Temporal and Frequency Multiplexing

• Temporal and frequency multiplexing are techniques that assign data to separate time and spectral channels, optimizing throughput and resource efficiency.
• They utilize rigorous frameworks like singular-value decomposition and three-wave mixing to achieve precise mode selectivity and effective channel discrimination.
• Applications include scalable quantum state generation, optical networking, and advanced metrology, while practical systems balance switching losses and noise trade-offs.

Temporal and frequency multiplexing are core strategies for encoding, manipulating, and extracting information in modern photonics, quantum optics, metrology, and signal processing. By judiciously assigning information to distinct time or frequency channels—sometimes in sophisticated combinations—they enable scaling of data rates, efficient resource utilization, and the realization of high-dimensional quantum protocols. This entry synthesizes rigorous theoretical and experimental advances, emphasizing the physical mechanisms, mathematical frameworks, and practical trade-offs at play in temporal and frequency multiplexing.

1. Fundamental Principles and Definitions

Temporal multiplexing allocates distinct information channels—optical, electrical, or quantum—into separate, non-overlapping temporal windows (time-bins), facilitating the serialization of parallel processes through shared hardware paths. Frequency multiplexing, conversely, encodes channels into orthogonal spectral bins—separate carrier frequencies or frequency combs—enabling simultaneous transmission over the same physical medium without temporal interleaving.

An important extension is the joint encoding and manipulation of quantum states or classical data in both time and frequency domains, often called frequency–time multiplexing, which leverages both degrees of freedom for resource-efficient, scalable, and high-capacity information handling (Fischer et al., 3 Mar 2026).

2. Theoretical Frameworks for Multiplexing

The mathematical formulation of multiplexing approaches is grounded in coupled-mode equations, Green-function mappings, and singular-value (Schmidt) decompositions.

Temporal/Frequency Demultiplexing via Three-Wave Mixing

Consider a uniform χ^2 nonlinear waveguide supporting three slow-varying envelopes: A_s(z, t) (signal), A_r(z, t) (idler), with a strong classical pump A_p at frequencies satisfying ωr = ω_s + ω_p and phase matching k_r – k_s – k_p = 0. Group-velocity mismatches (β_j ≡ ∂k/∂ω|ωj) lead to coupled partial differential equations:

$$(\partial_z + \beta_j \partial_t) A_j(z, t) = i \kappa e^{i\Delta k z} A_k(z, t) A_p(z, t)$$

for appropriate choices of j, k, and process (sum- or difference-frequency generation). This set governs temporal and spectral evolution, including pulse broadening and walk-off effects (Reddy et al., 2013).

Channel Discrimination via Green Function and Schmidt Decomposition

Temporal (and frequency) selectivity of the conversion process is captured by the Green function G(t_out, t_in), mapping input waveforms (temporal or frequency modes) at the signal port to output waveforms at the idler port. Singular-value (Schmidt) decomposition:

$$G(t_{\text{out}}, t_{\text{in}}) = \sum_j \lambda_j \chi_j(t_{\text{out}}) \phi_j^*(t_{\text{in}})$$

yields orthogonal input–output mode pairs. The degree to which conversion is restricted to a single mode is quantified by the selectivity figure of merit (S):

$S = \frac{\lambda_1^2}{\sum_j \lambda_j^2}$

Unit selectivity (S = 1) corresponds to perfect mode discrimination—all energy in the dominant mode (Reddy et al., 2013).

Frequency–Time Multiplexing for Quantum States

Joint time–frequency multiplexing creates n-photon states of the form

$$|\Psi_n\rangle = \prod_{k=1}^n a^\dagger(\omega_k, t_0) |0\rangle$$

by synchronizing temporal delay and frequency routing elements (e.g., optical quantum memory and cascaded fiber Bragg gratings). Properly engineered delays align all n frequency-encoded photons into a single temporal bin and spatial mode (Fischer et al., 3 Mar 2026).

3. Experimental Realizations and Architectures

Temporal Multiplexing Methdologies

Temporal multiplexing in single-photon sources utilizes repeated pumping of spontaneous parametric down-conversion (SPDC) sources across M time-bins, each associated with a low probability p for photon-pair generation. Heralding schemes and adjustable optical delay lines—often implemented via binary electronical switches and fiber loops—route any detected photon into a synchronized output time-slot. The total switching loss in cascaded architectures scales exponentially or logarithmically with M:

$$L_\mathrm{total}^\mathrm{temp}(M) = L_\mathrm{sw}^{M-1}$$

$$L_\mathrm{total}^\mathrm{tree}(M) = L_\mathrm{sw}^{\lceil \log_2 M \rceil}$$

where $L_\mathrm{sw}$ is single-switch transmissivity. Losses, speed, and noise from switch electronics and extended delay lines limit scalability (Joshi et al., 2017).

Frequency Multiplexing and Fixed-Loss Architectures

In frequency multiplexing, a broadband SPDC source’s output is partitioned into N distinct frequency channels and routing is accomplished by active frequency conversion (e.g., Bragg-scattering four-wave mixing, BS-FWM) in a single spatial mode. Crucially, all switching occurs in one fiber or integrated waveguide, making switching loss independent of N:

$L_\mathrm{total}^\mathrm{freq}(N) = L_\mathrm{sw}$

This approach supports high-purity, low-noise, and scalable channel multiplexing demonstrated with three frequency channels, yielding 220% rate enhancement over single channels at insertion loss of 1.3 dB (Joshi et al., 2017).

Frequency–Time Multiplexed Quantum State Generation

A hybrid architecture integrates temporal and frequency multiplexing. Heralded photons from an SPDC source are actively delayed by an optical quantum memory (switchable loop) to a target time bin; thereafter, an array of n fiber Bragg gratings, each tuned to a distinct frequency, introduce frequency-dependent delays to perfectly realign the n photons. For an optimally chosen number of time-bins per batch (m), realistic hardware yields 1 kHz rate for 8-photon outputs using only a single delay loop and passive frequency-routing hardware (Fischer et al., 3 Mar 2026).

4. Selectivity, Efficiency, and Scalability

Mode Selectivity in Nonlinear Conversion

Selectivity quantifies the ability of a system to transform only a specific input mode with unity efficiency while suppressing all others, essential for quantum information protocols using orthogonal temporal or spectral modes. Analytical solutions show a trade-off between mode selectivity and conversion efficiency in traveling-wave χ^2 media under group-velocity-mismatched regimes; maximal selectivity can be bounded and is pump-chirp independent (Reddy et al., 2013).

Cavity-based approaches—specifically, dichroic-finesse cavities with large finesse disparity—can approach unit selectivity and efficiency simultaneously by exploiting slow cavity build-up at the target frequency and rapid extraction at the signal frequency. The central parameter is the ratio $f_s = g^2/\gamma_s$ (coupling strength to cavity decay rate), which should be maximized for near-ideal performance (Reddy et al., 2017).

Scalability in Practical Systems

Frequency-multiplexed systems exhibit fixed loss as mode count increases, supporting large-N operation, which is vital for integrated quantum photonics. Frequency–time multiplexed single-photon sources, using quantum memories and frequency-routing, can increase multiphoton generation rates by several orders of magnitude without prohibitive loss scaling (Fischer et al., 3 Mar 2026).

5. Applications Across Physical Systems

Quantum Information Processing

Temporal and frequency multiplexing underpin high-dimensional quantum encoding, deterministic single- and multiphoton sources, and reconfigurable quantum networks. Orthogonality and mode discrimination in time/frequency domains directly impact quantum error correction, entanglement distribution, and deterministic gates (Reddy et al., 2017, Joshi et al., 2017, Fischer et al., 3 Mar 2026).

Optical Networks and Sensing

Transparent optical networks exploit temporal and frequency selectivity for efficient channel routing (“drop devices”). fMux readout architectures for TES bolometers in astrophysics use frequency-selective biasing with LC resonators to achieve 64:1 multiplexing on a single wire pair, reducing cryogenic complexity while maintaining low noise and dynamic range (Bender et al., 2014). Temporal multiplexing accelerates phase-shifting interferometric profilometry, enhancing SNR and coverage in 3D metrology (Servin et al., 2017).

THz Imaging and Reservoir Computing

In THz imaging, rapid alternation of pump conditions enables temporally-multiplexed, multi-frequency image acquisition at kHz rates, with precise crosstalk suppression via atomic transition selectivity (Downes et al., 15 Jul 2025). In physical reservoir computing with magnon systems, both frequency (modal) and temporal multiplexing realize high-dimensional virtual outputs; performance is determined by the number of features (virtual nodes), not modality—demonstrating modality equivalence for computational capacity (Heins et al., 4 Feb 2025).

6. Trade-offs, Limitations, and Future Prospects

A central trade-off in temporal multiplexing is exponential scaling of switching losses with time-bin count, and in frequency multiplexing, the requirement for sharply selective, low-loss frequency conversion or filtering components. Crosstalk is suppressed in both domains by mode orthogonality—temporal selectivity ensured by precise delay and gating, frequency selectivity by pulse shaping and phase matching.

In integrated photonics, frequency multiplexing is particularly suited due to fixed-path switching, compact comb-generation, and potential for on-chip nonlinearities and filtering. Frequency–time multiplexing further allows near-deterministic multiphoton sources with minimal overhead, holding promise for scalable quantum networking.

Table 1 summarizes the comparative scaling of losses:

Multiplexing Scheme	Loss Scaling with Mode Number	Limiting Factor
Temporal (feed-forward)	$L_\mathrm{sw}^{M-1}$	Switch/delay loss, speed
Frequency (single path)	$$G(t_{\text{out}}, t_{\text{in}}) = \sum_j \lambda_j \chi_j(t_{\text{out}}) \phi_j^*(t_{\text{in}})$$0	Conversion/filter fidelity
Frequency–Time (hybrid)	Polynomial (with components)	Memory, FBG, routing efficiency

Future improvements depend on reducing insertion loss in active switches and memories, extending dynamic range in frequency routers, and realizing high-finesse, low-loss cavities for near-ideal selectivity in integrated devices.

7. References to Key Literature

• Mode selective frequency conversion and selectivity bounds in χ^2 waveguides: (Reddy et al., 2013)
• Frequency-multiplexed, fixed-loss, low-noise single-photon sources: (Joshi et al., 2017)
• Hybrid frequency–time multiplexed n-photon state generation: (Fischer et al., 3 Mar 2026)
• Near-unity selectivity temporal-mode conversion in dichroic-finesse cavities: (Reddy et al., 2017)
• High-bandwidth fMux readout for millimeter-wave TES bolometers: (Bender et al., 2014)
• Temporal multiplexing for 3D profilometry: (Servin et al., 2017)
• Dual-frequency THz imaging with atomic vapor: (Downes et al., 15 Jul 2025)
• Modality-equivalent virtual node realization in reservoir computing: (Heins et al., 4 Feb 2025)