betaFeedback

WorksPhysics and Astronomy

Milliarcsecond-scale jet reorientations in GRS 1915+105 pointing towards a warped accretion disc

Abstract

The black hole X-ray binary GRS 1915+105 was persistently X-ray bright for over 26 years before entering a prolonged low X-ray flux state in 2018, while remaining bright at radio wavelengths. Multi-wavelength observations suggest this state results from a complex layer of absorbing material of unknown nature. Radio jet observations provide an unobstructed view of the physical changes during this phase, and recent Very Large Array and East Asian VLBI Network observations revealed substantial changes in the jet position angle (PA). We use higher-resolution VLBI observations to investigate the connection between the obscuration event and the jet behaviour close to the launch region. We observed GRS 1915+105 with the Very Long Baseline Array and European VLBI Network at four epochs during the early obscuration phase (2019-2021), sampling a diverse range of jet and accretion behaviours. We study the milliarcsecond jet morphology and PA changes and relate these to the reported X-ray behaviour. Many post-2018 PAs are significantly offset from the pre-obscuration jet axis, which has a weighted mean of 142∘±7∘142^\circ \pm 7^\circ on milliarcsecond scales. We measure a maximum offset of 57∘±8∘57^\circ \pm 8^\circ, the largest reported for GRS 1915+105 to date. Our epochs reveal a variety of jet behaviour, ranging from a stable, compact source with a large PA offset, to a bright, flaring source with a similar offset, to an extended jet with a PA closer to the pre-obscuration axis, and a non-detection. The lack of measurable PA changes within individual 8-hour epochs rules out rapid Lense-Thirring precession of the inner disc as the direct driver of the jet reorientation. Instead, the timescales involved, the system geometry, and the apparent connection to the X-ray obscuration favour a radiation-driven warped accretion disc, a configuration that could naturally explain the persistent obscuration.

Introduction

Black hole X-ray binaries (BHXBs) are binary systems containing a stellar mass black hole (BH) that accretes matter from a low-mass companion star. This accretion results in the formation of a disc around the BH that is heated through viscous processes and can emit in X-rays [100]. For decades, such systems provided the primary observational evidence for stellar-mass BHs, until the first detections of BH-BH mergers by gravitational-wave observatories [1], the discovery of an isolated stellar-mass BH through gravitational lensing [99], and the identification of optically-detected BH candidates in wide orbits [23, 24].

BHXBs are usually transient, varying on timescales much shorter than those of supermassive black holes (SMBHs) lying at the centre of most galaxies. BHXBs typically spend most of their time in quiescence, occasionally going into relatively short periods of outburst reaching luminosities of the order L∼1039L \sim10^{39} erg s−1^{-1} [e.g. 12]. These outbursts tend to show a cyclical transition between various accretion states that are accompanied by mechanical feedback in the form of winds and jets [e.g. 33, 31, 73].

Relativistic jets launched by black holes power some of the most energetic phenomena in the Universe. For example, gamma-ray bursts (GRBs) can produce jets with observed luminosities up to the order of ∼1054\sim10^{54} erg s−1^{-1} (Abdo et al. 2009; Greiner et al. 2009; Frederiks et al. 2013), jets from active galactic nuclei (AGN) can dominate the energy output of entire galaxies [6, 26], and their jet-driven outflows can extend over megaparsec scales [77]. The humanly-accessible timescales and repeated outbursts of BHXBs make them excellent laboratories for studying jet formation and evolution.

The states of an X-ray binary (XRB) outburst can span several orders of magnitude in luminosity in just weeks or months [92]. The literature divides this pattern into several states based on the spectral slope of the X-ray emission [92] and X-ray timing properties of the emission [8]. The ‘hard state’, dominated by optically thin, Comptonized emission peaking at ∼100\sim100 keV, exhibits strongly variable X-ray emission and is accompanied by steady radio jets. As the accretion flow becomes more radiatively efficient and the inner part of the disc moves inward, the source typically passes through intermediate states in which the Lorentz factor of the outflow is thought to increase and discrete, highly relativistic, optically thin ejecta are launched [33]. In the subsequent soft state, the emission is dominated by optically thick thermal radiation with characteristic photon energies of ∼1−2\sim1 - 2 keV. At this point, the jets have subsided and the main mode of mechanical feedback is in the form of disc winds.

Quasi-simultaneous X-ray and radio observations of BHXBs have established an empirical framework that links accretion flows to jet production [33]. In the hard state, this coupling manifests as a well-established radio-X-ray correlation [37, 14, 17], which extends to AGN once mass scaling is taken into account [56, 28]. Contemporaneous X-ray observations can therefore provide insights into the relationship between the accretion of inflowing matter and the launching of these jets.

Attempts to spatially resolve jets in XRBs have focused primarily on intermediate states, where discrete, large-scale ejecta can propagate to parsec scales from the core [44, 115, 98, 9]. In contrast, jets in the hard state are typically confined within ≤10\le10 AU of the black hole, and only a handful have been spatially resolved to date [20, 105, 93, 97, 108, 112], likely due to a combination of source proximity, jet persistence, and the availability of sufficiently sensitive monitoring campaigns. As a result, the structure and dynamics of hard-state jets remain poorly constrained, and it is unclear whether the small number of resolved systems is representative of the broader population.

High-angular-resolution imaging of resolved hard-state jets provides direct measurements of fundamental jet properties – such as opening angle, collimation, and brightness temperature. These complement constraints from radio and infrared timing studies [46, 51, 107, 108], astrometry and core-shift measurements [63, 82], and broadband spectral modelling and spectral-break studies [54, 13].

GRS 1915+105 is the largest known low-mass XRB (LMXB) and one of the best studied BHXBs in our Galaxy. It was first detected as a bright X-ray emitter in 1992 by the WATCH all-sky X-ray monitor on board the GRANAT space observatory [11] and soon after became the first Galactic source detected with apparently superluminal radio ejecta [64]. Trigonometric parallax measurements of the system place it at a distance of 9.4±1.49.4 \pm1.4 kpc with an inclination angle i=64∘±4∘i = 64^\circ\pm4^\circ [88] and hosting a stellar mass BH [40] with a mass of 12.4−1.8+2.0 M⊙12.4^{+2.0}_{-1.8}\,\mathrm{M}_{\odot} [90].

For over 26 years, GRS 1915+105 exhibited persistent, intense X-ray activity, until it entered a sudden prolonged period of low flux in July 2018 (∼\sim MJD 58300), followed by further X-ray dimming around the end of March 2019 (∼\sim MJD 58600) [74, 69]. Initially, this was interpreted as the end of GRS 1915+105’s 26-yr long outburst, but this hypothesis was refuted only days later on 14 May 2019 (MJD 58617) with the emergence of renewed flaring activity across different wavelengths.

Multi-wavelength observations of the X-rays, radio and near-infrared (NIR) winds have since indicated that this state of low X-ray flux is the result of enhanced obscuration, drawing similarities to the highly obscured Seyfert-2 AGNs [59, 5, 70] (Sánchez-Sierras et al. 2023). Several obscuring mechanisms have been proposed, including a puffed-up slim disc [3] at high inclination blocking our view of the inner accretion flow [75], failed disc winds enshrouding the central engine [59] (Sánchez-Sierras et al. 2023), and a warped outer disc occulting the inner flow [60]. However, no single scenario has yet been definitively established as dominant.

The presence of intrinsic absorption has been reported only for a handful of other XRBs. In 2015, the BHXB V404 Cyg showed clear signatures of heavy and variable intrinsic absorption during both outbursts monitored in the X-rays, which were both interpreted as the result of a puffed-up disc blocking the view to the inner accretion flow [68]. Similar disc-induced obscuration has also been inferred in the microquasar SS 433 [27, 58]. In both these systems, obscuration was the consequence of erratic (V404 Cyg) or sustained (SS 433) super-Eddington accretion, which in both cases is associated with extreme activity of the radio jets [103, 62].

Despite the suppressed X-ray flux, GRS 1915+105 has remained highly active at radio wavelengths, as shown by regular monitoring with the Arcminute Microkelvin Imager Large Array (AMI-LA) and MeerKAT [70, 71]. Very Large Array (VLA) images at 8.0 GHz (Sep 30+Oct 01 2023) and 5.5 GHz (May 25th 2024) have shown a surprising change in the jet orientation angle at arcsecond scales [95]. Additionally, East Asian VLBI network (EAVN) observations at 6.7 GHz and 43 GHz from 2023 April find a large misalignment between the discrete ejecta detected at 6.7 GHz (∼8∘\sim8^\circ) and the continuous jets seen at 43 GHz (∼152∘\sim152^\circ; [45]). Most recent multi-wavelength monitoring found GRS 1915+105 at its lowest recorded activity level since discovery, with non-detections in radio (L-band; [72]), X-ray [53], optical [22], and near-infrared [15], potentially indicating the onset of a genuine transition to quiescence after more than three decades of continuous activity.

In this paper, we present four epochs of Very Long Baseline Interferometry (VLBI) observations targeting GRS 1915+105 at X-band (8.4 GHz) and C band (4.9 GHz), taken with the Very Long Baseline Array (VLBA) and the European VLBI Network (EVN), respectively. Being the earliest and highest-resolution radio observations of GRS 1915+105 during the obscured phase, these data allow us to study the jet emission on milliarcsecond scales, closest to the launch region where the accretion–outflow coupling is strongest and environmental effects on jet propagation are minimized. In Section 2 we describe our data reduction and imaging procedure, in Section 3 we present the observational results, in Section 4 we detail the analysis and parameter estimation, and in Section 5 we discuss our findings. Finally, in Section 6 we summarize our main results and outline our conclusions.

Notes. a VLBA participating stations: Sc, Hn, Nl, Fd, La, Pt, Kp, Ov, Br, and Mk. b EVN participating stations: Jb, Ef, Wb, Mc, O8, T6, Ur, Tr, Yb, Hh, Sv, Zl, Bd, Ir, and e-MERLIN stations Cm, Da, De, Kn, and Pi. c Times denote on-source time and exclude the 30-min geodetic blocks at the start and end of the VLBA observation.

Observations and Calibration

GRS 1915+105 was observed across four 8-hour epochs using VLBI: one at X-band (8.4 GHz) with the VLBA (project code BM500, PI: S.E. Motta) and three at C-band (4.9 GHz) with the EVN (project code EM144, PI: S.E. Motta). Figure 1 shows the timing of these observations overlaid on the radio flux density, X-ray flux and hardness evolution of GRS 1915+105 over the last ∼10\sim10 years. The X-ray fluxes are publicly available on the web pages of the Monitor of All-sky X-ray Image (MAXI1; Matsuoka et al. [55]) and the Swift/BAT transient monitor (BAT2; Krimm et al. [47]). The observation log is presented in Table 1.

