Pith. sign in

REVIEW 2 major objections 6 minor 43 references

Clock-noise subtraction inside geometric TDI restores sensitivity and tightens source-parameter constraints for space-based gravitational-wave detectors.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

T0 review · grok-4.5

2026-07-13 03:47 UTC pith:62XX6QDC

load-bearing objection Clean geometric generalization of sideband clock subtraction that works for delay/advance TDI and demonstrably restores PE for a second-generation U-type channel. the 2 major comments →

arxiv 2607.09335 v1 pith:62XX6QDC submitted 2026-07-10 gr-qc

Clock-noise subtraction in geometric time-delay interferometry for space-based gravitational-wave parameter estimation

Rui Luo , Pan-Pan Wang , Zi-Jiang Yang , Wei-Liang Qian , Cheng-Gang Shao This is my paper
classification gr-qc PACS 04.80.Nn95.55.Ym07.60.Ly
keywords gravitational wavesgravitational-wave data analysisspace-based interferometerstime-delay interferometryclock noisegeometric TDIparameter estimation
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

Space-based gravitational-wave detectors such as LISA-type missions form time-delay interferometry (TDI) combinations to cancel laser phase noise when the arms are unequal and changing. After that cancellation, clock jitter from the onboard oscillators can still sit above the secondary-noise floor and distort the noise weighting used for detection and parameter estimation. This paper builds an explicit clock-noise subtraction scheme entirely inside the geometric-TDI framework. It introduces four generalized clock-noise observables that cover every combination of delay and time-advance operators, so that the residual clock contribution becomes algebraically parallel to the familiar laser residual and can be subtracted for any two-path geometric TDI combination. Time-domain simulations with LISA-like orbits and noise show that, for a representative modified second-generation U-type observable, the residual clock power falls below the injected monochromatic signal, the sensitivity curve is restored, and Fisher and MCMC uncertainties on amplitude, frequency and phase shrink. The practical claim is that clock calibration is not optional polish but a necessary step for precision data analysis of future millihertz detectors.

Core claim

Generalized clock-noise observables for the four space-time link structures that arise when both delay and time-advance operators are allowed make the residual clock contribution of any two-path geometric TDI combination algebraically parallel to the laser residual, yielding explicit subtraction terms that suppress clock noise below secondary levels and improve parameter constraints.

What carries the argument

The four generalized clock-noise observables r_{±i,±j} (built from carrier–sideband differences and extended to negative indices for time advances) that rewrite the geometric-TDI clock residual as a double sum of time-shifted r terms, exactly mirroring the laser residual and therefore providing the subtraction formula.

Load-bearing premise

The scheme assumes that the four carrier–sideband clock observables, together with perfect knowledge of the time-varying delays, are enough to cancel all residual clock contributions down to the secondary-noise floor, without higher-order orbital corrections or imperfect sideband measurements.

What would settle it

A time-domain simulation or laboratory optical-bench test of a modified second-generation U-type TDI combination that leaves residual clock power still above the secondary-noise floor or the monochromatic-signal peak after the prescribed subtraction terms are applied would falsify the claim.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • Any two-path geometric TDI observable, including those that use time advances, can be equipped with an explicit, data-stream-based clock subtraction term without redesigning the TDI algebra.
  • Likelihood evaluations that currently ignore residual clock noise will use an incorrect noise PSD and will therefore produce overstated uncertainties or biased posteriors; applying the subtraction restores the correct weighting.
  • Clock-noise calibration becomes a required preprocessing step in the precision data-analysis pipeline for LISA, TianQin and Taiji, on the same footing as laser-noise cancellation.
  • First- and second-generation Monitor, Michelson and U-type combinations already have concrete subtraction recipes that can be coded directly from the tabulated link-by-link expressions.
  • Parameter-estimation studies of monochromatic Galactic binaries (and by extension other continuous sources) must include the post-subtraction noise model if they claim realistic Fisher or MCMC uncertainties.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same four-link taxonomy should extend, with only bookkeeping changes, to geometric TDI combinations that use unequal numbers of links on the two virtual paths.
  • Because the subtraction is written entirely in terms of measured carrier, sideband, test-mass and reference streams, it can be validated on ground with existing optical-bench TDI demonstrators before flight.
  • Once residual clock noise is routinely removed, secondary-noise models (test-mass acceleration and optical-path) become the dominant remaining uncertainty in the likelihood, so improved acceleration-noise measurements will translate more directly into tighter astrophysical posteriors.
  • The algebraic parallelism between laser and clock residuals suggests that other secondary noises that enter through similar beat-note coefficients could be treated by analogous geometric subtractions.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Referee Report

