The X-ray corona turns into the relativistic jet in the micro quasar GRS 1915+105


 GRS 1915+1051 was the first stellar-mass black-hole in our Galaxy to display a superluminal radio jet2, similar to those observed in active galactic nuclei with a supermassive black hole at the centre3. It has been proposed that the radio emission in GRS 1915+105 is fed by instabilities in the accretion disc4 by which the inner parts of the accretion flow is ejected in the jet5–7. Here we show that there is a significant correlation between: (i) the radio flux, coming from the jet, and the flux of the iron emission line, coming from the disc and, (ii) the temperature of the corona that produces the high-energy part of the X-ray spectrum via inverse Compton scattering and the amplitude of a high-frequency variability component coming from the innermost part of the accretion flow. At the same time, the radio flux and the flux of the iron line are strongly anti-correlated with the temperature of the X-ray corona and the amplitude of the high-frequency variability component. These correlations persist over ~10 years, despite the highly variable X-ray and radio properties of the source in that period8,9. Our findings provide, for the first time, incontrovertible evidence that the energy that powers this black-hole system can be directed either to the X-ray corona or the jet. When this energy is used to power the corona, raising its temperature, there is less energy left to fuel the jet and the radio flux drops, and vice versa. These facts, plus the modelling of the variability in this source show conclusively that in GRS 1915+105 the X-ray corona morphs into the jet.

do not always show this high-frequency bump. At the same time, observations of GRS 1915+105 53 in which the spectrum is dominated by the corona are sometimes, but not always, accompanied by 54 high radio fluxes 21,22 . (We show the power-density spectra of two observations of GRS 1915+105 55 with the QPOs and the bump indicated in the Extended Data Figure 5.) 56 We studied a large dataset of 1800+ X-ray observations of GRS 1915+105 obtained with 57 the Rossi X-ray Timing Explorer (RXTE) between 1996 and 2012, combined with almost daily 58 observations of the source at 15 GHz with the Ryle telescope. Our final sample consists of 410 59 observations for which we have simultaneous data both in X-rays and radio. For each of these ob- 60 servations we have a measurement of (i) the radio flux density at 15 GHz, (ii) the X-ray hardness 61 ratio calculated as the ratio of the intensity in the 13−60 keV to the 2−7 keV band, (iii) the fre- 62 quency of the fundamental component of the type-C QPO, (iv) the phase lag at the QPO frequency 63 for photons in the 5.7−15 keV band with respect to those in the 2−5.7 keV band, (v) the 2−60 64 keV fractional rms amplitude of the high-frequency bump, and (vi) the best-fitting parameters to 65 the X-ray energy spectra of the source. (See the section Methods for details of the analysis and an 66 explanation of some of these quantities). These vastly different types of data, consisting of X-ray 67 and radio fluxes and spectral and timing properties of a single source, come from wavelengths that 68 are more than eight orders of magnitude apart and sample time scales that are more than eleven 69 orders of magnitude different, from ten milliseconds to a decade. 70 In Figure 1 we plot the X-ray hardness ratio as a function of the frequency of the QPO for  black-hole binaries, however, increases during periods in which the photon flux of the disc drops 106 and Compton cooling is less effective. This means that a source of power balances the inverse 107 Compton cooling and sets the temperature of the corona. Our findings show that in GRS 1915+105 108 the energy provided by this mechanism is split to either power the jet or heat the corona. 109 The thermal energy stored in a spherical corona of optical depth τ and size L around a black onto the disc before reaching the observer 28 . When this feedback fraction is low, time delays due 116 to Comptonisation dominate and the lags are positive; when this fraction is high, reprocessed disc 117 photons reach the observer later than those from the corona and the lags become negative. Taking 118 the corona sizes from fits with this model 29 , L≈10−1200R g , plus τ ≈1−6 and kT e ≈5−40 keV 119 from the spectral fits, we find that E th ≈10 31 −10 35 erg. If this energy is released over the time 120 scale of the high-frequency bump 1 , the thermal luminosity is 2 to 5 orders of magnitude lower 121 than the observed luminosity of the corona in GRS 1915+105. The alternative is that the corona 122 is powered by magnetic energy 27,30 , e.g., shear energy due to differential rotation of the magnetic-123 field lines that thread the accretion disc. This magnetic energy would also be responsible for the 124 synchrotron radio emission and the jet ejection mechanism in black-hole binaries. 125 We propose that in GRS 1915+105 the corona turns into the jet and that both are, at different 126 times, the same physical component. Based on the results shown here and fits with the model of 127 the lags that we present in the section Extended Results and Discussion, the process proceeds as 128 follows: (1) When the QPO frequency is ∼6 Hz the corona is relatively large and enshrouds the 129 inner parts of the accretion disc (Fig. 3a); as seen from the corona, the disc covers half of the sky 130 and hence there is a high probability that photons from the corona return to the disc leading to a 131 1 Note that this is the shortest variability time scale in the data; using any other time scale longer than this one to estimate the thermal luminosity would make the discrepancy even bigger. of the corona changes such that it becomes more prominent in the direction perpendicular to the 146 disc ( Fig. 3c); as seen from the corona, the disc now covers a much smaller area of the sky. As the 147 magnetic-field lines become more spatially coherent 31 there is less stochastic energy dissipation, 148 the corona temperature drops further and the material from the corona that was channeled off the 149 originally extended corona is expelled away from the accretion-disc plane and becomes the radio 150 jet (Fig. 3d). 151 2 The transition occurs when the lags fo the QPO change from negative to positive; although the exact value of the QPO frequency 19 at which this transition occurs is between 1.8 and 2 Hz, here we will always write 2 Hz for simplicity.
Our multi-wavelength correlations match the proposal that, during the initial parts of an 152 outburst, the X-ray corona of the black-hole binary MAXI J1820+070 contracts 32 and then re-153 expands 33 . Here we show conclusively that, as previously speculated 7,34,35 , in the case of GRS

154
1915+105 the contracting corona becomes the radio jet and that, at least part of the time, the 155 corona and the jet are actually one and the same physical component [36][37][38] .

156
The appearance of radio flares when the QPO frequency is below 2 Hz reinforces our inter-157 pretation. Figure 4 shows that in the periods that the frequency of the QPO moves more or less 158 stochastically between ∼2 Hz and ∼8 Hz (red circles) the radio flux (light blue curve) is low. Oc-159 casionally, the QPO frequency evolves in a more systematic way: it starts to decrease, crosses the 160 value of 2 Hz and moves up again; at the same time the radio flux increases sharply and a radio 161 flare lasting a few tens of days is observed. Figure 4 shows this behaviour over a period of about 162 500 days in which the source shows two radio flares. Extended Data Figure 9 shows that the same 163 behaviour repeats consistently over a period of 10 years and about a dozen radio flares.

164
The relation between the iron-line flux and the total flux is consistent with the above scenario.

165
The flux of the iron line depends more strongly upon X-ray flux when the temperature of the corona 166 is low, that is when the corona has turned into the jet 25 , than when the corona is more extended and 167 covers the inner parts of the accretion disc (Fig. 2) . This points to a lamp-post geometry 39 with 168 the corona, which is now the jet, illuminating the disc anisotropically 40,41 . Because the jet does not 169 cover the accretion disc, the flux of the iron line is not (or mildly) attenuated by the corona 42,43 .

170
The behaviour we observe in GRS 1915+105 could explain the deviations from a single 171 track in the radio/X-ray correlation of other accreting galactic black holes [44][45][46][47][48][49] If, as in the case of 172 GRS 1915+105, the energy powering these systems is used to either accelerate the jet or heat the 173 corona, different sources, or the same source at different times, will show lower radio fluxes at a 174 given X-ray flux (or, equivalently, higher X-ray fluxes at a given radio flux), depending on how 175 much energy is directed towards, respectively, launching the radio jet or heating the X-ray corona.   Competing Interests The authors declare that they have no competing financial interests.
Supplementary Information is available for this paper.
Correspondence Correspondence and requests for materials should be addressed to MM Words in the caption of Figure 1: 165 (max. should be around 100 or 250; unclear) Words in the caption of Figure 2: 128 (max. should be around 100 or 250; unclear) Words in the caption of Figure 3: 120 (max. should be around 100 or 250; unclear) Words in the caption of Figure 4: Words in the caption of Ext. Figure 5: 126 (max. should be around 100 or 250; unclear) Words in the caption of Ext. Figure 6: 135 (max. should be around 100 or 250; unclear) Words in the caption of Ext. Figure 7: 153 (max. should be around 100 or 250; unclear) Words in the caption of Ext. Figure 8: 144 (max. should be around 100 or 250; unclear) Words in the caption of Ext. Figure 9: 144 (max. should be around 100 or 250; unclear) Words in the caption of Ext. Figure 10: 172 (max. should be around 100 or 250; unclear) Total number of references in the Section Methods:

194
Based on the fitting results we only retained features that were detected at a significance greater 195 than 3-σ and had a Q factor, defined as the ratio of the QPO frequency to its width, of 2 or more. tons that enter into, and are up scattered in, the corona, kT seed , and a normalisation that gives the 234 flux density at 1 keV. We assumed that the source of seed soft photons is the accretion disc, and 235 3 We also fitted the RXTE/HEXTE (High Energy X-ray Timing Experiment) data of those observations in which the instrument was operational; the results were consistent with the ones of the RXTE/PCA, but since data of this instrument were not available for all observations, we did not use the RXTE/HEXTE data for the rest of the analysis.   Compared to the fits with a Gaussian, this model has six extra free parameters. Given that the 254 RXTE/PCA instrument has limited spectral resolution and does not extend below ∼3 keV, the 255 model is insensitive to some of the parameters of REXILLCP, and some of these parameters become 256 highly correlated with those of the DISC component. We therefore fixed the spin of the black hole 257 to a * = 0.95 (ref. [59]) and the inclination angle of the accretion disc to the line of sight to 258 i = 65 • (ref. [2]). We further linked the two emissivity indices during the fits, which eliminates 259 the parameter that gives the radius in the disc at which the power-law index of the emissivity profile 260 changes, and fixed the inner radius of the disc at the radius of the innermost stable circular orbit 59 .  for other details of the analysis of the radio data.

296
Finally, we cross-correlated all the X-ray and radio data based on the date of the observations, 297 which left us with a sample of 410 observations with simultaneous radio flux densities at 15 GHz 298 and X-ray energy, power-density and lag-frequency spectra and hardness ratios.

299
Having described the observations and analysis we used, it is worth noting that the mea-300 surements presented in this paper come from very different types of data and totally independent 301 analysis techniques: The hardness ratio, iron-line flux and kT e come from X-ray light curves and 302 time-averaged energy spectra, the frequency and lags of the QPO and the rms amplitude of the 303 high-frequency bump come from Fourier power spectra of high-time resolution data, and the radio 304 flux was measured independently in a totally different frequency band than the X-ray data.  In the observations in which the high-frequency bump is significantly detected the width of this 333 feature ranges from ∼30 Hz to ∼100 Hz. In those observations in which the high-frequency bump 334 was not significantly detected, to compute the upper limits we fixed the width of the Gaussian to 335 70 Hz, which is the average value that we obtained from the fits of the observations in which the 336 high-frequency bump was significantly detected. We report a detailed analysis of the properties of 337 the high-frequency bump in a separate paper 71 . it is apparent that the high-frequency bump is present when the temperature of the corona is high.
is higher when the temperature of the corona is higher, regardless of the values of the hardness ratio 350 and QPO frequency in each observation (see Figure 1).

351
Energy-dependent rms amplitude and phase lags of the QPO. The corona becomes the jet:  Figure 8). When the QPO frequency decreases below ∼2 Hz, the corona size increases again up 385 to L≈2×10 4 km (∼1200 R g ) at the lowest QPO frequency, the feedback fraction remains constant 386 close to zero and the lags become positive and increase with decreasing QPO frequency. Although 387 the size of the corona increases again when the QPO frequency is below ∼2 Hz, the fact that the 388 feedback fraction is close to zero whereas when the QPO frequency is above ∼2 Hz the feedback 389 fraction is high, implies that the corona does not cover the inner parts of the accretion disc. This in 390 turn demonstrates that the geometry of the corona has changed (see Fig. 3).

391
Extended Data Figure 9 shows the time evolution of the QPO frequency superimposed on 10 of the average QPO frequency in each observation, we decided to use these data for this Figure   399 because this allows us to have a denser coverage of the time evolution of the QPO frequency during 400 the radio monitoring campaign. Extended Data Figure 9 shows that the intervals of strong radio 401 flares 5 coincide with the times when the QPO frequency is below 2 Hz. This behaviour repeats 402 consistently for all the radio flares in these observations; there is no radio flare in which the QPO 403 frequency is not below 2 Hz, and there is no case of a QPO frequency below 2 Hz without a radio 404 flare. Given the data, we estimated that the probability of having a QPO frequency below 2 Hz and 405 a radio flare at the same time if the two phenomena were uncorrelated is less than 2×10 −10 . the feedback fraction 29,75 from ∼1 to ∼0 when the QPO frequency is, respectively, above or below 408 2 Hz are consistent with the scenario described earlier, in which the energy that is initially stored 409 in the X-ray corona is gradually released into the radio jet and, quite possibly, the X-ray corona 410 itself becomes the jet.      The points connected by a line show the time evolution of the frequency of the QPO in the X-ray power spectrum of GRS 1915+105. The light-blue curve (smoothed with a Gaussian kernel) shows the simultaneous measurements of the radio flux density at 15 GHz (y axis rescaled). A strong radio flare appears whenever the lags of the QPO turn from soft (red points) to hard (blue points), corresponding to the QPO frequency crossing from above to below ∼2 Hz. The horizontal band shows the range of QPO frequencies over which the transition occurs. The probability of having a positive QPO lag always and only during a radio flare if the two phenomena were uncorrelated is less than 2×10 −10 . (Fig. 4 in the main text shows a zoom-in of the first two flares.) The average error of the plotted quantities is shown at the upper right of the plot. The colour scale represents the phase lags, in units of (rad/2π), of the QPO (see Fig. 2 for an explanation). The triangles indicate upper limits of the ratio of the expected to the observed corona flux. The expected and observed fluxes are roughly the same (within a factor of ∼3) when the lags are negative, the corona is extended and hot, and it covers the inner parts of the accretion disc (colours towards red), whereas the expected flux drops significantly when the lags are positive, the geometry of the corona changes and its temperature drops as it turns into the jet (colours towards blue).