{"id":"86188fcd-b526-41b4-9a01-22bdce57b43c","arxiv_id":"2411.17641","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":1,"one_line_summary":"A four-path interferometer with GHz phase modulators switches light between cores of a multicore fiber in 0.7 microseconds and was tested on a real installed fiber network.","lead":"This paper builds a fast core-selective switch for multicore optical fibers using a reconfigurable interferometer, routing signals between cores in under 0.7 microseconds. It demonstrates the first field operation of such a switch, a step toward practical space-division multiplexing networks.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Field 'error-free' claim and WDM claim rest on measurements not shown; waveform-averaged monitoring cannot validate the 0.7 µs rise time or the 1 Gbps signal integrity.","rationale":"The reader's weakest assumption (path balance and stability) is fair and is indeed load-bearing: Eq. (2) assumes equal loss and polarization, and the 15 s stabilization window in Fig. 2a is short. However, I see the more acute gap in the evidence actually required to substantiate the headline claims. The 0.7 µs rise time in Fig. 2c is a single-sample edge with no averaging or error bar, and the field 'error-free transmission' assertion is made without any BER measurement being shown—the paper only reports the B2B BER curves in Fig. 3b and then switches to filtered power traces in Fig. 4b. Similarly, the WDM compatibility claim is supported by a visibility scan (Fig. 5c) but no crosstalk or BER data across the C-band. These are missing measurements, not contradictions, so they do not warrant rejection; they warrant conditional acceptance pending the requested data. The novelty and core interferometric concept are plausible, and the B2B BER penalty of 0.6 dB is a solid internal check. I therefore keep the CONDITIONAL verdict, noting that the required new evidence is targeted at the field and WDM claims rather than at the basic switching demonstration.","tokens_in":8382,"tokens_out":1811,"duration_ms":14722,"concrete_test":"In the field configuration of Fig. 4, run the PRBS-237 1 Gbps signal through the switch with the FPGA actively toggling between the 170 m and 1,305 m paths every 2.5 µs, and record (i) the SFP+ receiver's live BER counter and (ii) a synchronized oscilloscope trace of the recovered clock and data eye, for a 60 s window. Additionally, with the switch held on a single core for 60 s, log the output power in all four cores at the 0.8 MSps rate and compute the time-resolved IC-XT. For the WDM claim, repeat the C-band visibility scan of Fig. 5c while the switch is actively routing a WDM comb of four 100 GHz channels and report per-channel BER and IC-XT over the 1527-1569 nm range.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The central claim—that this is the first high-speed MCF switch operating under real-world field conditions with <0.7 µs switching and error-free 1 Gbps routing—depends on two load-bearing measurements whose evidence is not presented in the manuscript. First, the field test (Fig. 4) reports 'error-free transmission' with no BER curve, no received power vs. time with the switch actively held in a selected state, no mention of clock/data recovery or PRBS error count, and no indication that the switch was toggled during the 15 s window. The figure shows only two photodiode traces of filtered 1 Gbps pulses, from which a recovery penalty or error-free status cannot be inferred; the SFP+ receiver's own error detection is asserted but no BER or Q-factor is quoted. Second, the WDM compatibility claim rests on a single sentence in Methods: 'the device was stabilized using the laser centered at 1550 nm, while the visibility of the interferometer was evaluated for wavelengths between 1527 and 1569 nm.' The C-band visibility curve (Fig. 5c) is shown only down to ~0.90 around 1527 nm, and there is no statement of what happens to crosstalk or insertion loss at detuned wavelengths, nor any test with a modulated WDM comb. Moreover, the stability demonstration (Fig. 2a) is only 15 s of active stabilization; the control loop's ability to maintain visibility over thermally drifting paths for the full switching experiment (Fig. 2b,c) is not quantified. Each of these is addressable, but together they mean the 'first field demonstration' claim and the 'compatible with WDM' claim are not yet supported by the data shown.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper presents a multicore-fiber core-selective switch built as a four-path Mach-Zehnder interferometer, with two multicore-fiber beam splitters, lithium-niobate phase modulators, and an FPGA-based stabilization loop. The authors report a switching rise time below 0.7 microseconds, an average inter-core crosstalk of -16.25 dB, an insertion loss of 7.7 dB, and back-to-back transmission of a 1 Gbps PRBS signal at BER 1e-9 with about 0.6 dB sensitivity penalty. They also report routing a 1 Gbps signal over a field-installed multicore fiber network on a university campus and claim this is the first demonstration of a high-speed MCF switch operating under real-world field conditions, with speeds three orders of magnitude faster than prior MCF switches. The main results are experimental measurements rather than a new theoretical derivation, with the interferometer model in Eq. (1) borrowed from earlier MCF beam-splitter characterization.","tokens_in":8695,"tokens_out":3573,"duration_ms":31155,"significance":"If fully