Excitonic Energy Funnels

• Excitonic energy funnels are energetic landscapes that guide exciton transfer via spatial or potential gradients, leading to efficient exciton capture at reaction centers.
• They are achieved through resonant energy transfer, strain-induced bandgap modulation, and electrostatic potentials with design principles centered on spectral overlap and dipole alignment.
• Key implementations span molecular aggregates and semiconductor devices, demonstrating controlled transfer efficiencies and offering routes for optimized exciton harvesting.

Excitonic energy funnels are spatial or networked energy landscapes that bias exciton motion toward lower-energy sites, designated traps, reaction centers, or nanoscale emitters. In different material classes, the funnel can be realized as a sequence of resonant energy-transfer steps between chromophores, a monotonic electrostatic or strain-induced potential, a dielectric landscape that acts primarily on dark excitons, or a mixed-dimensional donor reservoir feeding a lower-dimensional acceptor. Across these realizations, directionality derives from favorable spectral overlap, band alignment, drift in a potential gradient, or bath-assisted relaxation, while the central performance question is how much of the initially generated exciton population reaches the target before radiative, nonradiative, or annihilation losses intervene (Saikin et al., 2013, Dorow et al., 2016, Fang et al., 2023).

1. Definition and physical scope

In molecular aggregates, excitonic funneling is tied to Frenkel excitons and to the fact that aggregate geometry and site energies can create downhill pathways toward selected traps. The review literature describes funnels in terms of spatially varying site energies $\varepsilon_n$, hierarchical antennas, and site-selective trapping, with transport occurring between fully coherent band transport and purely incoherent Förster hopping depending on disorder, vibronic coupling, and dephasing (Saikin et al., 2013). In this setting, a funnel is not merely a geometric arrangement of pigments; it is an energetic bias embedded in an excitonic Hamiltonian.

In semiconductor implementations, the same concept is recast as a real-space potential landscape. For indirect excitons in coupled quantum wells, the exciton energy obeys $E(x,y)=-ed\,F_z(x,y)$, so a lateral gradient in the perpendicular electric field produces a monotonic ramp that drives excitons downhill without any applied in-plane voltage gradient (Dorow et al., 2016). In monolayer transition-metal dichalcogenides, tensile strain lowers the bandgap and hence the exciton energy, so a local strain gradient creates a drift field toward the strain maximum (Moon et al., 2019). In bilayer WSe$_2$, dielectric nanobubbles can leave the bright exciton energy nearly unchanged while lowering the energy of momentum-indirect dark excitons, so the funnel is defined by a dark-state potential $U(r)$ rather than by the optically bright transition itself (Su et al., 2022).

This range of definitions implies that “excitonic energy funnel” denotes a functional principle rather than a single microscopic mechanism. In some systems the operative coordinate is an excitonic ladder in Hilbert space; in others it is a real-space scalar potential; in others still it is an interfacial reservoir-sink geometry in which diffusion feeds a localized acceptor (Fang et al., 2023). A plausible implication is that comparisons across platforms are most meaningful when made at the level of transfer efficiency, transport length, and loss channels rather than at the level of any single rate formula.

2. Microscopic transfer laws and transport formalisms

For multichromophore systems in the weak-coupling point-dipole limit, the baseline description is Förster-type resonant energy transfer. In the scanning-tunnelling-microscopy study of phthalocyanine assemblies, the transfer rate is written

$$k_{ET}=\frac{1}{\tau_D}\left(\frac{R_0}{R}\right)^6,$$

with

$R_0^6=8.79\times10^{-5}\kappa^2 n^{-4}\Phi_D J,$

and

$J=\int_0^\infty F_D(E)\,\epsilon_A(E)\,dE.$

Here $\tau_D$ is the donor lifetime, $R$ the donor-acceptor distance, $\kappa^2$ the dipole-orientation factor, $E(x,y)=-ed\,F_z(x,y)$0 the refractive index, $E(x,y)=-ed\,F_z(x,y)$1 the donor quantum yield, and $E(x,y)=-ed\,F_z(x,y)$2 the donor-acceptor spectral overlap. The same work also gives, for in-plane dipoles with $E(x,y)=-ed\,F_z(x,y)$3,

$E(x,y)=-ed\,F_z(x,y)$4

which makes explicit why dipole alignment is a primary control variable in molecular funnels (Cao et al., 2021).

