WorksPhysics and Astronomy
Lense-Thirring induced repeating blast trains in continuous gravitational waves
Abstract
We characterise gravitational lensing induced by the Lense-Thirring effect in a continuous gravitational-wave source orbiting a spinning supermassive black hole. This scenario is very likely to generate lensing events, as long as the source's orbital inclination is not aligned with the black hole's spin axis and the Lense-Thirring timescale is shorter than the observational window. The signal appears as short-duration () gravitational-wave "blasts" that repeat after every orbit () within a limited time window (). The blasts' signal strength depends on the system's geometry and not on the black hole mass. The signal will go silent and renew twice in a full Lense-Thirring precession period.
Introduction
The Lense–Thirring effect, also known as inertial frame dragging, makes the plane of an orbit precess about the spin axis of the central body [15, 23, 7, 8]. Near Sgr A*, the Lense–Thirring precession of stars and pulsars on orbits closer than the known S-stars has been proposed as a test of the Kerr metric [14], through stellar astrometry or radio pulsar timing [37, 22, 28]. Because of the Lense–Thirring effect, if an object follows an orbital motion around a massive body whose spin axis is misaligned with the orbit by an angle , the orbital plane will precess, while keeping its inclination fixed (see Figure 1). Searches for gravitational lensing of gravitational waves have so far focused on compact binary coalescences [12, 18, 20, 21] observed by Advanced LIGO [17], Advanced Virgo [1] and KAGRA [2]. The Galactic Centre is expected to host many neutron stars, and directed searches have looked for their continuous gravitational waves in Advanced LIGO and Advanced Virgo data [27, 19]. A neutron star behind Sgr A* would be strongly lensed into two interfering images, a signal that third-generation detectors may observe [5] and characterise even beyond the Einstein radius [30], while lenses within the Galaxy can also magnify continuous waves from Galactic sources (S. Suyamprakasam et al. 2025). Y.-Z. Li et al. (2026) showed that Lense–Thirring precession modulates the repeated lensing of continuous-wave sources orbiting supermassive black holes in AGN discs, where the precession period is much longer than the observation and slowly changes successive lensing events. Here we consider the opposite regime around Sgr A*, where the precession period is shorter than the observation. We show that if the orbiting body emits continuous gravitational waves, the Lense–Thirring effect will orient the system in such a way that the gravitational waves are lensed by the central massive body in a periodic repeating pattern, causing a ‘train’ of amplified gravitational-wave ‘blasts’ that switches on and off twice per Lense–Thirring period.
Figure 1. Illustration of how the Lense–Thirring effect produces lensing. (a) The orbital plane of a continuous gravitational-wave source inclined at precesses around a black hole’s spin axis (-axis) due to the Lense–Thirring effect until it is (near) edge-on with respect to the observer, when it is lensed once per orbit. The longitude of the ascending node (filled red circle) grows as , and the red arc shows how far the node has moved since . The filled and open red dots mark the ascending and descending nodes, respectively. (b) pictures the observer’s sky plane, and also defines the dimensionless image position used in the text. The dashed red circle () is the “Einstein radius”, which, approximately, marks where lensing occurs. (f) shows the time evolution of . (c), (d) and (e) give the orbital plane at given times, while the corresponding (g), (h) and (i) show what the observer would be seeing at these specific times. These are similar to (a) and (b). At (e) and (i), the orbit is edge-on and passes through the Einstein ring, producing a lensing blast once per orbit. These blasts repeat until the Lense–Thirring effect sweeps the orbit out of alignment on a timescale (b). Four timescales apply here: the Lense–Thirring timescale , the orbital timescale , the time for the source object to cross the Einstein radius once and produce a blast, , and the time for Lense–Thirring precession to sweep the orbit across the Einstein radius, . See also Figure 2.
Lense–Thirring precession near Sgr A*
In the weak-field limit, a point-like mass follows a quasi-circular orbit of radius around a black hole of mass with an angular velocity . Because of the Lense–Thirring effect, the line of nodes precesses around the central black hole’s spin axis with dimensionless spin at a nodal frequency . In the weak-field limit, these angular frequencies can be expressed as [15, 4 C. M. Will]
2008)
where we use geometrized units (). Without spin, the orbital plane is fixed: the continuous gravitational wave source either never passes behind the black hole, or does so on every orbit in the same way. With spin, the line of nodes precesses about the spin axis, and its longitude grows as (Figure 1). At any instant, the source still follows a quasi-circular path in a fixed plane at the current ; we call this plane the instantaneous orbit. Once every orbit, the source is behind the black hole, producing strong lensing (a “blast”).
Will lensing occur?
Whether Lense–Thirring induced lensing occurs depends on the timescale versus observation time. For a source orbiting gravitational radii () around a rapidly spinning black hole () with the mass of Sgr A* () ([11]), the orbital and Lense–Thirring periods are
so the orbital plane precesses once in about 560 orbits. To see the precession within an observing time , we need , which limits the orbital radius in gravitational radii to
For , our example covers about two Lense–Thirring periods, so the instantaneous orbit is perfectly edge-on to the line of sight about four times. We call each such instant the edge-on time. This does not mean that the source object is then perfectly behind the black hole: a dead-centre passage requires the edge-on time to coincide with a far-side node crossing, so the Lense–Thirring and orbital timescales must align. Under these conditions, the optical depth is
Indeed, in this parameter space, the Lense–Thirring effect practically guarantees that strong lensing occurs whenever the observer lies within of the black hole equatorial plane, the band of directions swept by the orbital plane (Appendix § 14).1 Precession similarly raises the lensing probability for sources in AGN discs (Y.-Z. Li et al. (2026)).
A gravitational-wave blast from an Einstein crossing
When the orbiting source object enters the Einstein radius of the lensing black hole, the waves are lensed strongly, producing a short “blast” of gravitational waves on a timescale . Given the large distances between the objects, we treat the lensing black hole as a point lens in the thin-lens approximation (P. Schneider et al. [31]), and the orbiting body producing continuous gravitational waves as a point source. The Einstein radius is (A. Einstein [9]; P. Schneider et al. [31])
where , and are the angular diameter distances from the observer to the lens and the source, and from the lens to the source, with for Sgr A* (GRAVITY Collaboration et al. [11]). Approximating the Einstein radius as a constant near the lensing region, such that the distance from the lens to the source object is approximately the orbital radius, ,
When the binary enters this radius (), the gravitational-wave amplitude is greatly amplified (by a factor of , where are the magnifications of the two images P. Schneider et al. [31]), producing a short-duration blast of gravitational waves of duration
where is the Einstein crossing time (B. Paczyński [25]).2 As a rough scaling, for Sgr A*,
The source spends a duration
inside the high-magnification region covering around 50% of the peak, .3 These relations estimate the duration of the fast gravitational-wave blasts near the caustic; we use them below for the strongest blast in a train.
Figure 2. Time scales of the blasts for the system of Figure 1, at . Top: the orbit as seen by the observer at , , and ; dark: behind the black hole; red circle: largest Einstein radius. (a) The root-mean-square amplification at the closest approach of each orbit, over one Lense–Thirring period, ; the two peaks are the two blast trains. (b) The first train: one line per orbit; shaded: , which lasts (72 blasts). Inset: five blasts around the best orbit, apart. (c) The best blast (): the root-mean-square (red) and the range of over the fringes, (blue); shaded: , . (d) Its path in the lens plane, in units of the local Einstein radius. (e) within 5 s of closest approach, showing the interference of the two images (blue) and the root-mean-square (red). Grey boxes and dashed lines show each zoom; the boxes in (b) and (c) are as wide as the next panel and so appear as lines.
From a single blast to a train of blasts
After the orbit crosses the Einstein radius, it resumes its motion until the next orbit and it crosses the Einstein radius again, producing another blast. This train of blast signals repeats over several cycles. A similar orbit-to-orbit modulation appears for sources in AGN discs (Y.-Z. Li et al. (2026)), although there an observation covers only a small part of a precession period. Each orbit crosses the Einstein ring a little further along (Figure 2), closer to a perfectly edge-on scenario, though unfortunately it is unlikely that the source ever passes dead-centre behind the black hole. The orbital node advances by a Lense–Thirring advance in the azimuthal direction per orbit. Converting this into the angle between the line of sight and the point where the orbiting gravitational-wave source passes closest to the dead-centre behind the black hole,4 we have the closest encounter at each far-side node crossing near the edge-on time,
where the nodal integer , choosing the time origin so that node crossings fall on integer multiples of the orbital time, while is the time at which
the “instantaneous orbit” is perfectly edge-on to the line of sight.5 Successive crossings happen once every orbit, so the closest encounter shifts by from one blast to the next.6 A single train of blasts contains the orbits inside the Einstein radius (), so the total number of blast cycles and the duration of one train is given by7
As a rough scaling relation,
The system can therefore be described as follows – On a single Lense–Thirring timescale, the orbiting source object will produce two trains of gravitational-wave blasts lasting , with each train containing blasts separated by orbital time and each lasting (Figure 2). The closest the source object gets to the black hole during a transit in dimensionless source position , which can be used to estimate the maximal magnification produced by the strongest blast in the train.
Fast gravitational-wave burst
At every orbit during a train of blasts, the source object sweeps past the dead-centre line of sight behind the black hole with a known nearest source position (Equation (12)). The loudest blast comes from the far-side node crossing closest to the edge-on time, offset from it by orbits, so its nearest source position is
which gives a peak duration of the loudest blast (enclosing 50% of the peak magnification) of
The very peak of the gravitational-wave blast depends sensitively on the orbit in the Lense–Thirring cycle. However, the duration of the very peak of the magnification / closest encounter is very brief, lasting only an order minute for typical parameters considered here. The timescale is, interestingly, independent of the orbital radius .
Lense–Thirring induced blast amplification
In geometric optics limit,8 the black hole splits the source object signal into two gravitational-wave “images” that in the continuous-wave limit interfere, such that the original gravitational-wave polarisations are convolved by an amplification factor ([34])
where the two images are time-shifted and Doppler-modulated with a frequency difference of ,9 such that the two images beat at a frequency of . During a blast, the two images interfere, so the magnification due to the amplification factor oscillates between the two different magnifications and (Figure 2). One can estimate roughly the average magnification at a given source position by taking the mean-square of the amplification factor (assuming high-magnification limit ):
The oscillation is slow at closest approach, where the time delay between the images is stationary, and reaches about 40 Hz away from it (Figure 2).
Successive lensing during a train of blasts
Successive orbits cross the lens plane on parallel lines spaced by . At the closest approach (), the mean-square of the amplification factor reaches its maximum
At typical frequencies (), and closest encounter of , the blast is still far below the diffraction scale, so diffraction effects are not relevant (Appendix § 13). For an observer in the black hole equatorial plane, the time spent per unit area of the lens plane is approximately uniform. For , the fraction of time spent at magnification (from the exact inversion of Equation (18))
For example, in the typical scenario that we have been considering,10 the fraction of time the source object spends inside the Einstein radius is around 1.3%. Integrating the exact magnification over the lens plane gives the time-averaged power gain,
which is 2.5% in our example. Interestingly, the fractional time spent in a blast (Equation (21)) and the time-averaged power gain (Equation (22)) do not depend on the black hole , spin or the gravitational-wave frequency . Therefore, given that the Lense–Thirring timescale is shorter than the observation time, the average magnification boost depends only on the orbital radius and inclination.
Does lensing boost detectability?
We have shown that a neutron star on an inclined orbit close to a spinning black hole produces fast gravitational-wave burst repeaters, with some of those bursts reaching magnifications up to 100 or even more (depending on the orbital and Lense–Thirring timescales), but the effect is mainly cancelled out by the short duration of each burst. Suppose that the source object is a neutron star with equatorial ellipticity and moment of inertia , with equal to twice its spin frequency, at distance . In that case, its strain amplitude
To a reasonable degree of accuracy, the amplitude and the inclinations are constant over the observation time . As for other lensed continuous-wave sources near the Galactic Centre (S. Basak et al. [5]), we consider a third-generation detector, the Einstein Telescope (M. Branchesi et al. [6]). Averaged over the star’s spin orientation and the Earth’s rotation, in the absence of lensing, the semi-coherent signal-to-noise ratio of the gravitational-wave emission recorded in the Einstein Telescope is
where is the noise power spectral density of one of the three interferometers and is the coherent integration time of each segment, with segments in total. Such semi-coherent searches analyse each segment coherently and add the segment powers without tracking the phase between segments, trading sensitivity for computational cost (R. Tenorio et al. [35]; K. Riles [29]; K. Wette [36]). Once we apply lensing, the semi-coherent signal magnitude changes such that
where the timing model is assumed to include the Doppler modulation between the two images. Therefore, a rough scaling for the lensed signal-to-noise ratio
That is, the lensing introduces a small correction to the integrated signal-to-noise ratio.11 In fact, the detection of these blasts may be more difficult than the detection of the actual continuous-wave signal due to the nature of the gravitational-wave analysis. Therefore, the most likely scenario is that the continuous-wave signal is detected first, and then the blasts are identified as lensed. These lensing effects do not make a typical continuous gravitational-wave source more detectable.
Example system: Sgr A* with a neutron star
We take the system shown in Figure 1 at distance , with gravitational-wave frequency , ellipticity , moment of inertia and observation time . This ellipticity is at the low end of the theoretical crustal maximum of about – ([10]; [24]). The observation contains about four trains and 300 blasts. Equation (23) to Equation (26) give , and , well above a threshold of . This is the best case, because it requires a phase-connected orbital solution across .
The orbital speed makes this an unusual continuous-wave source. The Doppler shift reaches over each orbit and drifts at near conjunction, which limits the coherence time to about 15 s without an orbital model, since the signal must stay within one frequency bin of width (K. Riles [29]). Directed searches of the Galactic Centre (O. J. Piccinni et al. [27]; LIGO Scientific Collaboration et al. 2022) cover frequency derivatives many orders of magnitude smaller than this, so a source on such an orbit would need a dedicated search that follows the orbit. At that coherence time, a semi-coherent stack of the blasts alone, which analyses short segments coherently and sums their power without tracking the phase between them, stays below 1 standard deviation. If an orbital model extends the coherence over a full blast of about 2.3 h, the same 北京快સ્માત stack reaches about 19 standard deviations, against an indicative threshold of 10 (Appendix § 15). No single blast is detectable, as even the loudest stays below at any time. Stacking all of with the same coherence time gives about 75 standard deviations, because the blasts cover only about 1% of the observation. The source would therefore be found first as an ordinary continuous-wave signal, and the blasts would then identify it as lensed. Sgr A is therefore within range for this signal, provided the orbit is known well enough to stay coherent across a blast.
Expected sources
Gravitational-wave inspiral from takes about 4 Myr (P. C. Peters [26]). With a neutron-star inspiral rate of order per galaxy ([13, 3]), we expect about 0.02 to 0.2 neutron stars this close to Sgr A, before requiring a fast spin or a nearly circular orbit.
Conclusions
If such a Lense–Thirring system exists around Sgr A, its signal may be observable. We focused on an observer in the black hole equatorial plane, for which any inclined orbit with becomes edge-on twice per Lense–Thirring period, giving two blast trains spaced by . Other observer orientations change the signal rhythm, giving unevenly spaced or merged trains (Appendix § 14). We have focused on the geometric optics case. Blasts that approach the diffraction limit are possible, but at 1 kHz they occur in only about of trains (Appendix § 13).
We also treated the source as a neutron star seen only in gravitational waves. If it were also a radio pulsar, pulse timing could supply the orbital solution ([22, 28]), making the fully coherent analysis realistic. However, scattering towards Sgr A makes such timing difficult. The same alignments would also lens the radio pulses, giving an independent check of the blast times. A detection would also probe the black hole itself, because the orbital period, the spacing of the trains and their length fix , , and , giving a gravitational-wave measurement of the spin of Sgr A.
Acknowledgments
We thank Aditya Sharma for reviewing the manuscript. [Funding for each author.] This material is based upon work supported by NSF’s LIGO Laboratory which is a major facility fully funded by the National Science Foundation. [Keep only if LIGO Laboratory computing was used:] The authors are grateful for computational resources provided by the LIGO Laboratory and supported by National Science Foundation Grants PHY-0757058 and PHY-0823459. This paper carries LIGO document number LIGO-P2600604.
Wave optics and the diffraction limit
The diffraction limit
The geometric-optics amplification in Equation (18) diverges as , but the signal’s wave nature keeps it finite. The relevant parameter is the dimensionless frequency . For and , the amplification of a point lens is [34]
where is a Bessel function. At this gives the diffraction limit,
For , Equation (27) oscillates between 0 and with a mean of , which reproduces the fringes and root-mean-square amplification of geometric optics at small . Wave effects therefore matter only for , which is at 1 kHz, far below the of the blasts in the main text.
Wave-optics blasts
Since grows linearly near the edge-on time (Equation (12)), the condition holds only in a short diffraction window,
As this is much shorter than one orbit, a wave-optics blast needs a special alignment. The source must cross the node on the far side of the black hole within of the edge-on time,
with and as in Equation (12). In our system no passage falls inside the window (Figure 3), as the closest one comes 0.6 d before , with . The number of orbits in the window is
Figure 3. The source object near the first edge-on time of Figure 2, at (). Blue: along the orbit, with dips apart. Grey: The of a passage at time . Dashed: and . Green: the diffraction window (, too narrow to resolve). The deepest dip, 0.6 day before the edge-on time, has .
A wave-optics blast therefore occurs in about of trains, or with a probability of about in 5 yr for this system. Every train reaches the wave-optics regime only if , that is, . For Sgr A* at this requires , far below the Einstein Telescope band.
A wave-optics blast would also be brief. The source crosses in at 1 kHz, about 16 gravitational-wave cycles. Using Equation (21), it spends a fraction of the time in this regime, so wave optics does not affect the results of Section 9.
Effect of black hole spin
Equation (26) assumes a point caustic. The spin of a Kerr black hole unfolds this point into a small astroid (M. Sereno & F. De Luca [32]), with a half-width in units of of about
for an observer in the black hole equatorial plane. This is 40 times larger than at 1 kHz, so the astroid is resolved at this frequency. A ring-diffraction estimate then gives a peak of about , so Equation (28) is an upper limit. The point-lens value applies only below about 25 Hz.
Observer orientation
The main text assumes the observer lies in the black hole equatorial plane. Here we let the observer lie at a latitude above that plane. The angle between the line of sight and the orbital plane then follows
which reduces to for . A far-side passage at angle has and , so Equation (7) gives .
Train pattern
For , the orbit is edge-on at , with
Successive trains are therefore separated alternately by and for , which gives the signal a short-long rhythm instead of the even spacing of . The angle also changes more slowly near each edge-on time, so each train lengthens,
and the number of blasts grows by the same factor. This holds to within about 10% for trains that remain separate. Across the short gap, reaches only . If this is below about , stays below 1 throughout the gap and the two trains merge into one long train per Lense–Thirring period. For , the orbit is never edge-on. The closest passage has , giving a single train per period with a floor on , and the blasts vanish once .
For the system of Figure 1, corresponds to . At , trains of 78 blasts are separated alternately by 0.94 yr and 1.40 yr. The trains merge above , so at there is one train of about 210 blasts per period, containing two edge-on times 0.47 yr apart. At , one train of about 130 blasts has , which limits the root-mean-square amplification to about 1.6. Above there are no blasts. The ratio rises from 0.12 at to 0.16 at because the trains last longer, then falls to 0.09 at . The equatorial observer of the main text is therefore a conservative choice for .
Likelihood of alignment
For a static orbit, reaching requires the line of sight to lie within of the orbital plane, which holds for a fraction of observers. This is only 0.3% for at . Precession sweeps the orbital plane through every direction with , a fraction of the sky, so each train reaches the same alignment for half of all observers at . Averaging over isotropic orbits gives , or 79%. The aligned case is therefore not fine-tuned.
Detection statistics
Coherent signal-to-noise ratio
For one L-shaped interferometer, averaging over sky position and source orientation gives . The triangular design has three interferometers with opening angles, so the response scales by , giving the factor in Equation (24). For a source at polar angle from the detector normal, the triangle response is proportional to . Averaged over a sidereal day towards Sgr A, this gives within 4% of the sky-averaged value at both candidate sites, so Equation (24) holds. Within a day, varies by a factor of 0.56 to 1.55, which matters for a single blast but averages out in a stack.
Semi-coherent stack
A stack divides the data into segments of length , analyses each coherently, and sums the power without tracking the phase between segments ([29, 35, 36]). Each segment contributes a power , and in noise the summed statistic has a variance of , so the significance is
For a stack of the blasts alone, and , where is the mean of over . Compared with a stack over all of at the same coherence time, this gives
for the system of Figure 1. The summed power of the example signal is about 830 for any , so sets the significance. With h, and . With s, and . A stack over all of with h gives .
The threshold is indicative and does not include trial factors from searching over the orbit. The stack also assumes a template that follows both images within each segment.
Approximations
Weak-field lensing. —At the Einstein radius the deflection angle is rad for , so the thin-lens, weak-field treatment is only approximate. The second-order deflection increases by a fraction , or about 7%, and by the same amount. This shifts the timescales but does not change the train structure.
Black hole spin in the lens. —In the weak-deflection limit, the spin displaces the lens centre by ([32]), which is in units of . This moves the deepest blast of a train by about one orbit. We discuss the astroid effect in Appendix § 13.
Quadrupole precession. —The spin-induced quadrupole changes by a fraction . We neglect it here, but a measurement of from would need to include it.
Orbital shape and decay. —We take the orbit to be circular. An eccentric orbit would undergo apsidal precession of per orbit, or with a period of 51 d, modulating the orbital radius at conjunction and thus the blast duration across a train. Gravitational-wave emission shrinks by only a fraction of about in 5 yr.
Spin of the star. —Geodetic precession rotates the star’s spin by per orbit, or with a period of 103 d, which is close to . Unless the spin is aligned with the orbit, this modulates the signal amplitude across a train. Gravitational-wave spin-down at and 1 kHz shifts by only about 3 mHz in 5 yr, so we take as constant.
References
- [1]Acernese, F., et al. 2015, Class. Quant. Grav., 32, 024001, doi: 10.1088/0264-9381/32/2/024001DOI
- [2]Akutsu, T., et al. 2021, PTEP, 2021, 05A101, doi: 10.1093/ptep/ptaa125DOI
- [3]Amaro-Seoane, P., & Preto, M. 2011, Class. Quant. Grav., 28, 094017, doi: 10.1088/0264-9381/28/9/094017DOI
- [4]Bardeen, J. M., Press, W. H., & Teukolsky, S. A. 1972, ApJ, 178, 347
- [5]Basak, S., Sharma, A. K., Kapadia, S. J., & Ajith, P. 2023, Astrophys. J. Lett., 942, L31, doi: 10.3847/2041-8213/acab50DOI
- [6]Branchesi, M., et al. 2023, JCAP, 07, 068, doi: 10.1088/1475-7516/2023/07/068DOI
- [7]Ciufolini, I., & Pavlis, E. C. 2004, Nature, 431, 958
- [8]Ciufolini, I., et al. 2023, Eur. Phys. J. Plus, 138, 1054, doi: 10.1140/epjplus/s13360-023-04696-6
- [9]Einstein, A. 1936, Science, 84, 506
- [10]Gittins, F., Andersson, N., & Jones, D. I. 2021, Mon. Not. Roy. Astron. Soc., 500, 5570, doi: 10.1093/mnras/staa3635DOI
- [11]GRAVITY Collaboration, Abuter, R., et al. 2022, A&A, 657, L12, doi: 10.1051/0004-6361/202142465
- [12]Hannuksela, O. A., Haris, K., Ng, K. K. Y., et al. 2019, Astrophys. J. Lett., 874, L2, doi: 10.3847/2041-8213/ab0c0fDOI
- [13]Hopman, C., & Alexander, T. 2006, Astrophys. J. Lett., 645, L133, doi: 10.1086/506273DOI
- [14]Kerr, R. P. 1963, Physical Review Letters, 11, 237, doi: 10.1103/PhysRevLett.11.237DOI
- [15]Lense, J., & Thirring, H. 1918, Physikalische Zeitschrift, 19, 156
- [17]LIGO Scientific Collaboration, Aasi, J., et al. 2015, Class. Quant. Grav., 32, 074001, doi: 10.1088/0264-9381/32/7/074001DOI
- [18]LIGO Scientific Collaboration, & Virgo Collaboration. 2021, Astrophys. J., 923, 14, doi: 10.3847/1538-4357/ac23db
- [19]LIGO Scientific Collaboration, Virgo Collaboration, & KAGRA Collaboration. 2022, Phys. Rev. D, 106, 042003, doi: 10.1103/PhysRevD.106.042003
- [20]LIGO Scientific Collaboration, Virgo Collaboration, & KAGRA Collaboration. 2024, Astrophys. J., 970, 191, doi: 10.3847/1538-4357/ad3e83
- [21]LIGO Scientific Collaboration, Virgo Collaboration, & KAGRA Collaboration. 2026, Astrophys. J., 1011, 240, doi: 10.3847/1538-4357/ae9744
- [22]Liu, K., Wex, N., Kramer, M., Cordes, J. M., & Lazio, T. J. W. 2012, ApJ, 747, 1, doi: 10.1088/0004-637X/747/1/1DOI
- [23]Mashhoon, B., Hehl, F. W., & Theiss, D. S. 1984, General Relativity and Gravitation, 16, 711, doi: 10.1007/BF00762913
- [24]Morales, J. A., & Horowitz, C. J. 2022, Mon. Not. Roy. Astron. Soc., 517, 5610, doi: 10.1093/mnras/stac3058DOI
- [25]Paczyński, B. 1986, ApJ, 304, 1
- [26]Peters, P. C. 1964, Phys. Rev., 136, B1224, doi: 10.1103/PhysRev.136.B1224DOI
- [27]Piccinni, O. J., Astone, P., D’Antonio, S., et al. 2020, Phys. Rev. D, 101, 082004, doi: 10.1103/PhysRevD.101.082004DOI
- [28]Psaltis, D., Wex, N., & Kramer, M. 2016, ApJ, 818, 121. https://arxiv.org/abs/1510.000394
- [29]Riles, K. 2023, Living Rev. Rel., 26, 3, doi: 10.1007/s41114-023-00044-3DOI
- [30]Savastano, S., Vernizzi, F., & Zumalacárregui, M. 2024, Phys. Rev. D, 109, 024064, doi: 10.1103/PhysRevD.109.024064DOI
- [31]Schneider, P., Ehlers, J., & Falco, E. E. 1992, Gravitational Lenses (Berlin: Springer-Verlag)
- [32]Sereno, M., & De Luca, F. 2008, Phys. Rev. D, 78, 023008, doi: 10.1103/PhysRevD.78.023008DOI
- [34]Takahashi, R., & Nakamura, T. 2003, ApJ, 595, 1039
- [35]Tenorio, R., Keitel, D., & Sintes, A. M. 2021, Universe, 7, 474, doi: 10.3390/universe7120474DOI
- [36]Wette, K. 2023, Astropart. Phys., 153, 102880, doi: 10.1016/j.astropartphys.2023.102880DOI
- [37]Will, C. M. 2008, ApJL, 674, L25. https://arxiv.org/abs/0711.1677


