Papers
Topics
Authors
Recent
Search
2000 character limit reached

Static Phase Reconfiguration

Updated 6 July 2026
  • Static phase reconfiguration is a method that sets persistent phase states via non-volatile material switching, precomputed schedules, or quasi-static hardware updates.
  • It spans diverse applications including optical metasurfaces, reconfigurable networks, distribution systems, and power converters, each interpreting 'phase' differently.
  • Techniques such as laser-induced phase changes, mechanical sliding of metasurfaces, and optimized topology scheduling enhance performance by reducing control complexity and static power consumption.

Static phase reconfiguration denotes a class of techniques in which a phase-related state is set by non-volatile material switching, a precomputed schedule, a one-time planning decision, or a slow or quasi-static hardware update, rather than by dense continuous tuning. In the cited literature, the phrase spans several technically distinct objects: optical phase profiles in phase-change metasurfaces, synchronized communication phases in optical reconfigurable networks, voltage phasors and user-to-phase assignments in distribution systems, and output-wire assignments or topology modes in reconfigurable converters (Karvounis et al., 2016, Juerss et al., 26 May 2026, 1804.02080, Kerckhove et al., 8 Jul 2025, Deakin et al., 2024). The unifying feature is that the phase-related configuration is deliberately chosen to persist over a useful interval, so that switching overhead, control complexity, or static power consumption is reduced.

1. Scope and terminology

A recurring source of ambiguity is that the word “phase” does not denote the same object across the literature. In metasurfaces it usually denotes the optical phase imparted by a unit cell or pixel; in optical reconfigurable networks it denotes synchronized communication phases in a collective schedule; in distribution engineering it can denote either the voltage phasor angle or the discrete connection of single-phase users to phases $1,2,3$; and in reconfigurable converters it appears through phase-to-wire assignments or three-phase topology changes (Karvounis et al., 2016, Juerss et al., 26 May 2026, 1804.02080, Li et al., 2017).

Domain Meaning of “phase” Reconfigured quantity
Metasurfaces and intelligent surfaces Optical phase or phase-shift profile Material state, geometry, or slide-position-dependent composite phase map
Optical reconfigurable networks Synchronized communication phases Sparse reusable topology states and reconfiguration schedule
Distribution systems Voltage phasor or feeder phase connection DER dispatch across switches or user phase assignment
Power converters Phase-to-wire or topology association Switch networks and selector matrices

“Static” is likewise domain-dependent. In GST and Sb2_2S3_3 devices, the written state is non-volatile and remains stable until another optical, thermal, or electrical stimulus is applied (Karvounis et al., 2016, Liu et al., 2024). In low-voltage distribution networks, static phase reconfiguration is explicitly a multi-period planning exercise in which selected consumers are manually re-phased during routine maintenance (Kerckhove et al., 8 Jul 2025). In movable intelligent surfaces, the phase shifts themselves are static, while reconfiguration is achieved by mechanical sliding of a secondary metasurface layer (Zheng et al., 21 Nov 2025). In reconfigurable converters for unbalanced systems, the reconfiguration stage is updated on a slow “static” timescale and is not part of the high-frequency modulation loop (Deakin et al., 2024).

2. Non-volatile optical phase control in phase-change metasurfaces

In the all-dielectric GST metasurface of Karvounis et al., the reconfigurable element is a $300$ nm-thick film of Ge2_2Sb2_2Te5_5 patterned into a one-dimensional subwavelength grating with period PP in the range $750$–$950$ nm, slot width 2_20 nm, nanowire width 2_21, and height 2_22 nm. Under normal incidence with TE polarization, these gratings support high-2_23 (2_24) Mie-type resonances in reflection and transmission. The optical response is written through the complex coefficients

2_25

with a rapid phase variation of order 2_26 across resonance. Switching GST from the amorphous to the crystalline state shifts the resonance frequency by 2_27, approximately 2_28 in practice, producing

2_29

Rigorous-coupled wave analysis or finite-element simulation yields 3_30 that peaks near the original resonance; in practice, 3_31 can approach 3_32 over a 3_33 nm bandwidth in reflection, with similar behavior in transmission (Karvounis et al., 2016).

The switching mechanism is laser-induced and non-volatile. The amorphous3_34crystalline transition uses a 3_35 nm CW laser with spot diameter 3_36m and intensity 3_37 mW/3_38m3_39, holding GST between $300$0C and $300$1C for $300$2s until crystallization nucleates. Once set, each phase is stable at room temperature, yielding reflection contrast up to $300$3 ($300$4 dB) and transmission contrast up to $300$5 ($300$6 dB). Static phase patterns can be created either by geometric tuning, for example varying $300$7, $300$8, or the resonator diameter, or by phase-state multiplexing that locally toggles GST between amorphous and crystalline states. In reflection the imposed phase is

$300$9

The reported phase-modulation depth is 2_20 per resonance shift of 2_21 nm, while full 2_22 coverage is described as achievable by cascading multiple resonances or by combining geometry and state switching (Karvounis et al., 2016).

