astro-ph0006086/ms.tex
1: \documentclass[manuscript]{aastex}
2: \slugcomment{Submitted to Ap.J.}
3: \shortauthors{Dhawan, Mirabel, \& Rodr\'\i guez}
4: \shorttitle{AU-Scale Jets in GRS~1915+105}                 
5: 
6: \begin{document}
7: 
8: \title{AU-Scale Synchrotron Jets and Superluminal Ejecta in GRS~1915+105.}
9: 
10: \author{V. Dhawan}
11: \email{vdhawan@nrao.edu} 
12: \affil{National Radio Astronomy Observatory, Socorro, NM 87801 
13: \altaffilmark{1}}
14: 
15: \author{I. F. Mirabel}
16: \email{mirabel@discovery.saclay.cea.fr}
17: \affil{CEA/DSM/DAPNIA/SAp, Centre d'Etudes de
18: Saclay, F-91191 Gif-sur-Yvette, France \& 
19: Instituto de Astronom\'\i a y F\'\i sica del Espacio, Buenos Aires, Argentina}
20: 
21: \and
22: 
23: \author{L. F. Rodr\'\i guez}
24: \email{luisfr@astrosmo.unam.mx}
25: \affil{Instituto de Astronom\'\i a, UNAM, Apdo. Postal 70-264, 04510,
26: M\'exico, DF, M\'exico}
27: 
28: \altaffiltext{1}{The NRAO is a facility of the National Science
29: Foundation operated under cooperative agreement by Associated
30: Universities, Inc.\ }
31: 
32: \begin{abstract}
33: 
34: Radio imaging of the microquasar GRS~1915+105 with the Very Long
35: Baseline Array (VLBA) over a range of wavelengths 
36: (13, 3.6, 2.0 and 0.7~cm), in different states of the black hole
37: binary, always resolves
38: the nucleus as a compact jet of length $\sim$10$\lambda_{\rm cm}$~AU.
39: The nucleus is best imaged at the shorter wavelengths, on scales of
40: 2.5~-~7~AU (0.2~-~0.6~mas resolution).  The brightness temperature of
41: the core is T$_{\rm B}\geq$10$^{9}$ K, 
42: and its properties are better fit by a conically expanding
43: synchrotron jet model, rather than a thermal jet.  The nuclear jet
44: varies in $\sim$30~min during minor X-ray/radio  outbursts, and
45: re-establishes within $\sim$18~hours of a major outburst, indicating
46: the robustness of the X-ray/radio (or disk/jet) system to disruption.
47: 
48: At lower resolution (80-240~AU), more extended ejecta are imaged at
49: $\sim$500 AU separation from the stationary core. 
50: Time-lapse images clearly detect the superluminal motion of the ejecta 
51: in a few  hours. The measured velocity is 1.5$\pm$0.1~c~(D/12~kpc) 
52: for the approaching
53: component, and is consistent with ballistic motion of the ejecta from
54: 500~AU outwards, perhaps even since birth.  The axis of the
55: ejecta differs by $\le$12$^{\circ}$ clockwise from the
56: axis of the AU-scale jet, measured in the same observation. Both axes
57: are stable in time ($\pm$5$^{\circ}$), the AU scale for two  years, and
58: the large scale for over four years.
59: Astrometry over two years relative to an extragalactic reference
60: locates the black hole to $\pm$1.5~mas, and  its secular parallax due
61: to Galactic rotation is 5.8$\pm$1.5~mas~yr$^{-1}$, consistent with a
62: distance of 12~kpc. Finally, a limit of $\le$100~km~s$^{-1}$ is placed
63: on its proper-motion with respect to its neighbourhood.
64: 
65: Some accreting black holes of stellar mass (e.g. Cyg~X-1,
66: 1E~1740-2942, GRS~1758-258, GX~339-4) and supermassive black holes
67: at the centre of galaxies (e.g. Sgr~A$^*$) lack evidence of large flares
68: and discrete transient ejecta, but have compact radio cores with
69: steady, flat-spectrum `plateau' states, like GRS~1915+105. Until now
70: GRS~1915+105 is the only system where both AU-scale
71: steady jets and large-scale superluminal ejections have been
72: unambiguously observed. Our observations suggest that the
73: unresolved flat-spectrum radio cores of accreting black
74: holes are compact quasi-continuous synchrotron jets.
75: 
76: \end{abstract}
77: 
78: \keywords{Subject headings: radio continuum: stars --- stars:
79: individual: (GRS~1915+105) --- X-rays: stars }
80: 
81: \section{Introduction}
82: 
83: The X-ray transient GRS~1915+105, discovered in 1992 by the \sl Granat
84: \rm satellite, \citep{cas94}, is one of a few Galactic sources
85: exhibiting superluminal radio ejecta (see \citet{araa99} for a
86: review).  This microquasar offers a nearby laboratory for the study of
87: black hole accretion, and the associated phenomena of jet formation,
88: collimation, and outflow.  Since the characteristic evolution time is
89: proportional to the mass of the central black hole (Rees, 1998; Sams et
90: al. 1996), microquasars can reveal in minutes a richness of phenomena
91: analogous to a 10$^8$~M$_{\odot}$ quasar observed over centuries.
92: 
93: In this paper, we present radio images of the microquasar GRS~1915+105,
94: made with the Very Long Baseline Array (VLBA, see \citet{vlba95}), 
95: in two distinct states
96: of the black hole binary, the {\bf plateau} and {\bf flare} states. The
97: {\bf plateau} state is characterized by a flat radio spectrum, compact
98: size of a few AU, and flux density 10-100~mJy. The XTE (2-12~keV) soft
99: X-rays are weak, and the BATSE (20-100~keV) emission is strong. In
100: contrast, during the {\bf flare} state, with rise-time $\le$1~day,
101: optically thin ejecta of up to 1~Jy at $\lambda$13~cm
102: (S$_{\nu}~\propto \nu^{-0.6}$), are expelled to thousands of AU and fade
103: over several days. The soft X-rays also flare and show extreme
104: variability, while the hard X-rays fade for a few days before
105: recovering.  Observations over a range of wavelengths (13~cm, 3.6~cm,
106: 2~cm and 0.7~cm), distinguish the intrinsically elongated nucleus, of
107: length $\sim$10$\lambda_{\rm cm}$~AU, from angular broadening due to
108: scattering by the line-of-sight ionized interstellar medium in the
109: Galactic plane.
110: 
111: GRS~1915+105 was also imaged in its distinct state of pronounced X-ray
112: dips, that repeat on timescales ranging from $\sim$12-60~min, with the
113: same repetition rate seen in the IR and radio pulses. The radio and
114: X-ray variability of GRS~1915+105 showed early evidence for recurrent
115: accretion and jet formation \citep{rod99,fos96}. There now exists a
116: wealth of data from X-ray, IR and radio observations, offering clues to
117: the coupled evolution of accretion disk, corona, and jet. On several
118: occasions, the flux of thermal X-rays from the inner accretion disk has
119: been observed to diminish, with simultaneous hardening of the power-law
120: spectrum from relativistic electrons, followed by progressively delayed
121: emission in the IR and radio,
122: \citep{mir98,mar99,bel97b,eik98,poo97,fen97,fen98}.  One scenario is to
123: have the inner part of the accretion disk and corona accelerated to
124: relativistic speeds, and ejected as synchrotron-emitting plasma.
125: Alternatively, advective infall of the disk past the black hole horizon
126: \citep{nar97}, could account for the X-ray dips, but not for the IR and
127: radio emission. A solution allowing simultaneous inflow and outflow 
128: is presented in the `ADIOS' (Advection Dominated Inflow-Outflow
129: Solution)  model of \citet{bla99},
130: but mechanisms to produce the observed X-ray variability without ADAF
131: (Advection Dominated Advection Flow) are discussed in \citet{nay00}.  
132: Though the mechanism is still debated, this particular source switches 
133: the jets on and off,
134: presenting repeated but unpredictable opportunities to observe the
135: ejecta evolve from  scales of a few to thousands of AU.
136: 
137: \section{VLBA Observations \label{sec-vlbaobs}}
138: 
139: Triggered by GBI \footnote{The Green Bank Interferometer is a facility
140: of the National Science Foundation operated by NRAO with support of the
141: NASA High Energy Astrophysics programs.}\ daily monitoring, we have
142: coordinated observations with X-ray  and IR telescopes, followed the
143: radio variations with the VLA \citep{mir98}, and imaged the radio
144: emission at multiple wavelengths with the VLBA (this paper).  Crucial
145: capabilities of the VLBA were the dynamic scheduling in
146: response to transients, and phase-referencing which permits astrometry
147: and the  detection of weak sources.
