Intermodal Quantum Key Distribution

• Intermodal quantum key distribution is a method that transmits secure key information across distinct optical channels, such as fiber and free-space links.
• It employs techniques like mode scrambling, adaptive optics, and hybrid encoding to overcome modal conversion challenges and ensure robust security.
• Key experiments demonstrate consistent secret-key rates and validate spatial-, temporal-, and hybrid-mode approaches in diverse network architectures.

Intermodal quantum key distribution denotes a family of QKD constructions in which security-relevant information is carried across distinct transmission media, optical mode spaces, or communication modes rather than within a single canonical qubit channel. In the literature summarized here, the term is used most directly for interoperability between deployed fiber and free-space links (Picciariello et al., 2023, Rossi et al., 18 Feb 2026), but closely related work also treats spatial high-dimensional QKD over multicore or multimode fiber (Cañas et al., 2016, Ding et al., 2016, Amitonova et al., 2018), hybrid encoding across orbital-angular-momentum and path modes (Jo et al., 2019), and, in a broader mode-based sense, retrospective pairing of temporal pulse modes in measurement-device-independent protocols (Zhu et al., 2022). This suggests that “intermodal” is best understood as an umbrella descriptor whose precise meaning depends on whether the relevant modes are transmission media, guided spatial channels, or postselected communication modes.

1. Terminological scope and conceptual boundaries

The most literal use of intermodal QKD in the present literature concerns heterogeneous physical transport, especially fiber–free-space interoperability. In the Padova field trial, intermodal QKD is defined as running the same QKD devices and protocol stack across a standard telecom fiber link and a free-space optical link whose received light is coupled into single-mode fiber so that the receiver sees an ordinary fiber input (Picciariello et al., 2023). The 18 km free-space demonstration adopts the same interpretation: a telecom-wavelength fiber-QKD platform is operated through a long atmospheric segment and then reintegrated into standard single-mode fiber and a fiber-compatible guided-wave analyzer (Rossi et al., 18 Feb 2026).

A second major meaning is spatial-mode or space-division-multiplexed QKD. In multicore-fiber HD-QKD, the quantum alphabet is defined by orthogonal core channels and coherent superpositions across them, so the relevant Hilbert space is explicitly spatial (Cañas et al., 2016, Ding et al., 2016). In the multimode-fiber key-establishment protocol, the information-bearing states are high-dimensional superpositions over many guided modes, and security depends on intermodal scrambling and calibrated inversion (Amitonova et al., 2018). These works are intermodal in the broad spatial-mode sense even when the physical carrier is better described as weakly coupled cores or a random multimode transmission matrix rather than LP eigenmodes of a few-mode fiber.

A broader hybrid-modal usage appears in "Efficient High-dimensional Quantum Key Distribution with Hybrid Encoding" (Jo et al., 2019). There, a single photon carries information jointly in orbital angular momentum and path. The paper itself does not use the term intermodal, and the most faithful label is hybrid-encoding or multi-degree-of-freedom QKD, but it is intermodal in the sense that the cryptographic variable is distributed across two distinct optical mode families.

Mode-pairing MDI-QKD occupies a further semantic boundary. "Experimental mode-pairing measurement-device-independent quantum key distribution without global phase-locking" (Zhu et al., 2022) is directly relevant only if intermodal is read in a temporal or optical-mode sense rather than a spatial-mode sense. Its key bit is recovered from the relation between two retrospectively selected communication modes, not from multimode-fiber or spatial-mode transport.

Several networking papers are adjacent rather than strict intermodal demonstrations. The O/C-band shared-receiver star network is a wavelength-division multiplexed prepare-and-measure system, not a spatial-mode experiment (Scalcon et al., 15 Jul 2025). The metropolitan optical network based on WDM addresses coexistence of quantum and classical channels and interoperability between access and backbone segments, not guided-mode intermodality (Ciurana et al., 2013). The metropolitan MDIQKD field network is conceptually important for heterogeneous networks because it secures communication through an untrusted central measurement node, but all demonstrated links are fiber-based and uniformly time-bin phase encoded (Tang et al., 2015). Finally, "Unconditionally secure key distribution without quantum channel" (Yin, 2024) is orthogonal to intermodal QKD: it claims secure key establishment with no quantum signal exchanged between the users, thereby bypassing rather than solving modality-bridging problems.