Figure 1. AMI-LA and MeerKAT data, MAXI data, and BAT data, spanning approximately 10 years. The colour-coding indicates the X-ray spectral hardness, calculated as the ratio of the BAT to MAXI fluxes and shown numerically in the bottom panel. The grey and purple dotted lines indicate the dates of our VLBA and EVN observations.

InstrumentProposal CodeDateTsourceT_{\mathrm{source}} (minutes)Time (UTC)νc\nu_{\mathrm{c}} (GHz)Bandwidth (MHz)Synthesized beam (mas ×\times mas, ° N to E)
VLBAaBM5002019-10-12260.920:48 – 03:48c8.42561.03 ×\times 2.39, 171.35
EVNbEM144A2020-10-30235.512:30 – 20:304.92561.39 ×\times 3.50, 6.84
EVNbEM144B2021-03-05235.504:00 – 12:004.92562.49 ×\times 3.11, 67.21
EVNbEM144C2021-06-17235.521:30 – 05:304.92561.49 ×\times 2.83, 9.69

Table 1. Observation log of VLBA and EVN observations. The total time on source is denoted by TsourceT_{\mathrm{source}}. The central frequency is denoted by νc\nu_{\mathrm{c}}.

VLBA

We observed GRS 1915+105 following the radio-brightening during its X-ray obscured state (Motta et al. [70]) for 8 hours on October 12th 2019 with the VLBA. The observations were carried out in X-band (8.4 GHz) with a recording rate of 2048 Mbps, utilizing a total bandwidth of 256 MHz, divided into eight intermediate-frequency (IF) pairs of 32 MHz each. All ten VLBA stations participated: Saint Croix, Hancock, North Liberty, Fort Davis, Los Alamos, Pie Town, Kitt Peak, Owens Valley, Brewster, and Mauna Kea, although Los Alamos was flagged during calibration due to poor fringe fitting solutions.

Of the total 8 hours of observations, 260.9 minutes were spent on target. We observed J1800+3848 (7.3 minutes) and J2253+1608 (2.9 minutes) as fringe finders, J1905+0952 (86.6 minutes) as a phase reference calibrator and J1913+0932 (8.5 minutes) as a check source. The phase reference calibrator is at a 2.58∘2.58^\circ separation from the target. We alternated between 180 second scans on the target and 60 second scans on the phase calibrator, J1905+0952. The check source, J1913+0932, was observed for 60 seconds approximately every 40 minutes. Furthermore, we used geodetic blocks ([89]) of ∼30\sim30 minutes at the start and end of the observations to correct for excess tropospheric delays and clock errors.

The data were correlated using the DiFX software correlator ([18, 19]), and calibrated following the standard procedures within the Astronomical Image Processing System (AIPS, version 31DEC25; [110, 42]). After applying standard external gain calibrations, we performed several iterations of self-calibration and imaging on the phase-reference calibrator to construct an optimal model. From this model, we derived time-varying phase, delay, and rate solutions, which were interpolated to the target source. The amplitude gain solutions obtained from the phase-reference source were also applied to the target to further improve the time-varying amplitude calibration.

EVN

GRS 1915+105 was observed for three epochs of 8 hours on October 30th 2020, March 5th 2021 and June 17th 2021 with the EVN. All these observations were performed in the C-band with a central frequency of 4.9 GHz and eight IF pairs of 32 MHz each. The participating EVN stations were Jodrell Bank (Lovell Telescope), Effelsberg, Medicina, Onsala, Toruń, Urumqi, Tian Ma, Yebes, Hartebeesthoek, Svetloe, Zelenchukskaya, Badary, and Irbene, together with the e-MERLIN stations Cambridge, Darnhall, Defford, Knockin, and Pickmere. The data were correlated with the SFXC software correlator at the Joint Institute for VLBI ERIC (JIVE).

The total on-source observation time was 235.5 minutes per epoch. Consistent with the VLBA observations, we used J1800+3848 (9.9 minutes) and J2253+1608 (6.0 minutes) as fringe finders, J1905+0952 (168.4 minutes) as a phase reference calibrator and J1913+0932 (22.2 minutes) as a check source. We alternated between 210 second scans on the target and 120 seconds scan on the phase calibrator. The check source was observed for 150 seconds every fifth cycle.

The pointing centre during the EVN observations did not account for the source’s proper motion. To correct for this, we applied a phase shift prior to fringe fitting using the AIPS task CLCOR. For the correction, we adopted the proper motion measured by Reid & Miller-Jones [88]: μα=−3.14±0.03\mu_{\alpha} = -3.14 \pm0.03 mas yr−1^{-1}, μδ=−6.23±0.04\mu_{\delta} = -6.23 \pm0.04 mas yr−1^{-1}. Application followed the standard procedures in AIPS similarly to the steps performed for the VLBA observation. To maximize sensitivity, we constructed the phase calibrator model from the combined calibrated visibilities of all three EVN observations before applying the model for phase referencing.

Results

The calibrated visibilities were imaged in DIFMAP ([102]) using natural weighting to maximize sensitivity to extended emission. Component fitting was performed directly on the visibilities using the MODELFIT task, fitting circular or elliptical Gaussian components to the brightest emission regions.

The results of the Gaussian fits to the calibrated data are detailed in Table 2 and shown in Figure 2 as contour plots with natural weighting. We add 10% systematic errors in quadrature with our statistical errors to account for inaccuracies of the absolute amplitude calibration. All detections show extended emission, indicating that the source is resolved on milliarcsecond scales.

Natural weighting contour plots of the three detections of GRS 1915+105

Figure 2. Natural weighting contour plots at the 3σ3\sigma level of the three detections of GRS 1915+105. Panel (b) shows only the stable-flux interval 12:30–17:30 UT (see § 3.2). The right ascension (R.A.) and declination (Dec.) offsets are relative to the best fit position. Notice that the imaged field of the EVN epochs is larger than the size of the imaged VLBA field. EVN E2 (panel (c)) has the Gaussian model components overlaid onto the image.

Epochθmajobs\theta_{\mathrm{maj}}^{\mathrm{obs}} (mas)θmajdeconv\theta_{\mathrm{maj}}^{\mathrm{deconv}} (mas)SpeakS_{\mathrm{peak}} (mJy beam−1^{-1})StotS_{\mathrm{tot}} (mJy)PA (°)TB,beamT_{\mathrm{B,beam}} (K)TB,scatT_{\mathrm{B,scat}} (K)
VLBA2.61.7±0.21.7 \pm0.22.20±0.042.20 \pm0.046.23±0.626.23 \pm0.6215.3±1.215.3 \pm1.2≥(6.2±1.0)×107\ge(6.2 \pm1.0) \times10^{7}≥(3.7±0.6)×107\ge(3.7 \pm0.6) \times10^{7}
EVN E17.95.6±0.65.6 \pm0.68.76±0.098.76 \pm0.0990.3±9.090.3 \pm9.018.61±0.0318.61 \pm0.03≥(6.8±1.0)×108\ge(6.8 \pm1.0) \times10^{8}≥(1.5±0.2)×108\ge(1.5 \pm0.2) \times10^{8}
EVN E210.89.2±0.39.2 \pm0.30.68±0.020.68 \pm0.028.12±0.818.12 \pm0.81170.9±0.3170.9 \pm0.3≥(1.3±0.1)×107\ge(1.3 \pm0.1) \times10^{7}≥(5.8±0.6)×106\ge(5.8 \pm0.6) \times10^{6}
EVN E3––0.016±0.0080.016 \pm0.008<0.04a< 0.04^{a}–––

Table 2. Gaussian fit parameters, peak flux density, and brightness temperature lower limits for the VLBA and EVN observations, measured in DIFMAP. θmajobs\theta_{\mathrm{maj}}^{\mathrm{obs}} and θmajdeconv\theta_{\mathrm{maj}}^{\mathrm{deconv}} represent the observed and scatter-broadening-corrected major-axis sizes of elliptical Gaussians fitted to the brightest central component. SpeakS_{\mathrm{peak}} is the peak flux density of the image, with error given by the image RMS noise; this reflects only the image-plane statistical noise and does not include the additional ∼10%\sim10\% flux-scale calibration uncertainty adopted elsewhere. StotS_{\mathrm{tot}} is the total integrated flux density of the fitted Gaussian, and PA the position angle of the fitted Gaussian. TB,beamT_{\mathrm{B,beam}} and TB,scatT_{\mathrm{B,scat}} are lower limits on the brightness temperature computed using the array beam minor axis and the scattering size as upper limits on θmin⁡\theta_{\min}, respectively. The statistical errors on θmajobs\theta_{\mathrm{maj}}^{\mathrm{obs}} are negligible compared to the assumed 10% errors on the scatter broadening size and are therefore omitted.

VLBA

The detection in the VLBA epoch is best described by a single elliptical Gaussian component; the major and minor axes differ at the ∼4σ\sim4\sigma level (2.55 ×\times 2.55 mas circular vs 2.55 ×\times 1.84 mas elliptical), though the improvement in reduced χ2\chi^2 is marginal (Δχν2≈10−4\Delta\chi^2_{\nu} \approx10^{-4}). The major axis is oriented at 15.3∘±1.2∘15.3^\circ\pm1.2^\circ and is 23.9∘23.9^\circ offset from the beam orientation (171.4∘171.4^\circ), confirming that the elongation reflects intrinsic source structure rather than beam smearing. The deconvolved major axis of 1.7±0.21.7 \pm0.2 mas implies the source is resolved at 8.4 GHz on milliarcsecond scales. Despite this resolved structure, the fitted Gaussian centroid provides a robust position estimate for the core: RA = 19h15m11.5448325s±0.10219^{\mathrm{h}}15^{\mathrm{m}}11.5448325^{\mathrm{s}} \pm0.102 mas, Dec = +10∘56′44.626020′′±0.121+10^\circ56'44.626020'' \pm0.121 mas.

Flaring activity during EVN epoch 1

During our first EVN epoch (EVN E1 hereafter), the source showed a dramatic increase in flux density, rising by over 150 mJy within 3 hours (Figure 3). Such rapid variability suggests that GRS 1915+105 was observed during a flaring event, typically associated with the ejection of discrete jets.