148: 
149: The VLBA observation epochs and parameters are presented in Table-1.
150: Observations are possible in any of nine wavelength bands, with band
151: changes requiring 20~s or less.  Simultaneous data can be obtained from
152: the 3.6~cm/13~cm receiver pair, by means of frequency-selective optics.
153: The choice of band was made just before each observation,  based on the
154: most recent flux and spectral information. Observations at 2~cm and
155: 7~mm are least affected by galactic electron scattering, and were used to
156: observe the AU scales when the radio-to-IR spectrum was flat. During
157: the steep-spectrum flares, we employed the 3.6~cm/13~cm pair to image
158: large-scale ejecta. Other bands were observed for $\sim$20-40~mins in
159: rotation.  We employed the highest data recording rates, (256 or
160: 128~Mb~s$^{-1}$, subject to scheduling constraints), for maximum
161: sensitivity in short time, since the source is known to be highly
162: variable. Ultimately, our image quality is limited not by thermal
163: noise, but by the restriction of instantaneous UV coverage to a maximum
164: of 45 baselines (Fourier components) from 10 antennas.
165: 
166: \placetable{table-1}
167: 
168: All the VLBA observations (except 7~mm) were phase-referenced, i.e.,
169: the antennas alternated between the target and one of several
170: extragalactic calibrators of known (sub-millarcsec) position. Cycle
171: times of three minutes were adequately fast to follow tropospheric
172: phase fluctuations at 2~cm and longer wavelengths. Thus the
173: interferometer array was rendered coherent for many hours, and the
174: resulting astrometric accuracy of $\sim$1.5~mas allows us to
175: discriminate between moving and fixed source components. In addition,
176: with frequent amplitude calibration, we avoid blind self-calibration on
177: the time-variable target.
178: 
179: We generated time-resolved images, with snap-shot intervals ranging
180: from 5 to 60~mins, depending on the beam size. Self-calibration was
181: used self-consistently, i.e. only over the snapshot interval.
182: Snap-shots were averaged in the image plane and UV data were combined
183: over longer intervals only if the snapshots were indistinguishable. In
184: practice, there is little effect at lower resolutions from the moving
185: ejecta - mainly a blurring of the moving component, and an increase in
186: residual artifacts, since self-calibration does not work perfectly if
187: the structure is varying. (See e.g., Fig.~\ref{fig-2}B, where the
188: extension of the SW component away from the core has occurred during
189: the `exposure time' of 5.1~hr.)
190: 
191: For the high-resolution images (e.g. in Fig.~\ref{fig-5}), the problem
192: of reliable imaging is more severe. In addition to snapshot imaging, a
193: careful search for time-variable or moving structure was done by
194: examining and model fitting to the closure phases over intervals
195: as short as 5~mins. Simultaneous flux measurement (from the VLA or GBI, when
196: available) was used to constrain the total flux while modelling the
197: AU-scale structure.  No evidence for rapidly moving or variable
198: structures was found. We can rule out discrete, moving ejecta on AU
199: scales - none of the snapshots or closure phases showed, e.g. double
200: sources or differently oriented structures.  We cannot at this time
201: rule out fast-moving features with low contrast, that would be blurred
202: out. The simulations can be used to place limits on the artifacts due
203: to the technique, to be discussed in a future paper.  We will continue
204: to pursue evidence for AU-scale motion in new data, as well as old data
205: after further refinement in software to model moving, time-variable
206: sources.
207: 
208: \section{Images During Two Flares}
209: 
210: These target-of-opportunity observations are subject to the constraints
211: of weather, scheduling, and source variability. For clarity, we present
212: individual images for two episodes of flare activity, and merge the
213: discussion of the common elements thereafter.
214: 
215: \subsection{1997 October}
216: 
217: The radio and X-ray behavior of GRS~1915+105 around this flare is
218: summarized in Fig.~\ref{fig-1}.  The AU-scale jet is imaged in the
219: nucleus (Fig~\ref{fig-2}A) on 1997 October 23, six days before a major
220: flare.  Images on October 31, centered at MJD~50752.02,
221: (Fig.~\ref{fig-2}B \& C) show the central core with a bright component
222: 47.5$\pm$0.3~mas away on the SE (approaching) side.  The core has a
223: flux density of $\sim$20~mJy with flat spectrum, while the SE ejected
224: component has S$_{\nu} \propto ~\nu^{-0.5}$.
225:  The position of the SE component corresponds to a separation rate of
226: 0.90$\pm$0.05~mas~hr$^{-1}$ in 53.3$\pm$2.4~hrs since the estimated
227: start of the flare at MJD=50749.8. A NW (receding) component is
228: marginally detected, at 17.4$\pm$0.3mas separation, consistent with it
229: being the counter-ejection, with velocity 0.33$\pm$0.02~mas~hr$^{-1}$
230: and flux density only $\sim$10\% of the SE component.
231: 
232: The time-lapse images, (Fig.~\ref{fig-3}), with interval of 2.5~hrs,
233: show a separation change of the SE ejection from the core of
234: 2.3$\pm$0.2~mas, or 0.92$\pm$0.08 mas~hr$^{-1}$.  Our start time for
235: the flare is derived by extrapolating the slopes of the X-ray and radio
236: data (see Fig.~\ref{fig-1}). With this start time the velocity derived
237: from the position of the ejecta after 53.3~hours agrees with the
238: velocity from the 2.5~hr time lapse images.  \citet{fen99} assume a
239: considerably different start at 50750.5 for the same flare.  However,
240: their velocities for the ejecta, derived from images several days after
241: the flare, at 500 to 5000~AU scales, agree with ours within the errors,
242: and imply ballistic motion from a few to 1000's of AU, certainly beyond
243: 500~AU.
244: 
245: \placefigure{fig-1}
246: 
247: \placefigure{fig-2}
248: 
249: \placefigure{fig-3}          
250: 
251: Astrometry (see Fig.~\ref{fig-7}) before and after the flare locates
252: the nucleus within 1.5~mas of the position we have measured for over
253: two years, after allowing for secular parallax. The 2~cm flux density
254: varies with the 30~min period seen in XTE dips, starting at 50750.5
255: \citep{fen99}. We show that the variable radio emission is from the
256: AU-scale jet, see  Fig.~\ref{fig-9}; (see also
257: \citet{mir98,poo97,fen97}).  We conclude that the nuclear jet that we
258: image has re-established itself within $\sim$18~hrs of the start of a
259: major outburst, if indeed it was disrupted at all.  The position angle
260: of the AU-scale jet in Fig.~\ref{fig-2}A is 157$\pm$2$^{\circ}$ at
261: 2~cm, whereas the large-scale ejecta show 133$\pm$3$^{\circ}$ at 13~cm,
262: in Fig.~\ref{fig-2}B; and 143$\pm$4$^{\circ}$ at 3.6~cm in
263: Fig.~\ref{fig-2}C.
264: 
265:   
266: \subsection{1998 April-May}
267: 
268: The radio and X-ray behavior of GRS~1915+105 around this flare is
269: summarized in Fig.~\ref{fig-4}.  As in the 1997 October event, there is
270: a jet in the nucleus during the plateau state, (see Fig.~\ref{fig-5}D),
271:  two days before the flare which started on MJD$\sim$50916. Figures 5E
272: and 5F also show compact jets 2 days after the start of a second flare
273: on MJD$\sim$50932.  The spectrum of the jet is essentially flat from
274: 2~cm to 0.7~cm.  After two intervening flares, on MJD~50935.5, the nucleus
275: is still within 1.0~mas of the expected position, after allowing for
276: secular parallax.  The position angles of the AU-scale jet are as
277: follows:  D:~155$\pm$2$^{\circ}$, at 2~cm; E:~154$\pm$4$^{\circ}$ again
278: at 2~cm; F:~145$\pm$6$^{\circ}$, at 0.7~cm for the overall structure,
279: although the innermost contours are rotated to 168$\pm$5$^{\circ}$,
280: perhaps indicating a bent jet.
