Yu et al., 2023, Zhao et al., 2024). LGADs introduce a shallow, highly doped gain layer beneath the main junction to create a controlled impact-ionization region (Kráčmar et al., 16 May 2026). The recent LGAD literature includes first-generation onsemi devices optimized for n-type substrate/epi wafers (Švihra et al., 12 Apr 2025) and the first fabricated segmented 4H-SiC LGADs, with strips and pixel arrays using both geometric separation and oxide-filled trenches (Kráčmar et al., 16 May 2026).

Graphene-contact devices represent a distinct fabrication direction. In one implementation, a CVD monolayer graphene transparent electrode was wet-transferred and patterned into a ring geometry on a $300$5 4H-SiC PIN detector, with Raman spectroscopy confirming a monolayer and low defect density (Xiao et al., 28 Mar 2026). In another, monolayer graphene replaced the central metal contact in a ring-electrode detector, preserving high CCE while reducing rise time and electrode shadowing for low-penetration radiation (Jiang et al., 28 May 2025). Epitaxially grown graphene has also been used as the front contact in synchrotron X-ray photodiodes to avoid high-$300$6 metal on the active surface (Paz et al., 2024).

3. Charge collection, transport, and field engineering

Charge collection in 4H-SiC detectors is strongly geometry-dependent, and the most explicit field-level picture comes from interdigitated IBIC mapping. In the interdigitated Schottky detector studied with focused proton microbeams, maps were taken at $300$7 and $300$8 for two readout configurations: back-ohmic readout and top-Schottky readout (Vittone et al., 2016). The experimental result was highly configuration-specific. With back readout, the full interdigit pattern appeared and active regions overlapped at higher bias. With top readout, signal appeared only under the sensitive comb and was null under the grounded comb. Line scans showed peak CCE approaching $300$9 for $490\ \text{W/m}\cdot\text{K}$0 protons at $490\ \text{W/m}\cdot\text{K}$1, whereas for $490\ \text{W/m}\cdot\text{K}$2 protons saturation was not reached up to $490\ \text{W/m}\cdot\text{K}$3 because generation extended beyond the depletion width (Vittone et al., 2016). In the top-readout configuration at low bias, the CCE valley at the finger-gap center could be below $490\ \text{W/m}\cdot\text{K}$4–$490\ \text{W/m}\cdot\text{K}$5.

That work is also notable for methodology. The detector response was modeled with two-dimensional finite elements in COMSOL/FEMLAB by solving adjoint carrier continuity equations with boundary conditions set by the chosen sensitive electrode. A minority-carrier diffusion length of $490\ \text{W/m}\cdot\text{K}$6, corresponding to $490\ \text{W/m}\cdot\text{K}$7 for $490\ \text{W/m}\cdot\text{K}$8, reproduced the data without further fitting once $490\ \text{W/m}\cdot\text{K}$9, mobilities, and $$\mu_{\text{low}}=950\ \text{cm}^2/\text{V}\cdot\text{s}$$0 were fixed from the literature (Vittone et al., 2016). The physical picture was that drift dominates in the depletion region, while minority carriers generated in the neutral region can diffuse over roughly $$\mu_{\text{low}}=950\ \text{cm}^2/\text{V}\cdot\text{s}$$1 before entering the depletion zone and being collected.

Planar diodes show the same dependence on where the ionization is produced. In neutron-irradiated $$\mu_{\text{low}}=950\ \text{cm}^2/\text{V}\cdot\text{s}$$2 p-in-n pad sensors measured with a $$\mu_{\text{low}}=950\ \text{cm}^2/\text{V}\cdot\text{s}$$3 $$\mu_{\text{low}}=950\ \text{cm}^2/\text{V}\cdot\text{s}$$4 source, the unirradiated detector reached a charge plateau near $$\mu_{\text{low}}=950\ \text{cm}^2/\text{V}\cdot\text{s}$$5 even though full depletion from capacitance and UV-TCT was around $$\mu_{\text{low}}=950\ \text{cm}^2/\text{V}\cdot\text{s}$$6, because the $$\mu_{\text{low}}=950\ \text{cm}^2/\text{V}\cdot\text{s}$$7 particles sampled only the first $$\mu_{\text{low}}=950\ \text{cm}^2/\text{V}\cdot\text{s}$$8 of the detector (Gaggl et al., 2022). This distinction between full epilayer depletion and full signal generation-depth depletion recurs throughout the literature and is important when comparing CCE curves from optical, $$\mu_{\text{low}}=950\ \text{cm}^2/\text{V}\cdot\text{s}$$9, proton, and MIP-like excitation.