Figure 3. Integrated source flux density evolution (left) and the on-sky best fit positions (right) of the highly varying radio emission during EVN E1 (EM144A). The observation has been split into time bins in a manner to allow for a maximum of ∼10%\sim10\% change in amplitudes. The integrated flux density of GRS 1915+105 is compared with the phase calibrator (J1905+0952) and check source (J1913+0932) to rule out calibration errors. Data points are colour-coded by time, allowing corresponding flux density and position measurements to be identified across the two panels. The nullpoint of the position plot is taken to be the mean of the stable flux density positions (before 17:30 UTC).

Given the high flux density of the source, we divided the observation into 5-minute time bins, ensuring that the amplitude variation between consecutive bins remains within ∼10%\sim10\%. We identified a stable-flux region (12:30–17:30 UTC) and a rising-flux region (17:30–20:30 UTC). This allows for imaging the stable part of the observation directly using conventional imaging techniques (Figure 2b). During this stable interval, we imaged the source in 30-minute segments and found that its position, flux density, and position angle remained constant within uncertainties.

The centroid positions from individual time bins were fitted using JMFIT (Figure 3). The integrated source flux density measurements are measured following self-calibration on 10 minute intervals, to optimize the amplitude calibration. The position fits are applied without self-calibration to retain absolute position accuracy.

Figure 3 (left) presents the integrated source flux density evolution of GRS 1915+105 within the 8-hour observation and compares this with the check source and phase calibrator flux density over time. The roughly constant flux density for both of these sources rules out amplitude calibration errors as the cause of the flux density increase in GRS 1915+105. The comparatively low flux density in the range 19:35–19:40 UTC is due to particularly poor uvuv-coverage within this time bin. The overall shape of the flux density evolution roughly follows a hockey-stick function. However, whether this reflects a linear flux density increase from a single component or a superposition of multiple independently Notes. (a)^{(a)} 5σ5\sigma upper limit on StotS_{\mathrm{tot}}, based on the image RMS of 0.008 mJy beam−1^{-1}.

varying components remains an open question, which we will pursue in a future paper.

Figure 3 (right) shows the best fit positions of the fitted components in each time bin. Systematic uncertainties were added following Pradel et al. [84] and Reid et al. [91]. We find a systematic shift in position along the axis of elongation, coinciding in time with the rise in flux density. This positional shift along the jet axis is consistent with measurable jet motion coincident with the onset of the radio flaring. No comparable shift is seen in the position of the check source over the same time bins, indicating that the observed motion is not an artifact of short timescale ionospheric variations or residual phase-referencing errors. However, splitting into smaller time bins severely limits the uvuv-coverage per data point, which greatly reduces our astrometric accuracy and could still, in principle, increase the risk of short timescale ionospheric variations affecting our positions.

A more detailed treatment of the implied jet velocity, and its relationship to the position angle behaviour discussed in § 5, will be the subject of a forthcoming study incorporating additional EVN monitoring of GRS 1915+105, and will allow a direct test of the recently proposed jet paradigm by Fender & Motta [29], which distinguishes between precessing, slow jets and fixed, fast jets, that arise from the inner disc locking onto the black hole spin-axis.

EVN epoch 2

EVN epoch 2 (hereby EVN E2) reveals an elongated core component with additional emission extending along the same orientation angle. We model the source with a total of five circular Gaussian components (overlaid on Figure 2c). All components are aligned with the brightest central emission, as well as the jet orientation angle observed with the VLA in 2023 (Rodríguez & Mirabel [95]). To rule out artefacts resulting from amplitude calibration errors we carried out tests including removing individual antennas, applying uvuv-tapers, varying the robustness parameter, and attempting self-calibration on several solution intervals.

EVN epoch 3

The third epoch of EVN observations (EVN E3) resulted in a non-detection and, therefore, is not shown in Figure 2. The calibrated data were also imaged using only the eMERLIN dishes, Westerbork, Jodrell Bank and Effelsberg to check for extended emission that might otherwise be resolved out by the full array, and to search for possible large-scale ejecta. No large-scale ejecta were detected. This non-detection places a 5σ5\sigma upper limit of ≤0.04\le0.04 mJy on the source’s flux density, indicating the source had faded by a factor of ≳220\gtrsim220 relative to its EVN E1 peak flux density eight months earlier. AMI-LA observations taken one day before and one day after EVN E3 measured a flux density of ∼0.15\sim0.15 mJy beam−1^{-1} at 15.5 GHz. Given AMI-LA’s much larger, tens-of-arcsecond-scale beam compared to our milliarcsecond-resolution VLBI observations, this is not necessarily inconsistent with our non-detection. Any diffuse or extended emission contributing to the AMI-LA flux density would be resolved out by the long VLBI baselines, and would not require a compact source above our detection threshold.

Analysis

Source morphology and extension

GRS 1915+105 lies close to the Galactic plane, where the line of sight intercepts a large column of ionized gas. This produces significant angular broadening and is partially responsible for the observed extension visible in our images. The NE2025 Galactic electron density model [76] predicts a scattering size of 1.4 mas at 5 GHz at the location of GRS 1915+105. Dhawan et al. [20] found a scattering size of ∼1.9\sim1.9 mas at 8.4 GHz and ∼5.6\sim5.6 mas at 4.9 GHz for GRS 1915+105. We prefer the direct, source-specific scatter-broadening measurements of Dhawan et al. [20] over the line-of-sight-averaged prediction from NE2025 for this analysis.

The Dhawan et al. [20] values agree with the observed extension of the minor axes fit by JMFIT in all epochs. We assume isotropic scattering and, consequently, account for the angular broadening by assuming a circular Gaussian such that θdeconv2=θobs2−θscat2\theta_{\mathrm{deconv}}^2 = \theta_{\mathrm{obs}}^2 - \theta_{\mathrm{scat}}^2. We adopt a conservative 10% uncertainty on the scattering size. After decoupling the measured source sizes from the expected scatter-broadening, the intrinsic source sizes still show significant elongation (Table 2; for beam sizes, see Table 1), suggesting the presence of continuous jet production. At the source distance of 9.4±1.49.4 \pm1.4 kpc [88], the deconvolved major-axis sizes correspond to projected extents of 16.0±3.016.0 \pm3.0 AU (VLBA, 8.4 GHz), 52.6±9.752.6 \pm9.7 AU (EVN E1, 4.9 GHz), and 86.5±13.286.5 \pm13.2 AU (EVN E2, 4.9 GHz). These are broadly consistent with the compact continuous jet sizes of a few to ∼100\sim100 AU reported by Dhawan et al. [20].

We estimate the brightness temperature based on the measured source size and flux density according to the Rayleigh-Jeans formalism [87, 48]

Tb≈1.22×1012Sν[Jy]ν2[GHz]θmaj[mas]θmin[mas] K(1)T_b \approx1.22 \times10^{12} \frac{S_\nu[\mathrm{Jy}]}{\nu^2[\mathrm{GHz}]\theta_{\mathrm{maj}}[\mathrm{mas}]\theta_{\mathrm{min}}[\mathrm{mas}]} \,\mathrm{K} \tag*{(1)}

The minor axes are limited by the synthesized beam and scatter broadening, yielding lower limits on the true brightness temperatures. The lower limits on the brightness temperatures, ranging from ∼6×106\sim6 \times10^6 K to ∼7×108\sim7 \times10^8 K across epochs (Table 2), are consistent with non-thermal, synchrotron emission, as expected for either a compact or extended radio jet.

Jet inclination angles

In addition to the changes in X-ray flux and jet position angle discussed above, if these position angle variations reflect precession of the jet axis, a corresponding change in the inclination angle would also be expected. Prior to obscuration, the inclination angle was consistent for compact and discrete ejecta over 25 years, with a best estimate of 64±4∘64 \pm4^\circ [88]. Recent measurements of discrete jet ejecta during the obscured phase suggest a nearly edge-on geometry, i=87±3∘i = 87 \pm3^\circ [95] and i=89±1∘i = 89 \pm1^\circ [45]. This motivates constraining ii using our data.

Assuming the approaching and receding jets are intrinsically symmetric and were launched simultaneously, we can use the relative distances from the core of corresponding jet components in EVN E2 to calculate possible combinations of the jet speed $\beta and inclination angle ii (e.g., Mirabel & Rodríguez [65]):

ΔraΔrr=1+βcos⁡i1−βcos⁡i.(2)\frac{\Delta r_{\mathrm{a}}}{\Delta r_{\mathrm{r}}} = \frac{1+\beta\cos i}{1-\beta\cos i}. \tag*{(2)}

Here Δra\Delta r_{\mathrm{a}} and Δrr\Delta r_{\mathrm{r}} are the angular separations from the core of the approaching and receding jet components, respectively, β=v/c\beta= v/c is the jet velocity, and ii is the inclination angle. For the jet component pair closest to the core we find Δra/Δrr=1.28±0.06\Delta r_{\mathrm{a}}/\Delta r_{\mathrm{r}} = 1.28 \pm0.06 while for the second, more distant pair we find Δra/Δrr=1.25±0.02\Delta r_{\mathrm{a}}/\Delta r_{\mathrm{r}} = 1.25 \pm0.02. In Figure 4 we show the allowed combinations of β\beta and ii for the jet component pairs (contours from Figure 2c). Error margins are propagated considering both statistical errors and systematic errors, following Pradel et al. [84] and Reid et al. [91]. The constraints from both jet component pairs overlap within error margins.

Figure 4. Constraints on the jet speed and inclinations of the two corresponding jet component pairs detected in EVN epoch 2. Shaded regions around the curves show the 1σ1\sigma uncertainties. The green shaded region indicates the inclination angle of GRS 1915+105 measured prior to the obscured phase (Reid & Miller-Jones [88]).

We show the pre-obscuration inclination angle of 64±4∘64 \pm4^\circ (Reid & Miller-Jones [88]) as a shaded region in Figure 4. More recent estimates during the obscured phase find higher inclinations: i=87±3∘i = 87 \pm3^\circ (Rodríguez & Mirabel [95]), who assumed the jet velocity remained constant at β=0.80±0.12\beta= 0.80 \pm0.12, based on the consistency of proper motions measured over many years prior to the obscured phase (Reid & Miller-Jones [88]). Jiang et al. [45] monitored the positions of discrete jet ejecta launched during the obscured phase over multiple days, and were able to directly constrain the velocity of these discrete ejecta to be β∼0.35\beta\sim0.35, consequently constraining the inclination angle to i=89±1∘i = 89 \pm1^\circ. This lower velocity measured by Jiang et al. [45] is inconsistent with the constant-velocity assumption of Rodríguez & Mirabel [95]; however, the authors note that their inclination estimate is largely insensitive to the assumed velocity. More recent observations with the EAVN from 2025 instead constrain the viewing angle of the jets to a smaller viewing angle, i≲78∘i \lesssim78^\circ, and an intrinsic jet speed of β≲0.37\beta\lesssim0.37 (Yan et al. [114]), broadly similar to the velocity found by Jiang et al. [45].