281: 
282: The phase-referenced image at 2~cm (Fig.~\ref{fig-6}), heavily tapered
283: to 100~AU beam size, shows the SE ejection at $\sim$650~AU from the
284: core.  The counter-ejection was not detected, with flux $\le$5.5\% of
285: the approaching component.  The separation from the core of the SE
286: component corresponds to a motion of 57.5$\pm$0.5~mas in 67$\pm$7~hrs,
287: or 0.87$\pm$0.10~mas~hr$^{-1}$. This is consistent with the velocity of
288: 0.93$\pm$0.07~mas~hr$^{-1}$, seen in the red and blue snapshots in
289: Fig.~\ref{fig-6}, which are 4.5~hrs apart.  The position angle of the
290: SE component is 148$\pm$4$^{\circ}$ from the core.
291: 
292: \placefigure{fig-4}
293: 
294: \placefigure{fig-5}
295: 
296: \placefigure{fig-6}              
297: 
298: \section{Astrometry \label{sec-astro}}
299: 
300: Astrometry is vital to register moving components from one epoch to the
301: next. The present astrometric accuracy is achieved with phase residuals
302: to the VLBA correlator model, and no special software.  The correlator
303: model used in these observations was in error by $\sim$30~mas for Earth
304: nutation, reduced by a factor of 0.1 due to the $\sim$5$^{\circ}$
305: distance to the calibrator.  We thus expect a maximum error in the
306: absolute position determination of 3~mas for a given epoch.  In fact,
307: when we compare our positions for secondary calibrators in these
308: observations to those obtained independently for the same calibrators
309: in the same reference frame by USNO, we find agreement within 1.5~mas,
310: (T.M. Eubanks, private communication).  The effects of nutation error
311: rate on proper motion are further reduced by the time-differencing, and
312: are $<$0.3~mas~yr$^{-1}$. Our errors in the secular parallax are
313: $\sim$1.5~mas~yr$^{-1}$, dominated by tropospheric `seeing', i.e.,
314: residual phase errors from the calibrator. This can be seen in
315: Fig.~\ref{fig-7} as the short-term scatter of positions over a few
316: hours on  a given date.
317: 
318: \placefigure{fig-7}
319: 
320: The J2000 position of the core is 19$^h$15$^m$11$^s$.54938$\pm$.00007,
321: 10$^{\circ}$56'44''.7585$\pm$.001, on 1998 May 02,in the ICRF reference
322: frame \citep{ma98}. The secular parallax is -5.4$\pm$1~mas~yr$^{-1}$
323: and -2.3$\pm$1~mas~yr$^{-1}$ in RA and Dec respectively. This level of
324: astrometric accuracy, though it can be improved by further analysis, is
325: adequate to demonstrate that:
326: 
327: 
328: (i) The identification of the stationary core and moving ejecta is
329: unambiguous. We note that the 1997 October 31 data at 3.6~cm are shifted
330: from the pre-flare position by $\sim$2~mas, probably due to the complex and
331: evolving structure during the flare. However, this error is not enough
332: to cause mis-identification of core with ejecta separated by 47.5~mas.
333: 
334: (ii) GRS1915 shows the galatic rotation expected from an object about
335: 12~kpc distant, \citep{dha00}, at the position
336: $l$~=~45$\rlap.{^\circ}$37, $b$~=~-0$\rlap.{^\circ}$22.  We find a
337: proper motion of 5.8$\pm$1.5~mas~yr$^{-1}$, which we ascribe to the
338: secular parallax of the core, within errors of $\pm$75~km~s$^{-1}$ in
339: the plane of the Galaxy (all errors are $\pm$3$\sigma$).  A
340: model-independent distance may perhaps be determined in the future by
341: measuring the annual trigonometric parallax of $\sim \pm$80
342: $\mu$arcsec, using closer calibrator sources, as attempted
343:  in the case of Sgr~A$^*$ \citep{reid99}.
344: 
345: (iii) The core is stationary on the sky to $\pm$1.5~mas once the
346: secular parallax is accounted for. The systematic drift perpendicular
347: to the Galactic plane over two years is consistent with
348: 0$\pm$50~km~s$^{-1}$.
349: 
350: Combining errors parallel and perpendicular to the galactic plane, we
351: can place an upper limit of $<$100~km~s$^{-1}$ for the velocity of
352: GRS~1915+105 on the sky. Note that velocity along the line of sight is
353: unconstrained by these observations, but could become evident in the
354: Doppler shift of periodic features, if any are detected in the future.
355: 
356: \section{Scattering in the ISM Towards GRS~1915+105. \label{sec-scat}}
357: 
358: GRS~1915+105 lies near the galactic plane, along a line of sight
359: tangential to a spiral arm and intercepting a large column of gas, both
360: neutral and ionized. The high dispersion measure of distant pulsars in
361: this direction \citep{tay93} leads us to expect considerable angular
362: broadening due to the inhomogeneous ionized interstellar medium, as is
363: indeed the case.
364: 
365: Observation over several octaves can separate the intrinsic structure
366: from the scatter-broadening, as shown in Fig.~\ref{fig-8}, and short
367: wavelengths (2~cm and 0.7~cm) were used to give an unscattered view into
368: the nucleus where the jet originates. The minor axis size scales as
369: $\lambda^2$, and we measure a  scattering size of 1.9$\pm$0.1mas at 
370: 3.6~cm, (135~mas at 30~cm=1~GHz).
371: Interestingly, the scattering seems to vary dramatically on scales of
372: $\sim$50~pc, seen by comparing our result with the 8.4~mas at 30~cm,
373: (measured by \citet{kem88}) for the OH maser in OH45.47+0.13, which
374: lies beyond the tangent point, as is the case for GRS~1915+105, and
375: only 10' away on the sky. A fraction of this increase by a factor of 16
376: in scattering  could be due to a putative
377: ionized circumstellar cocoon around the X-ray binary, as suggested by
378: recent infrared spectroscopy with the VLT, \citep{mart20}.
379: 
380: \placefigure{fig-8}
381: 
382: \section{Two Distinct Radio Emission Phenomena?}
383: 
384: We discuss in this section the evidence for two distinct radio emission
385: states, described most simply as attached to, and detached from the
386: nucleus.  They correspond to: \\
387: (a) A smooth, flat spectrum, continuous nuclear jet at AU scale, with
388: quiescent or oscillating radio flux; and \\
389: (b) Discrete ejecta, separating from the nucleus superluminally at
390: $\geq$1.3~c, during steep spectrum flares. 
391: 
392: The major outbursts appear qualitatively different from the 30~min
393: variations, and seem to have a different trigger.  For {\bf flares},
394: the onset of radio emission seems to correlate with the {\bf rise} of
395: 2-12~keV X-rays (see Figs.~\ref{fig-1}~\&~\ref{fig-4}). On the other
396: hand, the AU-scale (30~min) radio/IR oscillations are seen when
397: deep XTE {\bf dips} of $\le$5000 counts are present.  On the short
398: timescale of 10's of minutes, isolated (single) events have not been
399: observed to our knowledge - they are always part of a pulse-train,
400: accompanied by X-ray dips.  On the other hand, no periodicity or
401: regularity has been observed in the timing of the large flares, which
402: are always isolated events. More details of the two phenomena follow.
403: 
404: \subsection{The AU-Scale Jet in the Nucleus \label{sec-auscale}}
405: 
406: Our images are consistent with a conventional model of conical
407: expanding jet, (\citet{hj88,fal99}, see also Sec.\ref{sec-discussion})
408: i.e.,  incoherent synchrotron emission in an optically thick region of
409: size $\sim$10$\lambda_{\rm cm}$~AU,
410:  inclined at $\sim$66-70$^\circ$ to the line of sight. The brightness
411: temperature of the jet is  T$_{\rm B} \geq$10$^{9}$K at all
412: wavelengths. (T$_{\rm B}$=10$^{9}$K at 0.7~cm, where
413: the minor axis is marginally resolved because the the scattering is
414: least.  At 2~cm and longer wavelengths, we measure the same brightness 
415: temperature, but it is a lower limit because 
416: the scattering size dominates the beam resolution, and the intrinsic
417: width could be smaller.)
418: 
419: The images clearly identify the AU-scale jet with the {\bf plateau}
420: state and its quasi-periodic flux variations.  The radio light curve of
421: the jet is best described as a smoothed response, with time-constant of
422: $\sim$30~min, to the injection of relativistic plasma, presumably
423: generated during the X-ray dips, (Figs.~\ref{fig-9} \& ~\ref{fig-10}).
424: The injection interval is variable, 12-60~mins have been observed on
425: various occasions. The time delay between the shorter and longer radio
426: wavelengths is also variable, from 4~min to $\sim$30~min. Possible
427: causes include changes in the size, expansion rate, or optical depth of
428: the jet, or variable dynamics of accretion and jet formation. Variable
429: orientation appears not to be a major contributing factor.