At shorter distances, the pure Förster description breaks down. The same STM-resolved measurements show a mixed FRET/Dexter regime at sub-3 nm separations, with multipole terms and short-range exchange contributing and the effective rate following $E(x,y)=-ed\,F_z(x,y)$5 in the weak-coupling limit, where $E(x,y)=-ed\,F_z(x,y)$6 includes both Coulombic and exchange components (Cao et al., 2021). This matters because many experimentally useful funnels operate precisely in the regime where geometric compactness maximizes coupling but invalidates a strictly point-dipole treatment.

A complementary route to directionality is environmental rather than purely energetic. In the quantum-dot-chain analysis of directed exciton transfer, the key rate between exciton eigenstates takes the factored form

$E(x,y)=-ed\,F_z(x,y)$7

so directionality is determined jointly by the system factor $E(x,y)=-ed\,F_z(x,y)$8 and the bath frequency-correlation function $E(x,y)=-ed\,F_z(x,y)$9, itself set by the spectral density $_2$0 and thermal occupation. The paper’s central claim is that tailoring $_2$1 can produce a “spectral funnel,” including reversals of transfer direction without changing network topology or increasing electronic couplings (Perdomo et al., 2010).

For spatial funnels in semiconductors, the standard framework is drift-diffusion. Representative forms include

$_2$2

for strain-defined monolayer funnels, and

$_2$3

for more general potential landscapes, with the Einstein relation $_2$4 for neutral excitons (Moon et al., 2019, Su et al., 2022). In CQW ramps, interaction and disorder screening are added through a mean-field term $_2$5 and a thermionic diffusion coefficient

$_2$6

which directly ties funnel performance to density-dependent mobility enhancement (Dorow et al., 2016).

3. Molecular and supramolecular implementations

A particularly explicit molecular funnel was constructed from three phthalocyanine chromophores adsorbed on NaCl/Ag(111): PdPc as a high-gap donor with $_2$7 eV, ZnPc as an intermediate ancillary with $_2$8 eV, and H$_2$9Pc as a low-gap acceptor with $U(r)$0 eV and a weak $U(r)$1 eV. Local STM excitation of PdPc produces a cascaded transfer sequence PdPc $U(r)$2 ZnPc $U(r)$3 H$U(r)$4Pc, directly visualized by highly resolved fluorescence microscopy. The measured spectral overlaps rank as $U(r)$5 eV$U(r)$6 for PdPc$U(r)$7ZnPc, $U(r)$8 eV$U(r)$9 for ZnPc$$k_{ET}=\frac{1}{\tau_D}\left(\frac{R_0}{R}\right)^6,$$0H$$k_{ET}=\frac{1}{\tau_D}\left(\frac{R_0}{R}\right)^6,$$1Pc, and $$k_{ET}=\frac{1}{\tau_D}\left(\frac{R_0}{R}\right)^6,$$2 eV$$k_{ET}=\frac{1}{\tau_D}\left(\frac{R_0}{R}\right)^6,$$3 for PdPc$$k_{ET}=\frac{1}{\tau_D}\left(\frac{R_0}{R}\right)^6,$$4H$$k_{ET}=\frac{1}{\tau_D}\left(\frac{R_0}{R}\right)^6,$$5Pc, matching the efficiency ordering. For nearly colinear PdPc$$k_{ET}=\frac{1}{\tau_D}\left(\frac{R_0}{R}\right)^6,$$6ZnPc$$k_{ET}=\frac{1}{\tau_D}\left(\frac{R_0}{R}\right)^6,$$7 dipoles, the experiment reports $$k_{ET}=\frac{1}{\tau_D}\left(\frac{R_0}{R}\right)^6,$$8 and $$k_{ET}=\frac{1}{\tau_D}\left(\frac{R_0}{R}\right)^6,$$9; the end-to-end trimer funnel PdPc$R_0^6=8.79\times10^{-5}\kappa^2 n^{-4}\Phi_D J,$0H$R_0^6=8.79\times10^{-5}\kappa^2 n^{-4}\Phi_D J,$1Pc reaches $R_0^6=8.79\times10^{-5}\kappa^2 n^{-4}\Phi_D J,$2 up to $R_0^6=8.79\times10^{-5}\kappa^2 n^{-4}\Phi_D J,$3, compared with $R_0^6=8.79\times10^{-5}\kappa^2 n^{-4}\Phi_D J,$4 in the direct dimer (Cao et al., 2021).