Our own constraints do not reach the high inclinations found by Rodríguez & Mirabel [95] and Jiang et al. [45] even for β=1\beta= 1, suggesting lower inclination angles during EVN E2. If we instead consider a jet velocity of β∼0.35\beta\sim0.35, similar to that found by Jiang et al. [45] and consistent with the constraint from Yan et al. [114], our constraints give inclination angles broadly consistent with the pre-obscuration estimate of 64±4∘64 \pm4^\circ (Reid & Miller-Jones [88]), and with the upper limit of i≲78∘i \lesssim78^\circ found by Yan et al. [114]. However, we note that these estimates come from different epochs. If the position angle changes reported here indeed reflect jet precession, the inclination angle would be expected to vary between them as well, so close numerical agreement between epochs should not be taken as evidence for a single, unchanging inclination.

Variation in jet orientation

The position angle of the jets from GRS 1915+105 have been detected in different accretion states over decades (Rodríguez & Mirabel [94]; Fender et al. [32]; Dhawan et al. [20]; Rushton et al. [96]; Miller-Jones et al. [61]; Reid et al. [90]). Before the obscured phase, the position angles measured on arcsecond and milliarcsecond scales span a range of ∼30∘\sim30^\circ (from 130∘130^\circ to 159∘159^\circ; see Figure 5 and Tab. A.1), with a weighted mean of 144∘±7∘144^\circ\pm7^\circ, though the reduced χν2=25\chi^2_\nu= 25 suggests the formal uncertainties underestimate the true scatter, or that genuine intrinsic variability was present. Restricting to milliarcsecond-scale measurements alone, the pre-obscuration weighted mean is 142∘±7∘142^\circ\pm7^\circ, which we adopt as the reference axis for comparison with our own milliarcsecond-scale observations below.

Figure 5. (a) Jet position angles measured with the VLBA and EVN in this paper (2019–2021), compared with the means of archival VLBA, EVN, EAVN, VLA and MERLIN measurements. The line lengths of the archival data are not to scale. (b) Evolution of the measured position angles as a function of time (MJD). The position angle axis is folded with a 180∘180^\circ period, centred at 150∘150^\circ. The shaded region marks the X-ray obscuration phase beginning in July 2018 and continuing to the present. Individual position angle measurements and references are displayed in table A.1.

Our VLBI campaign, which began after the onset of the obscuration phase, shows a drastically different PA from the historical axis. In the first milliarcsecond-scale image of the GRS 1915+105 jets after 2015, our VLBA data show a PA of 15.3∘±1.2∘15.3^\circ\pm1.2^\circ. Our second milliarcsecond-scale image, taken with the EVN one year later in October 2020 (EVN E1), shows a similar PA of 18.61∘±0.03∘18.61^\circ\pm0.03^\circ, confirming that this large PA offset persists and is connected to the obscuration phase rather than being a one-off measurement. Our final detection, over four months after EVN E1, shows the PA at 170.9∘±0.3∘170.9^\circ\pm0.3^\circ (EVN E2), 28.9∘±7.0∘28.9^\circ\pm7.0^\circ from the pre-obscuration jet axis. These three detections fill an important gap in the jet orientation history between the onset of the obscuration phase in 2018 and the published milliarcsecond-scale position angles from 2023 [45], who subsequently measured a PA of 12∘12^\circ with the EAVN, a value not seen at milliarcsecond scales prior to the obscured phase. This suggests that the jet’s pointing direction is continuously changing throughout this phase.

The jet PA changes from 18.6∘18.6^\circ (EVN E1) to 170.9∘170.9^\circ (EVN E2) over 126 days, with a minimum angular displacement of 27.7∘27.7^\circ, implying a sky-projected slew rate PA˙≳0.2∘ d−1\dot{\mathrm{PA}} \gtrsim0.2^\circ\,\mathrm{d}^{-1}. The sparse temporal sampling prevents distinguishing whether the jet is steadily precessing or changing angle erratically. The position angles remain stable to within the fitting uncertainties (≲5∘\lesssim5^\circ) over the 8-hour EVN E1 observation (see Table B.1), implying a precession rate ≲5∘/8 hr=15∘ d−1\lesssim5^\circ/8\,\mathrm{hr} = 15^\circ\,\mathrm{d}^{-1}, assuming the position angle changes are due to stable precession. We discuss the possible mechanisms responsible for the changes in jet position angle in § 6 below.

Discussion

Our VLBI epochs reveal jet position angle (PA) variations up to 57∘±8∘57^\circ\pm8^\circ with respect to the weighted mean of VLBI observations before the X-ray obscured phase (see § 5). Arcsecond-scale VLA observations in 2023 also reported a change in jet PA, ΔPA=24∘\Delta\mathrm{PA}=24^\circ [95]. This measurement traces discrete ejecta at much larger distances from the core, whose position reflects the jet’s direction integrated over a long propagation time since launch. Our milliarcsecond-resolution VLBI observations, by contrast, probe the jet both spatially and temporally much closer to the point of launch, offering a more direct measurement of the jet-launching direction.

Placing this offset in the broader context of jet reorientation in X-ray binaries, the largest reported position-angle range is ∼110∘\sim110^\circ for the neutron-star XRB Circinus X-1 [16]. This measurement combines observations from different instruments and resolutions spanning 25 years, and reflects the total peak-to-peak wandering of the axis in both directions over that baseline. Our 57∘57^\circ offset, by contrast, is a deviation in a single direction from a fixed pre-obscuration axis, measured consistently at milliarcsecond resolution, and represents the largest such offset reported from milliarcsecond-scale VLBI observations of an XRB. Milliarcsecond-scale EAVN observations in 2025 have since found the jet on the opposite side of this axis, at a PA of 118∘±7∘118^\circ\pm7^\circ [113]. Together, these measurements extend the range of jet PAs observed in GRS 1915+105 during the obscured phase to ∼80∘\sim80^\circ, which is the largest range reported for a black hole XRB.

Due to sparse temporal sampling we are unable to identify whether the jets in GRS 1915+105 are steadily precessing or changing direction erratically. EAVN observations from 2023 show a 6∘6^\circ change within two days at 6.7 GHz (PA ∼3∘ d−1\sim3^\circ\,\mathrm{d}^{-1}), and a 40∘40^\circ difference between observations at 6.7 GHz and 43 GHz separated by two days (PA ∼20∘ d−1\sim20^\circ\,\mathrm{d}^{-1}) [45]. The latter comparison, however, points to two vastly different observing frequencies, so some or all of this difference could reflect a frequency-dependent core shift along a bent or precessing jet rather than a true fast change in jet direction over that interval. Since higher-frequency emission becomes optically thin closer to the black hole, the two frequencies probe different distances along the jet. If the jet is already curved due to ongoing precession, this alone could produce different, earlier-position angles at each frequency, without requiring the launching direction itself to have swung by 20∘20^\circ within the two-day interval.

A genuine 20∘ d−120^\circ\,\mathrm{d}^{-1} position-angle change would require the reorientation rate to have increased substantially beyond the ≲15∘ d−1\lesssim15^\circ\,\mathrm{d}^{-1} upper limit implied by the stability of our EVN E1 position angles over 8 hours (§ 5). This either points to a core-shift origin for at least part of the 40∘40^\circ difference, or, if the 40∘40^\circ difference is genuine, indicates that the reorientation rate had already increased considerably beyond our E1-epoch constraint by 2023. The same-frequency 6.7 GHz rate of ∼3∘ d−1\sim3^\circ\,\mathrm{d}^{-1}, by contrast, is consistent with both of our EVN-derived limits (≲15∘ d−1\lesssim15^\circ\,\mathrm{d}^{-1} and ≳0.2∘ d−1\gtrsim0.2^\circ\,\mathrm{d}^{-1}). The temporal sampling remains too sparse, however, to establish whether this reflects a continuous precession, whether steady or accelerating, or an erratic reorientation, whether through fluctuating rates or discontinuous jumps. We consider possible physical origins for both below.

Lense-Thirring precession

GRS 1915+105’s transition into an X-ray obscured state raises the possibility of a mechanism analogous to that invoked for the 2015 outburst of the BHXB V404 Cygni, in which a super-Eddington, X-ray obscured accretion episode was accompanied by rapid jet precession [62]. Motta et al. (2017b) explain the X-ray obscuration using a slim disc model [3], in which erratic or continuous super-Eddington accretion inflates the inner region of the accretion flow, hiding the innermost (and brightest) regions of the flow.

A slim disc scenario naturally gives rise to Lense–Thirring precession. Precession of the inner slim disc may drive jet precession, either through the magnetic field lines anchored in the disc, realignment of spin-powered jets via powerful inner disc outflows, or through direct influence of the precessing slim disc itself. For V404 Cygni’s 2015 outburst, Miller-Jones et al. (2019) estimated a Lense-Thirring precession timescale of 56 s.

The expected period of Lense-Thirring precession in the first order approximation is [34, 57]

Pprec=π3a∗GMc3rsph3[1−(rin/rsph)3ln⁡(rsph/rin)](3)P_{\mathrm{prec}} = \frac{\pi}{3a_*}\frac{GM}{c^3}r_{\mathrm{sph}}^3\left[\frac{1-(r_{\mathrm{in}}/r_{\mathrm{sph}})^3}{\ln(r_{\mathrm{sph}}/r_{\mathrm{in}})}\right] \tag*{(3)}

where MM is the mass of the central object; a∗a_* is the dimensionless black-hole spin Jc/GM2Jc/GM^2 (with JJ the spin angular momentum); GG is the gravitational constant; and rinr_{\mathrm{in}} and rsphr_{\mathrm{sph}} are the inner and outer radii (in this case the spherization radius) of the slim disc, with all radii given in units of the gravitational radius rg=GM/c2r_g = GM/c^2. We assume a Kerr black hole with a mass of 12.4−1.8+2.0M⊙12.4^{+2.0}_{-1.8}M_\odot [90], spin parameter a∗=0.703±0.034a_* = 0.703 \pm0.034 and rin=rISCO=3.36±0.15 rgr_{\mathrm{in}} = r_{\mathrm{ISCO}} = 3.36 \pm0.15\,r_g [66].

