Zero-Energy Devices: Concepts and Applications

• Zero-Energy Devices (ZEDs) are multifaceted systems that range from energy-harvesting, battery-less IoT endpoints to devices exploiting zero-energy modes in superconducting and non-Hermitian physics.
• In mobile networks, ZEDs leverage ambient backscatter and low-power RF modulation for applications like indoor localization, achieving significant detection and data reliability even in challenging conditions.
• Advanced ZED research integrates energy-aware protocols, TinyML for smart sensing, and robust diagnostic methods in nanowire devices to distinguish between trivial states and true zero-energy phenomena.

Zero-Energy Devices (ZEDs) is a polysemous term used in several technical literatures. In 6G and Ambient IoT research it denotes low-cost, resource-constrained, maintenance-free, and energy-harvesting IoT devices, often battery-less and backscatter-based (López et al., 2024). In semiconductor–superconductor nanowire research it is used for devices whose operation relies on controlled zero-energy modes (Wang et al., 2022). In related wave and quantum settings it appears in studies of robust zero energy extended states and of operations performed at zero energy cost (Ferdous et al., 2022, Chiribella et al., 2015). The cited papers therefore use the same expression for distinct technical objects, and careful disambiguation is necessary.

1. Terminology and scope

In the cited literature, the term “zero-energy” does not carry a single invariant meaning. In wireless systems it usually refers to operation powered by harvested energy and to near-zero-power RF transmission through backscatter. In condensed-matter physics it refers to zero-energy states in superconducting or topological spectra. In net-zero system design it refers to a zero net operating energy balance over a task or over long-term operation.

Domain	Meaning of ZED	Representative papers
6G / Ambient IoT	low-cost, resource-constrained, maintenance-free, energy-harvesting IoT devices; self-powered, backscatter-only, near-zero power RF transmitter	(López et al., 2024, Yang et al., 2024)
Semiconductor–superconductor devices	devices based on semiconductor–superconductor nanowires and controlled zero-energy modes	(Wang et al., 2022, Kobiałka et al., 2020)
Non-Hermitian / acoustic systems	devices exploiting robust zero energy extended states residing in the bulk	(Ferdous et al., 2022)
Net-zero platforms	net-zero-energy lifelogging system; zero net trip energy in principle	(Arita et al., 2022, Ahmad et al., 2011)

This terminological breadth is not superficial. It separates at least three technical questions: how to operate an IoT endpoint without battery replacement, how to engineer and diagnose zero-energy spectral states, and how to achieve zero net operating energy in a larger cyber-physical system.

2. Ambient-backscatter ZEDs in mobile networks

The dominant current use of ZEDs in mobile networking derives from the “Crowd-Detectable Zero-Energy Device” concept, shortened to ZED, and from related Ambient IoT work (Yang et al., 2024, Phan-Huy et al., 2021). In this sense, a ZED is self-powered, backscatter-only, and a near-zero power RF transmitter: it harvests ambient energy, does not generate its own RF carrier, and modulates existing cellular signals by switching its reflection state. Each device carries a unique identification number coded into a bit sequence, and later work generalizes the concept to battery-less, spectrum-agnostic, and future-proof XG-ZEDs across successive network generations (Amani et al., 12 Nov 2025).

The canonical architecture is an antenna plus a reconfigurable load or switch, a low-power digital controller, and an energy harvester such as a solar cell with small storage (Yang et al., 2024). Communication is achieved by altering the effective propagation channel seen by a nearby receiver. In the 4G localization formulation, transparent mode ideally reflects nothing, reflecting mode reflects everything, and the two effective channel states are $\boldsymbol{\Gamma}$ and $$\mathbf{G}=\boldsymbol{\Gamma}+\boldsymbol{\Phi}\odot\boldsymbol{\Lambda}$$, where $\boldsymbol{\Gamma}$ is the BS→SM direct channel and $\boldsymbol{\Phi}\odot\boldsymbol{\Lambda}$ is the backscatter component (Yang et al., 2024).

A central practical insight is that bursty cellular traffic degrades naive ambient backscatter reception, whereas standardized pilots remain steady. A real-time field demonstration therefore used only LTE Cell-specific Reference Signals from the Orange commercial 4G network as the ambient source, with FSK at $F_0=125$ Hz and $F_1=500$ Hz, 40 ms symbols, and a 120-bit frame composed of a 63-bit synchronization sequence and a 57-bit data sequence (Ndiaye et al., 2023). In deep indoor NLOS conditions, the reported detection ratios were 65.8% at average 4G SNR $\le 0$ dB and 96.27% at average 4G SNR $<4$ dB, with average data BER 0.1008 and 0.0439, respectively (Ndiaye et al., 2023). That result established field operation on a commercial network rather than on a base-station emulator.

Receiver simplicity remains a defining requirement. For Manchester-coded OOK downlinks to ZEDs, exact and approximate non-coherent soft-decision decoders have been derived for envelope-detection receivers (Zhang et al., 2024). In AWGN, at BER = 0.1%, soft decision with the low-complexity approximate LLR outperforms hard decision by about 1.5 dB and outperforms the uncoded case by about 4.6 dB; at BLER = 10%, the corresponding gains are ≈1.8 dB and ≈5.6 dB (Zhang et al., 2024). In block Rayleigh fading, interleaving raises the benefit further, with gains up to about 10 dB at BER = 1% for suitable interleaver sizes (Zhang et al., 2024). These results place ZED communications squarely in the regime of ultra-low-rate, ultra-low-complexity, energy-aware PHY design.

3. Detection, localization, and beacon protocols

Indoor localization is one of the most developed ZED application layers. In the 4G smartphone localization study, the infrastructure consists of an Orange commercial base station operating at 852–862 MHz with carrier $f_0=857$ MHz, bandwidth $BW=9$ MHz, transmit power 46 dBm, $$\mathbf{G}=\boldsymbol{\Gamma}+\boldsymbol{\Phi}\odot\boldsymbol{\Lambda}$$0, and $$\mathbf{G}=\boldsymbol{\Gamma}+\boldsymbol{\Phi}\odot\boldsymbol{\Lambda}$$1 pilot symbols per RB per TTI (Yang et al., 2024). The building contains $$\mathbf{G}=\boldsymbol{\Gamma}+\boldsymbol{\Phi}\odot\boldsymbol{\Lambda}$$2 ZEDs, generally at least one per room, positioned manually using QGIS. The smartphone knows the map of ZED positions and IDs, continuously measures pilots, decodes ZED transmissions from variations in the estimated downlink channel, estimates a ZED-specific SNR, and selects the ZED with highest SNR as the closest beacon (Yang et al., 2024).

The relevant physical-layer quantities are explicit. The backscatter power is

$$\mathbf{G}=\boldsymbol{\Gamma}+\boldsymbol{\Phi}\odot\boldsymbol{\Lambda}$$3

and the ZED SNR is

$$\mathbf{G}=\boldsymbol{\Gamma}+\boldsymbol{\Phi}\odot\boldsymbol{\Lambda}$$4

For coherent binary detection in AWGN, the Bit Error Probability is

$$\mathbf{G}=\boldsymbol{\Gamma}+\boldsymbol{\Phi}\odot\boldsymbol{\Lambda}$$5

and the coverage area of a ZED is the set of smartphone positions where $$\mathbf{G}=\boldsymbol{\Gamma}+\boldsymbol{\Phi}\odot\boldsymbol{\Lambda}$$6 (Yang et al., 2024). The reported result is qualitative but operationally important: room-level precision can be obtained with a basic localization scheme and a coherent detector in a realistic and challenging propagation scenario (Yang et al., 2024). Increasing $$\mathbf{G}=\boldsymbol{\Gamma}+\boldsymbol{\Phi}\odot\boldsymbol{\Lambda}$$7 enlarges coverage areas because the number of pilot observations per bit increases (Yang et al., 2024).

Later work formalized ZED detection as a Neyman–Pearson problem. For a single tag, the detector estimates the ZED path power $$\mathbf{G}=\boldsymbol{\Gamma}+\boldsymbol{\Phi}\odot\boldsymbol{\Lambda}$$8 from dual correlator outputs and combines statistics across subcarriers with

$$\mathbf{G}=\boldsymbol{\Gamma}+\boldsymbol{\Phi}\odot\boldsymbol{\Lambda}$$9

then sets the threshold from a target false-alarm probability through

$\boldsymbol{\Gamma}$0

with detection probability

$\boldsymbol{\Gamma}$1

(Yang et al., 14 Apr 2025). In a FIT/CorteXlab testbed, the reported peak-to-lobe ratio was about 11 dB (Yang et al., 14 Apr 2025). Multi-tag detection then replaced Barker coding with a Near-Perfect Code, improving the observed peak-to-sidelobe ratio from about 11 dB to about 22 dB and making concurrent-tag separability substantially better in practical indoor environments (Amani et al., 12 Nov 2025). This progression marks a shift from proof-of-concept beaconing to scalable ZED deployments under interference and synchronization uncertainty.

4. Net-zero operation, energy-aware protocols, and embedded intelligence

A broader ZED systems literature studies not just backscatter tags but complete self-sustaining embedded platforms. A general 6G overview describes ZEDs as relying on unconventional EH sources, multi-source EH, power management, energy storage solutions, backscattering, low-complexity receivers, TinyML, duty cycling, radio frequency wireless power transfer, and wake-up protocols (López et al., 2024). A more formal treatment decomposes ZED design into EI acquisition, task-level variability, and storage dynamics, with energy information obtained by comparator-based monitoring, information sampling, energy-integrated accumulation, or indirect EI from environmental sensors (López et al., 30 Jul 2025). The storage dynamics are written as

$\boldsymbol{\Gamma}$2

subject to $\boldsymbol{\Gamma}$3 (López et al., 30 Jul 2025). This framework shifts ZED protocol design away from idealized energy-neutral abstractions and toward explicit measurement cost, leakage, and task atomicity.

The lifelogging system ZEL provides a concrete net-zero wearable example (Arita et al., 2022). It is a 192-gram nametag-shaped device equipped with a dye-sensitized solar cell, an amorphous solar cell, a piezoelectric element used as a sensor, a 47 μF capacitor, a CR2032 coin cell as backup/timekeeping source, and two comparators that switch among operational states. In experiments with 11 participants, the person-dependent model achieved an 8-place recognition accuracy level of 87.2% weighted F-measure and a static/dynamic activities recognition accuracy level of 93.1% weighted F-measure (Arita et al., 2022). Over four 9-hour workdays, the average zero-energy rate was 99.57%, corresponding to a “not zero-energy” rate of 0.43%; the paper therefore reports a zero-energy operation rate of 99.6%, i.e., net-zero-energy operation (Arita et al., 2022). The result is important because it grounds “zero-energy” in measurable time-domain operability rather than in a purely conceptual energy balance.

Energy-aware TinyML extends this line to smart ZEDs. A solar-powered person-detection platform with a Panasonic AM-5608 cell, e-peas AEM10941 PMU, supercapacitor storage, Arduino Nano 33 BLE Sense, and Arducam camera used a multi-exit TinyML model so that low-energy inference is performed for easy instances and higher-precision inference is selectively employed for harder cases (Jahanbazi et al., 9 Mar 2026). Compared with a state-of-the-art energy-aware single-exit TinyML approach, the reported energy consumption reduction was approximately 29.6% (Jahanbazi et al., 9 Mar 2026). The paper also quantified subsystem costs: capture plus preprocess energy dropped from 110.442 mJ to 72.896 mJ with MOSFET-based camera power gating, while inference energy was 8.118 mJ at EX1 and 13.390 mJ at EX2 (Jahanbazi et al., 9 Mar 2026). This is a representative shift in ZED research from pure communication to joint sensing–inference–power-management co-design.

5. Zero-energy modes in condensed-matter devices

A separate ZED literature concerns semiconductor–superconductor platforms in which device functionality is tied to controlled zero-energy modes. Three-terminal InSb–Al nanowire devices use two normal probes and one superconducting drain to perform simultaneous tunneling spectroscopy from both ends of a proximitized $\boldsymbol{\Gamma}$4 nanowire segment (Wang et al., 2022). RF reflectometry at $\boldsymbol{\Gamma}$5 MHz accelerates data acquisition, the induced gap at zero field is $\boldsymbol{\Gamma}$6, and lowest-energy-state and zero-energy-state diagrams are mapped over $\boldsymbol{\Gamma}$7, $\boldsymbol{\Gamma}$8, and barrier transparencies (Wang et al., 2022). The major conclusion is negative but methodologically central: robust zero-bias features, including clusters, parabolic patterns, and oscillatory LES patterns, can be generated by trivial Andreev bound states, and the authors do not find conclusive evidence of an unbroken topological superconducting phase with correlated Majorana modes at both ends (Wang et al., 2022). For this reason, end-to-end correlation, barrier-gate dependence, and multi-parameter scans are presented as necessary diagnostics.

A complementary theoretical treatment models a semi-infinite superconductor/normal nanowire junction with Rashba spin–orbit coupling, Zeeman field $\boldsymbol{\Gamma}$9, pairing $\boldsymbol{\Phi}\odot\boldsymbol{\Lambda}$0, and gate-controlled chemical potential (Kobiałka et al., 2020). The topological phase boundary is

$\boldsymbol{\Phi}\odot\boldsymbol{\Lambda}$1

and in the topological regime a finite wire hosts a pair of nearly zero-energy Majorana bound states with splitting $\boldsymbol{\Phi}\odot\boldsymbol{\Lambda}$2 (Kobiałka et al., 2020). The zero-bias differential conductance is

$\boldsymbol{\Phi}\odot\boldsymbol{\Lambda}$3

and the ideal Majorana signature is $\boldsymbol{\Phi}\odot\boldsymbol{\Lambda}$4 (Kobiałka et al., 2020). The paper shows that $\boldsymbol{\Phi}\odot\boldsymbol{\Lambda}$5 can reproduce the topological phase diagram in magnetic field versus gate voltage.

A related experimental platform is a two-dimensional InAs/Al heterostructure patterned by top-down lithography (Suominen et al., 2017). The system exhibits a hard superconducting gap, ballistic tunneling contact, and in-plane critical fields up to 3 T (Suominen et al., 2017). Subgap Andreev states move toward zero energy with magnetic field, coalesce near $\boldsymbol{\Phi}\odot\boldsymbol{\Lambda}$6 T, and generate a stable zero-bias conductance peak over a field window $\boldsymbol{\Phi}\odot\boldsymbol{\Lambda}$7 T, behavior described as consistent with emerging Majorana zero modes (Suominen et al., 2017). The paper is cautious, however: the observed ZBCP is not claimed as definitive evidence, because trivial ABS scenarios remain plausible. The common misconception addressed across these nanowire papers is that a robust zero-bias peak alone suffices; the cited work argues that it does not.

6. Alternative zero-energy paradigms and conceptual boundaries

In non-Hermitian wave physics, zero-energy devices are associated not with harvested-power budgets but with robust spectral states. A one-dimensional chain of lossy coupled resonators based on a non-Hermitian SSH-type model exhibits a zero mode that becomes extended throughout the chain when the band gap is closed by setting $\boldsymbol{\Phi}\odot\boldsymbol{\Lambda}$8 (Ferdous et al., 2022). In the ideal $\boldsymbol{\Phi}\odot\boldsymbol{\Lambda}$9 case the eigenvector occupies all odd sites with alternating sign and vanishing even sites, and the robust bulk state appears experimentally in a 17-cavity acoustic waveguide near 256–257 Hz (Ferdous et al., 2022). The paper reports robustness to 20% coupling disorder in the model and 10% geometric disorder in the acoustic realization, and interprets the result as robust transport that is not limited to boundary phenomena (Ferdous et al., 2022). Here, “zero energy” means a band-center state whose real part remains pinned at zero, not a batteryless operating mode.

In quantum information, zero energy cost means no net energy exchange with the environment. Energy-preserving channels satisfy $F_0=125$0 for all $F_0=125$1, and deterministic state conversion between

$F_0=125$2

has maximal fidelity

$F_0=125$3

under energy-preserving operations (Chiribella et al., 2015). In the probabilistic setting, the maximal fidelity is

$F_0=125$4

with the corresponding optimal success probability specified by the minimum ratio $F_0=125$5 over the common support (Chiribella et al., 2015). This use of “zero-energy” is therefore operational and thermodynamic rather than wireless or topological.

A third distinct usage appears in transport and net-energy accounting. “Zero Energy Travel” defines zero energy as zero net energy over a trip, under the ideal conditions of zero friction and complete recovery of kinetic energy during braking (Ahmad et al., 2011). The argument is built around

$F_0=125$6

magnetic levitation, vacuum or low-pressure tunnels, and brushless permanent magnet generators that convert kinetic energy back into electrical energy (Ahmad et al., 2011). The paper explicitly states that absolute perfection is difficult and that even 99.5% efficiency is the practical target. Across the cited literatures, then, the main misconception is terminological: in IoT, zero-energy generally means energy-harvesting and maintenance-free operation; in topological and non-Hermitian physics, it denotes a zero-energy or zero-bias state; in quantum control, it denotes no energy exchange with the environment; and in transport it denotes zero net operating energy over a trip.