A related but electrically addressed implementation is the hybrid plasmonic–phase-change architecture of Abdollahramezani et al. Each pixel is a one-dimensional Au/GST/SiO2_23/Au stack with period 2_24 nm, Au nanoribbon width 2_25 nm, thickness 2_26 nm, GST stripe height 2_27 nm, and a 2_28 nm Au back-reflector. GST forms the dielectric gap under each Au ribbon, and its phase is switched by Joule heating. A short, high-amplitude voltage pulse of approximately 2_29 ns melts GST above 2_20C and rapid quenching re-amorphizes it, whereas a longer, lower-amplitude pulse of approximately 2_21s heats GST above its glass-transition/crystallization temperature of approximately 2_22C to nucleate and grow the crystalline phase. Thermal isolation by SiO2_23 ensures neighboring pixels remain below 2_24C, enabling individual pixel addressing (Abdollahramezani et al., 2018).

The phase modulation is governed by interference between a nearly constant non-resonant metallic reflection and a resonant Lorentzian term,

2_25

with 2_26 and 2_27. The GST crystallinity fraction 2_28 is described by a Lorentz–Lorenz effective-medium model,

2_29

At 5_50 nm, full-wave FEM/FIT simulations yield a reflected-phase sweep of nearly 5_51, from 5_52 to 5_53, with reflection amplitude above approximately 5_54 throughout. Eleven discrete states, 5_55, were simulated, with approximately 5_56 phase increment per step. The device supports a 5_57 nm bandwidth, approximately 5_58–5_59m, over which the phase can be reliably mapped to PP0, enabling focusing, beam steering, holography, amplitude modulation, and polarization conversion (Abdollahramezani et al., 2018).

A common misconception is that static optical phase reconfiguration is equivalent to a fixed-function device. These studies show the opposite: the written state is static only between updates, while the overall platform remains rewritable and non-volatile (Karvounis et al., 2016, Abdollahramezani et al., 2018).

3. Static phase shifts with mechanical or material-state reconfigurability

The movable intelligent surface framework extends static phase design by separating phase synthesis from reconfiguration actuation. The proposed MIS consists of two transmissive metasurfaces separated by a small air gap PP1: a fixed primary layer MS1 with PP2 elements and static phase shifts PP3, and a smaller secondary layer MS2 with PP4 elements and static phase shifts PP5, which slides discretely over MS1 along rows and columns. The number of slide positions is PP6, where PP7 and PP8. At slide position PP9, the equivalent phase-shift vector is

$750$0

where $750$1 is a binary selection matrix and $750$2 is a padding vector. Defining $750$3, the received signal at user $750$4 is

$750$5

The design problem jointly optimizes the static phase shifts and the slide-position selection, interpreted as beam-pattern scheduling (Zheng et al., 21 Nov 2025).

Two optimization classes are reported. For max–min rate, a penalty-assisted BCD-SCA method relaxes the binary scheduling variables, adds a penalty $750$6, and alternates over $750$7, $750$8, and $750$9. For sum rate, a Riemannian conjugate-gradient method operates on a product manifold formed by the complex-circle manifolds of $950$0 and $950$1 and the probability-simplex manifold of the scheduling matrix $950$2. Simulation results report that even a $950$3 MS2 increases min-rate by approximately $950$4–$950$5 bit/Hz over a static RIS, larger MS2 sizes such as $950$6 add up to $950$7–$950$8 bit/Hz, and element-wise MS2 mobility recovers approximately $950$9 of the gap between block-MIS and a fully dynamic RIS. At 2_200 dBm, the reported sum rates are approximately 2_201 bit/Hz for block-MIS, 2_202 bit/Hz for dynamic RIS, and approximately 2_203 bit/Hz for element-wise MIS. Reconfiguration is quasi-static, with switching latency on the order of seconds or minutes and control overhead of only 2_204 bits per re-pointing (Zheng et al., 21 Nov 2025).

Phase-change metasurfaces for reconfigurable image processing provide a different static mechanism. In the Sb2_205S2_206-based device of Liu et al., each nanobrick with height 2_207 nm, radius 2_208 nm, and lattice constant 2_209 nm supports magnetic-dipole and electric-dipole Mie resonances in the visible, observed at 2_210 nm and 2_211 nm, respectively. In the amorphous state, 2_212–2_213 and 2_214, yielding high-2_215, narrow Mie resonances and a transmission profile suitable for edge detection; in the crystalline state, 2_216–2_217 and 2_218, which blueshifts and broadens the resonances, lowers the 2_219-factor, and produces nearly angle-independent transmission. The target transfer functions are

2_220

Near resonance and for small absorption, the angular dependence satisfies

2_221

and the simulated transmission at 2_222 nm and 2_223 nm was fitted to 2_224, with 2_225 at 2_226. The reported modulation depth is 2_227, the high-2_228 transmission is 2_229 at 2_230 versus near 2_231 at 2_232, the numerical aperture is approximately 2_233, and the spatial resolution is approximately 2_234m at 2_235 nm (Liu et al., 2024).

These works show two distinct meanings of static phase reconfiguration in wave-control systems. In MIS, the phase values are static but the geometry changes; in Sb2_236S2_237 and GST metasurfaces, the geometry is fixed but the material state changes (Zheng et al., 21 Nov 2025, Liu et al., 2024).

4. Phase-matched topology states in optical reconfigurable networks

In optical reconfigurable networks, static phase reconfiguration concerns collective communication phases rather than electromagnetic phase delay. ReTri is designed for a minimal two-transceiver-per-node ORN, equivalently a ring-degree-two topology with 2_238 OCS ports. Assuming 2_239, each source block 2_240 is assigned a signed offset

2_241

which admits a unique balanced-ternary expansion

2_242

In phase 2_243, blocks with 2_244 are sent right, blocks with 2_245 are sent left, and blocks with 2_246 stay. This completes All-to-All in

2_247

synchronized phases, each served by a sparse reusable subring (Juerss et al., 26 May 2026).

The topology states are induced by residue classes

2_248

with bidirectional links 2_249 inside each 2_250. Each 2_251 is a simple ring of size 2_252. The paper states that no smaller subgraph can both connect the sender/receiver pairs for phase 2_253 and keep all future peers reachable. Reconfiguration is controlled by a binary schedule 2_254, where 2_255 means the OCS is reconfigured to the subrings 2_256 before phase 2_257, and 2_258 means the previous topology is reused. If the last reconfiguration occurred at phase 2_259, then each link is stretched by 2_260 in hop length, increasing propagation delay and per-link congestion (Juerss et al., 26 May 2026).

The cost model is an extended Hockney form,

2_261

where 2_262 is per-phase startup, 2_263 per-hop latency, 2_264 the per-byte transmission cost, 2_265 the message chunk size, 2_266 the maximum per-link congestion, and 2_267 the reconfiguration time. If ReTri reconfigures before every phase, then

2_268

If it never reconfigures, then

2_269

In practice the schedule is precomputed once from 2_270, 2_271, 2_272, 2_273, 2_274, and 2_275, and the optimal 2_276 is selected by evaluating 2_277 for 2_278 (Juerss et al., 26 May 2026).

In Astra-Sim/ns-3 experiments with 2_279 Gb/s links, 2_280s propagation per hop, 2_281s per-phase startup, and 2_282 swept from 2_283s to 2_284 ms, ReTri achieved up to 2_285 speed-up versus static All-to-All for small 2_286, with benefits persisting up to 2_287 ms for messages up to 2_288 MB and even 2_289 ms for 2_290 MB. It also improved reconfigurable Bruck by up to 2_291, with typical gains of 2_292–2_293 for small messages and 2_294–2_295 for larger messages. The central design principle is that topology optimization and communication-phase structure are co-designed rather than treated independently (Juerss et al., 26 May 2026).

A common misconception is that reconfigurable optical fabrics necessarily require per-phase topological reinstrumentation. ReTri is explicitly built around the opposite trade-off: sparse reconfiguration is used to amortize 2_296 across multiple phases, at the price of longer paths (Juerss et al., 26 May 2026).

5. Voltage phasors and user-phase assignment in distribution systems

One branch of the literature uses static phase reconfiguration to regulate voltage phasors before a switching action. In unbalanced distribution feeders, the linearized model developed for switch synchronization combines voltage magnitude and angle relations. For line 2_297,

2_298

and

2_299

where 3_300 and 3_301. Together they form the “full-phasor” linearization

3_302

An OPF then minimizes squared magnitude and angle differences across a target switch while keeping all voltages within 3_303–3_304 p.u. and respecting DER capacities. On modified IEEE 13-node feeders, the full phasor control case reduced the per-phase average maximum 3_305 from 3_306 p.u. in the no-control case to 3_307 p.u., and the maximum 3_308 from 3_309 to 3_310; the resulting switch flow decreased from approximately 3_311 p.u. to approximately 3_312 p.u. On the modified IEEE 37-node sequential switching case, 3_313 decreased from approximately 3_314 to 3_315 for the first switch and from approximately 3_316 to 3_317 for the second. Reported QP solution times were 3_318–3_319 s) on a standard PC (1804.02080).

A distinct planning-oriented use of the same phrase concerns the static re-phasing of low-voltage users. Here binary variables 3_320 assign single-phase user 3_321 to phase 3_322, subject to

3_323

The formulation can also limit the total number of switching actions through

3_324

Three optimization approaches are compared: an exact MINLP, an MIQP based on LinDist3Flow, and a GA. The exact MINLP supports actual imbalance metrics such as

3_325

while the MIQP uses proxy objectives such as 3_326 and 3_327. In a 3_328-user, 3_329-time-step case, MIQP reduced PVUR from 3_330 to 3_331 and 3_332 from 3_333 to 3_334, compared with 3_335 and 3_336 for the GA average. The MIQP solution times were approximately 3_337 min for the PVUR3_338 objective and approximately 3_339 min for the 3_340 objective on 3_341 threads, while the GA took approximately 3_342 min per run and the MINLP did not reach optimality within 3_343 h for the 3_344-node case. The study also reports that even 3_345 re-phasing, corresponding to 3_346, achieves approximately 3_347 of the maximum imbalance reduction (Kerckhove et al., 8 Jul 2025).

These two lines of work address different operational layers. The OPF-based method aligns full phasors across a switch immediately before a reconfiguration action, whereas the LV planning formulations choose a single phase-connection assignment that performs well over months to seasons (1804.02080, Kerckhove et al., 8 Jul 2025).

6. Reconfigurable power-electronic topologies and capability expansion

In reconfigurable power electronics, static phase reconfiguration is realized by changing circuit interconnections rather than continuously varying modulation parameters. One example is the automatic transformation between a three-phase 3-level cascaded H-bridge and a three-phase 2-level converter. Two topologies are presented. In Topology 1, ideal high-resistance switches reroute the midpoints; in Topology 2, a single full H-bridge replaces the separate PWM generators and the design uses two fewer ideal switches overall. A safe break-before-make sequence is specified: inhibit all PWM at 3_348, wait a deadtime margin, change the ideal-switch gates, wait for switch transition, then re-enable PWM on the newly selected bridge. The typical timing values given are 3_349–3_350s and 3_351s, with a representative total reconfiguration time of approximately 3_352s. In 3-level mode, the phase-to-neutral voltage is

3_353

with 3_354 V for 3_355 V sources; in 2-level mode,

3_356

The reported trade-offs are lower THD in 3-level mode, higher THD or higher switching frequency requirements in 2-level mode, and slightly higher reliability with Topology 2 because it uses fewer ideal switches (Li et al., 2017).

A second line of work concerns a low-cost reconfiguration stage at the output of balanced three-phase, multi-terminal ac/dc/ac converters for unbalanced distribution-system applications. The proposed four-wire VSC replaces fixed hard-wiring with 3_357 half-bridge legs and a 3_358 one-leg-to-four-wire selector matrix 3_359, where each column has exactly one 3_360. If 3_361 is the sum of all leg ratings and the leg capacities are 3_362 with 3_363, then the maximum current on wire 3_364 is

3_365

This allows leg capacity to be reassigned from one wire to another on a slow timescale, for example to support a heavily loaded phase or a large neutral current (Deakin et al., 2024).

Capability is described through a feasible injection set 3_366 subject to per-phase current constraints, neutral-current limits, and either standalone or interconnected dc-side power balance. In the conventional equal-leg four-wire design, the standalone active-power projection yields a regular hexagon with area

3_367

For an idealized fully reconfigurable converter,

3_368

For the proposed discrete reconfigurable converter, ignoring neutral clipping,

3_369

The paper reports that a conventional converter may need up to 3_370 greater capacity to match an idealized reconfigurable converter. For the optimally sized four-leg design 3_371 with 3_372, the required capacity increase is reported as approximately 3_373. Uniform 3_374-leg designs improve with 3_375, with reported standalone factors 3_376 for 3_377, 3_378 for 3_379, and 3_380 for 3_381, although the study notes diminishing marginal returns beyond 3_382–3_383 (Deakin et al., 2024).

Across converter studies, the principal trade-off is explicit: static reconfiguration increases feasible operating sets and converter utilization, but it adds selector-switch complexity, additional point-of-failure risk, and design constraints associated with neutral current and break-before-make operation (Li et al., 2017, Deakin et al., 2024).

Topic to Video (Beta)

No one has generated a video about this topic yet.

Whiteboard

No one has generated a whiteboard explanation for this topic yet.

Follow Topic

Get notified by email when new papers are published related to Static Phase Reconfiguration.