Miller-Jones et al. (2019) estimated a spherization radius of tens to hundreds of rgr_g for moderate spins for V404 Cyg, which is consistent with the maximum radius expected for rigid precession [67]. For GRS 1915+105, a spherization radius equal to rsph=100 rgr_{\mathrm{sph}} = 100\,r_g with these assumptions would result in a precession period of Pprec∼27P_{\mathrm{prec}} \sim27 s. A full derivation of rsphr_{\mathrm{sph}} following Poutanen et al. [81], using GRS 1915+105’s peak luminosity, gives a comparable or smaller spherization radius (of order 10 rg10\,r_g), implying an even shorter precession timescale.

A precession period of this order cannot explain the observed timescales of the position angle changes reported here, which instead suggest that if precession is happening, it must involve a much larger portion of the accretion flow reaching the outer disc. Lense–Thirring precession also requires a slim-disc geometry, which recent XRISM Resolve spectroscopy disfavors, instead supporting a standard, sub-Eddington disc (§ 6.2; Miller et al. [60]). The combination of an implausibly short predicted precession timescale and a disc geometry disfavoured by the best available spectroscopic data leads us to rule out Lense–Thirring precession as the primary driver of the observed jet reorientation.

Radiation-driven warping

As noted above (§ 6.1), Miller et al. [60] model the first XRISM Resolve spectrum of GRS 1915+105 with a standard, sub-Eddington accretion disc, disfavouring the super-Eddington, slim-disc geometry that had remained viable in lower-sensitivity Chandra data. This disc temperature, together with the properties of the obscuring gas and the radii and inclinations inferred from the neutral FeKα\alpha and broad and narrow ionized emission lines, points to occultation of the central engine by the outer accretion disc itself. The line spectrum is further fit by three photoionized emission zones with small but significantly different bulk velocities, which the authors suggest may indicate axial asymmetry from a large-scale warp in the outer disc. Radiation pressure on the outer disc provides a viable mechanism for such a warp (Pringle [85]) and the system’s high inclination angle makes it plausible that a warped outer disc could occult our view of the central engine.

Spectral analysis of AstroSat, NICER and NuStar data constrains the distance of the absorbing plasma to within ≤2×1010\le2 \times10^{10} cm from the black hole (Athulya & Nandi [4]). Earlier, Neilsen et al. [75] found an absorbing distance of ∼\sim few ×1011\times10^{11} cm from the black hole based on NICER data, arguing for a vertically extended stratified disc. Assuming a black hole mass of 12.4−1.8+2.0 M⊙12.4^{+2.0}_{-1.8}\,\mathrm{M}_{\odot} (Reid et al. [90]), a distance of 101010^{10}–101110^{11} cm corresponds to ∼103\sim10^3–104Rg10^4 R_g, placing it in the mid-disc region where slim discs and radiation-driven war…

Radiation-driven warped discs can precess on super-orbital timescales (e.g. Maloney & Begelman [50]). If the warp-propagation timescale is shorter than the precession timescale the warp can propagate across the entire disc and precess as a global structure (Fragile et al. [34]; ); Liska et al. [49]). As the disc and hence the jet launching region precess, this is expected to cause the jet to precess. Strong evidence for a radiation-driven warped, precessing disc in an X-ray binary was provided by the discovery of the relativistic precessing jets in SS 433 (Margon [52]). The systematic velocity variations in the optical jet emission, the radio jet morphology, and the optical photometry of the system all indicate a disc precession period of 164 days.

Maloney & Begelman [50] originally interpreted this as evidence for a global mode of precession, in which both the inner disc (explaining the jets) and the outer disc (explaining the photometry) precess together at the same rate. More recent work (Begelman et al. [7]), however, proposes a picture in which the jets remain fixed to the black hole spin axis and are instead bent into alignment with a precessing outflow launched from the warped outer disc, with the jets and photometric period coinciding for this reason rather than through direct co-precession of the jet-launching region. This mechanism specifically requires hyper-Eddington accretion, since the outflow is radiatively driven from well outside the black hole, and is therefore only expected to operate in this form during correspondingly high accretion states. Such a state is disfavoured for GRS 1915+105 by recent XRISM Resolve spectroscopy, which instead supports a standard, sub-Eddington disc (Miller et al. [60]).

In order for a radiation-driven instability to arise, the irradiating luminosity must exceed a critical fraction of the Eddington luminosity such that the radiation reaction torque overcomes viscous damping. Radiation-driven instability is strongest at larger radii where the disc intercepts more flux per unit surface mass, so the disc must extend to radii where the instability criterion is satisfied before being tidally truncated (Wijers & Pringle [111]). Only those X-ray binaries with sufficiently small mass ratios (q=M2/M1q = M_2/M_1; with M1M_1 the mass of the compact object and M2M_2 the mass of the companion star) and sufficiently large orbital separations fall into the region of parameter space where persistent, steadily precessing radiation-driven warps are expected (Ogilvie & Dubus [80]).

This makes GRS 1915+105 a strong candidate for radiation-driven warping: with an orbital period of 33.9 days (Steeghs et al. [104]), the longest of all LMXBs, its large orbital separation ensures the disc extends well into the instability regime. Furthermore, its small mass ratio (q∼0.07q \sim0.07; using M1=12.4−1.8+2.0 M⊙M_1 = 12.4^{+2.0}_{-1.8}\,\mathrm{M}_{\odot} (Reid et al. [90]) and M2=0.81±0.53 M⊙M_2 = 0.81 \pm0.53\,\mathrm{M}_{\odot} (Harlaftis & Greiner [43]) means the disc is only weakly tidally truncated, allowing it to extend to the large radii where the radiation-driven instability is strongest.

However, only a narrow range of orbital separations (rbr_b) can result in a steadily precessing, radiation-driven warped disc. Assuming a viscous parameter value of α=0.3\alpha= 0.3 and accretion efficiency ϵ=L∗/M˙c2=0.1\epsilon= L_*/\dot{M}c^2 = 0.1, Ogilvie & Dubus [80] determined the ranges of stability (defined as containing a single unstable mode) to be 1.82≤rb/(106×GM1/c2)≤2.791.82 \le r_b/(10^6 \times GM_1/c^2) \le2.79 and 1.82≤rb/(106×GM1/c2)≤2.631.82 \le r_b/(10^6 \times GM_1/c^2) \le2.63 for a constant and variable luminosity respectively. Using the best known donor star mass estimate of Md=0.81±0.53 M⊙M_d = 0.81 \pm0.53\,\mathrm{M}_{\odot} for GRS 1915+105 (Harlaftis & Greiner [43]) and α=0.3\alpha= 0.3 and ϵ=0.1\epsilon= 0.1, we get an rbr_b that lies slightly above the upper bound of the region of stability (rb∼4×106 GM1/c2r_b \sim4 \times10^6\,GM_1/c^2). This implies the existence of multiple interacting warping modes.

The exact range of stability is uncertain, as it depends sensitively on the assumed values of α\alpha and ϵ\epsilon. Lower values would shift the instability threshold to higher rbr_b, potentially placing GRS 1915+105 within the region of stable, single-mode warping. The donor star mass estimate, and therefore rbr_b itself, also carries significant uncertainty.

A genuine reduction in α\alpha and/or ϵ\epsilon around this time is independently supported by observations: Motta et al. [70] report that GRS 1915+105 entered a canonical low/hard accretion state in 2018, immediately preceding the onset of the X-ray obscured phase. This transition is characteristically associated with a less radiatively efficient, geometrically thicker inner accretion flow (e.g. Esin et al. [25]). The source’s continued decline toward its lowest recorded activity level in 2026 (Motta et al. [72]; ); Marino et al. [53]; ); Ducoin et al. [22]; ); Coti Zelati et al. [15]) is consistent with this same trend continuing. Given GRS 1915+105’s proximity to the region of stable, single-mode warping, a genuine hard-state transition provides a plausible trigger for the emergence of a stable warp. The specific character of the warp depends on the system’s mass ratio and the location of mass input to the disc, as discussed below.

In the linear stability analysis of Ogilvie & Dubus [80], radiation-driven disc warping is described in terms of discrete eigenmodes of the disc’s bending-wave equation, each characterized by its radial structure (the number of nodes at which the disc’s tilt reverses sign) and an associated growth rate and precession frequency. The lowest-order mode, mode 0, has no radial node: the disc tilts coherently in a single direction at all radii, and precesses retrogradely. This is the mode thought to operate in systems such as Her X-1 and SS 433. Mode 1, the next-lowest-order mode, has one radial node, so that the inner and outer disc tilt in opposite directions, and precesses prograde.

Which of these modes is unstable first depends on the mass ratio qq and on where mass is added to the disc. For mass added at the circularization radius, Ogilvie & Dubus [80] find that mode 1 becomes unstable before mode 0 for q≲0.2q \lesssim0.2. If instead mass is added at the disc’s outer radius, which is arguably more realistic given that the accretion stream is expected to impact the outer disc, mode 0 requires a much larger disc to become unstable at all, and mode 1 dominates over a substantially wider range of parameters. Ogilvie & Dubus [80] caution, however, that this result neglects the angular momentum deficit of mass added away from a circular orbit at ror_o, which would introduce an extra torque that might affect the stability properties.

With q∼0.07q \sim0.07 for GRS 1915+105, the system lies in the regime where mode 1 is expected to dominate under either assumption. Given the system’s high inclination, mode 1’s one-node structure could naturally place the inner and outer disc on opposite sides of the line of sight, potentially obscuring the inner disc regardless of the precession phase. This could naturally explain the persistent nature of the obscuration.

We also consider the scenario in which GRS 1915+105’s accretion disc contains multi-modal warping. This would result in quasi-periodic and potentially less coherent precession compared to systems such as SS 433 and Her X-1. In the linear regime, this can be represented by a superposition of global bending normal waves, each with their own growth rate and precession frequencies. However, once the warp amplitude becomes large, non-linear effects modify the internal torque, which can reach a critical amplitude beyond which the steepest part of the warp steepens further, potentially leading to disc breaking ([78, 79, 21]). Additionally, parametric instability can develop, in which inertial waves extract energy from the warp, driving it toward decay or a turbulent regime ([38, 35]).

If GRS 1915+105 sits in the multi-modal regime, the observed jet reorientation may reflect non-linear, irregular warping rather than a clean, steadily precessing global mode. This is broadly consistent with the increased rate of position angle change inferred from the 2023 EAVN campaign, though our data cannot distinguish this from an accelerating continuous precession or a frequency-dependent effect (§ 6). Multi-modal warping is, however, harder to reconcile with the persistent, multi-year nature of the obscuration observed since 2019, which would more naturally suggest a stable, long-lived disc structure.