430: 
431: \placefigure{fig-9}
432: 
433: \placefigure{fig-10}    
434: 
435: During a radio flare, one expectation might be that the radio core
436: would fade as the inner accretion disk was ejected and/or swallowed.
437: However, we find that the nuclear jet re-establishes itself within 18
438: hrs of the start of a major outburst, if it disappears at all.
439: 
440: 
441: The AU-scale structure is smooth, with little evidence for discrete,
442: moving ejecta. This can be currently ascribed to the difficulty of
443: Fourier synthesis imaging of moving, time-variable sources, rather than
444: to a true lack of relativistic flow. A static structure is unlikely,
445: given that we see superluminally moving ejecta on larger scales  {\em
446: during the same event.} A more likely explanation is a steady state
447: with fast, continuous flow and adiabatic losses which cause the jet to
448: fade rapidly with distance.  The travel-time along  the length of the
449: jet ($\sim$10~AU~hr$^{-1}$) at relativistic speed is comparable to the
450: adiabatic loss timescale ($\sim$30~min) over which new IR/radio emission
451: fades away after each X-ray injection event. Analogous compact jets may
452: be found in Cygnus X-1 and Cygnus X-3, as suggested by \cite{fen20}
453: from the flat radio-millimeter spectra of the cores in these X-ray
454: binaries.
455: 
456: Good evidence exists that the synchrotron spectrum extends up to 2$\mu$
457: in the IR \citep{eik98,mir98,fen98}. The IR light curves are quite
458: similar to the radio, believed to be caused by rapid adiabatic
459: expansion rather than synchrotron losses, as suggested by \citet{fen98}
460: and \citet{mir98}.  Maximal internal energy \citep{fal99} implies
461: expansion at $\sim$~0.6~c. The VLA measured expansion rate
462: \citep{rod95} is about 0.2~c on large scales.  On AU scales, we see an
463: elongated structure  with length about four times the width, which is
464: consistent with relativistic flow along the major axis and lateral
465: expansion at 0.2~c.
466: 
467: Considering next the position angles of the AU-scale images, no  significant
468: time variations are seen.  The
469: AU-scale position angle is stable within $\sim$5$^{\circ}$ for $\sim$2
470: years.  The 500~AU ejecta have the same P.A. for our two measurements
471: in 1997 and 1998. The large ejection reported in Mirabel \& Rodr\'\i
472: guez (1994) was along $\sim$150$^{\circ}$, about 7$^{\circ}$ CCW.  We
473: measure  a rotation of about 12$\pm$5$^{\circ}$ from the few-AU to
474: the  500~AU scale, though both seem to be stable in time.  This
475: rotation of position angle with size scale appears real. We suspect it
476: could be due to opacity effects in a conical jet, since the high
477: resolution is obtained at shorter wavelengths - however we do not have
478: a clear explanation for it.
479: 
480: In summary, a synchrotron jet (see also Discussion) accounts in a
481: unified, consistent  way for the following phenomena:
482: \begin{itemize}
483: \item The flat radio spectrum and high brightness temperature.
484: \item The elongation of the core along the axis of arcsecond-scale
485: 	superluminal ejecta, ($\sim$155$^{\circ}$). 
486: \item The peak emission is
487: 	progressively delayed at longer wavelengths, as they emanate
488: 	from further along the expanding jet;
489: \item There is progressively less variation of the flux at longer
490:  	 wavelengths, due to convolution over a larger region;
491: \item The decay time of flux variations is consistent with the  
492: 	travel time of relativistic plasma along the jet.
493: \end{itemize}
494: 
495: 
496: However, we must point out that the ratio of integrated flux density
497: of  approaching to receding parts  of the continuous jet are only
498: 1.15$\pm$0.04, 1.20$\pm$0.05, and  1.10$\pm$0.08, for the three images
499: of Fig.~\ref{fig-5}A, B, C respectively. (Assuming we are seeing
500: both sides of the jet; see the next paragraph for a possible exception).  
501: Assuming an inclination
502: $\theta$=70$^{\circ}$, a flat spectral index ($\alpha$=0), and a
503: continuous jet (k=2), we obtain a mildly relativistic speed of
504: $\beta_{\rm F}$~=~0.1 for the AU-scale jet, from the flux ratio (see,
505: e.g., \citet{bod95}, eqn.4, for a stationary jet, with pattern speed
506: $\beta_{\rm S}$~=~0, and bulk flow at speed $\beta_{\rm F}$):
507: 
508: \begin{equation}
509: \left(  {\rm S}_{\rm approaching} 
510: \over   {\rm S}_{\rm receding}  \right)  = 
511: \left(1 +\beta_{\rm F} ~{\rm cos}\theta \over  1 - \beta_{\rm F} 
512: ~{\rm cos}\theta   \right) ^{k-\alpha}
513: \end{equation} 
514: 
515: This result of 0.1c from the jet/counter-jet asymmetry is puzzling,
516: in light of the expected relativistic flow in the jet model.
517: We offer three possible explanations, none completely 
518: satisfactory: 
519: 
520: \paragraph{Slow Mini-Jet?}
521:  The jet/counter-jet ratio on AU scales is really due to slow jet 
522: velocity of 0.1c. This contradicts
523: some of the data, and we do not favour it. First,
524: the measured speed of 0.9c at 500~AU
525: implies acceleration of ejecta from 0.1c to 0.9c between 20~AU
526: and 500~AU,  very far from 
527: the black hole ($>$10$^8$R$_{\rm s})$. This seems very unlikely. 
528: Secondly, as discussed  above,
529: the expansion speed of the synchrotron cloud must be fast, in at least 
530: one dimension, to account for the rapid fading of emission from adiabatic loss.
531: (synchrotron radiative decay would take years, not 30~min as observed). 
532: The size and elongation of the jet, $\sim$20~AU, implies the speed
533: is $\sim$c, in at least one direction. For jet speed of 0.1~c, the
534: isotropic expansion would dominate, so the AU-scale images would 
535: be circular, and endure for over 10~hrs.
536: 
537: \paragraph{Fast Mini-Jet?}
538: On the other hand, \footnote{We acknowledge discussion with Peter
539: Goldreich.} the brightness temperature of
540: an optically thick synchrotron source should be about 10$^{11}$~K, 
541: (predicted by the Falcke model as well) for a bulk Doppler factor
542: $\sim$1, whereas we measure 10$^{9}$~K at 0.7~cm. 
543: If we assume a fast mini-jet, with bulk $\beta_{\rm F} \geq$0.99,
544: then, because of the large inclination of 70$^{\circ}$, 
545: the (approaching) doppler factor is $\leq$0.2 and  the
546: observed brightness temperature is reduced.
547: This argues for a highly relativistic mini-jet on scales of few
548: AU, that slows down as it reaches 1000 AU. Unfortunately, there 
549: is no observational evidence for deceleration, such as brightening due 
550: to hitting a density enhancement, at that radius. The jet/counter jet
551: ratio would require explanation as in the next paragraph.
552: 
553: \paragraph{Hidden Counter-Jet?} 
554: We can postulate that the (receding) counter-jet is hidden by 
555: free-free absorption in the inclined disk, and only the
556: approaching mini-jet is observed, with nearly symmetric brightness
557: profile. This absorption 
558:  geometry has been clearly 
559: demonstrated in 3C84, \citep{wal00}. In GRS~1915+105, the 
560: absorption must occur
561: within 1000~AU of the core, because the counter-ejection is
562: seen outside that radius, \citep{fen99,mir94,rod99} 
563: We detect no counter-ejection at $<$500~AU in 1998 May 02 
564: ($\tau >$1.5 at 2~cm), and
565: on 1997 October 31 there is a weak counter-ejection at a flux reduced from 
566: the expected beaming model, \citep{mir94} indicating $\tau \sim$0.5 at 
567: 3.6~cm, (sec.3.1 and 3.2). Different opacity for each ejection could be
568: due to variable ionizing illumination from the disk, or variable
569: mass-loss, for both of which ample evidence exists.
570: Thus free-free opacity hiding the counter-jet on AU-scales, and hiding 
571: the counter-ejecta 
572: within a few 100~AU of the core, is consistent with the images. 
573: Assuming $\tau_{\rm free-free} \sim$1,  L$\sim$100~AU, we derive an
574: electron density n$_{\rm e}$=5~10$^{5}$~cm$^{-3}$, emission measure
575: EM$\sim$2~10$^8$~cm$^{-6}$~pc,   and 
576: column~density~$\sim$10$^{21}$~cm$^{-2}$, for an electron temperature 
577: T$_e$=10$^4$~K.
578: 
579: Furthermore, we might explain both the enhanced scattering
580: (Section-\ref{sec-scat}) and the
581: free-free opacity as signatures of the ionized envelope 
582: detected by \citet{mart20}.
583: However, the scattering is measured against the core, 
584: and the absorption is invoked against the counter jet, \em not 
585: \rm the core, so special geometry is needed. We leave this as a speculation.
586: The test would be to actually
587: observe the counter jet and measure its spectrum which should be strongly
588: rising with frequency, since the opacity is exponentially less at shorter
589: wavelength, \citep{wal00}.  Sensitive, technically difficult 
590: observations would be required at 0.7~cm or 0.3~cm.
591: 
592: \subsection{The Ejecta at $\sim$500~AU \label{sec-largescale}}
593: 
594: Images during flares show rapidly decaying superluminal ejecta.
595: Snapshots taken a few hours apart allow us to measure, for the
596: approaching ejection, a velocity of 1.28$\pm$0.07~c~$({\rm D}/10~{\rm
597: kpc})$ at 500 to 600~AU away from the nucleus,  consistent with the
598: velocity of \citet{fen99} at 5000~AU  for  the same event.  This 
599: implies ballistic motion from
600: 500~AU to 5000~AU.  Phase-referencing clearly distinguishes the
601: stationary core from the moving ejecta, unlike previous observations.
602: Next, back-extrapolation of the position of the ejecta (over 2-3
603: days), using the velocity measured over a few hours, gives a position
604: coincident with the core at the  start time of the flares. Thus, on
605: average, this is consistent with ballistic motion from the few-AU scale
606: outwards.  The lateral expansion rate of the ejecta is about 0.14-0.2c,
607: and is not well constrained in the VLBA images due to the lack of
608: sensitivity to extended structure. The value of 0.2~c was measured in
609: the VLA images on larger scales, \citep{rod99}.
610: 
611: \section{Discussion \label{sec-discussion}}
612: 
613: In this section we discuss a possible thermal origin for the emission,
614: and decide in favor of a synchrotron jet. We also find good consistency
615: between the synchrotron model of \citet{fal99} and the disk instability
616: model of \citet{bel97a,bel97b}  for the source of the relativistic
617: plasma in the baby jets. The large ejecta require $\sim$10 times more
618: mass, and may be qualitatively different from the AU-scale jets as well.
619: 
620: \subsection{Thermal Jet}
621: 
622: We rule out a thermal origin for the emission in favor of a synchrotron
623: jet by the following argument.  If we assume that the radio emission is
624: due to free-free (thermal) radiation from gas with an electron
625: temperature of $\sim$10$^{10}$K, and follow the formulation of Reynolds
626: (1986) for the mass loss rate in a thermal jet, we roughly estimate
627: 
628: \begin{equation} 
629: \left( \dot{\rm M} \over {\rm M}_{\odot}~{\rm yr}^{-1} \right) 
630: = 4 \times 10^{-4} 
631: \left( {\rm v} \over 10^4~{\rm km~s}^{-1} \right) 
632: \end{equation} 
633: where v is the velocity of the outflowing gas. If this velocity were of
634: the order of the speed of sound in a gas of 10$^{10}$K,
635: v=10$^4$~km~s$^{-1}$, then a mass loss rate of
636: $\sim$4x10$^{-4}$M$_{\odot}$~yr$^{-1}$, a mechanical power of
637: 3$\times$10$^6~{\rm L}_{\odot}$, and an X-ray luminosity of the jet of
638: $\sim10^7~{\rm L}_{\odot}$ are derived. These powers are similar to the
639: Eddington luminosity of a $\sim 100~{\rm M}_{\odot}$ black hole, and
640: appear excessive for what is known of GRS~1915+105.
641: 
642: Finally, the QPO's observed in the radio and X-rays suggest that the
643: material is travelling and expanding at relativistic speeds.  This
644: would make the estimated mechanical power and X-ray luminosity of the
645: jet increase by a further large factor. We hence favour the synchrotron
646: emission mechanism for the AU-scale jet.
647: 
648: \subsection{Conical Synchrotron Jet}
649: 
650: A jet model with minimum free parameters is discussed in
651: \citet{fal99}.
652:  Considering the coupled disk-jet system, subject to the assumption of
653: total equipartition of energy between disk, jet kinetic energy, jet
654: internal energy (thermal energy of proton/electron plasma) and magnetic
655: field, expanding freely into vacuum, they derive a relativistic jet of
656: terminal $\beta ~ \sim$0.96. (A higher velocity would require
657: acceleration by a mechanism other than pressure gradient.) The only
658: free parameter is the measured disk X-ray luminosity.  In the case of
659: GRS~1915+105, with X-ray luminosity of 10$^{39}$~erg~s$^{-1}$, (0.3 to
660: 3 times this have been observed) the following properties are expected
661: for the jet at 2~cm.  The mechanical power of the jet is of the order of
662: Q$_{\rm jet}$=10$^{39}$~erg~s$^{-1}$, which indicates a mass accretion
663: rate of 10$^{19}$~g~s$^{-1}$, assuming a conversion efficiency of 10\%
664: from gravitational energy. Note that this model assumes a maximally
665: efficient radio jet for a given disk luminosity, i.e. a radio-loud
666: jet.  This  implies, as pointed out in \citet{fal99}, that
667: 10$^{39}$~erg~s$^{-1}$ is a {\bf lower limit} to the mechanical power
668: of the jet. The model then predicts a (steady state) flux density of
669: about 20~mJy, and jet length of about 0.4~mas, with expected wavelength
670: dependence of jet size
671:  $\Theta~\propto \nu^{-1}$, and spectrum S$_{\nu}~\propto \nu^{0.2}$.  Other
672: derived parameters are a magnetic field  B of 0.4~gauss at 10~AU from
673: the black hole, and an electron Lorentz factor $\gamma _{\rm e}$ in the
674: range of 200-800.
675: 
676: Thus, in the steady state, the model yields roughly the correct size,
677: spectrum, and flux of the  AU-scale nuclear jet. Time variable jet
678: length and flux density, can perhaps be accomodated as changes in mass
679: accretion rate. Allowing for the complex time-variability in this
680: system, we judge the agreement with our observations to be
681: satisfactory.  Next, given that the jet is already maximally radio
682: loud, it seems to us necessary to invoke an additional mechanism, to
683: explain the transient release of excess synchrotron emission in the
684: flares, with spectrum S$_{\nu}~\propto \nu^{-0.6}$. The dependence of flux
685: density on orientation, (as in AGN unification schemes), does not apply
686: here since we observe the axis position angle does not change
687: sufficiently from small to large scales.
688: 
689: \subsection{Disk Instability as the Source of Relativistic Plasma}
690: 
691: GRS~1915+105 shows extremely complex behavior in X-rays, see e.g.
692: Fig.~\ref{fig-10}, and \citet{mar99}. The  X-ray variability is
693: explained by \citep{bel97a}, and \citep{bel97b}, in terms of the
694: evacuation and refilling of the inner hot accretion disk due to
695: thermal-viscous instability. They attempt to reduce the complexity of
696: the light curve to essentially one parameter, the radius of the missing
697: inner disk.  The light curve, and hardness ratio variations, are
698: decomposed into a superposition of an inner, 2.2~keV disk component, of
699: variable radius ($\sim$20-100~km), along with an outer constant disk of
700: radius over 300~km and temperature 0.5~keV. The instability
701: (emptying/refilling of the disk) occurs in the region where radiation
702: pressure dominates the gas pressure. The time required to refill the
703: disk with X-ray emitting gas is set by the viscous/thermal  timescale.
704: Assuming a viscocity parameter $\alpha \sim$0.01, they estimate this
705: timescale as
706: 
707: \begin{equation} 
708: t_{\rm viscous} = 3 \alpha^{-1} 
709: 	\left( {\rm M}_{\rm BH} \over 10 {\rm ~M_{\odot}}  \right)
710: 	\left( {\rm R} \over 10^{7}{\rm cm}  \right) ^{7/2}
711: 	\left( {\rm \dot{M}} \over 10^{18}{\rm ~g~s}^{-1}  \right)^{-2} ~
712:  {\rm seconds}
713: \end{equation}       
714: 
715: For the 30~min dips in the X-rays (the timescale of the radio
716: oscillations) the radius of the missing disk is of the order of 180~km.
717: From spectral fitting to the disk/black body, they also derive a mass
718: accretion rate of $\sim$10$^{18}$ to $\sim$10$^{20}$g~s$^{-1}$,
719: depending on whether the black hole is maximally rotating or not.  The
720: highest accretion rates are  inversely correlated with the X-ray
721: luminosity, which they take as evidence that the majority of the mass
722: may be eliminated from view by an ADAF-type flow into the black hole.
723: 
724: From our AU-scale images, we find good 
725: agreement between the mass of synchrotron-emitting plasma in the AU-scale
726: `mini-jet',  and the mass loss from the missing radius of the inner
727: accretion disk. The timescales and radio/soft X-ray correlations are
728: consistent with the disk instability model. In contrast, the large 
729: flares  eject an order of magnitude more mass,  and the recovery 
730: timescale of $\sim$18~hrs indicates a radius of $\sim$600~km for 
731: the missing disk. This is so large as to lie outside the radiation-
732: pressure dominated region, so the instability (oscillation) would not work.
733: Indeed, the large-scale events never repeat, unlike the AU-scale 
734: 30~min cycles.  Once again, we take this as evidence of a qualitative 
735: difference between the AU-scale emission and the extreme
736: ejections during flares. Thus the Belloni model works for mini-jets, 
737: but not for large flares.
738: 
739:                                                          
740: \section{Conclusions}
741: 
742: From VLBA images of GRS~1915+105 in 1997 and 1998 we conclude the
743: following:
744: 
745: \begin{enumerate}
746: 
747: \item  The core is always resolved as a compact collimated
748: jet of $\sim$10$\lambda$$_{\rm cm}$~AU, that has a peak brightness
749: temperature  T$_{\rm B}\geq$10$^9$K.
750: 
751: 
752: \item  The compact jet is observed in the {\bf plateau} state that can
753: last from days up to several months, which consists of a steady flat
754: spectrum at radio wavelengths, persistently bright hard X-ray
755: (20-100~keV) flux, and a faint flux in the softer X-rays (2-12~keV)
756: (see Figs.~\ref{fig-1} \& \ref{fig-2} and
757:      Figs.~\ref{fig-4} \& \ref{fig-5}).
758: 
759: \item  The compact jet is observed a few hours before and a few hours
760: after major flare/ejection events (Figs.~\ref{fig-2} \& \ref{fig-5}).
761: The nuclear jet is re-established in $\leq$18 hours after a major
762: outburst. In minor outbursts the jet flux varies in $\sim$30~min
763: (Figs.~\ref{fig-9} \& \ref{fig-10}), with the same period seen in the
764: infrared and in the 2-12~KeV X-rays, (e.g.  Mirabel et al. 1998; Fender
765: et al. 1998; Eikenberry et al. 1998).
766: 
767: \item  Although we could not measure the proper motion of matter in the
768: quasi-steady compact jet, its length of $\sim$20~AU (e.g.
769: Fig.~\ref{fig-5}D \& E) is comparable to the distance that would
770: be traversed at relativistic speed during the
771: time-scale of the oscillations ($\sim$30~min).
772: 
773: \item  The compact jet is optically thick synchrotron emission. This is
774: suggested by the high degree of collimation, the high brightness
775: temperature, the flat radio spectrum, and the inferred relativistic
776: bulk motions. The mass outflow assuming a synchrotron jet is
777: $\geq$10$^{-8}$M$_{\odot}$~yr$^{-1}$, consistent with the accretion
778: rate estimated  from the X-ray luminosity of the black hole binary.
779: Thermal processes alone cannot account for the radio emission because
780: they are relatively inefficient radiative mechanisms that would require
781: an accretion rate {\bf greater} than 10$^{-5}$~M$_{\odot}$~yr$^{-1}$,
782: which is unreasonably large.
783: 
784: \item  The large mass ejection events with measured superluminal
785: velocities take place at the time of abrupt changes in the X-ray state
786: of the source.  Within $\pm$5$^{\circ}$ the same position angles were
787: observed for major ejections in 1994-1996 \citep{rod99}, and 1997-1998
788: (this paper).  The change from radio-quiet to
789: radio-loud states cannot be caused by change in  orientation, 
790: since the position angle of the ejecta does not change
791: sufficiently from small to large scales. The axis of the ejecta at
792: large scales ($\geq$500~AU) appear to be rotated clockwise by
793: $\le$12$^{\circ}$ relative to the jet axis at $\leq$100~AU, both
794: measured at the same time.  The flux variations must be intrinsic.
795: 
796: \item  Time-lapse images allow us to detect the motions of the
797: large-scale ejecta within a few hours, with apparent speeds of
798: $\geq$1.3~c, consistent with previous observations (Mirabel \& Rodr\'\i
799: guez, 1999).  Knowing the time for the onset of the radio outburst, it
800: is inferred that the acceleration to terminal speeds must take place
801: within 500~AU of the black hole.
802: 
803: \item  By astrometry over two years relative to an extragalactic
804: reference, we locate the black hole binary within $\sim$1.5~mas, and
805: follow its secular parallax due to Galactic rotation. The black hole is
806: stationary on the sky to $\le$100~km~s$^{-1}$,  once the secular
807: parallax is accounted for.
808: 
809: \item  Some accreting black holes of stellar mass (e.g. Cyg~X-1,
810: 1E~1740-2942, GRS~1758-258, GX~339-4) and the supermassive black hole
811: Sgr~A$^*$, show compact radio cores with a steady flat-spectrum plateau
812: state, in common with the core of GRS~1915+105. It is possible that
813: high resolution images of the other systems will also reveal compact
814: jets. So far GRS~1915+105 is the only galactic source where both,
815: AU-sized jets and large-scale superluminal jets have been unambigously
816: observed.
817: 
818: \end{enumerate}
819: 
820: \acknowledgements
821: 
822: We thank the ASM/RXTE team for providing us with quick-look results.
823: LFR acknowledges support from CONACyT, M\'exico. IFM acknowledges
824: support from CONICET, Argentina.
825: 
826: \clearpage \newpage
827: 
828: \begin{thebibliography}{}
829: 
830: \bibitem[Belloni et al. (1997a)]{bel97a}  Belloni, T., Mendez, M.,
831: King, A. R., van der Klis, M., \& Van Paradijs, J.  1997, ApJ, 488,
832: L109
833: 
834: \bibitem[Belloni et al. (1997b)]{bel97b} Belloni, T., Mendez, M., King,
835: A. R., van der Klis, M., \& van Paradijs, J.  1997, ApJ, 479, L145
836: 
837: \bibitem[Blandford \& Begelman (1999)]{bla99}
838:  Blandford, R. D. \& Begelman, M. C.  1999, MNRAS, 303, L1
839: 
840: \bibitem[Bodo \& Ghisellini (1995)]{bod95} Bodo, G., \& Ghisellini, G.,
841: 1995, ApJ, 441, L69
842: 
843: \bibitem[Castro-Tirado et al. (1994)]{cas94} Castro-Tirado, A. J.,
844: Brandt, S., Lund, N., Lapshov, I., Sunyaev, R. A., Shlyapnikov, A. A.,
845: Guziy, S., \& Pavlenko, E. P. 1994, ApJS, 92, 469
846: 
847: \bibitem[Dhawan et al. (2000)]{dha00} Dhawan, V., Goss, W. M., \&
848: Rodr\'\i guez, L.F. 2000, ApJ, 540, in press.
849: 
850: \bibitem[Eikenberry et al. (1998)]{eik98}Eikenberry, S. S., Matthews,
851: K., Morgan, E. H., Remillard, R. A., Nelson, R. W.  1998, ApJ, 494, L61
852: 
853: \bibitem[Falcke \& Biermann (1999)]{fal99}Falcke, H. \& Biermann, P.
854: L.  1999, A\&A, 342, 49
855: 
856: \bibitem[Fender et al. (1999)]{fen99} Fender, R.P., Garrington, S. T.,
857: McKay, D. J., Muxlow, T. W. B., Pooley, G. G., Spencer, R. E.,
858: Stirling, A. M., \& Waltman, E. B.  1999, MNRAS, 304, 865
859: 
860: \bibitem[Fender et al. (1997)]{fen97}Fender, R. P., Pooley, G. G.,
861: Brocksopp, C., \& Newell, S. J.  1997, MNRAS, 290, L65
862: 
863: \bibitem[Fender et al. (1998)]{fen98} Fender, R.P., \& Pooley, G.G.
864: 1998, MNRAS, 300, 573
865: 
866: \bibitem[Fender et al. (2000)]{fen20} Fender, R.P., \& Pooley, G.G.,
867: Durouchoux, P., Tilanus, R.P.J. \& Brocksopp, C.  2000, MNRAS, in press
868: 
869: \bibitem[Foster et al. (1996)]{fos96} Foster, R. S., Waltman, E. B.,
870: Tavani, M., Harmon, B. A., Zhang, S. N., Paciesas, W. S., \& Ghigo, F.
871: D. 1996, ApJ, 467, L81
872: 
873: \bibitem[Hjellming \& Johnston (1988)]{hj88} Hjellming, R. M.  \&
874: Johnston, K. J. 1988, ApJ, 328, 600
875: 
876: \bibitem[Kemball et al. (1988)]{kem88}Kemball, A. J., Diamond, P. J.,
877:  \& Mantovani, F. A., 1988, MNRAS, 234, 713
878: 
879: \bibitem[Ma, et al. (1998)]{ma98}Ma, C., Arias, E. F., Eubanks, T. M.,
880: Fey, A. L., Gontier, A.-M., Jacobs, C. S., Sovers, O. J., Archinal, B.
881: A., Charlot, P., 1998, AJ, 116, 516
882: 
883: \bibitem[Markwardt et al. (1999)]{mar99} Markwardt, C. B., Swank, J.
884: H., \& Taam, R. E.  1999, ApJ, 513, L37
885: 
886: \bibitem[Mart\'\i\ et al. (2000)]{mart20} Mart\'\i\, J., Mirabel, I.F., Chaty,
887: S. \& Rodr\'\i guez  2000, ApJ, in press
888: 
889: \bibitem[Mirabel et al. (1998)]{mir98} Mirabel, I.  F., Dhawan, V.,
890: Chaty, S., Rodr\'\i guez, L. F., Mart\'\i\, J., Robinson, C. R., Swank, J.,
891: \& Geballe, T.  1998, A\&A, 330, L9
892: 
893: \bibitem[Mirabel \& Rodr\'\i guez (1994)]{mir94} Mirabel, I. F. \&
894: Rodr\'\i guez, L. F. 1994, Nature, 371, 46
895: 
896: \bibitem[Mirabel \& Rodr\'\i guez (1999)]{araa99} Mirabel, I. F. \&
897: Rodr\'\i guez, L. F. 1999, ARA\&A, 37, 409
898: 
899: \bibitem[Narayan et al. (1997)]{nar97}Narayan, R., Kato, S., \& Honma,
900: F.  1997, ApJ, 476, 49
901: 
902: \bibitem[Nayakshin et al. (1999)]{nay00}Nayakshin, S.,  Rappaport, S.,
903: Melia, F.  1999, AAS HEAD meeting,  31, 2801N    (ApJ 2000, submitted)
904: 
905: \bibitem[Pooley \& Fender (1997)]{poo97} Pooley, G.  G. \& Fender, R.
906: P. 1997, MNRAS, 292, 925
907: 
908: \bibitem[Rees (1998)]{rees98}Rees, M. J. 1998, in Black Holes and
909: Relativistic Stars, Wald, R. M., ed., University of Chicago Press, p.
910: 79
911: 
912: \bibitem[Reid et al. (1999)]{reid99} Reid, M. J., Readhead, A. C. S.,
913: Vermeulen, R. C., \& Treuhaft, R. N. 1999, ApJ, 524, 816
914: 
915: \bibitem[Reynolds, S. P. (1986)]{rey86} Reynolds, 1986, ApJ, 304, 713
916: 
917: \bibitem[Rodr\'\i guez et al.(1995)]{rod95} Rodr\'\i guez, L. F.,
918: Gerard, E., Mirabel, I. F., G\'omez, Y., and Vel\'azquez, A., 1995,
919: ApJS, 101, 173
920: 
921: \bibitem[Rodr\'\i guez \& Mirabel (1999)]{rod99} Rodr\'\i guez, L. F.
922: \& Mirabel, I. F. 1999, ApJ, 511, 398
923: 
924: \bibitem[Sams et al.(1996)]{sams96}Sams, B. J., Eckart, A., \& Sunyaev,
925: R. 1996, Nature, 382, 47
926: 
927: \bibitem[Taylor \& Cordes (1993)]{tay93}Taylor, J. H. \& Cordes, J. M.
928: 1993, ApJ, 411, 674
929: 
930: \bibitem[Walker et al.(2000)]{wal00}Walker, R. C., Dhawan, V., Romney, 
931: J. D., Kellermann, K. I. \&  Vermeulen, R. C., 2000, ApJ, 530, 233
932: 
933: \bibitem[Zensus, Diamond, \& Napier (1995)]{vlba95}Zensus, J. A.,
934: Diamond, P.J., \& Napier. P. J. 1995, Very Long Baseline Interferometry
935: and the VLBA, ASP Conf. Series, Vol.82
936: 
937: \end{thebibliography}
938: 
939: \clearpage \newpage
940: 
941: \figcaption[fig-1.eps]{X-ray and Radio data for 1997 October 14 to
942: November 08 (MJD=50735 to 50760).\protect\\[2mm]
943: Top:~[red]~RXTE All-Sky Monitor counts, 2-12~keV; quick-look results
944: provided by the ASM/RXTE team; ~~~[blue]~BATSE~fluxes, 20-100~keV from
945: occultation measurements, (photons~cm$^{-2}$~s$^{-1}$, multiplied by
946: 2000 for plotting).   \protect\\[2mm]
947: Middle:~GBI~radio~flux densities at 13 and 3.6~cm.  \protect\\[2mm] 
948: Bottom:~GBI~radio~spectral index.  The radio data is typically $\sim$10
949: scans per day of 5~min each, separated by 30-60~min.  Note the
950: correlation between the start of the soft X-ray and radio flares, and
951: the similar decay times of the BATSE and radio fluxes.  The VLBA
952: observed in the periods indicated by the black V's labelled A, B, \& C,
953: with the corresponding images in Fig.\ref{fig-2}. The vertical green
954: line is our estimate of the start of the flare at
955: MJD=50749.8$\pm$0.1~day.
956: \label{fig-1}} 
957: 
958: \figcaption[fig-2.eps]{VLBA images during 1997 October, corresponding
959: to epochs A, B, \& C in Fig.~\ref{fig-1}. Contours are at -5\%, 5, 10,
960: 20, 40, 60, \& 90\% of peak.  \protect\\[2mm]
961: (A)~2~cm~image with 7.5~AU resolution, during a plateau state, showing
962: the jet in the nucleus, 6 days before the flare in B \& C;
963:  peak~=~15.7~mJy~beam$^{-1}$.  \protect\\[2mm] (B)~13~cm~image with
964: 220~AU resolution; peak~=~93.7~mJy~beam$^{-1}$.  \protect\\[2mm]
965: (C)~3.6~cm~image, with 80~AU resolution; peak~=~19.7~mJy~beam$^{-1}$.
966: \protect\\[2mm] B and C are simultaneous data, centered at epoch
967: MJD=50752.02, 53.3$\pm$2.4 hours after the estimated start of the
968: flare.  The position of {\bf A}  is the same as the central component
969: of {\bf C}  within 1.5mas.  The black `*' marks  the astrometric
970: position of the nucleus, here and in
971:  Fig.~\ref{fig-7} A, item 5.
972: \label{fig-2}} 
973: 
974: \figcaption[fig-3.eps]{Images at 3.6~cm, of the same data as in
975: Fig.\ref{fig-2}C, epoch 1997 Oct 31. Contours are at -4\%, 4, 8, 16,
976: 32, 64, 96\% of peak.  The SE component has a proper motion of
977: 2.3$\pm$0.2~mas during 2.5~hrs, or 0.92$\pm$0.08~mas~hr$^{-1}$. The
978: mean position of the SE component implies a motion of 47.5$\pm$0.3~mas
979: in 53.3$\pm$2.4~hrs since the start of the flare, or
980: 0.90$\pm$0.05~mas~hr$^{-1}$.
981: \label{fig-3}} 
982: 
983: 
984: \figcaption[fig-4.eps]{X-ray and radio data for 1998 April 07 to May 12
985: (MJD=50910 to 50945).  Details are as in Fig.~\ref{fig-1}. Images from
986: VLBA data, indicated by the black V's labelled D, E, \& F, are shown in
987: Fig.~\ref{fig-5}. The vertical green line is our estimate of the start
988: of the flare at MJD=50932.7$\pm0.3$~days.
989: \label{fig-4}} 
990: 
991: \figcaption[fig-5.eps]{VLBA images during 1998 April-May, corresponding
992: to epochs D, E, and F in Fig.~\ref{fig-4}.  Contours are at -2\%, 2, 4,
993: 8, 16, 32, 64, 96\% of the peak. \protect\\[2mm]
994: (D)~2~cm~image, 7.5~AU resolution, showing the nuclear jet during a
995: plateau state, 2~days before the flare on MJD~50916; 
996: peak~=~48~mJy~beam$^{-1}$.  \protect\\[2mm]
997: (E)~2cm~image, 7.5~AU resolution; peak~=~64~mJy~beam$^{-1}$.  
998: \protect\\[2mm]
999: (F)~0.7cm~image, 2.5~AU resolution; peak~=~66.4~mJy~beam$^{-1}$.  
1000: \protect\\[2mm]
1001: E and F are quasi-simultaneous data (see Table-\ref{table-1}), centered
1002: at MJD~50935.50. The position of {\bf E} is  identical to {\bf D}
1003: within 1.0~mas. The black `+' marks the astrometric position,  here and
1004: in Fig.~\ref{fig-7}~A, item 7.
1005: \label{fig-5}} 
1006: 
1007: \figcaption[fig-6.eps]{Lower resolution images at 2~cm, of the same data
1008: as in Fig.\ref{fig-5}, epoch 1998 May 02 at 75~AU resolution.  Contours
1009: are at -2\%, 2, 3, 4, 6, 8, 16, 32, 64, 96\% of peak. The blue and red
1010: contours show time-resolved images 4.5~hrs apart. The SE component has
1011: a proper motion of 4.2$\pm$0.3mas, or 0.93$\pm$0.07~mas~hr$^{-1}$. The
1012: mean position of the SE component implies a motion of 57.5$\pm$0.5~mas
1013: in 67$\pm$7~hrs since the start of the flare, or
1014: 0.87$\pm$0.10~mas~hr$^{-1}$.
1015: \label{fig-6}} 
1016: 
1017: \clearpage \newpage
1018: 
1019: \figcaption[fig-7.eps]{A:~Astrometric positions from 1996 May, to 1998
1020: May, showing that the motion of GRS1915+105 lies mainly in the galactic
1021: plane. Key:  \protect\\[2mm]
1022: [1]=1996 May 23, 3.6~cm.          [2]=1996 Aug 04, 3.6~cm.
1023: [3]=1997 May 15, 3.6~cm, rise.    [4]=1997 May 15, 3.6~cm, set.
1024: [5]=1997 Oct 23, 2~cm, pre-flare. [6]=1997 Oct 31, 3.6~cm, during flare. 
1025: [7]=1998 Apr 11, 2~cm, pre-flare. [8]=1998 May 02, 2~cm, during flare. 
1026: The sloping line is parallel to the Galactic plane.  \protect\\[2mm]
1027: B:~Expected~secular~parallax from galactic rotation vs. distance, in
1028: the direction of GRS1915+105, assuming 220~km~s$^{-1}$ rotation, (solid
1029: line).  If the source were at 12~kpc, and stationary with respect to
1030: the galactic surroundings, 5.3~mas~yr$^{-1}$ is expected, compared to
1031: the 5.8$\pm$1.5~mas~yr$^{-1}$ measured. The effect of a $\pm$10\%
1032: ($\pm$22~km~s$^{-1}$) variation in rotation velocity is shown as the
1033: dashed lines.
1034: \label{fig-7}} 
1035: 
1036: \figcaption[fig-8.eps]{Deconvolved core size (Gaussian FWHM) vs.
1037: wavelength. Galactic electron scattering is fitted to the intrinsically
1038: unresolved minor axis, (lower line) $\Theta_{\rm min}=(0.15\lambda_{\rm
1039: cm}^{2}$)mas.  \protect\\[2mm]
1040: An intrinsically elongated source, plus scattering, is independently
1041: fit to the major axis,  (upper line), $\Theta_{\rm Maj}=(1.0\lambda_{\rm cm}+0.14\lambda_{\rm cm}^{2}$)mas. \protect\\[2mm] 
1042: The solid lines are fitted to data from a single day (1996 Aug 01) when
1043: the flux from the core was steady. Dashed lines extend to newer data
1044: (1998 May 02) at 2~cm and 0.7~cm. The Y-axis at right shows the linear size
1045: at an assumed distance of 12~kpc. The jet length is variable by up to a
1046: factor of $\sim$2 at different epochs.
1047: \label{fig-8}} 
1048: 
1049: \clearpage \newpage
1050: 
1051: \figcaption[fig-9.eps]{Radio flux density variations at three
1052: wavelengths, monitored by the VLA during the plateau state on
1053: 1997~May~15, 06-16~UTC. Inset is the VLBA at 2~cm, from observations
1054: during the gap in the VLA coverage from 10-12~UTC.
1055: \label{fig-9}}
1056: 
1057: \figcaption[fig-10.eps]{Top: XTE 2-12keV counts on 1997 Sep 05,
1058: 06-10~UTC. Note the dips every $\sim$12~min.  \protect\\[2mm]
1059: Bottom:~VLA~flux density at 3.6~cm, with the same periodicity (within
1060: errors) as the X-ray dips.  Fig.~\ref{fig-9} and Fig.~\ref{fig-10}
1061: taken together, support the idea that the radio oscillations arise in a
1062: synchrotron jet of (Gaussian FWHM) size $\sim$10$\lambda$~AU, which is
1063: fed by plasma injected during the dips in the soft X-ray flux.
1064: \label{fig-10}}
1065: 
1066: \clearpage \newpage
1067: 
1068: \begin{table}
1069: \begin{tabular}[]{lllllllll}
1070: \tableline               
1071: \tableline               
1072: 
1073:  & ~~~MJD & ~~~~~~~~~~UT & Fig. & ~$\lambda$ & Bit rate 
1074: & rms, M. & rms, T. & beam \\ 
1075: 
1076:  &   &  &   & (cm)    & (Mb~s$^{-1}$) 
1077: & ($\mu$Jy) & ($\mu$Jy)  & (mas)  \\
1078: 
1079: \tableline               
1080: A & 50744 & 97 Oct 23 22:15-03:23 & \ref{fig-2} A & 2.0 & 256
1081: & ~210 & ~85 & 0.62 \\
1082: 
1083: B & 50751 & 97 Oct 30 23:00-04:08 & \ref{fig-2} B & 13  & 128$^B$
1084: & ~180 & ~75 & 20. \\
1085: 
1086: C & 50751 & 97 Oct 30 23:00-04:08 & \ref{fig-2} C & 3.6 & 128$^C$
1087: & ~200 & ~75 & 7.0 \\
1088: 
1089: D & 50914 & 98 Apr 11 09:00-14:08 & \ref{fig-5} D & 2.0 & 256
1090: & ~230 & ~85 & 0.62 \\
1091: 
1092: E & 50935 & 98 May 02 07:35-15:55 & \ref{fig-5} E & 2.0 & 128$^E$
1093: & ~260 & 120  & 0.62 \\
1094: 
1095: F & 50935 & 98 May 02 08:26-16:22 & \ref{fig-5} F & 0.7 & 256$^F$
1096: & ~510 & 300  & 0.21 \\
1097: \tableline               
1098: \end{tabular} 
1099: 
1100: \vspace{5mm}
1101: 
1102: B, \&  C in Fig. \ref{fig-2}  were  simultaneous observations,
1103: 	128~Mb/s was recorded at each of 3.6~cm and 13~cm. \\
1104: E, \&  F in Fig. \ref{fig-5}  were quasi-simultaneous observations,
1105: 	the sequence (128~Mb/s for 44~min at 2~cm, 5~min dead time, 
1106: 	then 256~Mb/s for 22~min at 7~mm) was repeated 7 times.
1107: 
1108: \vspace{2mm}
1109: 
1110: \caption{VLBA observations. The columns are: Observation Epoch; 
1111: Figure reference; Wavelength; Bit
1112: rate, megabits per second; rms noise in image, microJansky per beam
1113: area, measured; RMS noise, theoretical; and Synthesized Beam FWHM, mas.}
1114: \label{table-1}    
1115: \end{table}
1116: 
1117: \end{document}
1118: 
1119: 
1120: