{"id":"9dce7529-e2e0-433a-8859-1ca815a2e7bc","arxiv_id":"2506.04910","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"A 19-ring-core fiber carrying 266 OAM modes achieves a GMI-estimated 25.24 Pb/s capacity and a record 1935.6 bit/s/Hz spectral efficiency with only 4x4 MIMO.","lead":"This paper reports a record spectral efficiency of 1935.6 bit/s/Hz in a 19-ring-core optical fiber using 266 orbital angular momentum modes, with an estimated total capacity of 25.24 Pb/s. It shows that many spatial channels can be packed into a single fiber while keeping signal processing complexity low.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"25.24 Pb/s and 1935.6 bit/s/Hz are bidirectional sums; halving gives 12.62 Pb/s and ~967.8 bit/s/Hz per direction, undermining the 'record' claim.","rationale":"The reader correctly identifies the ASE-dummy-channel extrapolation as a weakness, but that concern affects the precision of the GMI estimate, not the validity of the headline metric. The bidirectional double-counting is more fundamental: it changes the numerical value of the central claim by a factor of two and makes the 'record' incomparable to the cited literature. The paper's own conclusion discloses the bidirectional framing, but the title and abstract do not, and Fig.1 does not flag which points are bidirectional. A simple arithmetic check from the stated channel count and symbol rate proves that the aggregate capacity cannot be one-way, because the implied GMI of 6.38 bit/symbol exceeds the 4 bit/symbol limit of probabilistically shaped 16QAM. Therefore the only reasonable reading is that the capacity is summed over the two directions. On a per-direction basis the SE becomes ~967.8 bit/s/Hz, which may still be high but must be re-evaluated against unidirectional records. The appropriate verdict remains conditional: the authors should provide a per-direction breakdown, justify the bidirectional metric if they wish to retain it, and re-run the comparison in Fig.1 on a fair basis. If the per-direction value does not set a record, the 'record-breaking' claim in the title and abstract must be withdrawn. This is a more load-bearing issue than the dummy-channel approximation because it affects the core assertion of the paper, not just the uncertainty of the capacity estimate.","tokens_in":5056,"tokens_out":16559,"duration_ms":173298,"concrete_test":"Use the paper's own numbers: compute the per-channel GMI implied by 25.24 Pb/s over 329,840 channels at 12 GBaud. If it exceeds 4 bit/symbol, the capacity must be a sum over two directions. Then subtract 3 dB: halve the reported SE and capacity and compare the resulting one-way values (≈967.8 bit/s/Hz, 12.62 Pb/s) against the unidirectional comparison points plotted in Fig.1. If the one-way SE does not exceed the prior maximum, the record claim is an artifact of bidirectional counting.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The headline SE and capacity are reported without the word 'bidirectional' in the title or abstract, yet arithmetic shows they are sums over two propagation directions. With 1240 wavelengths × 266 mode channels = 329,840 channels at 12 GBaud, a one-way 25.24 Pb/s aggregate would require an average GMI of 25.24e15/(329,840×12e9) ≈ 6.38 bit/symbol per channel, exceeding the 4 bit/symbol entropy limit of PS-16QAM. Halving to 12.62 Pb/s gives 3.19 bit/symbol, a plausible value. The conclusion confirms this by stating 'bidirectional C+L bands transmission achieves a SE of 1935.6 bit/s/Hz' and an SED based on bidirectional transmission. Thus the claimed record SE is 2× the per-direction SE (~967.8 bit/s/Hz). Since the comparison points in Fig.1 are unidirectional SDM experiments (the standard in the field), the 'record-breaking' claim is inflated by a factor of two. A bidirectional system can double SE by using the same spectrum in both directions, so this metric is not comparable to the cited unidirectional records. The central claim as stated in the title and abstract is therefore not supported unless the authors explicitly restrict it to bidirectional aggregate and justify that as a meaningful record. The dummy-channel issue raised by the reader is real but secondary; even if ASE dummies perfectly reproduce a loaded WDM environment, the factor-of-two framing remains.","agreement_with_reader":"disagree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper reports an experimental