2. Fiber–free-space intermodal architectures

The clearest operational definition of intermodal QKD appears in the 2023 Padova field trial. The network comprised two polarization-based transmitters and a single receiver, with the active channel alternately switched between a 620 m free-space link and a 17 km deployed metropolitan fiber (Picciariello et al., 2023). Both systems used polarization encoding at $1550.12\ \mathrm{nm}$, efficient BB84 with 3 states and 1 decoy, a passive decoder with one single-photon detector at Bob, and Qubit4Sync. The optical switch was a custom $2\times 2$ fiber switch with insertion loss of approximately $-1\ \mathrm{dB}$, integrated into the control software for manual or automatic switching. In the field trial, the channels alternated in 15-minute intervals, and the switching overhead was about $20$–$30\ \mathrm{s}$. The mean SKR was approximately $1.6\ \mathrm{kbps}$ on the fiber link and $1.5\ \mathrm{kbps}$ on the free-space link, with average QBERs of approximately $1.4\%$ and $2.5\%$, respectively. The central result was that the same QKD hardware and software required no different strategies to work over the two channels.

The free-space terminal in that experiment was designed around single-mode-fiber recoupling, which is the essential intermodal interface. The receiver used a $6\times$ beam reducer, a fast steering mirror, a dichroic separation stage, a position-sensitive detector for beacon-based tracking, and SMF coupling followed by a 100 GHz WDM filter (Picciariello et al., 2023). The link was characterized against atmospheric turbulence via the beacon-coupling histogram. The trial reported $2\times 2$0, $2\times 2$1, $2\times 2$2, and $2\times 2$3, which the authors classified as moderate turbulence. This suggests that the intermodal abstraction in that system is not merely software reuse; it is a physical mode-conversion strategy in which the free-space receiver restores a fiber-compatible spatial mode.

The 2026 field trial extends this architecture to long-range free-space operation with explicit high-order adaptive optics. The end-to-end link connected a remote transmitter to an urban optical ground station over an 18 km horizontal free-space path, followed by SMF coupling, DWDM separation, and a deployed 0.5 km fiber link to a physically separate QKD receiver (Rossi et al., 18 Feb 2026). The telescope aperture was $2\times 2$4, the quantum channel used $2\times 2$5, and the AO beacon used $2\times 2$6. The protocol was polarization-encoded 3-state 1-decoy efficient BB84 with Qubit4Sync.

In that system, the receiver-side SMF-coupling model was written as

$2\times 2$7

with $2\times 2$8 (Rossi et al., 18 Feb 2026). The AO bench used a Shack-Hartmann wavefront sensor, a 64-actuator deformable mirror, and correction up to $2\times 2$9 Zernike modes, although the effective AO rejection bandwidth was limited to $-1\ \mathrm{dB}$0. The measured average signal detection rates were $-1\ \mathrm{dB}$1 with SNSPDs and $-1\ \mathrm{dB}$2 with room-temperature SPADs, at an average noise rate of approximately $-1\ \mathrm{dB}$3. The average SKR was $-1\ \mathrm{dB}$4 with $-1\ \mathrm{dB}$5-efficiency SNSPDs and $-1\ \mathrm{dB}$6 with $-1\ \mathrm{dB}$7-efficiency room-temperature InGaAs SPADs, at overall channel losses of approximately $-1\ \mathrm{dB}$8–$-1\ \mathrm{dB}$9. This establishes that intermodal QKD can be implemented as a telecom-compatible fiber/free-space/fiber chain rather than as a specialized bulk-optics endpoint.

3. Spatial-mode and multicore-fiber high-dimensional QKD

A core strand of intermodal QKD research treats the spatial transmission basis itself as the quantum alphabet. In the four-core decoy-state system, the high-dimensional states are defined by the multiple core modes of a telecom multicore fiber, and the protocol is a 4-dimensional BB84-like prepare-and-measure decoy-state scheme (Cañas et al., 2016). The fiducial multicore superposition is

$20$0

where $20$1 denotes the state of the photon transmitted by the $20$2-th core mode. The implemented mutually unbiased bases are equal-amplitude four-core superpositions with relative phases $20$3 or $20$4, and Bob’s overlap-based projection rule is

$20$5

The secret-key formula is

$20$6

with

$20$7

Experimentally, the system demonstrated a stable and secure high-dimensional decoy-state QKD session over a 0.3 km four-core telecommunication fiber, with average QBER $20$8 over more than 20 hours of continuous operation, and for the weak-decoy-plus-vacuum run at 0.3 km it reported $20$9 per pulse. The paper further states that the decoy-state analysis enables a positive secret key generation rate up to 25 km of fiber propagation, but that claim is model-based rather than an end-to-end experimental transmission.

The silicon-photonic multicore experiment implements a related spatial Hilbert space with integrated state preparation and analysis (Ding et al., 2016). Four cores of a 7-core MCF define the basis states $30\ \mathrm{s}$0, and three MUBs in $30\ \mathrm{s}$1 are realized as coherent two-core superpositions such as

$30\ \mathrm{s}$2

The mutual information is written as

$30\ \mathrm{s}$3

with $30\ \mathrm{s}$4, and the theoretical information content is $30\ \mathrm{s}$5. The experiment used a 3 m MCF link, achieved an average QBER of approximately $30\ \mathrm{s}$6 over more than 10 minutes of operation, and remained below the quoted $30\ \mathrm{s}$7 thresholds of $30\ \mathrm{s}$8 for individual attacks with 2 MUBs, $30\ \mathrm{s}$9 for individual attacks with 5 MUBs, and $1.6\ \mathrm{kbps}$0 for coherent attacks. This work is path encoding across weakly coupled cores rather than few-mode-fiber eigenmode transport, but from the quantum-information perspective it is squarely spatial-mode HD-QKD.

The multimode-fiber key-establishment protocol pushes the spatial-mode idea into strongly scrambled intermodal propagation (Amitonova et al., 2018). Alice encodes one symbol $1.6\ \mathrm{kbps}$1 by preparing an input wavefront that, after transmission through a 50 $1.6\ \mathrm{kbps}$2-core step-index MMF with approximately $1.6\ \mathrm{kbps}$3 guided modes, focuses onto one of $1.6\ \mathrm{kbps}$4 designated output positions at Bob. The field relation is linear, and Eve’s intermediate-tap statistics are described by $1.6\ \mathrm{kbps}$5. The paper derives Bob’s ideal alphabet capacity $1.6\ \mathrm{kbps}$6, while Eve’s single-photon information is bounded by

$1.6\ \mathrm{kbps}$7

For weak coherent light, Eve’s wavefront-estimation fidelity is

$1.6\ \mathrm{kbps}$8

This is not standard BB84; the paper explicitly frames it as quantum key establishment rather than orthodox basis-switched QKD. Nonetheless, it is one of the clearest examples in which intermodal scrambling itself is the cryptographic resource.