The same study is notable because it distinguishes ancillary relays from passive bridges. A near-resonant ZnPc intermediary extends the transfer range with only modest loss, whereas a passive high-gap PdPc bridge can enhance ZnPc$R_0^6=8.79\times10^{-5}\kappa^2 n^{-4}\Phi_D J,$5H$R_0^6=8.79\times10^{-5}\kappa^2 n^{-4}\Phi_D J,$6Pc transfer across a $R_0^6=8.79\times10^{-5}\kappa^2 n^{-4}\Phi_D J,$7 nm separation where the vacuum-bridged case is negligible. The proposed mechanisms are a three-body dipolar enhancement via ac-polarizability and a superexchange pathway that increases $R_0^6=8.79\times10^{-5}\kappa^2 n^{-4}\Phi_D J,$8 (Cao et al., 2021). This suggests that efficient funnels need not be strict energy staircases; nonresonant units can be useful if they reshape coupling pathways.

At a larger supramolecular scale, double-walled C8S3 nanotubes realize a hierarchical outer-to-inner funnel. The outer wall absorbs at 589 nm ($R_0^6=8.79\times10^{-5}\kappa^2 n^{-4}\Phi_D J,$9 cm$J=\int_0^\infty F_D(E)\,\epsilon_A(E)\,dE.$0) and the inner wall at 599 nm ($J=\int_0^\infty F_D(E)\,\epsilon_A(E)\,dE.$1 cm$J=\int_0^\infty F_D(E)\,\epsilon_A(E)\,dE.$2), producing a downhill offset $J=\int_0^\infty F_D(E)\,\epsilon_A(E)\,dE.$3 cm$J=\int_0^\infty F_D(E)\,\epsilon_A(E)\,dE.$4. Ultrafast 2D spectroscopy resolves outer$J=\int_0^\infty F_D(E)\,\epsilon_A(E)\,dE.$5inner transfer with $J=\int_0^\infty F_D(E)\,\epsilon_A(E)\,dE.$6–$J=\int_0^\infty F_D(E)\,\epsilon_A(E)\,dE.$7 fs, while intralayer diffusion is described by an effective $J=\int_0^\infty F_D(E)\,\epsilon_A(E)\,dE.$8 nm$J=\int_0^\infty F_D(E)\,\epsilon_A(E)\,dE.$9 ps$\tau_D$0 and Haken–Strobl–Reineker diffusion tensors with axial components $\tau_D$1 nm$\tau_D$2 ps$\tau_D$3 for the inner wall and $\tau_D$4 nm$\tau_D$5 ps$\tau_D$6 for the outer wall. At low exciton density the outer wall acts as an antenna supplying excitons to the inner tube; at high density outer-wall annihilation throttles transfer and protects the inner tube from overburning (Kriete et al., 2019).

Natural light-harvesting systems provide the historical template for these artificial examples. The molecular-aggregate review identifies hierarchical energy flow such as chlorosome $\tau_D$7 baseplate $\tau_D$8 FMO $\tau_D$9 reaction center, with nearest-neighbor chlorosome couplings on the order of 100 meV and staged spectral tuning across subunits (Saikin et al., 2013). In purple bacteria, however, counterfactual modeling shows that energetic funneling is more important than delocalization-induced supertransfer: after energy optimization, efficiencies become high even when delocalization is strongly reduced, whereas weakening the downhill energy landscape sharply degrades transfer (Baghbanzadeh et al., 2015).