substantiated, the result is significant for space-division-multiplexed networks: a core-selective switch with sub-microsecond switching would be a large practical improvement over millisecond-class mechanical and beam-steering switches. The paper's strengths are its direct experimental approach, the use of commercial components, the back-to-back BER characterization, and the ecological validity of the field-installed MCF test. The reported interference visibilities of 0.97-0.98 support the core unitary-combiner assumption. However, several headline claims are currently backed by incomplete evidence, so the significance is conditional on the missing measurements being supplied.","major_comments":[{"comment":"The abstract states 'less than -18 dB of average inter-core crosstalk,' while the Core switching section reports an average IC-XT of -16.25 dB with a range of -12 to -22 dB. These numbers are mutually inconsistent; since average crosstalk is a headline specification, the abstract must be corrected to match the measured value or the measurement must be re-derived.","section":"Abstract / 'Core switching' Results"},{"comment":"The field-test claim of 'error-free transmission' is not supported by the evidence shown. Fig. 4b presents only low-bandwidth photodiode traces of the received signal; there is no BER curve, no PRBS error count, no statement about clock/data recovery, and no indication of whether the switch was actively toggled during the 15 s window. Without a quantified error measurement and a description of the switching pattern during the field test, the central claim of a field demonstration with error-free 1 Gbps routing is not established.","section":"Operation in optical networks, Fig. 4"},{"comment":"The WDM compatibility claim rests on a single visibility scan over wavelengths from 1527 to 1569 nm, with no crosstalk, insertion loss, or BER data reported for detuned wavelengths and no test with a modulated WDM comb. The abstract's statement that the design is 'fully compatible with standard multiplexing techniques' therefore goes beyond the presented evidence; either add a WDM transmission test or soften the claim.","section":"Core switching / Methods, WDM compatibility"},{"comment":"The 0.7 microsecond rise time appears to be extracted from a single oscilloscope trace (Fig. 2c shows 0.68125 us) with no definition of the rise-time threshold, no repeated switching events, and no statistics across the four output cores. Because the three-orders-of-magnitude speed advantage is a core claim, the rise time should be reported as a distribution over many switching events with a clear 10-90% definition.","section":"Core switching, Fig. 2c"},{"comment":"The active-stabilization demonstration in Fig. 2a covers only 15 s, while the switching and field experiments do not report how long the feedback loop maintained the selected output state. Since the device requires balanced optical paths and the control loop operates at 0.8 MSps, the absence of a stability timescale leaves the long-term field operability claim under-supported.","section":"Active phase stabilization, Fig. 2a"}],"minor_comments":[{"comment":"There are several typographical and terminology issues: 'wave-division multiplexing' should be 'wavelength-division multiplexing', '1GBs' should be '1 Gbps', and the sentence 'this is the first time a a signal is directed...' contains a duplicated article.","section":"Main text, Abstract and Introduction"},{"comment":"The figure axis for the C-band visibility plot shows 1530-1560 nm, while the text states wavelengths between 1527 and 1569 nm; please make the figure and text consistent.","section":"Fig. 5c and Core switching text"},{"comment":"The phase-stabilization description would benefit from a circuit-level block diagram or a more explicit timing analysis; the stated 0.8 MSps conversion rate and the mention of RF filters are not sufficient to assess the loop bandwidth and settling behavior.","section":"Methods, Active phase stabilization"},{"comment":"Eq. (1) uses the same symbol MBS4x4 for the first and second multicore beam splitter without noting whether they are nominally identical devices; a sentence confirming the assumption would clarify the model.","section":"Equation (1) and (2)"}],"recommendation":"major_revision","confidential_remarks":"The hardware demonstration appears credible in its core measurements, but the abstract and conclusion overstate what the presented data show, particularly regarding the field test and WDM compatibility. The missing BER and rise-time statistics are addressable and should be requested before acceptance."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Read it. The device is real and the core result stands: a four-path Mach-Zehnder built from multicore-fiber beam splitters and LiNbO3 phase modulators, switched in under 0.7 us and routed over an installed campus MCF link. That speed is three orders faster than the LPG, beam-steering, and mechanical alternatives they cite, and the field test is legitimately new.\n\nCredit where due: the paper shows interference visibilities of 0.97 and 0.98, BER curves for all four output cores with 0.4 to 0.9 dB penalty, crosstalk between -12 and -22 dB, and a 15-second active-stabilization trace. The architecture has no hidden free parameters: the phase settings are measured, and Eq. (1) takes the MCF-BS matrix from the earlier component characterization. The use of internal polarizers and per-path polarization control is a sensible way to make a coherent interferometer out of telecom parts.\n\nSoft spots, in proportion. First, the abstract says average crosstalk is below -18 dB, but the body reports -16.25 dB. That is a small but real inconsistency. Second, the field \"error-free\" claim is not supported by the evidence shown. Figure 4b is two averaged photodiode traces at 150 MHz bandwidth; there is no BER curve, no PRBS error count, and no indication the switch was toggled at speed during the 15-second window. The stress-test note lands here. This is the biggest gap. Third, WDM compatibility is shown only as visibility versus wavelength. There is no crosstalk or insertion-loss data across the band and no test with a modulated WDM comb, so \"fully compatible\" is too strong. The visibility also dips to about 0.90 at 1527 nm, so the range is narrower than implied. Fourth, the 0.7 us rise time comes from a single trace in Fig. 2c; a second measurement or an error bar would help. Fifth, the stability demonstration is only 15 seconds; long-term thermal drift under field conditions is not quantified. That is not fatal, because the feedback loop is the right instrument, but an hour-long trace would make the claim solid.\n\nThe math and the citation pattern look clean. Equations (1) and (2) are standard unitary interferometer algebra, and the MCF-BS matrix is borrowed from a prior characterization, which is fine. The load-bearing premise is that the four paths stay balanced over time, and the measured visibilities support that premise for the windows shown.\n\nWho gets value from this: people working on MCF networking, SDM switching, and field trials of spatial-division devices. It deserves a serious referee. I would send it out, with a request that the field BER evidence and the WDM crosstalk-versus-wavelength data be added before acceptance.","headline":"A genuinely faster MCF core-selective switch with a real field demo; the evidence supports the architecture, but the abstract's crosstalk number and the field 'error-free' claim are ahead of the data shown.","tokens_in":9298,"tokens_out":2426,"would_cite":true,"duration_ms":22742,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"A four-path interference switch built from multicore fiber beam splitters routes optical signals between cores of a multicore fiber in under 0.7 microseconds, and the paper demonstrates it error-free on a field-installed multicore fiber…","keywords":["multicore fiber","core-selective switch","space-division multiplexing","four-path Mach-Zehnder interferometer","phase stabilization","optical switching","field-installed fiber network","WDM compatibility"],"falsifier":"Leave the switch running for an hour in an environment with ordinary temperature drift and monitor per-core output power at a fixed phase setting; if the feedback loop cannot hold the four-path interference visibility near its measured 0.97–0.98 level and power leaks into non-selected cores beyond the reported −12 to −22 dB range, the balancing premise fails in practice.","tokens_in":125,"feed_emoji":"🔀","tokens_out":9008,"duration_ms":121821,"temperature":0.7,"pith_summary":"The paper is trying to establish that a core-selective switch for multicore fibers can be built from a four-path Mach–Zehnder interferometer that uses multicore fiber beam splitters to split and recombine light, with electro-optic phase modulators choosing which output core receives the signal. The claimed result is a switching time below 0.7 µs, roughly three orders of magnitude faster than existing multicore fiber switches, with average inter-core crosstalk around −16 dB and insertion loss of 7.7 dB. This matters because high-speed core switching is a missing piece for space-division-multiplexed networks; the paper further reports routing a 1 Gbps telecom signal error-free through the switch in a back-to-back test and over an actual campus-installed multicore fiber network. A sympathetic reader would take the central claim to be that MCF switching can be made fast enough for burst switching and data-center networking while remaining compatible with standard WDM.","feed_headline":"Multicore fiber switch routes signals in under 0.7 µs","feed_subtitle":"Four-path interference switch is the first core selector proven on a field multicore network; 1 Gbps ran error-free.","key_machinery":"The central object is a four-path Mach–Zehnder interferometer formed by two four-core multicore fiber beam splitters (MCF-BS), with three electro-optic phase modulators and one reference path in between. A digital FPGA control system applies phases via the modulators to select the output core and runs a perturb-and-observe algorithm to cancel environmental phase fluctuations; the unitary matrix description in Eqs. (1)–(2) treats the beam splitter as a fixed unitary $M_{\\mathrm{BS}}^{4\\times4}$ and the phase shifts as a diagonal matrix $M_\\theta$, so output core selection reduces to choosing relative phases $\\theta_i$.","core_discovery":"The central discovery is that a four-core multicore fiber beam splitter can serve as the splitting and recombining element of a fast optical switch: the first MCF beam splitter divides the input into four paths, three of which contain GHz-bandwidth lithium-niobate phase modulators, and the second recombines them so that interference directs all power to the chosen core. With an FPGA feedback loop applying perturb-and-observe phase corrections, the interferometer remains stabilized and can hop among all four output cores every 2.5 µs with a rise time under 0.7 µs. The authors report 19.8 dB extinction, −16.25 dB average inter-core crosstalk (range −12 to −22 dB), and 0.97–0.98 interference visibility, and they show the device works with a commercial SFP+ transceiver at 1 Gbps in the lab and over 170 m and 1,305 m field-installed multicore links with error-free transmission. To the authors' knowledge this is the first demonstration of a high-speed MCF switch operating in real-world conditions.","pith_inferences":["The insertion-loss bottleneck is the commercial lithium-niobate modulators (about 3.3 dB); replacing them with lower-loss phase shifters could bring total loss closer to the 2.2 dB level of the passive splitter/recombiner stages.","The 0.8 MSps feedback update rate sets an upper bound on how fast environmental perturbations can be tracked; faster feedback or feed-forward phase control would be needed for outdoor or vibration-heavy deployments.","The same four-path interferometer could be repurposed for quantum information tasks, such as actively routing single photons between multicore fiber cores, since interference visibility of 0.97–0.98 is already suitable for quantum interference.","Combining this switch with few-mode multicore fibers may extend core-selective routing into spatial-mode-selective routing, going from one dimension of SDM switching to two."],"forward_implications":["MCF core selection can be carried out at burst-switching timescales, making multicore fiber viable for data-center and high-capacity metro networks that require sub-10 µs reconfiguration.","Because the switch operates on all wavelengths together and maintains >0.99 visibility over roughly 1540–1560 nm, it can be inserted into WDM systems without per-wavelength switching.","Error-free 1 Gbps routing over installed campus MCF links with crosstalk of −18 to −22 dB indicates the switch does not add a penalty beyond expected attenuation.","The architecture uses standard fiber components and an FPGA control loop, so it can be packaged in a standard 19-inch rack shelf compatible with existing telecom infrastructure."],"supporting_citations":[{"why":"Supplies the multicore fiber beam splitter implementing the four-path splitter/recombiner.","marker":"17"},{"why":"Supplies the commercial lithium-niobate phase modulators whose GHz bandwidth sets the switching speed.","marker":"18"},{"why":"Supplies the perturb-and-observe algorithm the FPGA uses to stabilize phases.","marker":"21"},{"why":"Defines the sub-10 µs switching-speed target for data-center and datacom networks.","marker":"16"},{"why":"Prior long-period-grating MCF switch, the millisecond-speed baseline the new switch exceeds.","marker":"12"},{"why":"Beam-steering MCF switch, an alternative approach with lower loss and crosstalk but millisecond speeds.","marker":"14"},{"why":"Mechanical rotating-core MCF switch, providing the comparison point for loss, crosstalk, and speed.","marker":"15"}],"fun_headline_variants":["Core-selective switch hits sub-microsecond speeds","First field-tested multicore fiber switch in under 0.7 µs","Fastest core switch: 3 orders faster than commercial","Multicore fiber switch: 0.7 µs, field-proven","High-speed core selector for multicore networks"],"cache_read_input_tokens":11264,"weakest_assumption_plain":"The four interferometer paths must stay balanced (equal optical path length, equal loss, equal polarization response) so the second multicore beam splitter combines them coherently; if drift outruns the feedback loop, interference visibility, crosstalk, and core selection all degrade.","fun_headline_variants_meta":{"raw":{"variants":["Core-selective switch hits sub-microsecond speeds","First field-tested multicore fiber switch in under 0.7 µs","Fastest core switch: 3 orders faster than commercial","Multicore fiber switch: 0.7 µs, field-proven","High-speed core selector for multicore networks"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000265,"raw_usage":{"total_tokens":1632,"prompt_tokens":997,"completion_tokens":635,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":613,"completion_tokens_details":{"reasoning_tokens":549}},"tokens_in":613,"tokens_out":635,"duration_ms":5344,"temperature":1.0,"reasoning_tokens":549,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T11:53:54.117313+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Leave the switch running for an hour in an environment with ordinary temperature drift and monitor per-core output power at a fixed phase setting; if the feedback loop cannot hold the four-path interference visibility near its measured 0.97–0.98 level and power leaks into non-selected cores beyond the reported −12 to −22 dB range, the balancing premise fails in practice.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the multicore fiber beam splitter implementing the four-path splitter/recombiner."},{"cited_title":"LN65S-FC - Lithium Niobate Modulator","cited_arxiv_id":null,"evidence_quote":"Supplies the commercial lithium-niobate phase modulators whose GHz bandwidth sets the switching speed."},{"cited_title":"T., Jusoh A","cited_arxiv_id":null,"evidence_quote":"Supplies the perturb-and-observe algorithm the FPGA uses to stabilize phases."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Prior long-period-grating MCF switch, the millisecond-speed baseline the new switch exceeds."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Beam-steering MCF switch, an alternative approach with lower loss and crosstalk but millisecond speeds."}],"review_version":1}