4. Hybrid and mode-based encodings beyond spatial multiplexing

Intermodal structure also appears when the quantum alphabet spans different photonic degrees of freedom rather than multiple guided channels of the same fiber. In the high-dimensional hybrid protocol, Alice encodes a $1.6\ \mathrm{kbps}$9-ary symbol in OAM and Bob encodes a $1.5\ \mathrm{kbps}$0-ary symbol in path, so the single-photon Hilbert space is

$1.5\ \mathrm{kbps}$1

(Jo et al., 2019). The receiver projects onto hybrid states

$1.5\ \mathrm{kbps}$2

which are analogous to generalized Bell states built from two degrees of freedom of the same photon. Against collective attacks, the sifted-signal key rate is

$1.5\ \mathrm{kbps}$3

In the ideal case $1.5\ \mathrm{kbps}$4, this reduces to $1.5\ \mathrm{kbps}$5. The protocol is explicitly not fully detector-device-independent, so its significance lies less in trustlessness than in showing how hybrid cross-DoF encoding can raise the alphabet dimension.

Mode pairing provides a different broad interpretation of intermodality. In MP-MDI-QKD, Alice and Bob encode information into single pulses and pair successful rounds only after Charlie announces which pulses clicked, so the logical variable lives in the relation between two selected temporal modes rather than a predetermined pulse pair (Zhu et al., 2022). The long-distance motivation is to retain the $1.5\ \mathrm{kbps}$6 scaling associated with single-photon-interference families while avoiding global phase locking between remote lasers. The experiment used two independent off-the-shelf lasers at $1.5\ \mathrm{kbps}$7, achieved secure transmission over 304 km commercial fiber and 407 km ultralow-loss fiber, and reported finite-size secret-key rates per second of $1.5\ \mathrm{kbps}$8, $1.5\ \mathrm{kbps}$9, $1.4\%$0, and $1.4\%$1 bits/s at 101, 202, 304, and 407 km, respectively. The paper is explicit that this is intermodal only in a temporal/optical-mode sense and not in the spatial-mode or multimode-fiber sense.

A further adjacent development is the HM-QCT protocol, which is multimode in a physically explicit sense but lies outside standard unconditional QKD (Mazzoncini et al., 2023). Alice sends $1.4\%$2 copies of the $1.4\%$3-mode state

$1.4\%$4

and Bob decodes Hidden Matching parities by interfering mode pairs selected by his matching. The paper claims everlasting security in the Quantum Computational Timelock model against arbitrary i.i.d. attacks and states that the scheme remains secure with up to

$1.4\%$5

input photons per channel use. This is not conventional intermodal QKD in the unconditional-security sense, but it is a clear example of intermodal interference used as a cryptographic primitive.

5. Security models and trust assumptions

The security architecture of intermodal and adjacent mode-based QKD is heterogeneous. At one end lies MDI networking, which removes trust in the measurement device. The metropolitan MDIQKD network deployed in Hefei used a three-user, four-node star topology over a 200 square kilometers metropolitan area, with user-to-relay links of 17 km, 25 km, and 30 km, and an untrusted central relay that performed the Bell-state measurement (Tang et al., 2015). The network ran continuously for one week. The secure-key-rate formula was given as

$1.4\%$6

with failure probability $1.4\%$7. Although this work is not intermodal in the physical-link sense, it supplies the untrusted-relay architecture that heterogeneous metropolitan networks often require.

At the other end are trusted shared-receiver architectures. The O/C-band WDM-QKD star network used three simultaneous prepare-and-measure links—two in the C band and one in the O band—decoded by a single broadband shared receiver (Scalcon et al., 15 Jul 2025). The links ran simultaneously for about 6 hours, generating approximately $1.4\%$8, $1.4\%$9, and $2.5\%$0 of secret key for the 1550.12 nm, 1549.32 nm, and 1310 nm channels, respectively, with average QBERs of $2.5\%$1, $2.5\%$2, and $2.5\%$3 in the $2.5\%$4-basis and $2.5\%$5 in the $2.5\%$6-basis for the 1310 nm link. Its security follows the finite-key three-state BB84 framework of Rusca et al., but the central node remains trusted. This is not strict intermodal QKD, yet it is structurally relevant because it demonstrates how heterogeneous quantum channels can share a broadband measurement front end.

A more radical departure is "Unconditionally secure key distribution without quantum channel" (Yin, 2024). The proposed probability key distribution protocol claims unconditional security with no quantum signal exchanged between the users, using local quantum state preparation and measurement, pre-shared secret keys $2.5\%$7 and $2.5\%$8, and authenticated classical communication. The paper gives the final secret-key length as

$2.5\%$9

up to the obvious typographical normalization in the provided text. Its own positioning is orthogonal to intermodal QKD: it does not bridge modalities, but rather removes the inter-user quantum channel entirely. A plausible implication is that it should be treated as a comparative alternative rather than as part of the core intermodal-QKD literature.

6. Engineering constraints, performance envelopes, and network outlook

Across these papers, the dominant engineering bottleneck is the preservation or restoration of the relevant optical mode structure. In fiber–free-space intermodal QKD, the limiting step is SMF coupling after atmospheric propagation. The 18 km trial writes the end-to-end efficiency as

$6\times$0

with $6\times$1 and $6\times$2, and models turbulence via the Fried parameter

$6\times$3

(Rossi et al., 18 Feb 2026). The Padova switching experiment reached similar kbps-scale SKRs without high-order AO at 620 m, but it also showed that free-space links do not always produce a complete post-processing block within each scheduled slot because of fluctuating coupling and loss (Picciariello et al., 2023). Together these results imply that long-range intermodal deployment is less a question of protocol redesign than of mode-recovery optics, tracking, and AO bandwidth.

In spatial-mode fiber systems, the dominant impairments are differential phase drift, polarization-path leakage, and balanced-loss control across channels. The multicore decoy-state experiment required active feedback, a quantum eraser stage, periodic checks every 30 s, and a threshold of $6\times$4 QBER for feedback re-engagement (Cañas et al., 2016). The silicon-photonic multicore system identified independent phase and polarization drift in each core as the major long-distance obstacle and proposed per-core polarization tuning, on-chip polarization-diversity circuits, cascaded delays, and active feedback loops as remedies (Ding et al., 2016). In the multimode-fiber key-establishment experiment, channel characterization and stability are central because the decoding basis is the private transmission matrix itself; the calibration phase therefore becomes part of the security perimeter (Amitonova et al., 2018).

Temporal-mode intermodality introduces a different constraint: phase estimation across retrospectively paired rounds. MP-MDI-QKD avoids global optical phase locking by estimating the slowly varying laser-frequency difference from strong reference pulses in software, but its performance still depends on the pairing window $6\times$5, detector dead time, and the growth of $6\times$6-basis error with pairing length (Zhu et al., 2022). This clarifies why the protocol realizes clean $6\times$7 behavior most clearly in metropolitan and intercity regimes rather than indefinitely.

At the network level, several papers indicate how intermodal systems may share infrastructure even when the multiplexed degrees of freedom are not themselves spatial modes. The metropolitan WDM architecture separates quantum channels in the O band from service channels in the C/L bands and reports compatibility with aggregate service-band input power up to approximately $6\times$8, corresponding to more than 32 simultaneous service channels at $6\times$9 each (Ciurana et al., 2013). The O/C-band shared-receiver network shows that nearly all Bob-side decoding hardware except the SPADs can be shared across simultaneous heterogeneous quantum channels (Scalcon et al., 15 Jul 2025). This suggests that future intermodal networks may be architected as hybrids of mode-selective front ends, shared relay or receiver resources, and telecom-style wavelength planning rather than as isolated one-link experiments.

The literature therefore points toward a layered picture of intermodal QKD. At the physical layer, intermodality may mean free-space/fiber transitions, multicore or multimode spatial Hilbert spaces, or cross-DoF hybrid encodings. At the protocol layer, it may also include temporal mode pairing and multimode interference primitives. At the network layer, shared-receiver and shared-infrastructure architectures become decisive. The common requirement across these layers is not a single universal protocol, but high-fidelity control over the mapping between a distributed optical mode structure and the trusted degrees of freedom used for key extraction.