Photoinduced Modulation Doping
- Photoinduced modulation doping is a light-activated process where a remote charge reservoir modulates carrier density in 2D materials without chemical substitution.
- It employs various platforms, such as MoS2/ITO and graphene/BN, to achieve rewritable, nonlocal, or persistent doping via optical excitation.
- This technique offers versatile tuning of electronic properties, impacting 2D electronics, optoelectronics, and nanoscale device engineering.
Photoinduced modulation doping is a light-activated analogue of conventional modulation doping in which the states that donate, accept, or trap charge are spatially separated from the transport channel, so optical excitation changes carrier density without conventional chemical substitution of the channel lattice. In reported implementations, the remote charge reservoir may be a defect-rich dielectric, a nanocrystal layer, a gate electrode, or an adjacent van der Waals acceptor; the channel may be graphene, monolayer MoS, a graphene nanoelectromechanical membrane, or a twisted WSe bilayer. The resulting doping can be contactless, rewritable, quasi-permanent, strongly local, or nonlocal over tens of micrometers, depending on the heterostructure and the mechanism that stabilizes the transferred charge (Kriegel et al., 2018, Ju et al., 2014, Tiberj et al., 2013, Miller et al., 2019, Gadelha et al., 2020, Li et al., 23 Oct 2025).
1. Conceptual foundation and relation to conventional modulation doping
In conventional modulation doping, dopant charges are placed in a nearby layer rather than inside the semiconductor channel, so carriers are transferred electrostatically while impurity scattering is reduced. Recent work on 2D electronics reformulates this in terms of workfunction-mediated charge transfer across van der Waals interfaces, especially under type-III (“broken-gap”) alignment, with vacuum-referenced band-edge descriptors
Within this framework, high-electron-affinity materials such as -RuCl, MoO, and VO are identified as strong p-type dopants for 2D channels, because they pull electrons from the semiconductor without introducing defects into the channel itself (Arora et al., 2024).
A closely related non-photonic implementation relocates the dopant function into a dielectric: aluminium in SiO forms acceptor states that can capture electrons from adjacent Si through tunnelling, leaving holes in silicon while keeping the Si active volume essentially intrinsic (König et al., 2016). In a correlated oxide variant, modulation-doped VO-based heterostructures achieve carrier densities greater than 0 without measurable structural changes, explicitly separating filling control from direct lattice modification (Mondal et al., 2023).
Photoinduced modulation doping preserves the same remote-doping logic but adds an optical trigger. Light either excites defect transitions, charges nanocapacitors, photoionizes traps, or generates hot carriers in remote electrodes; the key outcome is that a nearby layer or interface becomes the charge reservoir that electrostatically dopes the channel. This suggests that photoinduced modulation doping is best understood not as a single microscopic process but as a family of optically activated remote charge-transfer and charge-storage phenomena.
2. Mechanistic classes
Several distinct microscopic realizations have been demonstrated. In monolayer MoS1/ITO nanocrystal hybrids, UV illumination at 2 nm excites the nanocrystals above their bandgap; the photo-generated valence-band holes in the nanocrystals are filled by MoS3 electrons, leaving stored electrons in the nanocrystals and effectively p-type photodoping the MoS4. In graphene/BN heterostructures, visible light excites defect states in BN; under gate bias, the photoexcited carriers move and the resulting fixed charge in BN screens the gate field and shifts the graphene doping. In monolayer MoS5 transistors on SiO6/Si, laser illumination plus back-gate bias creates persistent electron accumulation tied to the gate-insulator interface. In graphene NEMS, a focused laser and applied bias photoionize defect states in the heterostructure and oxide, and the trapped charge then shifts the mechanical charge-neutrality point 7. In dual-gated twisted WSe8 bilayers, optical pumping heats carriers in graphite gates, and holes are then thermionically injected across thick hBN into the moiré bilayer (Kriegel et al., 2018, Ju et al., 2014, Gadelha et al., 2020, Miller et al., 2019, Li et al., 23 Oct 2025).
| Platform | Photoactive reservoir | Reported effect |
|---|---|---|
| Monolayer MoS9/ITO nanocrystals | UV-excited ITO nanocrystals | 0; 1 electrons per nanocrystal |
| Graphene/BN | BN defect states under visible light and gate bias | CNP pinning; writing and erasing of doping; 2-3 junction formation |
| Monolayer MoS4 on SiO5/Si | Gate-bias-assisted photodoping tied to gate-insulator interface | 6 shifts from 7 V to 8 V |
| Graphene NEMS | Trapped charge written by focused laser plus bias | Persistent resonance-frequency tuning |
| Twisted WSe9 bilayer | Photo-thermionic hole injection from graphite gates | Gate-voltage shifts of correlated-insulator signatures |
These mechanisms separate naturally into four categories. One is defect-mediated photodoping, exemplified by graphene/BN and graphene on hydrophilic SiO0/Si, where optically active states in a neighboring dielectric or interfacial environment trap charge and thereby dope graphene (Ju et al., 2014, Tiberj et al., 2013). A second is interfacial photoinduced charge transfer, as in MoS1/ITO hybrids, where the illuminated remote layer stores charge capacitively and extracts carriers from the channel (Kriegel et al., 2018). A third is bias-assisted local trap writing, as in MoS2 transistors and graphene NEMS, where illumination and an applied field together create long-lived trapped charge distributions (Gadelha et al., 2020, Miller et al., 2019). A fourth is photo-thermionic remote injection, where the photonic absorber and the doped quantum material are different layers, as in graphite-gated twisted WSe3 (Li et al., 23 Oct 2025).
A distinct, more chemically mediated variant appears in exfoliated graphene under intense 4 nm, 5 ps laser pulses. There, the stable local increase in hole doping is accompanied by a reduction in compressive strain, and the proposed mechanisms are photo-induced oxygenation together with buckling or slippage of the graphene; the effect is local, stable for months, and reversible by solvent soaking (Alexeev et al., 2013).
3. Experimental signatures and quantitative observables
The observables used to identify photoinduced modulation doping depend on the channel. In n-type monolayer MoS6, reduced electron density is diagnosed optically by a decrease in trion photoluminescence, an increase in neutral-exciton photoluminescence, a PL blue-shift, and a reduction of the exciton–trion energy separation toward the intrinsic value. Using a mass-action / Boltzmann equilibrium model for the exciton–trion–electron system, the MoS7/ITO study extracted a carrier-density reduction of approximately 8 and a final carrier density of about 9, corresponding to roughly 0 electrons stored per nanocrystal and an attofarad-range capacitor-like response (Kriegel et al., 2018).
In monolayer MoS1 transistors, threshold-voltage shifts provide the carrier-density measure:
2
The reported threshold voltage moved from 3 V to 4 V, corresponding to 5. The decay was weak enough that after 6 hours the photodoping had barely decreased, and an exponential fit predicted that 7 of the initial photodoping remains long-term (Gadelha et al., 2020).
In graphene/BN devices, the fundamental transport signature is the anomalous 8 response under illumination: with light on and negative 9, resistance rises to the charge-neutrality point and becomes pinned, indicating that the BN charges screen the field. After controlled writing, the entire 0 curve shifts in the dark so that the CNP lands at the selected write voltage. The photo-induced charge density in BN acts like an effective built-in gate, consistent with the usual field-effect relation
1
A further signature of remote-dopant behavior is that the extracted electron mobility stays nearly constant over the full photo-doping range, while the charge-density fluctuations near the CNP change only weakly (Ju et al., 2014).
Raman spectroscopy is the main diagnostic in graphene systems exposed directly to light. On hydrophilic SiO2/Si, visible illumination at 3 nm reversibly tuned graphene from 4 at 5 mW to 6 at 7 mW, with the neutral point marked by maximum 8, maximum 9, and minimum 0; no D band appeared (Tiberj et al., 2013). Under intense picosecond irradiation, correlated Raman shifts of the G and 2D peaks were decomposed using the characteristic slopes 1 for uniaxial strain and 2 for hole doping above 3, leading to the conclusion that photoexcitation causes both increased p-doping and reduced compressive strain (Alexeev et al., 2013).
4. Spatial extent, locality, and persistence
Spatial behavior is strongly platform dependent. In the MoS4/ITO hybrid, the UV excitation was localized to about 5, yet photoluminescence changes proliferated tens of micrometers away; the largest reported extent was up to 6 from the excitation spot, and in some maps spectral-median shifts of up to 7 meV appeared over areas around 8. The authors attributed this to carrier redistribution in MoS9 driven by local band-structure variations, particularly in regions with lower initial PL energy, higher initial trion content, edges and grain boundaries, and strain or defect variations (Kriegel et al., 2018).
By contrast, the monolayer MoS0 transistor study emphasized locality. Local writing used a 1 nm laser with a spot size of about 2, and after illumination only the selected region showed photocurrent suppression in later SPCM images; the rest of the device remained photoactive. In that system, the photodoping was persistent but micrometer-scale and tied to the gate-insulator interface rather than to long-range redistribution across the flake (Gadelha et al., 2020).
Graphene/BN photodoping combined long persistence with spatial programmability. Charges in BN remained trapped for long times, so the induced doping persisted for days in the dark, while erasure by illumination near 3 usually required much higher light dose. Patterned illumination under negative gate bias wrote inhomogeneous doping profiles and produced a graphene 4-5 junction whose dark 6 trace exhibited two resistance peaks of similar height separated by about 7 V (Ju et al., 2014). In graphene on hydrophilic SiO8/Si, the response time was preliminarily evaluated to be less than 9 s, whereas suspended graphene remained neutral over the full power sweep, showing that the interface rather than the graphene alone sets the dynamical scale (Tiberj et al., 2013).
Graphene NEMS and twisted WSe0 bilayers introduce two additional time regimes. In NEMS, 1 saturated in about 2 ms, persistence extended for many days with a measured slow drift of about 3/hour after an initial small transient, and the spatial resolution was about 4, roughly the laser spot size (Miller et al., 2019). In graphite-gated twisted WSe5, injection occurs on sub-picosecond to picosecond timescales, the transient reflectance signal persists at a delay of 6s, pre-time-zero features indicate pulse-to-pulse accumulation, and the estimated electrostatic discharge time is 7 ms (Li et al., 23 Oct 2025).
5. Functional consequences and device implementations
The principal functional consequence of photoinduced modulation doping is that carrier density becomes an optically writable state variable. In monolayer MoS8/ITO, this enabled all-optical carrier-density control, contactless and quasi-permanent p-type photodoping, remote tuning over distances up to 9, and an energy-storage analogy in which the ITO nanocrystals act as optically charged nanocapacitors (Kriegel et al., 2018).
In graphene/BN heterostructures, the preserved high mobility and rewritable doping profiles allowed light-written and light-erased 0-1 junctions, suggesting photoresist-free photolithography of carrier-density landscapes (Ju et al., 2014). In graphene on hydrophilic SiO2/Si, the reversible ambipolar tuning under moderate laser power showed that visible light can function as an optical gate, but it also established that Raman spectroscopy is not always non-invasive when the electronic state is the quantity of interest (Tiberj et al., 2013).
In graphene NEMS, the modulation-doped state directly controlled mechanics through the effective voltage
3
The trapped charge produced a localized electrostatic strain and therefore a persistent frequency shift. Reported tuning ranged from about 4 MHz to 5 MHz in one device and from 6 MHz to 7 MHz in another, corresponding to nearly 8 tuning range. The write–erase cycle was repeated for 9 cycles with 00 repeatability (Miller et al., 2019).
In graphite-gated twisted WSe01, the functional effect was filling-factor control of correlated states. Photo-induced hole injection shifted the gate voltages at which optical signatures of correlated insulators appeared, generated persistent microsecond-scale signals consistent with charge accumulation, and produced photoinduced absorption interpreted as a transient move toward the 02 correlated-insulating regime (Li et al., 23 Oct 2025).
A recent graphene implementation in hBN/graphene/hBN/SiO03 heterostructures extended the device scope from carrier-density programming to in situ disorder control. Standard white light and a back-gate write voltage tuned the CNP over a 04 V range, corresponding to induced electron densities up to about 05. Drude and Landauer analyses showed mobility tunability between about 06 and 07 and a mean scattering length at 08 that could be tuned from about 09 to 10, enabling reversible switching between diffusive and quasi-ballistic transport regimes. The same procedure made quantum Hall isospin ferromagnetic states observable in a device whose initial quality would otherwise leave such states unobservable (Sanborn et al., 21 May 2026).
6. Misconceptions, limitations, and broader context
A recurring misconception is that photodoping is simply direct photoexcitation of the channel. The detailed literature shows otherwise. In MoS11/ITO, UV illumination is essential because it excites the ITO nanocrystals above their bandgap; without ITO nanocrystals, UV illumination does not produce the same MoS12 PL evolution (Kriegel et al., 2018). In monolayer MoS13 transistors, comparison between MoS14 on SiO15/Si and MoS16 on SiO17 glass showed that persistent photocurrent is not an intrinsic property of MoS18 alone but depends strongly on the gate-insulator interface (Gadelha et al., 2020). In graphene/BN, the BN-thickness dependence demonstrated that the relevant optical excitations occur in the BN bulk rather than merely in interfacial traps (Ju et al., 2014).
A second misconception is that all photoinduced modulation doping is either strictly local or necessarily long-range. The data show both limits. Micrometer-scale writing appears in SPCM studies of MoS19 and in graphene NEMS, whereas nonlocal redistribution over up to 20 occurs in MoS21/ITO hybrids (Gadelha et al., 2020, Miller et al., 2019, Kriegel et al., 2018). The distinction arises from the electronic landscape and the nature of the remote reservoir rather than from the optical trigger alone.
A third limitation is mechanistic specificity. In graphene on SiO22/Si, the effect requires hydrophilic substrates, is significantly affected by the substrate cleaning method, vanishes in suspended graphene, and was interpreted as a substrate-assisted electrochemical or photo-redox process whose detailed microscopic pathway remained unresolved (Tiberj et al., 2013). In picosecond-laser-modified graphene, the stable p-type change is likely tied to photo-induced oxygenation and buckling rather than to a purely electronic reservoir model (Alexeev et al., 2013). These cases indicate that “photoinduced modulation doping” covers both purely electrostatic remote charging and optically induced changes in the local chemical environment, provided the dopant function remains spatially separated from the main transport layer.
The broader non-photonic modulation-doping literature provides the equilibrium framework within which many photoinduced schemes can be interpreted. A single-layer 23-RuCl24 acceptor, for example, produces short-ranged lateral doping 25 nm, high homogeneity, and hole densities of 26 in monolayer graphene and 27 in bilayer graphene, while charge transfer through about 28 nm of hBN is reduced but not eliminated (Wang et al., 2020). First-principles screening then generalizes the design rule: type-III band alignment and large electron affinity are the central criteria for strong remote p-type doping in 2D electronics (Arora et al., 2024). A plausible implication is that future photoinduced modulation-doping platforms will increasingly combine these equilibrium band-engineering principles with optically addressable reservoirs, so that light controls not only nonequilibrium excitation but also the quasi-static carrier density landscape of low-dimensional devices.