Direct carrier-transport metrology has now reached the point where electron and hole drift can be separated in a $2$00 PIN detector using a high-frequency board with effective bandwidth of approximately $2$01 and $2$02 10–90% rise time (Gsponer et al., 15 Sep 2025). Using UV-TCT, $2$03 particles, and a $2$04 proton beam, the transient current was decomposed into electron and hole contributions, allowing drift times $2$05 and $2$06 and thus

$2$07

to be extracted directly (Gsponer et al., 15 Sep 2025). The resulting transport parameters provide an essential input for TCAD and for hybrid simulation chains that combine Synopsys Sentaurus, Allpix$2$08, and SPICE/QUCS.

A general design consequence is that 4H-SiC detectors are unusually sensitive to field shaping rather than only to depletion depth. Interdigitated weighting potentials localize sensing under selected electrodes (Vittone et al., 2016); graphene electrodes can homogenize the in-plane field (Jiang et al., 28 May 2025); and 3D columnar structures reduce drift length independently of substrate thickness (Tan et al., 2021). This suggests that, within the 4H-SiC platform, transport engineering is at least as central as bulk material quality.

4. Ultrafast timing, internal gain, and single-photon operation

Timing performance is one of the strongest motives for 4H-SiC detector development. A general decomposition used in laser-TCT work writes the time resolution as

$2$09

with $2$10 negligible for well-controlled laser injection (Xiao et al., 28 Mar 2026). In graphene-optimized ring-electrode PIN detectors, TCT measurements showed that replacing the conventional windowed metal structure with a transparent graphene electrode strongly improved timing uniformity: the reference ring-electrode detector degraded from $2$11 to $2$12 across the scan, whereas the graphene-integrated device degraded only from $2$13 to $2$14, corresponding to an $2$15 stability improvement (Xiao et al., 28 Mar 2026). The same work reported noise reduction from $2$16 to $2$17 at maximum scan distance and attributed the improvement to a more uniform vertical field and high-speed lateral conduction in graphene.

Rise-time studies point in the same direction. In a graphene-optimized $2$18 4H-SiC PIN detector, the mean rise time at $2$19 was $2$20 versus $2$21 for a ring-electrode detector, a $2$22 reduction, while CCE reached $2$23 at the same bias (Jiang et al., 28 May 2025). The SICAR PIN detector maintained a rise time of $2$24 at $2$25, together with leakage current below $2$26 at $2$27 and no degradation in depletion capacitance or charge collection (Li et al., 18 Jul 2025). At the simulation level, RASER studies of 3D columnar devices predicted $2$28 for an optimized $2$29 cell at $2$30 with $2$31 thickness and $2$32 pitch (Tan et al., 2021).

LGADs address the lower primary charge yield of SiC by adding internal avalanche gain. First-generation onsemi 4H-SiC LGADs showed $2$33 at $2$34, $2$35 at $2$36, and $2$37 at $2$38 in TCT, with $2$39-source timing in the $2$40–$2$41 range at the highest bias (Švihra et al., 12 Apr 2025). Segmented 4H-SiC LGADs extended this concept to $2$42-pitch strips and $2$43/$2$44-pitch pixels, where two-photon absorption TCT demonstrated clear charge separation between adjacent strips with internal gain $2$45–$2$46 (Kráčmar et al., 16 May 2026). This is the first fabricated and characterized segmented 4H-SiC LGAD generation reported in the supplied literature (Kráčmar et al., 16 May 2026).

Single-photon detection has developed in parallel. A free-running 4H-SiC UVSPD based on a beveled-mesa SPAD and passive quenching with active reset achieved a typical performance of $2$47 PDE, $2$48 dark count rate, and $2$49 afterpulse probability at $2$50, at $2$51 with $2$52 (Yu et al., 2023). A later room-temperature implementation using an actively quenched SPAD with $2$53 quenching time reached $2$54 PDE at $2$55, $2$56 dark count rate, $2$57 maximum count rate, and $2$58 afterpulse probability, and was then deployed in a two-channel ozone DIAL instrument (Zhao et al., 2024). Together these results show that the 4H-SiC platform now spans both analog ultrafast timing and Geiger-mode ultraviolet photon counting.

5. Radiation effects, elevated temperature, and experimental caveats

Radiation-hardness studies consistently show that 4H-SiC remains functional after fluences that significantly degrade signal collection in silicon-like operating modes, but the degradation mechanisms are architecture- and bias-dependent. In $2$59 p-in-n diodes irradiated with neutrons, the most-probable CCE at $2$60 fell to approximately $2$61 at $2$62, $2$63 at $2$64, and $2$65 at $2$66; the highest-fluence curve rose almost linearly with reverse bias and showed no saturation within safe bias (Gaggl et al., 2022). For the sample irradiated to $2$67, no detectable $2$68 signals were seen below approximately $2$69 (Gaggl et al., 2022).

Defect spectroscopy and lifetime measurements under proton irradiation have clarified part of the mechanism. In $2$70 proton-irradiated 4H-SiC PIN devices, DLTS identified three electron-trap levels, including EH$2$71 at $2$72, and TRPL showed minority-carrier lifetimes decreasing from $2$73 before irradiation to $2$74 at $2$75 and $2$76 at $2$77 (Li et al., 12 Mar 2025). A deep-level compensation model in RASER used these trap parameters to explain two otherwise counterintuitive observations: a decrease of leakage current with rising radiation intensity and a nearly constant capacitance under proton irradiation up to $2$78 (Li et al., 12 Mar 2025). In that interpretation, EH$2$79 acceptor-like traps compensate donors, flatten the electric field, and reduce trap-assisted tunneling.

A separate and more controversial phenomenon is forward-bias charge enhancement in irradiated PiN detectors. UV-TCT measurements on IMB-CNM-CSIC devices irradiated up to $2$80 showed that, under forward bias with a tightly focused laser, CCE could exceed $2$81: up to $2$82 at $2$83 for $2$84 and around $2$85 for $2$86 (Gsponer et al., 2024). With a defocused beam, the same samples showed ordinary saturation around $2$87 or $2$88, matching reverse-bias behavior; at $2$89, the overshoot disappeared (Gsponer et al., 2024). The paper explicitly states that the mechanism is still an ongoing topic of study and links the effect to local carrier density and pulse frequency. This does not support the simplistic reading that the detector intrinsically collects more than the injected charge in the usual sense; rather, the reported hypothesis is that transient forward-injection current contributes to the measured signal (Gsponer et al., 2024).

High-temperature operation is another area where 4H-SiC departs from silicon. An n-type 4H-SiC Schottky diode detector was operated with $2$90 $2$91 particles up to $2$92, with reverse bias adjusted around $2$93 (Garcia et al., 2013). The work found that leakage current rose from approximately $2$94 at $2$95 to approximately $2$96 at $2$97, and that FWHM increased exponentially above about $2$98 as shot noise from leakage began to dominate (Garcia et al., 2013). By contrast, the more recent SICAR PIN detector remained below $2$99 leakage at $3\times10^6\ \text{V/cm}$00 and maintained essentially $3\times10^6\ \text{V/cm}$01 CCE over $3\times10^6\ \text{V/cm}$02 (Li et al., 18 Jul 2025). The difference reflects not a contradiction in material behavior but differences in device structure, field management, and temperature range.

6. Applications, measurement regimes, and design trade-offs

The application landscape is broad and increasingly system-level. In collider instrumentation, a CEPC fast luminosity monitor based on an array of $3\times10^6\ \text{V/cm}$03-SiC PIN diodes used RASER simulations to optimize the detector height. The selected layout employed two identical $3\times10^6\ \text{V/cm}$04-pixel modules, each $3\times10^6\ \text{V/cm}$05 tall and $3\times10^6\ \text{V/cm}$06 along the beam, placed at $3\times10^6\ \text{V/cm}$07 from the beam-pipe center (Li et al., 31 Jul 2025). For $3\times10^6\ \text{V/cm}$08 with $3\times10^6\ \text{V/cm}$09, the design reached $3\times10^6\ \text{V/cm}$10 per millisecond and $3\times10^6\ \text{V/cm}$11 relative precision at $3\times10^6\ \text{V/cm}$12; the Total Sample Current showed a near-linear correlation with luminosity attenuation, with $3\times10^6\ \text{V/cm}$13 (Li et al., 31 Jul 2025). The same study quoted a system time resolution of approximately $3\times10^6\ \text{V/cm}$14 after adding electronics jitter in quadrature to the intrinsic device contribution (Li et al., 31 Jul 2025).

For intense ion-beam diagnostics, 4H-SiC has been explored at MedAustron in both pad and strip formats. The strip sensor read out with APV25 electronics recovered MIP-like signals with $3\times10^6\ \text{V/cm}$15, and the measured yield was $3\times10^6\ \text{V/cm}$16 electron–hole pairs per micrometre (Christanell et al., 2021). The discrete GHz-bandwidth board used with single-channel sensors did not reliably detect MIPs, which illustrates a recurring 4H-SiC trade-off: the material enables fast, low-leakage devices, but the smaller primary charge relative to silicon makes front-end electronics a decisive part of detector performance (Christanell et al., 2021).

In atmospheric sensing, 4H-SiC SPADs have already moved from component characterization to field deployment. In the ozone DIAL demonstration, two room-temperature 4H-SiC SPDs, one per wavelength, were used for ozone retrieval with $3\times10^6\ \text{V/cm}$17 and $3\times10^6\ \text{V/cm}$18 (Zhao et al., 2024). After correcting for hold-off, afterpulse, and dark counts, the measured ozone concentrations at altitudes of $3\times10^6\ \text{V/cm}$19–$3\times10^6\ \text{V/cm}$20 agreed well with a commercial ozone DIAL, with the average difference described as within experimental noise at $3\times10^6\ \text{V/cm}$21 (Zhao et al., 2024).

Low-energy X-ray dosimetry and medical instrumentation emphasize another trade-off: front-side metal can distort the response. Synchrotron measurements at $3\times10^6\ \text{V/cm}$22 on mesa-type 4H-SiC photodiodes showed good linearity at $3\times10^6\ \text{V/cm}$23 and good spatial uniformity, even for under-depleted $3\times10^6\ \text{V/cm}$24 devices, but also an over-response of $3\times10^6\ \text{V/cm}$25–$3\times10^6\ \text{V/cm}$26 in metallized regions due to metal dose enhancement (Paz et al., 2024). This directly motivates graphene-contact designs for low-energy X-rays, UV, and low-penetration particles, where removal of high-$3\times10^6\ \text{V/cm}$27 material from the active surface improves tissue equivalence and optical access (Paz et al., 2024, Jiang et al., 28 May 2025).

Finally, several design trade-offs recur across the entire 4H-SiC detector literature. Interdigitated back readout maximizes uniformity but sacrifices pixelation, whereas front readout preserves pixel discrimination at the cost of low-bias gap losses (Vittone et al., 2016). SPAD hold-off must be long enough to suppress afterpulses but then limits count rate (Yu et al., 2023). LGAD segmentation introduces a no-gain fill-factor penalty, although trenches or geometric gaps of at least $3\times10^6\ \text{V/cm}$28 recover more than $3\times10^6\ \text{V/cm}$29 active fill factor in the reported segmented devices (Kráčmar et al., 16 May 2026). Thin detectors improve transit time but reduce signal amplitude; thicker detectors increase charge but demand careful field engineering or internal gain (Tan et al., 2021, Švihra et al., 12 Apr 2025). Taken together, the literature suggests that 4H-SiC detectors are no longer a single device class but a platform technology whose performance is set by the co-design of epitaxy, field profile, contact architecture, and readout chain.