Independent evidence for a warped disc in GRS 1915+105 itself predates the current obscured phase. Rau et al. [86] reported a 590-day periodicity in the source’s hard X-ray and radio flux from over four years of RXTE monitoring, and put forth precession of a radiation-driven warp as an explanation, by analogy with the super-orbital periods seen in other warped-disc systems. This periodicity, however, corresponds to only a few cycles, and could plausibly instead reflect red noise. Stochastic, intrinsic source variability could mimic a periodic signal when only a small number of cycles are sampled (Vaughan [109]). The modulation was notably absent from the soft, disc-dominated X-ray flux, unlike Her X-1, SMC X-1, and LMC X-4, where the soft X-ray flux carries the super-orbital signal. The authors therefore only tentatively identified the periodicity’s physical origin with disc warping. If a fixed-period, precessing warp was indeed already present prior to 2018, the hard-state transition discussed above would not have established single-mode warping for the first time. Instead, it would have driven an increase in the warp’s amplitude, growing it from a level insufficient to occult the central engine into the large-amplitude warp required to produce the persistent obscuration and large PA variations.

Radiative warping therefore offers a plausible mechanism for the observed jet precession and obscuration in GRS 1915+105. The system’s proximity to the single-mode stability boundary, combined with the plausible reduction in α\alpha and ϵ\epsilon associated with the source’s hard-state transition and subsequent decline, makes a stable, single-mode warp a viable candidate, and one that naturally accounts for the persistent nature of the obscuration. However, we cannot exclude a more irregular, multi-modal warp, and GRS 1915+105’s proximity to the stability boundary may equally place it in a regime that is not a clean single-mode. Furthermore, mode 1 radiative warping has not been confirmed in any X-ray binary, and dedicated simulations in this regime have yet to be performed.

Other explanations

Other processes have been proposed to explain the apparently sudden onset of jet precession in GRS 1915+105. Previous studies have invoked interactions with an external perturber ([95, 45]). Recent population studies suggest that triple evolution is a significant formation channel for BH-LMXBs ([101]). However, the only current evidence of a BH-LMXB in a triple system is V404 Cyg whose tertiary companion is inferred to be at least 3500 astronomical units (AU) away ([10]). Simulations find that classical triple-channel formation without a common envelope typically results in outer separations of 10310^3 – 10410^4 AU, while a common envelope phase produces less constrained separations, generally in the range ∼100\sim100 – 10001000 AU. However, wider inner separations can result in closer outer separations ([101]), potentially bringing a tertiary companion close enough to be dynamically relevant. Nevertheless, GRS 1915+105’s decades-long X-ray brightness would require any tertiary companion to reside in a stable, wide orbit, making a dynamical perturbation sufficient to produce the observed precession onset difficult to motivate without additional evidence. Additionally, if the recent multi-wavelength non-detections ([72, 53, 22, 15]) indeed indicate a genuine transition into a low-accretion state, rather than a further deepening of the obscuration, this would favour an intrinsic, accretion-driven origin for the jet reorientation over an external perturbation, since the latter would not obviously predict a large-scale drop in luminosity.

Other studies have suggested that the obscuration could be the result of disc winds enshrouding the central emitting region ([59, 39]). This explanation would be consistent with the predictions of Fender & Motta [30], who found a correlation between the presence of absorption and the observations of slow and precessing jets. One possible physical link is mechanical wind-induced bending of the jets, which has been detected in the high mass X-ray binary Cygnus X-1 (Prabu et al. 2026). However, this is for winds expelled by the companion star, and simulations show that sufficiently strong stellar-wind bending is expected to disrupt the jet into a diffuse, no longer collimated outflow [116]. Jiang et al. [45] invoke this stellar-wind interaction physics by analogy to argue against a disc-wind origin for GRS 1915+105, based on the largely ballistic motion of the observed discrete ejecta. Disc-wind-driven jet bending, distinct from the stellar-wind case, has also been proposed directly for the precessing microquasar SS 433, where a powerful wind driven off a hyper-Eddington outer disc bends otherwise spin-aligned jets into alignment with the disc’s precession [7], as discussed in § 6.2. However, this mechanism specifically requires hyper-Eddington accretion, in tension with the sub-Eddington, standard-disc geometry favoured by recent XRISM spectroscopy of GRS 1915+105 (§ 6.1; [60]), and its applicability to the current, lower-accretion-rate obscured phase therefore remains uncertain.

Conclusions

We observed GRS 1915+105 for four epochs using VLBI during its X-ray obscured state between 2019–2021. As the earliest and highest-resolution radio observations of GRS 1915+105 during this phase, these data resolve the jet on milliarcsecond scales, closer to launch where accretion-outflow coupling is stronger.

The jet position angle evolved from 15.3∘±1.2∘15.3^\circ\pm1.2^\circ (VLBA) to 18.61∘±0.03∘18.61^\circ\pm0.03^\circ (EVN E1), before shifting by a minimum of 27.7∘27.7^\circ to 170.9∘±0.3∘170.9^\circ\pm0.3^\circ (EVN E2) over the following 126 days with no source detected at EVN E3. This sequence corresponds to a maximum offset of 57∘±8∘57^\circ\pm8^\circ from the weighted mean of jet angles measured prior to the X-ray obscured phase. We find no evidence for jet orientation changes within our 8-hour EVN E1 observation, implying a precession rate ≲15∘d−1\lesssim15^\circ\mathrm{d}^{-1} if the changes are due to stable precession.

We rule out Lense-Thirring precession of the inner slim disc as the direct driver. The predicted precession timescale is orders of magnitude shorter than observed, and the slim-disc geometry it requires is disfavoured by recent XRISM Resolve spectroscopy, which instead supports a standard, sub-Eddington disc.

We instead find that radiation-driven disc warping, previously proposed to explain the persistent X-ray obscuration of GRS 1915+105 [60], also plausibly accounts for the observed jet reorientation and its apparent coincidence with the onset of obscuration. GRS 1915+105 lies close to the boundary of the regime in which a single, steadily precessing mode 1 warp is expected, which would naturally explain the persistent, multi-year obscuration. We suggest that the source’s transition into a low/hard accretion state, and its continued decline toward its lowest recorded activity level in 2026, may have driven the emergence or growth of such a warp. However, the parameters placing GRS 1915+105 near this boundary carry significant uncertainties and mode 1 warping has not been confirmed in any X-ray binary. Sparse temporal sampling also prevents us from establishing whether the jet reorientation is smooth or erratic, so we cannot exclude a warp in which multiple modes, precessing at different rates, are superposed rather than one coherent structure dominating.

Acknowledgements. LIR would like to thank Benito Marcote for useful discussions on VLBI imaging. PA and LIR are supported by the WISE fellowship program, which is financed by NWO. SEM acknowledges support from the INAF Fundamental Research Grant (2022) EJ ECTA. RPF acknowledges support from the UKRI, European Research Council Synergy Grant ’BlackHolistic’ (grant agreement number 101071643) and The Hintze Family Charitable Foundation. The European VLBI Network is a joint facility of independent European, African, Asian, and North American radio astronomy institutes. Scientific results from data presented in this publication are derived from the following EVN project code: EM144. We acknowledge support from the Astrophysics Center for Multi-Messenger Studies in Europe (ACME), funded under the European Union’s Horizon Europe Research and Innovation Program, Grant Agreement No. 101131928. The National Radio Astronomy Observatory is a facility of the National Science Foundation operated under cooperative agreement by Associated Universities, Inc. This work made use of the Swinburne University of Technology software correlator, developed as part of the Australian Major National Research Facilities Programme and operated under licence. We thank the staff of the Mullard Radio Astronomy Observatory, University of Cambridge, for their support in the maintenance and operation of AMI. We thank the staff at the South African Radio Astronomy Observatory (SARAO) for scheduling the MeerKAT observations. The MeerKAT telescope is operated by the South African Radio Astronomy Observatory, which is a facility of the National Research Foundation, an agency of the Department of Science and Innovation. This research has made use of the MAXI data provided by RIKEN, JAXA and the MAXI team [55]. The authors wish to recognize and acknowledge the very significant cultural role and reverence that the summit of Maunakea has always had within the indigenous Hawaiian community. We are most fortunate to have the opportunity to conduct observations from this mountain.

Data Availability

The uncalibrated VLBA visibilities presented in this paper are publicly available in the National Radio Astronomy Observatory archive under project code BM500. The uncalibrated visibilities of the three EVN epochs are publicly available in the EVN archive under project code EM144 (https://doi.org/10.48717/y0rm-hd58). The uncalibrated MeerKAT visibility data are publicly available in the South African Radio Astronomy Observatory archive at https://archive.sarao.ac.za, subject to a standard proprietary period of one year. The AMI-LA data are available upon reasonable request to the corresponding author. Data from Swift/BAT are publicly available in NASA’s HEASARC archive (https://heasarc.gsfc.nasa.gov). The MAXI/GSC data are provided by RIKEN, JAXA, and the MAXI team (http://maxi.riken.jp). Data not available through public archives, and all source code, will be shared upon reasonable request to the corresponding author.

Position angle compilation

Table A.1: Compiled position angle (PA) measurements of the GRS 1915+105 jet. PA values are measured east of north and are defined modulo 180∘180^\circ, reflecting the undirected orientation of the fitted jet axis. Where both approaching and receding components were detected, the PA represents the average of the two.

EpochMJDPA (∘^\circ)Instrumentνc\nu_c (GHz)References
1994 Jan 2949381159±8159 \pm8VLA8.4Rodríguez & Mirabel [94]
1994 Feb 1949402157±6157 \pm6VLA8.4Rodríguez & Mirabel [94]
1994 Mar 1949430149±4149 \pm4VLA8.4Rodríguez & Mirabel [94]
1994 Apr 2149463147±6147 \pm6VLA8.4Rodríguez & Mirabel [94]
1994 Apr 2349465149.7±0.7149.7 \pm0.7VLA8.4Rodríguez & Mirabel [94]
1995 Aug 1049939140±10140 \pm10VLA8.4Rodríguez & Mirabel [94]
1997 Oct 2350744157±2157 \pm2VLBA15Dhawan et al. [20]
1997 Oct 3150752133±2133 \pm2VLBA2.3Dhawan et al. [20]
1997 Nov 650758142±2142 \pm2MERLIN5.0Fender et al. [32]
1998 Apr 1150914155±2155 \pm2VLBA15Dhawan et al. [20]
1998 May 250935154±4154 \pm4VLBA15Dhawan et al. [20]
2006 Feb 853794144.1±1144.1 \pm1VLBA8.4Miller-Jones et al. [61]
2006 Feb 853794140.1±1140.1 \pm1VLBA2.3Miller-Jones et al. [61]
2006 Feb 1253798143.4±1143.4 \pm1VLBA8.4Miller-Jones et al. [61]
2006 Feb 1453800145.3±1145.3 \pm1VLBA8.4Miller-Jones et al. [61]
2006 Feb 1453800138.5±1138.5 \pm1VLBA2.3Miller-Jones et al. [61]
2006 Feb 1753803150.7±1150.7 \pm1VLBA8.4Miller-Jones et al. [61]
2006 Apr 2153846140±2140 \pm2eEVN5.0Rushton et al. [96]
2013 May 2456436130±1130 \pm1VLBA22Reid et al. [90]
2015 Mar 557086140±8140 \pm8VLA2215A-460 (PI: A. King)
2019 Oct 125876815.3±1.215.3 \pm1.2VLBA8.4This paper
2020 Oct 305915218.61±0.0318.61 \pm0.03EVN4.9This paper
2021 Mar 559278170.9±0.3170.9 \pm0.3EVN4.9This paper
2023 Apr 156004912±112 \pm1EAVN6.7Jiang et al. [45]
2023 Apr 1660050158±7158 \pm7EAVN43Jiang et al. [45]
2023 Apr 17600516±16 \pm1EAVN6.7Jiang et al. [45]
2023 Apr 1960053146±4146 \pm4EAVN43Jiang et al. [45]
2023 Apr 23600576±16 \pm1EAVN6.7Jiang et al. [45]
2023 May 2360088138±9138 \pm9VLA5.5SS192168 (PI: A. Shaw)
2023 Sep 3060217148±3148 \pm3VLA10Jiang et al. [45]
2023 Sep 3060217.5174.3±0.5174.3 \pm0.5VLA8.023B-314 (PI: S.E. Motta)
2024 May 2560455152±1152 \pm1VLA5.524A-474 (PI: S.E. Motta)
2024 Jun 2360484149±1149 \pm1VLA5.524A-474 (PI: S.E. Motta)
2025 Jan 1860693118±7118 \pm7EAVN6.7Yan et al. [113]
2025 Jan 2560700131±6131 \pm6EAVN6.7Yan et al. [113]
2025 Feb 2460730152±2152 \pm2EAVN6.7Yan et al. [113]

Table A.1. Compiled position angle (PA) measurements of the GRS 1915+105 jet.

Time binning results from EVN E1

Time Range (UTC)Peak Flux Density (mJy beam−1^{-1})Total Flux Density (mJy)PA (∘^\circ)
12:30–13:006.23 ±\pm 0.09102.8 ±\pm 1.620.9 ±\pm 2.0
13:00–13:306.16 ±\pm 0.1098.4 ±\pm 1.821.3 ±\pm 2.1
13:30–14:005.50 ±\pm 0.0993.8 ±\pm 1.719.9 ±\pm 2.1
14:00–14:306.28 ±\pm 0.1495.5 ±\pm 2.219.1 ±\pm 2.7
14:30–15:0011.67 ±\pm 0.2196.6 ±\pm 1.918.8 ±\pm 2.8
15:00–15:307.86 ±\pm 0.1693.7 ±\pm 2.118.3 ±\pm 2.6
15:30–16:009.04 ±\pm 0.2095.6 ±\pm 2.320.4 ±\pm 2.9
16:00–16:3010.31 ±\pm 0.2299.7 ±\pm 2.419.9 ±\pm 3.0
16:30–17:0011.28 ±\pm 0.1997.0 ±\pm 1.820.6 ±\pm 2.5
17:00–17:3013.07 ±\pm 0.4597.4 ±\pm 3.721.6 ±\pm 7.9
17:30–17:4015.91 ±\pm 0.20105.0 ±\pm 1.522.3 ±\pm 2.5
17:40–17:5016.94 ±\pm 0.19108.4 ±\pm 1.422.0 ±\pm 2.4
17:50–18:0017.85 ±\pm 0.24111.8 ±\pm 1.821.7 ±\pm 2.9
18:00–18:0521.32 ±\pm 0.31115.1 ±\pm 1.924.5 ±\pm 3.0
18:05–18:1020.01 ±\pm 0.23117.1 ±\pm 1.522.2 ±\pm 2.5
18:10–18:1522.13 ±\pm 0.26123.8 ±\pm 1.721.4 ±\pm 2.6
18:15–18:2023.73 ±\pm 0.30132.1 ±\pm 1.921.1 ±\pm 3.0
18:20–18:2525.09 ±\pm 0.31135.7 ±\pm 1.922.1 ±\pm 3.1
18:25–18:3026.53 ±\pm 0.30138.4 ±\pm 1.820.7 ±\pm 2.9
18:30–18:3527.44 ±\pm 0.27142.2 ±\pm 1.721.4 ±\pm 2.5
18:35–18:4028.49 ±\pm 0.33148.5 ±\pm 2.021.0 ±\pm 3.1
18:40–18:4530.90 ±\pm 0.33156.4 ±\pm 2.021.0 ±\pm 3.0
18:45–18:5032.15 ±\pm 0.30161.5 ±\pm 1.821.4 ±\pm 2.6
18:55–19:0036.06 ±\pm 0.32173.1 ±\pm 1.821.8 ±\pm 2.6
19:00–19:0536.87 ±\pm 0.30172.2 ±\pm 1.721.7 ±\pm 2.5
19:05–19:1038.55 ±\pm 0.27178.7 ±\pm 1.522.1 ±\pm 2.2
19:10–19:1542.63 ±\pm 0.32184.4 ±\pm 1.720.4 ±\pm 2.4
19:15–19:2045.09 ±\pm 0.36191.1 ±\pm 1.920.1 ±\pm 2.6
19:20–19:2547.16 ±\pm 0.42205.2 ±\pm 2.223.0 ±\pm 3.2
19:25–19:3066.81 ±\pm 0.74213.0 ±\pm 3.025.4 ±\pm 4.4
19:30–19:35112.56 ±\pm 0.59216.3 ±\pm 1.620.8 ±\pm 1.5
19:35–19:4093.63 ±\pm 3.06190.8 ±\pm 8.89.1 ±\pm 15.7
19:40–19:45138.81 ±\pm 0.96228.9 ±\pm 2.416.9 ±\pm 4.0
19:45–19:50143.61 ±\pm 0.99237.2 ±\pm 2.417.5 ±\pm 3.8
19:50–19:55144.89 ±\pm 0.74247.1 ±\pm 1.917.6 ±\pm 3.1
19:55–20:00146.29 ±\pm 0.70254.6 ±\pm 1.816.5 ±\pm 3.0
20:00–20:05148.20 ±\pm 0.76257.7 ±\pm 1.914.4 ±\pm 3.2
20:05–20:10157.00 ±\pm 0.61268.6 ±\pm 1.515.9 ±\pm 2.6
20:10–20:15158.92 ±\pm 0.63272.5 ±\pm 1.615.5 ±\pm 2.8

Table B.1. Flux and position angle (PA) measurements of the time bins in EVN E1. Values correspond to Figure 3. The peak and total flux density measurements and position angle (PA) measurements are derived following self-calibration on 10 minute intervals, to improve the amplitude calibration.