4. Semiconductor potential landscapes: electrostatic, strain, and dielectric funnels

In coupled quantum wells, funneling can be implemented electrically rather than chemically. A perforated top electrode at constant voltage creates a spatial gradient in the perpendicular field $R$0 and thus a ramp potential $R$1 for indirect excitons. In the reported GaAs/Al$R$2Ga$R$3As device, the electron-hole layer separation is $R$4 nm, the CQWs lie 100 nm above an n$R$5 ground plane inside a 1 $R$6m intrinsic layer, and the perforation-induced fine-scale modulation is only $R$7–$R$8 meV, below the intrinsic disorder scale of $R$9 meV. The average transport distance is quantified by

$\kappa^2$0

and at $\kappa^2$1 K and $\kappa^2$2 V the measured $\kappa^2$3 increases markedly with excitation power up to $\kappa^2$4W because repulsive exciton-exciton interactions screen disorder and increase both $\kappa^2$5 and $\kappa^2$6 (Dorow et al., 2016).

Strain-defined funnels in monolayer WSe$\kappa^2$7 use the bandgap as the control parameter. In suspended membranes indented by a nanoscale tip, the exciton energy follows approximately $\kappa^2$8 with $\kappa^2$9–$E(x,y)=-ed\,F_z(x,y)$00 meV/\%. At 4 K, a z-piezo displacement of 50 V corresponds to a membrane-center deflection $E(x,y)=-ed\,F_z(x,y)$01 nm, applied force $E(x,y)=-ed\,F_z(x,y)$02 nN, maximum local biaxial strain $E(x,y)=-ed\,F_z(x,y)$03, and an optically averaged central strain $E(x,y)=-ed\,F_z(x,y)$04, producing a measured redshift of about 10–12 meV. At room temperature, time-resolved photoluminescence gives $E(x,y)=-ed\,F_z(x,y)$05 cm$E(x,y)=-ed\,F_z(x,y)$06/s and $E(x,y)=-ed\,F_z(x,y)$07 ns, so $E(x,y)=-ed\,F_z(x,y)$08m; the measured funneled-intensity decay lengths are $E(x,y)=-ed\,F_z(x,y)$09 nm and $E(x,y)=-ed\,F_z(x,y)$10 nm, and about $E(x,y)=-ed\,F_z(x,y)$11 of excitons can be collected at the tip from $E(x,y)=-ed\,F_z(x,y)$12m away (Moon et al., 2019).

A different strain geometry, based on Au nano-gaps and hyperspectral TEPL imaging, pushes the gradient to the nanoscale and thereby increases the drift fraction. In monolayer WSe$E(x,y)=-ed\,F_z(x,y)$13 and MoS$E(x,y)=-ed\,F_z(x,y)$14 suspended over $E(x,y)=-ed\,F_z(x,y)$15–300 nm gaps, the drift-diffusion analysis defines a funneling efficiency

$E(x,y)=-ed\,F_z(x,y)$16

For WSe$E(x,y)=-ed\,F_z(x,y)$17 at $E(x,y)=-ed\,F_z(x,y)$18, the model gives $E(x,y)=-ed\,F_z(x,y)$19 for a 300 nm nano-gap, compared with $E(x,y)=-ed\,F_z(x,y)$20 for a 3 $E(x,y)=-ed\,F_z(x,y)$21m micro-gap under the same strain; TEPL intensity increases by about $E(x,y)=-ed\,F_z(x,y)$22 at natural wrinkles and about $E(x,y)=-ed\,F_z(x,y)$23 at deterministic nano-gap centers (Lee et al., 2021). The authors attribute this gain to the fact that the drift-dominant region spans about 100 nm, more than 60% of the strain-gradient area.

