Quantum Fans in Research: Disambiguation & Applications

• Quantum Fans are context-dependent constructs in quantum research that encompass devices for heat redirection, extended finite-temperature critical regions, and multi-target entangling primitives.
• They manifest in diverse fields such as mesoscopic thermodynamics, condensed-matter physics, and quantum computing, each with unique operational and experimental implications.
• The term also extends to quantum generative modeling through QFAN, highlighting a unified theme of distributing limited quantum resources across broader applications.

Searching arXiv for recent and foundational usages of “quantum fan” and closely related terms. “Quantum Fans” denotes several distinct constructions in contemporary research rather than a single canonical object. In mesoscopic thermodynamics, the phrase is used informally for autonomous heat-moving devices that redirect one heat flow using another, without converting the driving resource into external work (Manikandan et al., 2020). In condensed-matter theory, “quantum critical fans” are the finite-temperature crossover regions controlled by nearby zero-temperature criticality, including the case of critical lines rather than isolated critical points (Yu et al., 2023). In quantum computing, “quantum fan-out” is a simultaneous shared-control entangling primitive enabled by global interactions and used for circuit optimization and memory design (Gokhale et al., 2020). The expression also appears by analogy in forced cooling by quantum fluids, where a subsonic flow can act like a “quantum fan” only after a heater creates a thin shell of nonsuperfluid helium around itself (Diribarne et al., 2021), and as the acronym QFAN, the “Quantum Feature Amplification Network,” an autoregressive quantum generative model (Slim et al., 15 May 2026). Taken together, these usages suggest a family of context-dependent meanings centered on directed transport, amplification, or finite-temperature spreading of quantum effects.

1. Terminological scope and disambiguation

The term “fan” is not uniquely quantum-mechanical. In the present literature, some uses are physical and operational, whereas others are purely geometric or combinatorial. The papers “Skeletons and fans of logarithmic structures” (Abramovich et al., 2015), “Fans and polytopes in tilting theory II: $g$-fans of rank 2” (Aoki et al., 2023), and “Fans and polytopes in tilting theory III: Classification of convex $g$-fans of rank 3” (Aoki et al., 26 Aug 2025) explicitly concern fans as collections of cones or sign-coherent combinatorial structures rather than quantum-physical objects. By contrast, the usages relevant here arise in quantum thermodynamics, condensed-matter criticality, quantum hardware, quantum-fluid transport, and quantum machine learning.

This distinction matters because identical vocabulary encodes very different objects. In the thermodynamic and fluid-mechanical usages, “fan” is an analogy for heat redirection; in condensed matter it refers to a region in a phase diagram; in circuit theory it names a multi-target shared-control primitive; and in QFAN it is an acronym. A plausible implication is that “Quantum Fans” is best treated as a disambiguation category rather than a single technical doctrine.

2. Autonomous heat redirection in mesoscopic thermodynamics

A particularly explicit physical use of the “quantum fan” analogy appears in the autonomous absorption refrigerator of resonant-tunneling quantum dots (Manikandan et al., 2020). The device is a three-terminal fermionic setup with a cold left reservoir $L$, a hot right reservoir $R$, and a still hotter fermionic cavity $H$, with

$T_L < T_R < T_H.$

The left and right reservoirs are tunnel-coupled to the central cavity through resonant quantum dots with energies $E_L$ and $E_R$. Because the dots are energy filters, an electron moving from left to right typically enters at $E_L$, leaves at $E_R$, and absorbs

$g$0

from the hot cavity.

The transport theory is formulated in a Landauer-Büttiker description. The particle and energy currents out of lead $g$1 are

$g$2

$g$3

with Lorentzian transmission

$g$4

Steady state requires particle conservation $g$5 and energy conservation $g$6, where $g$7 is the heat current out of the hot cavity.

In the narrow-linewidth regime,

$g$8

transport becomes tight-coupled. Then

$g$9

and for the absorption-refrigerator regime with $L$0,

$L$1

If $L$2 and $L$3, then $L$4, $L$5, and $L$6: heat is extracted from $L$7, dumped into $L$8, and powered solely by heat absorbed from $L$9. This is why the setup is described as autonomous and absorptive rather than work-driven.

The thermodynamic performance is characterized by

$R$0

with Carnot bound

$R$1

In the tight-coupling limit,

$R$2

The paper emphasizes that Carnot coefficient of performance is reached only in the reversible limit where cooling power vanishes.

A notable feature is the particle-hole symmetry of the cooling mechanism. Cooling occurs either in the electron regime

$R$3

or in the hole regime

$R$4

The heat transport is invariant under the corresponding particle-hole transformation. This is the context in which the “quantum fan” analogy becomes useful: the device does not create cold or produce work, but redirects fermionic heat flow by energy filtering. A plausible implication is that the analogy is best understood as a nanoscale heat-current redirector rather than as a generic refrigerator metaphor.

3. Quantum critical fans and finite-temperature crossover structure

In condensed-matter theory, a “quantum critical fan” is the finite-temperature region controlled by zero-temperature criticality (Yu et al., 2023). The exactly solved example studied in the transverse-field Ising model with an added three-spin interaction has Hamiltonian

$R$5

specialized to uniform $R$6. After Jordan–Wigner transformation and Bogoliubov diagonalization, the exact dispersion is

$R$7

The zero-temperature phase boundaries occur where the gap closes. The model has three critical lines: $R$8

$R$9

$H$0

The last is an incommensurate critical line, with gap closing at

$H$1

Generic points on these lines have $H$2, whereas the multicritical point $H$3 has quadratic dispersion and hence $H$4.

The fan is diagnosed by the competition between the gap scale and temperature. The crossover scale obeys

$H$5

with the quantum-critical regime defined by

$H$6

so that

$H$7

On the disordered side, for $H$8, the correlation length saturates,

$H$9

while on the ordered side the renormalized classical regime exhibits

$T_L < T_R < T_H.$0

The distinctive point of this work is that the fan originates from a critical line rather than a single quantum critical point. Instead of a cone emerging from one point in parameter space, the finite-temperature region is extended along the entire critical line, which the authors describe schematically as a “valley.” Near $T_L < T_R < T_H.$1, the extracted exponents are $T_L < T_R < T_H.$2 and $T_L < T_R < T_H.$3; near $T_L < T_R < T_H.$4, they are $T_L < T_R < T_H.$5 and $T_L < T_R < T_H.$6. In both cases,

$T_L < T_R < T_H.$7

so the crossover boundary is linear in distance,

$T_L < T_R < T_H.$8

The same structure appears dynamically through the finite-temperature dynamical structure factor $T_L < T_R < T_H.$9, whose low-temperature features broaden thermally inside the fan. This suggests that “quantum fan” in this context denotes not a device or primitive but an extended finite-temperature domain in which critical scaling rather than phase-specific gapped physics controls observables.

4. Quantum fan-out as a circuit and architectural primitive

In quantum computing, “quantum fan-out” has a sharply defined operational meaning: one control qubit acts on many targets in a single physical step through global interactions (Gokhale et al., 2020). The paper distinguishes logical fan-out, normally represented as several CNOTs with a shared control, from physical fan-out, where overlapping interactions are executed simultaneously. For classical input states with targets initialized to $E_L$0, the state of the control gets copied to the targets; more generally, the primitive is a simultaneous shared-control entangling operation rather than arbitrary cloning.

This distinction matters because standard schedulers assume “exclusive activation,” meaning a qubit can be used in at most one gate per timestep. The paper argues that this assumption is unnecessarily restrictive on hardware that supports global interactions. Using simultaneous fan-out changes the complexity of controlled operations. For a controlled-$E_L$1 operation with $E_L$2 of depth $E_L$3 and width $E_L$4, the effective depth under serialization is stated as $E_L$5, whereas the fan-out-based synthesis achieves $E_L$6 depth with $E_L$7 ancilla.

The mechanism is illustrated by compiler identities that allow shared-control controlled-$E_L$8 gates to be rearranged into constant-depth structures. For shared-control single-qubit gates, the resulting decomposition has constant 5-layer depth, independent of width. For shared-control Toffoli layers, the decomposition has constant 12-layer depth, again independent of width. The paper states that the SWAP test can be reduced from $E_L$9 depth to $E_R$0, with an optimized constant of 14 layers.

These reductions are not purely asymptotic. Under realistic simulations for trapped-ion quantum computers, the paper reports 7–24% reduction in error, and it also demonstrates a superconducting proof-of-concept using simultaneous cross-resonance drives. The hardware motivation is explicit: global interactions permit simultaneous operations on overlapping qubits, so fan-out becomes both a compiler optimization and a technology-modeling problem. The same primitive underlies proposed explicit and implicit quantum memory architectures, where memory-access depth is reduced by simultaneous controlled-SWAP or broadcast patterns. A plausible implication is that “quantum fan” here denotes a hardware-enabled broadcast-like entangling resource whose significance is architectural rather than metaphorical.

5. Subsonic quantum-fluid flow as a “quantum fan”

A further analogy appears in forced cooling by flowing $E_R$1He (Diribarne et al., 2021). The paper shows that a subsonic flow of quantum fluid can behave like a kind of “quantum fan,” but only in a restricted regime. The central result is that forced flow does not generally improve heat transfer in He II simply because He II is already an efficient conductor via thermal counterflow. A velocity dependence appears only when the heater is driven hard enough to create a thin shell of nonsuperfluid helium around itself.

The experiments use wire, film, and chip heaters in pressurized liquid $E_R$2He, with bath temperatures from $E_R$3 to $E_R$4, superfluid fraction from $E_R$5 down to $E_R$6, and imposed mean velocity up to $E_R$7. The cleanest threshold appears for the chip heater: velocity sensitivity begins around

$E_R$8

identified with local interface temperature above $E_R$9, producing a thin He I layer. Below this threshold, no measurable dependence on external velocity is seen.

The mean excess power is fit by the classical-like square-root law

$E_L$0

with fitted prefactors

$E_L$1

in $E_L$2 from $E_L$3 down to $E_L$4. The prefactor is maximal around $E_L$5, i.e. at intermediate superfluid fraction. This is not explained by a purely monotone decomposition of the heat flux into He I and He II contributions.

The physical interpretation is local and hybrid. Very near the surface, the overheated heater produces a supercritical He I shell. In zero flow, the shell obeys Fourier conduction while the outer He II region carries heat by nonlinear counterflow. At finite velocity, the external co-flow modifies primarily the He I-shell transport, leading to an effective Nusselt-type correction

$E_L$6

with effective momentum velocity

$E_L$7

Thus the local velocity dependence is classical-like, but it arises only because the heater has first broken superfluidity in a thin shell.

The paper also identifies two nontrivial quantum-fluid effects. First, at the largest superfluid fraction ($E_L$8, $E_L$9), a new regime appears at non-null velocities and is typically 10% less conductive than at zero velocity, with two metastable states coexisting for $E_R$0. Second, the time series exhibit short-lived “cooling glitches,” producing a broad spectral bump around

$E_R$1

with velocity-dependent characteristic time. The observed peaking is argued to be quantitatively consistent with a von Kármán vortex street in the wake of an effective obstacle much larger than the wire itself. Outer scales

$E_R$2

can reach roughly $E_R$3, about two decades larger than the wire radius in typical conditions.

This suggests that the “quantum fan” analogy is conditional and double-edged. Flow can enhance cooling when the heater has already created a local normal shell, but in strongly superfluid conditions it can reduce mean conductivity by about 10%. The paper’s bottom line is therefore not that He II supports ordinary forced convection, but that a quantum two-fluid environment can support fan-like cooling only through a locally generated classical shell embedded in a fundamentally quantum flow.

6. QFAN: “Quantum Feature Amplification Network”

The acronym QFAN names a different usage altogether: the “Quantum Feature Amplification Network,” an autoregressive quantum generative model for calorimeter shower simulation (Slim et al., 15 May 2026). The central architectural claim is that direct-register quantum generative models tie the quantum output dimension to the full image dimension, whereas QFAN generates an image as a sequence of blocks so that the required register size is fixed by block size rather than full image size.

The model factorizes

$E_R$4

and compresses the generated prefix into a fixed-size count-sketch vector $E_R$5, updated by

$E_R$6

After a near-identity mixing layer,

$E_R$7

the sketch is projected to circuit angles

$E_R$8

The same parameterized quantum circuit $E_R$9 is then reused across all blocks.

For the main demonstration, the circuit uses $g$00 qubits, $g$01 layers, and

$g$02

shared trainable quantum parameters. The measured feature family is

$g$03

with feature count

$g$04

For $g$05, this yields $g$06. All features are estimated from just two tensor-product measurement settings, all-$g$07 and all-$g$08, so the number of measurement groups is $g$09.

Each block is decoded by ridge regression,

$g$10

followed by a post-hoc residual sampler. The per-step circuit-count theorem states that the number of circuits per SPSA step is

$g$11

hence

$g$12

for the present observable family. This is independent of the full image size $g$13. The paper also derives a conservative worst-case bound on shot-noise propagation. If decoder gain is

$g$14

then after $g$15 blocks with $g$16, a sufficient condition for expected sketch signal-to-noise ratio at least $g$17 is

$g$18

so

$g$19

is a conservative worst-case scaling.

Empirically, the model uses an effective decoder-capacity ratio

$g$20

with a practical threshold

$g$21

at the present scale. The main experiments use $g$22, $g$23, $g$24, $g$25, and 512 shots per circuit. On held-out CLIC electromagnetic shower data, the simulator and IBM hardware runs reproduce per-pixel intensity distributions, inter-pixel correlations, and total deposited energy distributions, with mean per-pixel Wasserstein-1 distances

$g$26

for simulator and

$g$27

for hardware. The paper states that, at this scale, the hardware-simulator gap is consistent with optimization-budget limits dominating over device noise, although the experiments do not causally separate these effects.

Here “fan” is neither a thermal analogy nor a phase-diagram region. It is an acronym for a small-circuit feature generator whose outputs are amplified by classical decoders into larger real-valued image blocks. A plausible implication is that QFAN extends the semantic range of “Quantum Fans” from transport and hardware primitives to quantum generative modeling, while retaining the common motif of distributing limited quantum resources across a wider effective output space.