SDM transmission over a 10-km 19-ring-core fiber supporting 266 OAM modes. Using 1240 WDM carriers in the S, C, and L bands, with PS-16QAM in C+L and QPSK in S, the authors measure GMI on three sliding test wavelengths and derive an aggregate GMI-estimated capacity of 25.24 Pb/s and a spectral efficiency of 1935.6 bit/s/Hz in the C+L bands, achieved with 4x4 time-domain MIMO with 35 taps. The paper claims this is a record SE/SED and the first high-SE, high-SED demonstration in a >100-mode fiber with low-complexity MIMO.","tokens_in":5266,"tokens_out":5634,"duration_ms":60313,"significance":"The result is potentially significant for SDM if validated: it suggests that dense spatial multiplexing can be achieved without large MIMO matrices when inter-core and inter-mode crosstalk are suppressed. The authors disclose that the capacity is GMI-estimated rather than decoded, and the sliding-test-channel methodology is standard in high-capacity experiments. However, the significance is substantially tempered by the bidirectional aggregation of the headline metrics, which are not comparable to the unidirectional results in Fig. 1, and by the use of ASE dummy channels for the vast majority of the WDM load.","major_comments":[{"comment":"The headline '1935.6 bit/s/Hz' and '25.24 Pb/s' are bidirectional aggregates, yet the title and abstract present them without the qualifier 'bidirectional'. The Conclusion states that 'bidirectional C+L bands transmission achieves a SE of 1935.6 bit/s/Hz.' With 329,840 channels (1240 wavelengths × 266 modes) at 12 GBaud, a one-way capacity of 25.24 Pb/s would require an average GMI of about 6.38 bit/symbol per channel, exceeding the 4 bit/symbol entropy limit of PS-16QAM; halving to 12.62 Pb/s yields ~3.19 bit/symbol, which is plausible. Thus the per-direction SE is ~967.8 bit/s/Hz, not 1935.6 bit/s/Hz. Since Fig. 1 compares against unidirectional SDM experiments from Refs. [4-11], the 'record-breaking' claim is inflated by a factor of two. The authors should either report per-direction values or explicitly restrict the claim to bidirectional aggregate capacity and benchmark against bidirectional systems.","section":"Abstract / Conclusion / Fig. 1"},{"comment":"The dummy channels used to fill the remaining 1237 of 1240 WDM slots are generated from ASE and shaped by optical processors, rather than modulated data. The paper assumes these reproduce the crosstalk and nonlinear environment of a fully loaded WDM system, but no evidence is given that ASE loading induces the same inter-modal or inter-core penalties as real modulated neighbors. Since the claimed capacity is a summation of GMI over all 1240 wavelengths, an optimistic GMI from incomplete loading would propagate directly into the headline number. The authors should justify this assumption quantitatively (e.g., by comparing GMI with and without modulated neighboring channels, or by measuring the crosstalk statistics).","section":"Experimental setup, second paragraph"},{"comment":"The GMI-based capacity is reported without any decoded bit-error-rate (pre-FEC or post-FEC) validation, error bars, or repeated measurements. For a 'record-breaking' claim, GMI estimates are standard in the literature, but the absence of any companion BER measurement for the measured test channels and no indication of the number of independent measurements make it difficult to assess the margin against the PS-16QAM entropy limit. At minimum, the authors should report the distribution of GMI across repeated acquisitions for the primary test channel and across the 266 modes.","section":"Results, Fig. 3"}],"minor_comments":[{"comment":"The sentence 'GMI-estimated capacity of 25.24 Pb/s are transmitted' is ungrammatical ('capacity' is singular) and the phrase 'are transmitted' overstates what is an estimate; consider 'is' and 'was achieved'.","section":"Abstract"},{"comment":"The paper does not state the exact wavelength range used for the C+L band SE calculation; specify the total optical bandwidth (e.g., 1530-1625 nm) so the reader can verify the SE.","section":"Results, SE calculation"},{"comment":"The aggregate capacity 25.24 Pb/s includes the S-band (491.8 Tb/s), while the headline SE (1935.6 bit/s/Hz) refers only to C+L bands; clarify the band definitions in the abstract to avoid ambiguity.","section":"Abstract / Conclusion"},{"comment":"There is a typo in the vertical axis label: 'data rata' should be 'data rate'.","section":"Fig. 3(b)"},{"comment":"It is not explained how GMI is obtained for all 1240 wavelengths from only three sliding test channels; if the curves in Fig. 3(a) are measurements at the edges of bands, clarify the interpolation or measurement procedure for intermediate channels.","section":"Experimental setup / Results"},{"comment":"The paper relies on prior work [12] for negligible inter-core crosstalk but does not provide a measured crosstalk matrix for this fiber; a brief quantitative statement would strengthen the claim that the 4x4 MIMO is sufficient.","section":"Experimental setup, inter-core crosstalk"}],"recommendation":"major_revision","confidential_remarks":"The bidirectional aggregation issue is likely the main point of dispute; the authors should be asked to make the bidirectional nature clear in the title or abstract if they wish to claim a record. Also, the novelty relative to Ref. [3], which already demonstrated low-complexity MIMO SDM over a multi-core OAM fiber, should be clarified in the introduction."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"First thing you should know: the 1935.6 bit/s/Hz and 25.24 Pb/s are bidirectional sums. The conclusion says so, but the title and abstract don't. Halve them and you get ~967.8 bit/s/Hz and ~12.62 Pb/s per direction. That arithmetic also checks out—329,840 channels at 12 GBaud can't carry 25.24 Pb/s one-way with PS-16QAM's 4 bit/symbol limit. So the 'record-breaking' framing is misleading unless the authors explicitly benchmark bidirectional aggregate SE against other bidirectional systems, which they don't. That's a load-bearing flaw in the headline claim.\n\nThe underlying experiment is solid and interesting. A 19-ring-core fiber supporting 266 OAM modes, with only 4x4 MIMO and 35-tap TDE, is a real engineering step. The GMI-based capacity is labeled as such, and the methodology follows standard practice in high-capacity SDM: sliding test channels, ASE dummy channels, per-mode-group demultiplexing. The fiber loss and mode consistency data look plausible, and the SED figure is a useful addition. This deserves credit.\n\nThe soft spots beyond the bidirectional issue: no decoded BER or FEC results, no error bars, and the dummy-channel assumption—that ASE-loaded dummy channels reproduce a fully loaded WDM environment—is asserted rather than validated. The reader's concern about dummy channels is real but secondary; even if that assumption holds, the factor-of-two framing remains. The paper does explicitly say 'GMI-estimated', so it's not presenting a demonstrated end-to-end throughput, but the 'record' language should be qualified more carefully.\n\nMy take: the paper is worth a serious referee, but not as-is. The authors need to either state bidirectional in the title/abstract, report per-direction SE, and compare like-for-like with the unidirectional records in Fig. 1, or drop the 'record-breaking' claim. I'd also ask for error bars and a more detailed justification of the dummy-channel loading. If those conditions are met, it's a publishable contribution to the SDM literature.","headline":"The headline record is a bidirectional aggregate, so the 'record-breaking' claim is inflated by 2x; the underlying fiber and MIMO demo is still worth a look.","tokens_in":5949,"tokens_out":2951,"would_cite":false,"duration_ms":32624,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"A 19-core fiber carrying 266 OAM modes reaches 1935.6 bit/s/Hz and an estimated 25.24 Pb/s using only 4x4 MIMO.","keywords":["space-division multiplexing","orbital angular momentum","19-ring-core fiber","4x4 MIMO","spectral efficiency","generalized mutual information","C+L bands","spatial efficiency density"],"falsifier":"Load all 1240 wavelength channels with real modulated signals (or a statistically representative subset) through the same 10-km 19-ring-core fiber, measure per-channel GMI, and sum the capacities; if the total falls materially below 25.24 Pb/s or the C+L spectral efficiency below 1935.6 bit/s/Hz, the ASE-filled dummy channels did not reproduce the full-load environment. Independently, launch a single strong mode in one ring core and detect in neighbouring cores and mode groups to test whether inter-core and inter-mode-group crosstalk stays below what the 4x4 MIMO can equalize.","tokens_in":4789,"feed_emoji":"🌀","tokens_out":7545,"duration_ms":79613,"temperature":0.7,"pith_summary":"The paper claims a record spectral efficiency of 1935.6 bit/s/Hz in the C+L bands and a GMI-estimated aggregate capacity of 25.24 Pb/s after 10 km of a 19-ring-core fiber that supports 266 orbital-angular-momentum mode channels. Its point is that dense spatial multiplexing does not have to be paid for with large MIMO matrices: because the fiber suppresses inter-core and inter-mode-group crosstalk, equalizing just four modes at a time with a 35-tap 4x4 MIMO is enough. A sympathetic reader would care because it suggests a path to petabit-scale single-fiber links whose DSP cost stays low. The capacity and spectral-efficiency figures are estimated from generalized mutual information with the WDM comb mostly filled by ASE dummy channels, so they are upper bounds awaiting a fully loaded, decoded demonstration.","feed_headline":"4x4 MIMO carries estimated 25.24 Pb/s over 19-core fiber","feed_subtitle":"Record 1935.6 bit/s/Hz C+L spectral efficiency comes from 35-tap equalization, not huge MIMO matrices.","key_machinery":"The central object is the 19-ring-core fiber (19-RCF), whose 19 separate ring cores each carry orbital angular momentum (OAM) modes with topological charges $|l| = 0, 1, 2, 3$, producing 7 OAM modes per core and 266 mode channels after polarization multiplexing. The mechanism carrying the argument is a crosstalk hierarchy: the ring-core geometry keeps inter-core and inter-mode-group crosstalk weak enough that only the four modes within one mode group need joint equalization, so a 4x4 MIMO with 35-tap time-domain equalization suffices. Around this sit three experimental supports: sliding test wavelengths at high OSNR, ASE-derived dummy channels that populate the rest of the WDM comb, and GMI soft-decision estimation that converts received constellations into capacity numbers.","core_discovery":"The paper reports a 10-km transmission experiment in which 266 mode channels, formed from 19 ring cores times 7 OAM modes times 2 polarizations, carry 1240 wavelength channels across the S, C, and L bands. The central claim is that the C+L bands achieve 1935.6 bit/s/Hz spectral efficiency and a spatial efficiency density of $3.94\\times10^{-2}$ bit/s/Hz/µm², with a GMI-estimated aggregate capacity of 25.24 Pb/s, while using only 4x4 MIMO with 35-tap time-domain equalization. GMI-estimated means the capacity is computed from per-channel generalized mutual information before forward-error correction, not measured as decoded error-free throughput. The authors argue this is the first experimental SDM system to combine high spectral efficiency, high spatial efficiency density, and more than 100 mode channels in one fiber while keeping MIMO complexity low.","pith_inferences":["A natural test the paper does not run is to replace the ASE dummy channels on a subset of wavelengths with real modulated channels and check whether neighboring-channel GMI changes; if it does, the 25.24 Pb/s estimate would need revision.","Because the demonstration is only 10 km long, the finding that 4x4 MIMO suffices may not survive longer links where inter-mode-group coupling accumulates; extending the same fiber to 100 km or more would test that scaling.","The SED metric rewards a small fiber cross-section, and the 250 µm cladding here suggests a trade-off with standard-cladding compatibility that the paper leaves implicit.","The wavelength uniformity claim rests on evaluating performance at 4 nm intervals plus sliding test channels; denser sampling or a fully loaded comb would show whether uniformity holds across all 1240 channels."],"forward_implications":["If the crosstalk hierarchy holds, future SDM fibers can add spatial channels without scaling MIMO size, keeping the DSP cost per bit near that of a 4x4 system.","A fully populated WDM comb with decoded forward-error correction would be the next test; GMI estimates are pre-FEC upper bounds, so realized error-free capacity would be lower after coding overhead.","The S-band contribution of about 491.8 Tb/s is limited by available amplifiers, not by the fiber, so improving S-band amplification should raise the aggregate capacity.","The same low-complexity approach could transfer to other fibers with negligible inter-core crosstalk, making SE and SED comparisons against strongly coupled multi-mode systems more direct."],"supporting_citations":[{"why":"Defines the ring-core OAM fiber platform and supplies the prior claim that inter-core crosstalk is negligible, the premise that lets only 4x4 MIMO be used.","marker":"[12]"},{"why":"Establishes the prior 1-Pb/s OAM fiber transmission baseline that the 19-core, S+C+L-band result extends.","marker":"[2]"},{"why":"Supplies the bidirectional transmission approach and the prior finding that Rayleigh backscattering is negligible, justifying the doubled capacity count and low-complexity MIMO.","marker":"[3]"},{"why":"Provides the dense SDM/WDM baseline over 6-mode 19-core fiber in the C+L band that this work compares against for SE and SED.","marker":"[5]"},{"why":"Reports 22.9 Pb/s via extreme space-wavelength multiplexing, the aggregate-capacity comparison point the paper's record SE/SED claim is set against.","marker":"[10]"},{"why":"Supplies the randomly coupled 19-core MCF with standard cladding as a comparison for spatial efficiency density.","marker":"[11]"}],"fun_headline_variants":["19-ring-core fiber breaks spectral efficiency record at 1935.6 bit/s/Hz","266-mode fiber transmission hits 25 Pb/s with low-complexity 4x4 MIMO","Record 1935.6 bit/s/Hz from 19-core fiber using simple 4x4 MIMO","GMI-estimated 25.24 Pb/s over 19-core fiber with 4x4 MIMO","Spectral efficiency record in 19-ring fiber: 1935.6 bit/s/Hz via 4x4 MIMO"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The result stands on the assumption that ASE-based dummy channels filling 1237 of the 1240 wavelengths reproduce the crosstalk and nonlinearity that real modulated channels would create, so the per-channel GMI values and the summed 25.24 Pb/s are only as good as that substitution.","fun_headline_variants_meta":{"raw":{"variants":["19-ring-core fiber breaks spectral efficiency record at 1935.6 bit/s/Hz","266-mode fiber transmission hits 25 Pb/s with low-complexity 4x4 MIMO","Record 1935.6 bit/s/Hz from 19-core fiber using simple 4x4 MIMO","GMI-estimated 25.24 Pb/s over 19-core fiber with 4x4 MIMO","Spectral efficiency record in 19-ring fiber: 1935.6 bit/s/Hz via 4x4 MIMO"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000234,"raw_usage":{"total_tokens":1433,"prompt_tokens":821,"completion_tokens":612,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":437,"completion_tokens_details":{"reasoning_tokens":491}},"tokens_in":437,"tokens_out":612,"duration_ms":6695,"temperature":1.0,"reasoning_tokens":491,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-07T10:31:36.600007+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Load all 1240 wavelength channels with real modulated signals (or a statistically representative subset) through the same 10-km 19-ring-core fiber, measure per-channel GMI, and sum the capacities; if the total falls materially below 25.24 Pb/s or the C+L spectral efficiency below 1935.6 bit/s/Hz, the ASE-filled dummy channels did not reproduce the full-load environment. Independently, launch a single strong mode in one ring core and detect in neighbouring cores and mode groups to test whether inter-core and inter-mode-group crosstalk stays below what the 4x4 MIMO can equalize.","supporting_citations":[{"cited_title":"and Lin, S., “1-Pbps orbital Fig3","cited_arxiv_id":null,"evidence_quote":"Establishes the prior 1-Pb/s OAM fiber transmission baseline that the 19-core, S+C+L-band result extends."},{"cited_title":"High spectral- efficiency, ultra-low MIMO SDM transmission over a field-deployed multi-core OAM fiber,","cited_arxiv_id":null,"evidence_quote":"Supplies the bidirectional transmission approach and the prior finding that Rayleigh backscattering is negligible, justifying the doubled capacity count and low-complexity MIMO."},{"cited_title":"10.16-Peta- B/s dense SDM/WDM transmission over 6-mode 19- core fiber across the C+ L band","cited_arxiv_id":null,"evidence_quote":"Provides the dense SDM/WDM baseline over 6-mode 19-core fiber in the C+L band that this work compares against for SE and SED."},{"cited_title":"22.9 Pb/s data-rate by extreme space-wavelength multiplexing,","cited_arxiv_id":null,"evidence_quote":"Reports 22.9 Pb/s via extreme space-wavelength multiplexing, the aggregate-capacity comparison point the paper's record SE/SED claim is set against."},{"cited_title":"Randomly coupled 19-core multi-core fiber with standard cladding diameter,","cited_arxiv_id":null,"evidence_quote":"Supplies the randomly coupled 19-core MCF with standard cladding as a comparison for spatial efficiency density."}],"review_version":1}