2 major / 6 minor

Summary. The paper formulates a clock-noise subtraction scheme inside the geometric-TDI framework for space-based interferometers. It introduces four generalized clock-noise observables r_{±i,±j} associated with the delay/advance link topologies that appear when both time-delay and time-advance operators are allowed, writes the clock residual in a form algebraically parallel to the laser residual (Eqs. 23 and 38–45), and supplies explicit subtraction terms for arbitrary two-path geometric TDI combinations. The construction is illustrated on Monitor-E, the modified second-generation Michelson [X]^{16}_{1}, and the modified second-generation U-type [U]^{16}_{3}. Time-domain simulations with LISA-like orbits and noise levels show that, for [U]^{16}_{3}, the residual clock PSD falls below the secondary-noise floor and the injected monochromatic signal, restoring the expected sensitivity. A follow-up Fisher/MCMC study then shows improved constraints on source amplitude, frequency and phase once the effective noise PSD in the likelihood is reduced.

Significance. If the construction holds under the stated approximations, the paper supplies a compact, algorithmic prescription for clock-noise calibration of arbitrary two-path geometric TDI observables, including those with time advances. That is a useful organizational advance relative to earlier sideband-based and algebraic treatments, and the explicit link to likelihood weighting and parameter-estimation posteriors correctly frames clock calibration as part of precision data analysis rather than only sensitivity-curve presentation. The algebraic parallel between laser and clock residuals, the four exhaustive link topologies, and the matching of analytic residual PSDs to time-domain simulations are concrete strengths. The PE demonstration, though simplified, makes the data-analysis consequence of residual clock noise quantitative.

major comments (2)
  1. Sec. 6, Eqs. (84)–(85) and the accompanying MCMC setup: the parameter-estimation comparison uses ϵ_clk = 0 (ideal subtraction) rather than the residual clock PSD actually left by the algorithm in Sec. 5. The Sec. 5 simulations already show the residual below the secondary floor around the injected frequency, so the qualitative conclusion is plausible, but the reported Fisher/MCMC gains are those of perfect removal, not of the calibrated residual. The PE section should either recompute the likelihood with the post-subtraction residual PSD measured from the same time series, or clearly label the PE results as an ideal-subtraction upper bound and state how large a residual would re-broaden the posteriors.
  2. Secs. 2.2 and 3.2, Eqs. (11) and (35)–(37): the generalized clock observables are built from carrier–sideband differences with Doppler neglected, and the residual cancellation (Eqs. 40–45) assumes perfect knowledge of the time-varying delays. These are standard idealizations in the literature and are consistent with the paper’s internal derivation, but they are load-bearing for the claim that the residual is suppressed “below the secondary-noise floor” for precision PE. A short quantitative discussion (or a single additional simulation) of residual leakage from Doppler terms and from realistic delay-estimation error would make the PE conclusions more robust; without it the central claim remains conditional on those approximations.
minor comments (6)
  1. Abstract and Sec. 6: the abstract advertises both Fisher and MCMC constraints, but Sec. 6 is written almost entirely around the MCMC likelihood and one-sigma uncertainty curves (Fig. 9). Either add an explicit Fisher-matrix calculation or soften the abstract wording to match what is shown.
  2. Tables 2–5 in the manuscript text appear as headers only (content is present for Tables 6–7 and for the summary Table 1). Ensure the full per-link expressions for Monitor-E and [X]^{16}_{1} are printed so that the algebraic examples are reproducible without reverse-engineering the space-time diagrams.
  3. Eq. (83): the time-domain single-frequency likelihood is a reasonable monochromatic approximation, but it should be stated that it neglects frequency-dependent weighting across the band and any non-stationarity from the orbits; a one-sentence caveat would prevent over-reading the posterior widths.
  4. Notation: the dual use of D^{−}_{ji} for time advance and of negative subscripts on q and r is clear once explained, but a short notation table early in Sec. 2 would help readers who jump to the residual formulae.
  5. Figs. 7–8: the legend text is dense; distinguishing simulation vs theory and pre- vs post-subtraction with line styles (in addition to color) would improve readability in print.
  6. Related work: Refs. [33–35] already treat geometric/second-generation clock noise; a sentence clarifying what is new (four-topology residual parallel to laser residual + PE impact) would help place the contribution.

Circularity Check

0 steps flagged

No significant circularity: clock-noise residual and subtraction terms are algebraically derived from measured streams and parallel the laser residual by construction, then validated on independent simulations.

full rationale

The paper's load-bearing chain (Secs. 2–4) defines intermediary variables η from carrier/test-mass/reference streams (Eqs. 8–10), clock observables r_ij from carrier–sideband differences (Eq. 11), and three generalized r_±i,±j for advance operators (Eqs. 35–37). It then rewrites the geometric-TDI residual (Eq. 23) by isolating clock pieces (Eqs. 25–34), obtains the double-sum residual gTDI_q (Eq. 43) that is algebraically parallel to the laser residual, and subtracts it (Eq. 45). The four link cases (Fig. 4, Table 1) and explicit Monitor-E / [X]^{16}_{1} / [U]^{16}_{3} tables follow directly from those definitions; no parameter is fitted to data and then re-predicted. Numerical PSDs, sensitivity curves and MCMC posteriors (Secs. 5–6) use literature LISA-like ASDs and orbits solely for illustration; the improvement after subtraction is the expected consequence of lowering S_U, not a forced fit. Self-citations to prior geometric-TDI work supply the two-path framework but are not used as uniqueness theorems that close the argument. The derivation is therefore self-contained against its own inputs.

Axiom & Free-Parameter Ledger

2 free parameters · 4 axioms · 1 invented entities

The central cancellation rests on the standard geometric-TDI representation of two virtual paths, the sideband encoding of clock differences, and the algebraic identity that the sum of effective beat-note frequencies around a closed path vanishes. No new free parameters are fitted to data; simulation numbers are conventional LISA values. The only invented objects are the four generalized clock observables needed to handle time advances.

free parameters (2)
  • beat-note frequency coefficients a,b,c (and derived ã)
    Set by hand to 5–20 MHz for the numerical examples; they scale the clock residual but are not fitted to the GW signal.
  • injected monochromatic amplitude H and frequency f_GW
    Chosen for illustration (H ~ 10^{-18}, f_GW = 1 mHz, T = 2 days); used only to demonstrate residual suppression and PE improvement, not to calibrate the subtraction algorithm.
axioms (4)
  • domain assumption A valid geometric TDI observable is the interference of two virtual equal-arm optical paths whose lowest-order light-travel times cancel (Eq. 24).
    Standard geometric-TDI premise used throughout Sec. 3; taken from Vallisneri, Muratore et al.
  • domain assumption Carrier–sideband differences yield pure clock observables r_ij = D_ij q_j – q_i (Doppler neglected).
    Eq. 11 and surrounding text; inherited from Hellings/Tinto/Hartwig sideband literature.
  • domain assumption Time-advance operators can be realized from delay-only measured streams via the identity η_{–ij} = –D^{–}_{ij} η_ji.
    Eqs. 19–20; required for second-generation combinations that contain advances.
  • standard math The sum of effective beat-note frequencies around any closed geometric path vanishes (Eq. 40).
    Direct consequence of the equal-arm condition (Eq. 24) applied to the ã coefficients.
invented entities (1)
  • Four generalized clock-noise observables r_{±i,±j} covering delay/advance link topologies (a–d) independent evidence
    purpose: Make the clock residual algebraically parallel to the laser residual so that explicit subtraction terms exist for arbitrary two-path geometric TDI.
    Defined in Eqs. 11, 35–37 and Fig. 4; they are linear combinations of ordinary measured r_ij and therefore have independent experimental handles via sideband data.

pith-pipeline@v1.1.0-grok45 · 29599 in / 3004 out tokens · 34351 ms · 2026-07-13T03:47:52.870433+00:00 · methodology

0 comments
read the original abstract

Millihertz gravitational-wave observations with space-based interferometers require time-delay interferometry (TDI) observables whose residual instrumental noise is sufficiently controlled for both detection and parameter inference. Although TDI suppresses laser phase noise in unequal and time-dependent arms, clock jitter from onboard ultra-stable oscillators can remain above the secondary-noise floor and bias the effective noise weighting used in data analysis. We formulate a clock-noise subtraction scheme directly in the geometric-TDI framework. The construction introduces generalized clock-noise observables for the four space-time link structures that arise when both delay and time-advance operators are allowed. This makes the clock-noise residual algebraically parallel to the laser-noise residual and yields explicit subtraction terms for arbitrary two-path geometric TDI observables. We illustrate the method with representative first- and second-generation geometric TDI combinations, and test it with time-domain simulations using LISA-like orbits and noise levels. For a modified second-generation U-type observable, the subtraction suppresses the clock-noise residual below the signal region, restores the expected sensitivity to a monochromatic source, and improves the Fisher and Markov-chain Monte Carlo parameter constraints on the source amplitude, frequency and phase. These results show that clock-noise calibration is a necessary component of precision data analysis for future space-based gravitational-wave detectors.

Figures

Figures reproduced from arXiv: 2607.09335 by Cheng-Gang Shao, Pan-Pan Wang, Rui Luo, Wei-Liang Qian, Zi-Jiang Yang.

Figure 1
Figure 1. Figure 1: The experimental layout of the constellation consists of three spacecraft and six optical benches. SC: spacecraft; USO: ultra-stable oscillators; OB: optical bench; TM:test mass; PM: phaseme￾ters; PD: photodetector; ADC: analog-to-digital converter. The space-based GW detector consists of three spacecraft labeled SC1, SC2, and SC3, respec￾tively, in the clockwise direction depicted in [PITH_FULL_IMAGE:fig… view at source ↗
Figure 2
Figure 2. Figure 2: The space-time diagram of the eight-link Michelson combination X1 (left) and the sixteen￾link Michelson combination [X] 16 1 (right). The red-solid and blue-dashed line segments correspond to the two synthesized routes, which might contain both forward and backward propagation in time. The black squares and black dots indicate the initial and terminal grids of the two routes, and the reference time t = 0 i… view at source ↗
Figure 3
Figure 3. Figure 3: Equivalent beat-note frequencies are illustrated as vector sums. The arrowed line segments in red are the original beat-note frequencies defined in Eqs. (7), and the effective ones given in Eqs. (29) are indicated by arrowed line segments in blue. Moreover, to accommodate the “time advance” operations, we separate the terms relevant to the clock noise from Eq. (20) and find η q −ij = −D−ijη q ji = −D−ij  … view at source ↗
Figure 4
Figure 4. Figure 4: Noise propagation in the space-time diagram for four exhaustive scenarios (a-d) for a given geometric TDI solution. The laser propagations in question primarily concern the link ik−1ik, but the preceding one ik−2ik−1 is also involved. The arrowed line segments in black indicate the propagations along the time direction, and the ones in magenta show those in the opposite direction of time. the subfigure (b)… view at source ↗
Figure 5
Figure 5. Figure 5: The spacetime diagrams of the first-generation eight-link Monitor combination E1 (left) and the sixteen-link Relay combination [U]16 3 (right). The red-solid and blue-dashed line segments correspond to the two synthesized routes, which might contain both forward and backward propagation in time. The black squares and black dots indicate the initial and terminal grids of the two routes, and the reference ti… view at source ↗
Figure 6
Figure 6. Figure 6: Flowchart of the numerical simulations for the laser phase and clock-noise reduction scheme. ISI: interspacecraft interferometer (scientific carrier data stream); SBI: sideband interferometer (scientific sideband data stream); TMI: test-mass interferometer (test mass data stream); RFI: reference interfer￾ometer (reference data stream). The flowchart of the numerical simulation is depicted in [PITH_FULL_IM… view at source ↗
Figure 7
Figure 7. Figure 7: The residual noise PSD of a monochromatic GW source located at the north celestial pole (w.r.t. the detector plane) when the TDI combination [U] 16 3 is applied. The curves correspond to scenarios before and after implementing the proposed clock-noise suppression algorithm. The red and green curves represent the noise residual after applying the TDI algorithm but not the clock-noise reduction procedure. Th… view at source ↗
Figure 8
Figure 8. Figure 8: The sensitivity curve of a monochromatic GW source located at the north celestial pole (w.r.t. the detector plane) when the TDI combination [U] 16 3 is applied. The curves correspond to scenarios before and after implementing the proposed clock-noise suppression algorithm. The legend of the plot is the same as [PITH_FULL_IMAGE:figures/full_fig_p022_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Parameter-estimation uncertainty for a monochromatic source observed with the [U] 16 3 TDI combination. The dashed curves are obtained before clock-noise subtraction, while the solid curves correspond to the clock-noise-subtracted case. The panels show the one-sigma uncertainties of ln H, fGW and ϕ0, together with the corresponding SNR, as functions of the injected GW frequency. Clock-noise subtraction low… view at source ↗
Figure 10
Figure 10. Figure 10: MCMC posterior samples for the parameters (ln H, fGW, ϕ0) after clock-noise subtraction. The vertical and horizontal black lines indicate the injected values. The posterior is localized around the injection, showing that the clock-noise-subtracted [U] 16 3 observable retains the phase and amplitude information required for parameter inference. The comparison demonstrates that residual clock noise has a di… view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

43 extracted references · 20 linked inside Pith

  1. [1]

    Amaro-Seoane Pet al.2017 (Preprint1702.00786)

    Pith/arXiv arXiv 2017

  2. [2]

    Danzmann K 1997Class. Quant. Grav.141399–1404

  3. [3]

    Luo Jet al.2016Class. Quant. Grav.33035010 (Preprint1512.02076)

    Pith/arXiv arXiv

  4. [4]

    Hu W R and Wu Y L 2017Natl. Sci. Rev.4685–686

  5. [5]

    Tinto M and Armstrong J 1999Phys. Rev. D59102003

  6. [6]

    Armstrong J W, Estabrook F B and Tinto M 1999The Astrophysical Journal527814–826 URLhttps://doi.org/10.1086/308110

    doi:10.1086/308110

  7. [7]

    Tinto M and Dhurandhar S V 2021Living Rev. Rel.241

  8. [8]

    Estabrook F B, Tinto M and Armstrong J W 2000Phys. Rev. D62042002

  9. [9]

    Armstrong J W, Estabrook F B and Tinto M 2001Class. Quant. Grav.184059–4065

  10. [10]

    Larson S L, Hellings R W and Hiscock W A 2002Phys. Rev. D66062001 (Preprint gr-qc/0206081) 25

    Pith/arXiv arXiv

  11. [11]

    Dhurandhar S V, Rajesh Nayak K and Vinet J Y 2002Phys. Rev. D65102002 (Preprint gr-qc/0112059)

    Pith/arXiv arXiv

  12. [12]

    Tinto M, Shaddock D A, Sylvestre J and Armstrong J W 2003Phys. Rev. D67122003 (Preprintgr-qc/0303013)

    arXiv

  13. [13]

    Shaddock D A, Tinto M, Estabrook F B and Armstrong J 2003Phys. Rev. D68061303 (Preprintgr-qc/0307080)

    arXiv

  14. [14]

    Cornish N J and Hellings R W 2003Class. Quant. Grav.204851–4860 (Preprintgr-qc/ 0306096)

  15. [15]

    Tinto M, Estabrook F B and Armstrong J W 2004Phys. Rev. D69082001 (Preprint gr-qc/0310017)

    Pith/arXiv arXiv

  16. [16]

    Vallisneri M 2005Phys. Rev. D71022001 (Preprintgr-qc/0407102)

    arXiv

  17. [17]

    Petiteau A, Auger G, Halloin H, Jeannin O, Plagnol E, Pireaux S, Regimbau T and Vinet J Y 2008Phys. Rev. D77023002 (Preprint0802.2023)

    Pith/arXiv arXiv 2023

  18. [18]

    Dhurandhar S V, Nayak K R and Vinet J Y 2010Class. Quant. Grav.27135013 (Preprint 1001.4911)

    Pith/arXiv arXiv

  19. [19]

    thesis LeibnizUniversit¨ atHannover

    Otto M 2015Time-Delay Interferometry Simulations for the Laser Interferometer Space AntennaPh.D. thesis LeibnizUniversit¨ atHannover

  20. [20]

    Bayle J B, Lilley M, Petiteau A and Halloin H 2019Phys. Rev. D99084023 (Preprint 1811.01575)

    Pith/arXiv arXiv

  21. [21]

    Wu Z Q, Wang P P, Qian W L and Shao C G 2023Phys. Rev. D107024042 (Preprint 2210.07801)

    Pith/arXiv arXiv

  22. [22]

    Qian W L, Wang P P, Wu Z Q, Shao C G, Wang B and Yue R H 2023Phys. Rev. D108 022002 (Preprint2301.00814)

    Pith/arXiv arXiv

  23. [23]

    de Vine G, Ware B, McKenzie K, Spero R E, Klipstein W M and Shaddock D A 2010Phys. Rev. Lett.104211103 (Preprint1005.2176)

    Pith/arXiv arXiv

  24. [24]

    Vinckier Q, Tinto M, Grudinin I, Rieländer D and Yu N 2020Phys. Rev. D102062002

  25. [25]

    Vallisneri M 2005Phys. Rev. D72042003 [Erratum: Phys.Rev.D 76, 109903 (2007)] (Preprintgr-qc/0504145)

    arXiv 2007

  26. [26]

    Muratore M, Vetrugno D and Vitale S 2020Class. Quant. Grav.37185019 (Preprint 2001.11221)

    Pith/arXiv arXiv 2001

  27. [27]

    Muratore M, Vetrugno D, Vitale S and Hartwig O 2022Phys. Rev. D105023009 (Preprint 2108.02738)

    Pith/arXiv arXiv

  28. [28]

    Wang P P, Qian W L, Tan Y J, Wu H Z and Shao C G 2022Phys. Rev. D106024003 (Preprint2205.08709)

    Pith/arXiv arXiv

  29. [29]

    Hellings R W 2001Phys. Rev. D64022002

  30. [30]

    Tinto M, Estabrook F B and Armstrong J W 2002Phys. Rev. D65082003

  31. [31]

    Otto M, Heinzel G and Danzmann K 2012Class. Quant. Grav.29205003

  32. [32]

    Tinto M and Yu N 2015Phys. Rev. D92042002 (Preprint1502.06651) 26

    Pith/arXiv arXiv

  33. [33]

    Tinto M and Hartwig O 2018Phys. Rev. D98042003 (Preprint1807.02594)

    Pith/arXiv arXiv

  34. [34]

    Hartwig O and Bayle J B 2021Phys. Rev. D103123027 (Preprint2005.02430)

    Pith/arXiv arXiv

  35. [35]

    Wang P P, Tan Y J, Qian W L and Shao C G 2021Phys. Rev. D104082002

  36. [36]

    Hellings R, Giampieri G, Maleki L, Tinto M, Danzmann K, Hough J and Robertson D 1996 Optics Communications124313–320 ISSN 0030-4018

    1996

  37. [37]

    Dhurandhar S, Rajesh Nayak K and Vinet J 2002Phys. Rev. D65102002 (Preprintgr-qc/ 0112059)

  38. [38]

    Bayle J B and Hartwig O 2023Phys. Rev. D107083019 (Preprint2212.05351)

    Pith/arXiv arXiv

  39. [39]

    Babak S, Petiteau A and Hewitson M 2021 (Preprint2108.01167)

    Pith/arXiv arXiv 2021

  40. [40]

    Martens W and Joffre E 2021The Journal of the Astronautical Sciences68402–443 (Preprint2101.03040)

    Pith/arXiv arXiv

  41. [41]

    Cornish N J and Larson S L 2001Class. Quant. Grav.183473–3496 (Preprintgr-qc/ 0103075)

  42. [42]

    Thorne K S 1980Gravity: Newtonian, Post-Newtonian, Relativistic(Princeton University Press)

  43. [43]

    Wang P P, Qian W L, Wu Z Q, Chen H K, Huang W, Wu H Z, Tan Y J and Shao C G 2023Phys. Rev. D108044075 27