Dielectric funnels show that the bright exciton is not always the relevant degree of freedom. In bilayer WSe$E(x,y)=-ed\,F_z(x,y)$24, dielectric nanobubbles in the hBN cladding create local screening reductions that hardly perturb the bright $E(x,y)=-ed\,F_z(x,y)$25–$E(x,y)=-ed\,F_z(x,y)$26 resonance but produce a low-energy landscape for momentum-indirect dark excitons. stroboSCAT directly images superdiffusive drift toward these bubbles, with a drift velocity $E(x,y)=-ed\,F_z(x,y)$27 nm/ns over roughly $E(x,y)=-ed\,F_z(x,y)$28m. In fully encapsulated bilayers the diffusivity is $E(x,y)=-ed\,F_z(x,y)$29 cm$E(x,y)=-ed\,F_z(x,y)$30/s; at representative dielectric bubbles the trap lifetime reaches $E(x,y)=-ed\,F_z(x,y)$31 ns, about $E(x,y)=-ed\,F_z(x,y)$32 longer than the flat-region lifetime, corresponding via a thermally activated estimate to $E(x,y)=-ed\,F_z(x,y)$33 meV (Su et al., 2022). This is a qualitatively different funnel from a strain-defined bright-state sink.

5. Competing mechanisms, controversies, and limiting factors

A persistent question is whether directional transport in light-harvesting systems is governed mainly by coherent delocalization or by the energy landscape itself. In purple-bacterial antennas, counterfactual models that compare natural and “trimmed” geometries show that energetic funneling is decisive, while supertransfer provides only a limited rate enhancement. With original site energies, the natural geometry gives $E(x,y)=-ed\,F_z(x,y)$34 for the S parameter set and $E(x,y)=-ed\,F_z(x,y)$35 for the R set, but trimming LH2 can improve efficiency, and after optimizing site energies to reinforce the downhill landscape all S geometries reach $E(x,y)=-ed\,F_z(x,y)$36 while all R geometries reach $E(x,y)=-ed\,F_z(x,y)$37, regardless of the presence or absence of strong delocalization (Baghbanzadeh et al., 2015). The paper’s interpretation is that supertransfer is at most a constant-factor gain, whereas spectral-overlap penalties from poor energy alignment are exponential.

A second controversy concerns the range of validity of simple point-dipole transfer models. In the phthalocyanine dimers and trimers, RET efficiency decreases monotonically with donor-acceptor distance and vanishes for $E(x,y)=-ed\,F_z(x,y)$38 nm on NaCl/Ag(111), while the measured contrast between inline and parallel dipole geometries is smaller than predicted by pure Förster theory. The authors interpret this as evidence for a short-range mixed FRET/Dexter regime and for the breakdown of a pure point-dipole model at small separations (Cao et al., 2021). Similar caveats apply to interfacial funnels in mixed-dimensional heterostructures, where a Förster interpretation is disfavored because transfer strength varies by orders of magnitude across CNT chiralities with similar spectral overlap and instead tracks band offsets (Fang et al., 2023).

Theoretical work has also shown that funnels can be improved by counterintuitive energy landscapes. In one-dimensional excitonic wires with an intrinsic energy gradient, periodic on-site barriers partition the chain into blocks with one bright state at the top and multiple lower dark states. Vibrationally mediated transitions then move population predominantly through dark subspaces, suppressing radiative recombination. For $E(x,y)=-ed\,F_z(x,y)$39 and room-temperature phonons, an optimized barrier configuration yields an approximately $E(x,y)=-ed\,F_z(x,y)$40 increase in steady-state power relative to a simple linear gradient, and in radiatively dominated regimes the improvement can reach about seven orders of magnitude (Davidson et al., 2020). This is not a rejection of funneling; it is a redefinition of what counts as an optimal funnel architecture.

Practical limits remain severe in several geometries. In non-uniformly strained monolayer TMDCs, realistic funneling efficiencies are predicted to remain below $E(x,y)=-ed\,F_z(x,y)$41 both at room temperature and at low temperature because diffusion dominates at room temperature while monolayer exciton lifetimes become too short at cryogenic temperature. By contrast, in TMDC heterostructures with long-lived interlayer excitons, the efficiency approaches a thermodynamic limit of about $E(x,y)=-ed\,F_z(x,y)$42 at room temperature for $E(x,y)=-ed\,F_z(x,y)$43, and Auger recombination becomes the main limitation under intense illumination (Harats et al., 2020). This suggests that lifetime engineering can be more consequential than further steepening an already strong gradient.

6. Design principles and emerging architectures

Across the literature, several design rules recur. In multichromophore funnels, small downhill energy steps maximize spectral overlap at each hop while limiting energetic loss; the phthalocyanine example $E(x,y)=-ed\,F_z(x,y)$44 eV is the clearest explicit case. Distance and orientation remain critical: on NaCl/Ag(111), keeping $E(x,y)=-ed\,F_z(x,y)$45 nm and targeting near-colinear dipole geometries maximizes transfer, while ancillary near-resonant chromophores and passive polarizable bridges can preserve efficiency over larger spans (Cao et al., 2021). In electrically defined CQW funnels, the imposed gradient must remain smooth relative to disorder, and the perforated-electrode geometry is advantageous because it decouples channel width from ramp slope while keeping fine-scale modulation below the disorder amplitude (Dorow et al., 2016).

For 2D semiconductors, the dominant lesson is geometric: drift becomes useful only when the energy gradient is concentrated over a length scale comparable to or shorter than the diffusion length. The nano-gap TEPL study makes this explicit, with a drift-dominant region of about 100 nm and $E(x,y)=-ed\,F_z(x,y)$46 at only $E(x,y)=-ed\,F_z(x,y)$47 strain, whereas microscale strain gradients typically yield efficiencies below $E(x,y)=-ed\,F_z(x,y)$48 at room temperature because diffusion dominates (Lee et al., 2021). The 2025 dielectric-nanochannel platform pushes this principle further by defining sub-10 nm-wide hBN nanochannels under MoSe$E(x,y)=-ed\,F_z(x,y)$49, creating a quasi-1D dark-exciton funnel and guide. There the transport length exceeds $E(x,y)=-ed\,F_z(x,y)$50m at room temperature, the mean-squared displacement scales as $E(x,y)=-ed\,F_z(x,y)$51 with $E(x,y)=-ed\,F_z(x,y)$52, the propagation speed is about $E(x,y)=-ed\,F_z(x,y)$53 m/s, and the slow lifetime component inside the channel is $E(x,y)=-ed\,F_z(x,y)$54 ns, compared with $E(x,y)=-ed\,F_z(x,y)$55 ns outside (Wang et al., 3 Aug 2025). This suggests that sufficiently smooth dielectric boundaries can convert a funnel from a mere concentrator into a transport channel.

Mixed-dimensional heterostructures add a further principle: the acceptor need not absorb strongly if it is fed by a long-lived donor reservoir. In WSe$E(x,y)=-ed\,F_z(x,y)$56/CNT funnels, room-temperature transfer is strongest at resonant band alignment, where Monte Carlo fits give a local interfacial transfer time $E(x,y)=-ed\,F_z(x,y)$57 ps and thus $E(x,y)=-ed\,F_z(x,y)$58 ps$E(x,y)=-ed\,F_z(x,y)$59. Since the WSe$E(x,y)=-ed\,F_z(x,y)$60 donor lifetime can be about 500 ps, the local transfer efficiency for excitons that reach the interface is

$E(x,y)=-ed\,F_z(x,y)$61

while the global excitation-enhancement factor $E(x,y)=-ed\,F_z(x,y)$62 reaches about $E(x,y)=-ed\,F_z(x,y)$63 at low power for resonant devices (Fang et al., 2023). The practical implication is that a large-area 2D donor can overcome the absorption-area, spectral, and polarization constraints of a 1D emitter without sacrificing the emitter’s optical selectivity.

Coherent-state engineering offers a more speculative but conceptually consistent extension of the funnel idea. Shaped laser pulses can create excitonic wave packets with prescribed speed, direction, and spectral content, allowing selective passing, rejection, dissociation, and remote stimulated-emission removal. In tight-binding and RT-TDDFT simulations, the group velocity follows the excitonic dispersion, packet speeds are tunable by more than a factor of five, and terminal annihilation by a time-reversed field reaches a reported RMS error $E(x,y)=-ed\,F_z(x,y)$64 (Zang et al., 2016). This suggests that future “funnel” architectures may be defined not only by static potentials and rate hierarchies, but also by actively synthesized excitonic initial states.