References

  1. [1]Abbott, B. P., Abbott, R., Abbott, T. D., et al. 2016, Phys. Rev. Lett., 116, 061102
  2. [3]Abramowicz, M. A., Czerny, B., Lasota, J. P., & Szuszkiewicz, E. 1988, ApJ, 332, 646
  3. [4]Athulya, M. P. & Nandi, A. 2023, MNRAS, 525, 489
  4. [5]Balakrishnan, M., Miller, J. M., Reynolds, M. T., et al. 2021, ApJ, 909, 41
  5. [6]Begelman, M. C., Blandford, R. D., & Rees, M. J. 1984, Reviews of Modern Physics, 56, 255
  6. [7]Begelman, M. C., King, A. R., & Pringle, J. E. 2006, MNRAS, 370, 399
  7. [8]Belloni, T. 2009, States and Transitions in Black Hole Binaries (Springer Berlin Heidelberg), 53–84
  8. [9]Bright, J. S., Fender, R. P., Motta, S. E., et al. 2020, Nature Astronomy, 4, 697
  9. [10]Burdge, K. B., El-Badry, K., Kara, E., et al. 2024, Nature, 635, 316
  10. [11]Castro-Tirado, A. J., Brandt, S., & Lund, N. 1992, IAU Circ., 5590, 2
  11. [12]Chen, W., Shrader, C. R., & Livio, M. 1997, ApJ, 491, 312
  12. [13]Corbel, S., Fender, R. P., Tzioumis, A. K., et al. 2002, Science, 298, 196
  13. [14]Corbel, S., Nowak, M. A., Fender, R. P., Tzioumis, A. K., & Markoff, S. 2003, A&A, 400, 1007
  14. [15]Coti Zelati, F., Baglio, M. C., Motta, S., et al. 2026, The Astronomer’s Telegram, 17883, 1
  15. [16]Cowie, F. J., Fender, R. P., Heywood, I., et al. 2025, MNRAS, 544, L37
  16. [17]Crook-Mansour, J., Fender, R., Hughes, A., et al. 2026, MNRAS, 551, stag1022
  17. [18]Deller, A. T., Tingay, S. J., Bailes, M., & West, M. 2007, PASP, 119, 318
  18. [19]Deller, A. T., Brisken, W. F., Phillips, C. J., et al. 2011, PASP, 123, 275
  19. [20]Dhawan, V., Mirabel, I. F., & Rodríguez, L. F. 2000, ApJ, 543, 373
  20. [21]Doğan, S., Nixon, C. J., King, A. R., & Pringle, J. E. 2018, MNRAS, 476, 1519
  21. [22]Ducoin, J.-G., Magnani, F., Dornic, D., et al. 2026, The Astronomer’s Telegram, 17882, 1
  22. [23]El-Badry, K., Rix, H.-W., Quataert, E., et al. 2022, MNRAS, 518, 1057
  23. [24]El-Badry, K., Rix, H.-W., Cendes, Y., et al. 2023, MNRAS, 521, 4323
  24. [25]Esin, A. M., McClintock, J. E., & Narayan, R. 1997, ApJ, 489, 865
  25. [26]Fabian, A. C. 2012, ARA&A, 50, 455
  26. [27]Fabrika, S. 2004, Astrophys. Space Phys. Res., 12, 1
  27. [28]Falcke, H., Körding, E., & Markoff, S. 2004, A&A, 414, 895
  28. [29]Fender, R. P. & Motta, S. E. 2025, Nature Astronomy
  29. [30]Fender, R. P. & Motta, S. E. 2026, MNRAS, 551, stagn 112
  30. [31]Fender, R. P. & Muñoz-Darias, T. 2016, in Lecture Notes in Physics, Berlin Springer Verlag, ed. F. Haardt, V. Gorini, U. Moschella, A. Treves, & M. Colpi, Vol. 905, 65
  31. [32]Fender, R. P., Garrington, S. T., McKay, D. J., et al. 1999, MNRAS, 304, 865
  32. [33]Fender, R. P., Belloni, T. M., & Gallo, E. 2004, MNRAS, 355, 1105
  33. [34]Fragile, P. C., Blaes, O. M., Anninos, P., & Salmonson, J. D. 2007, ApJ, 668, 417
  34. [35]Fragile, P. C., Ingram, A., Musoke, G., & Ogilvie, G. I. 2026, Space Sci. Rev., 222, 25
  35. [37]Gallo, E., Fender, R. P., & Pooley, G. G. 2003, MNRAS, 344, 60
  36. [38]Gammie, C. F., Goodman, J., & Ogilvie, G. I. 2000, MNRAS, 318, 1005
  37. [39]Gandhi, P., Borowski, E. S., Byrom, J., et al. 2025, MNRAS, 537, 1385
  38. [40]Greiner, J., Cub y, J. G., & McCaughrean, M. J. 2001, Nature, 414, 522
  39. [42]Greisen, E. W. 2003, in Astrophysics and Space Science Library, Vol. 285, Information Handling in Astronomy — Historical Vistas, ed. A. Heck (Dordrecht / Boston: Kluwer Academic Publishers / Springer), 109–125
  40. [43]Harlaftis, E. T. & Greiner, J. 2004, A&A, 414, L13
  41. [44]Hjellming, R. M., Rupen, M. P., & Mioduszewski, A. J. 1998, Bulletin of the American Astronomical Society, 30, 1405
  42. [45]Jiang, W., Yan, X., Yan, Z., et al. 2026, ApJ, 1000, L45
  43. [46]Kalamkar, M., Casella, P., Uttley, P., et al. 2016, MNRAS, 460, 3284
  44. [47]Krimm, H. A., Holland, S. T., Corbert, R. H. D., et al. 2013, ApJS, 209, 14
  45. [48]Lee, S.-S., Lobanov, A. P., Krichbaum, T. P., et al. 2008, AJ, 136, 159
  46. [49]Liska, M. T. P., Hesp, C., Tchekhovskoy, A., et al. 2018, MNRAS, 474, L81
  47. [50]Maloney, P. R. & Begelman, M. C. 1997, ApJ, 491, L43
  48. [51]Malzac, J., Kalamkar, M., Vincentelli, F., et al. 2018, MNRAS, 480, 2054
  49. [52]Margon, B. 1984, ARA&A, 22, 507
  50. [53]Marino, A., Rea, N., Coti Zelati, F., et al. 2026, The Astronomer’s Telegram, 17869, 1
  51. [54]Markoff, S., Falcke, H., & Fender, R. 2001, A&A, 372, L25
  52. [55]Matsuoka, M., Kawasaki, K., Ueno, S., et al. 2009, PASJ, 61, 999
  53. [56]Merloni, A., Heinz, S., & di Matteo, T. 2003, MNRAS, 345, 1057
  54. [57]Middleton, M. J., Fragile, P. C., Bachetti, M., et al. 2017, MNRAS, 475, 154
  55. [58]Middleton, M. J., Walton, D. J., Alston, W., et al. 2021, MNRAS, 506, 1045
  56. [59]Miller, J. M., Zoghbi, A., Raymond, J., et al. 2020, ApJ, 904, 30
  57. [60]Miller, J. M., Gu, L., Raymond, J., et al. 2025, ApJ, 995, L14
  58. [61]Miller-Jones, J. C. A., Rupen, M. P., Fender, R. P., et al. 2007, MNRAS, 375, 1087
  59. [62]Miller-Jones, J. C. A., Tetarenko, A. J., Sivakoff, G. R., et al. 2019, Nature, 569, 374
  60. [63]Miller-Jones, J. C. A., Bahramian, A., Orosz, J. A., et al. 2021, Science, 371, 1046
  61. [64]Mirabel, I. F. & Rodríguez, L. F. 1994, Nature, 371, 46
  62. [65]Mirabel, I. F. & Rodríguez, L. F. 1999, A&A, 37, 409
  63. [66]Motta, S. E. & Belloni, T. M. 2024, A&A, 684, A209
  64. [67]Motta, S. E., Franchini, A., Lodato, G., & Mastroserio, G. 2017a, MNRAS, 473, 431
  65. [68]Motta, S. E., Kajava, J. J. E., Sánchez-Fernández, C., et al. 2017b, MNRAS, 471, 1797
  66. [69]Motta, S. E., Williams, D., Fender, R., et al. 2019, The Astronomer’s Telegram, 12773, 1
  67. [70]Motta, S. E., Kajava, J. J. E., Giustini, M., et al. 2021, MNRAS, 503, 152
  68. [71]Motta, S. E., Atri, P., Matthews, J. H., et al. 2025, A&A, 696, A222
  69. [72]Motta, S. E., Fender, R., Marino, A., et al. 2026, The Astronomer’s Telegram, 17865, 1
  70. [73]Muñoz-Darias, T., Díaz Trigo, M., Done, C., Ponti, G., & Tomaru, R. 2026, Space Sci. Rev., 222
  71. [74]Negoro, H. et al. 2018, The Astronomer’s Telegram, 11828, 1
  72. [75]Neilsen, J., Homan, J., Steiner, J. F., et al. 2020, ApJ, 902, 152
  73. [76]Ocker, S. K. & Cordes, J. M. 2026, ApJ, 1002, 3
  74. [77]Oei, M. S. L., Hardcastle, M. J., Timmerman, R., et al. 2024, Nature, 633, 537
  75. [78]Ogilvie, G. I. 1999, MNRAS, 304, 557
  76. [79]Ogilvie, G. I. 2000, MNRAS, 317, 607
  77. [80]Ogilvie, G. I. & Dubus, G. 2001, MNRAS, 320, 485
  78. [81]Poutanen, J., Lipunova, G., Fabrika, S., Butkevich, A. G., & Abolmasov, P. 2007, MNRAS, 377, 1187
  79. [82]Prabu, S., Miller-Jones, J. C. A., Bahramian, A., et al. 2023, MNRAS, 525, 4426
  80. [84]Pradel, Charlot, P., & Lestrade, J.-F. 2006, A&A, 452, 1099
  81. [85]Pringle, J. E. 1996, MNRAS, 281, 357
  82. [86]Rau, A., Greiner, J., & McCollough, M. L. 2003, ApJ, 590, L37
  83. [87]Readhead, A. C. S. 1994, ApJ, 426, 51
  84. [88]Reid, M. J. & Miller-Jones, J. C. A. 2023, ApJ, 959, 85
  85. [89]Reid, M. J., Menten, K. M., Brunthaler, A., et al. 2009, ApJ, 693, 397
  86. [90]Reid, M. J., McClintock, J. E., Steiner, J. F., et al. 2014, ApJ, 796, 2
  87. [91]Reid, M. J., Brunthaler, A., Menten, K. M., et al. 2017, AJ, 154, 63
  88. [92]Remillard, R. A. & McClintock, J. E. 2006, ARA&A, 44, 49
  89. [93]Ribó, M., Dhawan, V., & Mirabel, I. F. 2004, in European VLBI Network on New Developments in VLBI Science and Technology, ed. R. Bachiller, F. Colomer, J. F. Desmurs, & P. de Vicente, 111–112
  90. [94]Rodríguez, L. F. & Mirabel, I. F. 1999, ApJ, 511, 398
  91. [95]Rodríguez, L. F. & Mirabel, I. F. 2025, ApJ, 986, 108
  92. [96]Rushton, A., Spencer, R. E., Strong, M., et al. 2007, MNRAS, 374, L47
  93. [97]Russell, T. D., Miller-Jones, J. C. A., Curran, P. A., et al. 2015, MNRAS, 450, 1745
  94. [98]Russell, T. D., Tetarenko, A. J., Miller-Jones, J. C. A., et al. 2019, ApJ, 883, 198
  95. [99]Sahu, K. C., Anderson, J., Casertano, S., et al. 2022, ApJ, 933, 83
  96. [100]Shakura, N. I. & Sunyaev, R. A. 1973, A&A, 24, 337
  97. [101]Shariat, C., Naoz, S., El-Badry, K., et al. 2025, ApJ, 983, 115
  98. [102]Shepherd, M. C., Pearson, T. J., & Taylor, G. B. 1994, in Bulletin of the American Astronomical Society, Vol. 26, 987, proceedings of the 8th Astronomical Data Analysis Software and Systems (ADASS) Conference
  99. [103]Spencer, R. E. 1979, Nature, 282, 483
  100. [104]Steeghs, D., McClintock, J. E., Parsons, S. G., et al. 2013, ApJ, 768, 185
  101. [105]Stirling, A. M., Spencer, R. E., de la Force, C. J., et al. 2001, MNRAS, 327, 1273
  102. [107]Tetarenko, A. J., Casella, P., Miller-Jones, J., et al. 2019, in AAS/High Energy Astrophysics Division, Vol. 17, AAS/High Energy Astrophysics Division, 112.25
  103. [108]Tetarenko, A. J., Casella, P., Miller-Jones, J., et al. 2021, in American Astronomical Society Meeting Abstracts, Vol. 238, American Astronomical Society Meeting Abstracts #238, 125.02
  104. [109]Vaughan, S. 2005, A&A, 431, 391
  105. [110]Wells, D. C. 1985, in Ettore Majorana International Science Series, Vol. 24, Data Analysis in Astronomy, ed. V. di Gesù, L. Scarsi, P. Crane, J. H. Friedman, & S. Levialdi (New York: Springer), 195, in Data Analysis in Astronomy
  106. [111]Wijers, R. A. M. J. & Pringle, J. E. 1999, MNRAS, 308, 207
  107. [112]Wood, C. M., Miller-Jones, J. C. A., Bahramian, A., et al. 2024, ApJ, 971, L9
  108. [113]Yan, X., Cui, L., Jiang, W., et al. 2026a, ApJ, 1006, L15
  109. [114]Yan, X., Cui, L., Trushkin, S., et al. 2026b, arXiv e-prints, arXiv:2607.27688arxiv.org/abs/2607.27688
  110. [115]Yang, J., Brockso pp, C., Corbel, S., et al. 2010, MNRAS, 409, L64
  111. [116]Yoon, D. & Heinz, S. 2015, ApJ, 801, 55

Paper details

vlnc:W_2gaxhvu5
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 11

downloads · 2

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents