astro-ph0107296/ms.tex
1: \documentclass[12pt,preprint]{aastex}
2: 
3: \begin{document}
4: 
5: \title{Twenty Years of Timing SS433}
6: 
7: \author{S.S. Eikenberry, P.B. Cameron, B.W. Fierce, D.M. Kull,
8: D.H. Dror, J.R. Houck}
9: \affil{Astronomy Department, Cornell University, Ithaca, NY
10: 14853} 
11: 
12: \author{B. Margon}
13: \affil{Space Telescope Science Institute, 3700 San Martin Drive, Baltimore, MD  21218}
14: 
15: \begin{abstract}
16: 
17: 	We present observations of the optical ``moving lines'' in
18: spectra of the Galactic relativistic jet source SS433 spread over a
19: twenty year baseline from 1979 to 1999.  The red/blue-shifts of the
20: lines reveal the apparent precession of the jet axis in SS433, and we
21: present a new determination of the precession parameters based on these
22: data.  We investigate the amplitude and nature of time- and
23: phase-dependent deviations from the kinematic model for the jet
24: precession, including an upper limit on any precessional period
25: derivative of $\dot P < 5 \times 10^{-5}$.  We also dicuss the
26: implications of these results for the origins of the relativistic jets
27: in SS433.
28: 
29: \end{abstract}
30: 
31: 
32: \keywords{binaries - stars: individual (SS433)}
33: 
34: \section{Introduction}
35: 
36: 	SS433 is the first known example of a Galactic relativistic
37: jet source, and thus the forerunner of modern microquasar astrophysics.
38: The optical spectrum of this object shows a number of strong, broad
39: emission lines of the Balmer and HeI series, as well as several lines
40: at unusual wavelengths.  These latter have been identified as
41: red/blue-shifted Balmer and HeI emission from collimated jets with
42: intrinsic velocities of $v \simeq 0.26c$ \citep{AbellMargon}.
43: Furthermore, the Doppler shifts of these features change with time in
44: a cosinusoidal manner, leading to the label of ``moving lines''.  This
45: behavior is now widely accepted to be a symptom of precession of the
46: jet axis in SS433 on a timescale of $\sim 164$ days \citep{Margon84}.
47: 
48: 	Early studies of the precession in SS433 indicated possible
49: instabilities or drifts in the precessional clock \citep{Anderson},
50: which could give considerable insight into the accretion processes
51: which must provide the precessional torque.  However,
52: \citet{MargonAnderson} reviewed ten years of SS433 timing data and
53: concluded that while significant deviations from cosinusoidal behavior
54: exist in SS433, the evidence for systematic long-term drifts
55: (e.g. precessional period derivative, $\dot P$) remained inconclusive.
56: 
57: 	In this paper, we take the data set considered by
58: \citet{MargonAnderson} and add to it more than 50 Doppler shift
59: measurements spread over 10 years, including 9 Doppler shifts measured
60: in 1999.  Combined, these observations span more than 20 years, and
61: thus provide an excellent data set for constraining long-term drifts
62: in SS433's precessional clock.  We discuss the observations in Section
63: 2.  In Section 3, we present analyses of the entire data set in the
64: context of the ``kinematic model'' for SS433's precessing jets.  In
65: Section 4, we discuss the results of these analyses, and in Section 5
66: we present our conclusions.
67: 
68: \section{Observations}
69: 
70: 	The primary observations used here are optical spectroscopic
71: observations of SS433, from a wide range of telescopes and instruments
72: (see \citet{MargonAnderson} and references therein for details).  The
73: net result of these observations, spread over the period from June
74: 1978 to July 1992, is the measurement of 433 Doppler shifts for the
75: ``receding jet'' ($z_1$) and 482 Doppler shifts for the ``approaching
76: jet'' ($z_2$) for the optical moving lines in SS433.
77: 
78: 	We obtained further spectra of SS433 in July, 1999 using the
79: Hartung-Boothroyd Observatory (HBO) 24-inch telescope and optical CCD
80: spectrograph.  We used a 600 lines/mm grating and $6 \arcsec$ slit
81: providing a resolution of $R \sim 800$ ($6 \ {\rm \AA /pix}$).  We
82: present a typical spectrum in Figure 1.
83: 
84: 	We determined the Doppler shift of each HBO spectrum using
85: only the moving $H \alpha$ lines, and we did so by fitting a Gaussian
86: profile to the red and blue components separately.  Note that the
87: profiles of the moving lines are broad, time-variable, and often
88: asymmetric.  This is due to the time overlap of multiple discrete
89: emission components, commonly referred to as ``bullets''
90: \citep{Vermeulen}, with typical lifetimes of $\sim 3$ days.  These
91: systematic deviations introduce a relatively large uncertainty in the
92: Doppler shift determination.  Based upon examination of many spectra,
93: we find that the typical full-width at half-maximum (FWHM) is $\Delta z
94: \sim 0.003$, and we adopt this as our uncertainty in the Doppler shift
95: determination $\sigma_z$.
96: 
97: \section{Analysis}
98: 
99: \subsection{The Kinematic Model}
100: 
101: 	Throughout our analysis of these data, we adopt the
102: ``kinematic model'' for the moving lines, which assumes that the
103: changing Doppler shifts arise from the precession of the jet axis in
104: SS433.  The simplest form of the kinematic model takes into account
105: five components: the jet velocity $\beta = {v \over{c}}$; the jet
106: angle from the precessional axis $\theta$, the inclination angle of
107: the system with respect to the observer's line of sight $i$, the
108: precession period $P$, and the epoch of zero precessional phase $t_0$.
109: The period and zero-phase epoch combine to give the precessional phase
110: $\phi = {{t - t_0} \over{P}}$.  The resulting Doppler shifts obey the
111: equation
112: 
113: $$ z_{1,2} = 1 - \gamma [1 \pm \ \beta \ \sin \theta \ \sin i \ \cos \phi \ \pm \ \beta \ \cos \theta \ \cos i]$$
114: 
115: \noindent where $\gamma = (1- \beta^2)^{-1/2}$.  SS433 exhibits
116: ``nodding'' of the jets on a $\sim 6.5$-day period \citep{Katz} due to
117: the $\sim13$-day binary motion of SS433 \citep{Crampton}, which are
118: not accounted for in this model.  However, the effects of this nodding
119: are essentially negligible for long timescale studies of the jets such
120: as ours.  We further mitigate the impact of nodding by applying a
121: 7-day boxcar smoothing filter to the individual Doppler shifts
122: determined above.  We then used chi-squared minimization to find the
123: best-fit parameters for the kinematic model (Table 1).  The resulting
124: model fit is plotted versus time along with the data and residuals in
125: Figures 2-3 for $z_1$ and $z_2$.  We plot the same model fit, data,
126: and residuals versus precessional phase in Figure 4.
127: 
128: 	The resulting fit has a chi-squared residual per degree of
129: freedom of $\chi^2_{\nu} = 8.9$, indicating the presence of
130: statistically significant residuals.  However, we can still use this
131: fit to estimate uncertainties in the kinematic model parameters as
132: follows.  First, we scale all of the $\sigma_z$ values by
133: $\sqrt{8.9}$, so that the residuals have $\chi^2_{\nu-fix} = 1.0$,
134: essentially by fiat.  We then take the uncertainties to be the range
135: of a model parameter which introduces a total change of $\Delta
136: \chi^2_{fix} = 1.0$.  We also report these values in Table 1.  This
137: rescaling approach for deriving the model parameter uncertainties is
138: statistically valid in a strict sense only if the residuals are
139: consistent with Gaussian noise and are not correlated with any model
140: parameters in a systamtic way.  If so, then the residuals would simply
141: indicate that we have ignored one or more sources of noise in the
142: system when estimating the uncertainties in the individual Doppler
143: shifts.  As we show below, this is largely true, though we see some
144: evidence of small (but statistically-significant) systematic
145: deviations from the kinematic model.  Thus, the uncertainties in the
146: model parameters in Table 1 are likely to be good, but not perfect,
147: statistical estimates.  For the remainder of the paper, we adopt the
148: best-fit model parameters presented in Table 1.
149: 
150: 
151: \subsection{Doppler Shift Residuals}
152: 
153: 	One obvious feature of Figures 2-4 is that the residuals to
154: the model fit greatly exceed the uncertainties in the Doppler shift
155: determinations (as also shown by the large value of $\chi^2_{\nu}$
156: above).  We also notice no obvious trend in the residuals versus time
157: as would be expected for systematic timing effects, such as a constant
158: precessional period time derivative $\dot P$.  Such large, apparently
159: random residuals have been noticed in previous timing studies of SS433
160: (\citet{Anderson}; \citet{MargonAnderson}).
161: 
162: \subsubsection{Correlations in Residuals}
163: 
164: 	Previous studies have also noticed that the velocity residuals
165: in SS433 show a pattern of correlation between $z_1$ and $z_2$
166: \citep{MargonAnderson}.  Specifically, when we plot the residuals of
167: $z_1(obs)-z_1(mod)$ versus $z_2(obs)-z_2(mod)$, we find that most of
168: the points lie in the second and fourth quadrants (Figure 5).  In
169: other words, when the absolute value of $z_1$ is greater than expected,
170: the absolute vlaue of $z_2$ is also greater than expected, and vice
171: versa.  The number of data points with $z_1$ and $z_2$ residuals in
172: quadrants 2 and 4 is $271 \pm 16$, while in quadrants 1 and 3 the
173: number is $110 \pm 10$ -- a $>8 \sigma$ difference.
174: 
175: 	The linear correlation coefficient between the residuals is $r
176: = -0.69 \pm 0.02$.  We estimate the uncertainty from a Monte Carlo
177: simulation as follows.  We take the 381 pairs of $z_1$ and $z_2$
178: residuals and add to each a random number drawn from a Gaussian
179: distribution with mean of zero and a standard deviation of 0.003 --
180: the typical uncertainty in the Doppler shift measurements.  We then
181: calculate the correlation coefficient of the resulting simulated
182: distribution.  We repeat this procedure 1000 times and then take the
183: standard deviation in the correlation coefficient as the uncertainty
184: above, $\sigma_r = \pm 0.02$.
185: 
186: 	This correlation pattern could have several physical sources.
187: The effect considered most commonly in previous studies
188: (e.g. \citet{MargonAnderson}) is that of phase noise in the
189: precessional motion, with strict symmetry between $z_1$ and $z_2$.  As
190: the jet precessional phase either lags or leads the model ephemeris,
191: the projected velocity amplitudes of the jets on the observers line of
192: sight will either exceed or fall short of the model prediction.
193: Another possible physical explanation is modulation of the velocity
194: amplitude -- ``$\beta$-noise'' -- in a system which otherwise follows
195: the 5-parameter kinematic model ideally (e.g. \citet{Milgrom}).
196: 
197: \subsubsection{Phase-Dependence of Residuals}
198: 
199: 	Another factor which could impact the $z_1/z_2$ residual
200: correlation are phase-dependent residuals to the kinematic model.
201: \citet{MargonAnderson} found no evidence for such phase-dependence in
202: their data.  In analyzing this data set, we divided the data into 10
203: evenly-spaced phase intervals and calculated the average and standard
204: deviation of the residual Doppler shifts in each bin (Table 2).  None
205: of the average residuals from the kinematic model is as large as the
206: rms deviation for residuals in that phase bin, indicating that
207: phase-dependent deviations from the kinematic model do not dominate
208: the residuals.  Furthermore, if we compare the average residuals of
209: the individual phase bins to the rms scatter of all the phase bins,
210: none of them are more than $2 \sigma$ outliers.  Thus, the amplitude
211: of any average deviation from the model velocity does not seem to be
212: phase-dependent.
213: 
214: 	However, when we calculate the uncertainty in the average
215: residual for each phase bin, equal to the standard deviation in the
216: residuals divided by the square-root of the number of points, we find
217: that the deviations are in fact statistically significant (Table 2).
218: That is, while the scatter around the kinematic model in any given
219: phase bin is not dominated by systematic deviations from the model,
220: such deviations are present in the data set.  The nature of these
221: systematic deviations are not clearly determined.  Figures 2-3 show
222: that the velocity residuals show some correlation on timescales of
223: weeks or months.  Thus, sparse sampling of the velocities combined
224: with such correlations in the residuals could be one explanation for
225: the apparent systematic deviations.
226: 
227: 	We used the average residuals from the kinematic model as
228: correction factors for the quadrant analysis presented above and in
229: Figure 5, taking the observations and subtracting both the kinematic
230: model and the average deviation for all points within $\Delta \phi =
231: \pm 0.05$ cycles of each data point.  As a result, the number of
232: $z_1$,$z_2$ residual pairs in quadrants 2 and 4 does decrease, but only
233: to 261 (with 120 pairs in quadrants 1 and 3).  Thus the correlation
234: between $z_1$ and $z_2$ residuals remains highly statistically
235: significant even after correction for the systematic deviations.
236: 
237: 	We also note that the relative phase-independence of the
238: residuals raises questions regarding the nature of the residuals.  As
239: can be seen from the equation for Doppler shifts in the kinematic
240: model, every parameter which {\it could} be time-variable ($\beta, P,
241: i, \theta, $) either feeds directly into the phase $\phi$ or is
242: multiplied by $\cos \phi$.  Thus, noise in these terms or in $\phi$
243: itself should result in a cosinusoidal modulation in the RMS of the
244: Doppler shift residuals with $\phi$.
245: 
246: \subsubsection{The Phase Noise Model}
247: 
248: 	As mentioned above, the most commonly-invoked physical model
249: for the velocity residuals in SS433 is ``phase jitter'' in the jet
250: precession.  Since the precession phase affects both jets similarly,
251: it naturally explains the correlation between the $z_1$ and $z_2$
252: residuals.  If such jitter can occur over timescales of weeks or
253: months, it can also explain the long-term residual correlations
254: evident in Figures 2-3.
255: 
256: 	We analyzed this SS433 data set following the example of
257: \citet{MargonAnderson}, determining phase errors from the velocity
258: residuals above.  We simply defined the phase error to be the phase
259: difference between the actual phase of the observation given its epoch
260: and the kinematic model parameters in Table 2 and the closest model
261: point with the same observed velocity.  As can be seen in Figure 4,
262: some observed velocity amplitudes exceed the maximum model velocity
263: amplitude, and such points were dropped from this analysis.  We then
264: divided the data set into 10-day intervals and calculated the average
265: and standard deviation of the phase errors from all phase measurements
266: doing that interval (including both $z_1$ and $z_2$).  For 10-day
267: intervals with only 1 phase measurement we have no estimate of the
268: standard deviation, and thus dropped such intervals from the analysis.
269: We plot the resulting phase noise measurements in Figure 6.  We
270: repeated this same analysis using a 30-day interval for averaging,
271: with the results shown in Figure 7.
272: 
273: 	We note that while there are occasional trends in the
274: residuals on timescales of several hundred days, no obvious trend is
275: apparent over the full time span in either panel of Figure 7.  The
276: 1999 data are marginally inconsistent with zero phase residual (at the
277: $2.8 \sigma$ level for one of the two data points in Figure 6b).
278: However, it is clear that this phase residual is less than many prior
279: apprently secular deviations from the kinematic model in Figures 6-7.
280: If we assume that some period derivative is present in SS433 over the
281: span of our observations, these secular deviations could mask its
282: effects up to $\Delta \phi \sim 0.05$ cycles.  Given the span of our
283: observations, this corresponds to an upper limit on the period
284: derivative of $\dot P < 5 \times 10^{-5}$.
285: 
286: \subsubsection{The Velocity Amplitude Noise Model}
287: 
288: 	As mentioned above, an alternate physical explanation for the
289: velocity residuals in Figures 2-4 is noise in the intrinsic velocity
290: of the jets.  To investigate this possibility further, we calculated
291: the intrinsic jet velocity necessary to match each observed Doppler
292: shift, given $\theta, \ i , \ t_0 , \ {\rm and} \ p$ from the best-fit
293: parameter set in Table 1.  We plot the corresponding values for $\beta
294: = {v \over{c}}$ versus time in Figure 8 and versus precessional phase
295: in Figure 9.
296: 
297: 	The average value of $\beta$ we find is 0.254 with a standard
298: deviation of $0.024$.  Given 507 independent measurements of $\beta$
299: in this way, we arrive at an average value of $\beta_{ave} = 0.254 \pm
300: 0.0011$.  While this value is only $\sim 4 \%$ lower than the value
301: given in Table 1, the difference is statistically significant at the
302: $7.9 \sigma$ level.  This may indicate that ``noise'' in the Doppler
303: shifts may in fact be impacting the parameter estimates for the
304: kinematic model in a systematic way, as discussed in Section 3.1.
305: 
306: \section{Discussion}
307: 
308: \subsection{Phase Noise}
309: 
310: 	As noted above, the upper limit on precessional period
311: derivative of $\dot P < 5 \times 10^{-5}$ shows that there is no large
312: long-term drift in the precessional timing properties of SS433.  The
313: presence of jitter in the system implies some ``torque noise'' in the
314: process driving the precession, according to the phase noise model.
315: However, if this were the case, that noise must average out over
316: timescales of $\sim 20$ years.  We can also see from Figure 7 that
317: there are fairly large phase deviations of $\Delta \phi \sim 0.1$
318: cycles over timescales as short as $\sim 10$ days.  This implies that
319: the torque noise $\Delta \tau$ has a maximum relative amplitude of at
320: least
321: 
322: $$ {\Delta \tau_{max} \over{\tau}} \simeq {{(\Delta \phi / \Delta t)} \over{(\partial \phi / \partial t)}}$$
323: 
324: $$	\simeq 1.6$$
325: 
326: \noindent Thus, the variation in torque can in fact exceed the
327: time-averaged torque driving the precession.  This may be a problem
328: for certain physical models of the precession and timing noise in
329: SS433.
330: 
331: 	Finally, we note that the phase noise model is incapable of
332: producing the observed Doppler shifts which exceed the maximum
333: amplitude predicted by the kinematic model.  We have considered the
334: possibility that the phase noise itself causes the $\chi$-squared
335: fitting procedure used to determine the model parameters to
336: systematically underestimate the true jet velocity, and thus
337: ``undershoot'' the maxima.  However, Monte Carlo simulations of data
338: sets with higher true velocities and phase noise identical to that
339: observed here fail to produce such undershooting.  Therefore, we
340: conclude that phase noise model cannot reproduce the observed Doppler
341: shift residuals near the maximum projected velocities.
342: 
343: 
344: \subsection{Velocity Noise}
345: 
346: 	The alternate ``$\beta$-noise'' model, on the other hand, can
347: clearly explain the excess velocity at the extrema (and any other
348: precessional phase) by the changing jet velocity amplitude.  Such a
349: model also has a physical basis, given recent advances in the modeling
350: of relativistic jet production.  \citet{meier} dicuss a scenario where
351: such jets are launched by a magnetic accretion disk instability around
352: a black hole (or other compact object).  Variations in the accretion
353: flow onto the compact object (i.e. $\dot M$, intrinsic magnetic field,
354: etc.) can alter the radius at which the magnetic field saturates and
355: the jet is launched, and thus the jet velocity.  The relation between
356: jet velocity and launch radius for a non-rotating black hole follows:
357: 
358: $$ \beta (R) = \sqrt{{2 R_g \over{R}}} $$
359: 
360: \noindent where $\beta = {v \over{c}}$, and $R_g$ is the gravitational
361: radius of the black hole (one-half of the Schwarzschild radius).
362: 
363: \subsection{Jitter models and phase-dependence of residuals}
364: 
365: 	It is also interesting to view these model in light of the
366: apparent lack of phase-dependence in the Doppler shift residuals noted
367: above.  By differentiating the equation for Doppler shifts in the
368: kinematic model (eqn. 1) with respect to the potentially time-varying
369: model components ($\beta$, $i$, $\theta$, and $\phi$) we can see the
370: relative phase-dependence of Doppler shift residuals on deviations in
371: each term.  In the phase noise model, we would expect the following
372: dependence:
373: 
374: $$ \Delta z = (\gamma \ \beta \ \sin \theta \ \sin i \ \sin \phi) \ \Delta \phi $$
375: $$	\simeq 0.1 \sin \phi \ \Delta \phi $$
376: 
377: \noindent Thus, we would expect the amplitude of the Doppler shift
378: residuals to be sinusoidally modulated with phase, in apparent
379: contradiction with our analyses above.
380: 
381: For the ``$\beta$-noise'' model, we have:
382: 
383: $$ \Delta z = (\gamma \ \sin \theta \ \sin i \ \cos \phi \ - \ \gamma \ \cos \theta \ \cos i) \ \Delta \beta $$
384: $$	\simeq (0.35 \cos \phi \ - \ 0.2) \ \Delta \beta $$
385: 
386: \noindent (with the approximation that $\partial \gamma / \partial
387: \beta \simeq 0$).  Again, we have a modulation of the Doppler shift
388: residual amplitude dominated by a term varying cosinusoidally with
389: respect to phase.
390: 
391: 	For variations in the angle between the jet axis and the
392: precessional axis, $\theta$, we have:
393: 
394: $$ \Delta z = (\gamma \ \beta \ \cos \theta \ \sin i \ \cos \phi \ + \ \gamma \ \beta \ \sin \theta \ \cos i) \ \Delta \theta $$
395: $$	\simeq (0.24 \cos \phi \ + \ 0.09) \ \Delta {\theta} $$
396: 
397: \noindent again dominated by a $\cos \phi$ term.
398: 
399: 	Finally, for variations in the system inclination angle, $i$,
400: we  have:
401: 
402: $$ \Delta z = (\gamma \ \beta \ \sin \theta \ \cos i \ \cos \phi \ + \ \gamma \ \beta \ \cos \theta \ \sin i) \ \Delta i $$
403: $$	\simeq (0.02 \cos \phi \ + \ 0.24) \ \Delta i $$
404: 
405: \noindent Interestingly, the Doppler shift residual amplitudes for
406: ``$i$-noise'' would be dominated by a constant term, with only a small
407: dependence on phase.  Thus, this is the only parameter in the
408: kinematic model for which variations causing the Doppler shift
409: residuals are consistent with their observed phase-independence.
410: Unfortunately, we know of no physical model for such variability at
411: this time.  Furthermore, based on the mass estimates for the compact
412: object and companion star ($\sim 10 M_{\odot}$ total), the known
413: 13.5-day binary period, and the presence of deviations as large as
414: 0.01 rad/day in inclination angle, the change in rotational energy
415: would require average powers of $\dot E \sim 10^3 \ L_{Edd}$, which
416: seems implausible.
417: 
418: 	One other possible explanation is that the Doppler shift
419: residuals are due to variations in both precessional phase and one (or
420: more) of the other model parameters.  In that case, the residual
421: amplitude would depend on a sum of $\cos \phi$ and $\sin \phi$ terms
422: which could potentially smooth out any phase-dependence in the
423: residuals.
424: 
425: 
426: \section{Conclusions}
427: 
428: 	We have presented observations of the Doppler-shifted optical
429: moving lines in SS433 spanning over 20 years.  We draw the following
430: conclusions based on the data:
431: 
432: \begin{itemize}
433: 
434: \item We find parameters for the kinematic model for the jet
435: precession which are similar to those found by previous authors
436: (e.g. \citet{MargonAnderson}).
437: 
438: \item We find a strong correlation between residuals to the models
439: fits for the two jets $z_1$ and $z_2$, with a linear correlation
440: coefficient of $r = -0.69 \pm 0.02$.
441: 
442: \item We find that the residuals to the kinematic model fit are {\it
443: not} dominated by systematic phase-dependent deviations from the
444: model.  However, systematic phase-dependent deviations from the
445: kinematic model {\it are} seen in the data set at a low level.
446: 
447: \item If we adopt a ``phase noise'' model for the velocity residuals,
448: we find correlated deviations over timescales of months to years, but
449: no long-term trend over the full data set.  We place a limit on the
450: precessional period derivative of $\dot P < 5 \times 10^{-5}$.
451: 
452: \item Noise in any single parameter of the kinematic model seems
453: unable to explain the observed phase-independence of the velocity
454: residuals in SS433.  However, variations in both phase and one of the
455: other parameters would vary as the weighted sum of $\cos \phi$ and
456: $\sin \phi$ terms, which could smooth out any phase-dependence of the
457: Doppler shift residuals.
458: 
459: \end{itemize}
460: 
461: 
462: \acknowledgments The authors thank D. Lai and I. Wasserman for helpful
463: discussions of these results.  SSE is supported in part at Cornell by
464: an NSF CAREER award (NSF-9983830), and PBC was partially supported by
465: this grant.  DMK was supported at Cornell by a NASA Space Grant summer
466: research fellowship.
467: 
468: \begin{thebibliography}
469: 
470: \bibitem[Abell \& Margon(1979)]{AbellMargon} Abell, G.O. \& Margon, B. 1979, \nat, 279, 701
471: 
472: \bibitem[Anderson et al.(1983)]{Anderson} Anderson, S.F., Margon, B., Grandi, S.A. 1983, \apj, 273, 697
473: 
474: \bibitem[Crampton et al.(1980)]{Crampton} Crampton, D., Cowley, A.P., Hutchings, J.B. 1980, \apj, 235, L131
475: 
476: \bibitem[Katz et al.(1982)]{Katz} Katz, J.I., Anderson, S.F., Margon, B., Grandi, S.A. 1982, \apj, 260, 780
477: 
478: \bibitem[Margon(1984)]{Margon84} Margon, B. 1984, \araa, 22, 507
479: 
480: \bibitem[Margon \& Anderson(1989)]{MargonAnderson} Margon, B. \&  Anderson, S.F. 1989, \apj, 347, 448
481: 
482: \bibitem[Milgrom et al.(1982)]{Milgrom} Milgrom, M. et al. 1982, \apj, 256 222
483: 
484: \bibitem[Meier et al.(2001)]{meier} Meier, D.L., Koide, S., Uchida,
485: Y. 2001, {\it Science}, 291, 84
486: 
487: \bibitem[Vermeulen et al.(1993)]{Vermeulen} Vermeulen, R.C., et al. 1993, \aap, 270, 204
488: 
489: \end{thebibliography}
490: 
491: \vfill \eject
492: 
493: \begin{deluxetable}{lccccc}
494: \tablecaption{Best-fit parameters for the kinematic jet precession model}
495: \startdata
496: Parameter & $\beta$ & $\theta$ & $i$ & $P$ & $t_0$ \\
497:   & & (deg) & (deg) & (days) & (TJD) \\
498: \hline
499: Value & 0.2647 & 20.92 & 78.05 & 162.375 & 3563.23 \\
500: Uncertainty & $\pm 0.0008$ & $\pm 0.08$ & $\pm 0.05$ & $\pm 0.011$ & $\pm 0.11$ \\
501: \enddata
502: \end{deluxetable}
503: 
504: \begin{deluxetable}{lcccccc}
505: \tablecaption{Average residuals in Doppler shift versus phase}
506: \startdata
507: Phase Interval & $<\Delta z_1>$ & $\Delta z_1$ & $\Delta z_1$ & $<\Delta z_2>$ & $\Delta z_2$ & $\Delta z_2$\\
508: (cycles) & & RMS & Uncert. & & RMS & Uncert. \\
509: \hline
510: 0.0-0.1 & -0.0035 & 0.0074 & 0.0010 & 0.0040 & 0.0067 & 0.0008 \\
511: 0.1-0.2 & 0.0017 & 0.0082 & 0.0011 & -0.0024 & 0.0083 & 0.0010 \\
512: 0.2-0.3 & -0.0005 & 0.0101 & 0.0014 & -0.0016 & 0.0093 & 0.0014 \\
513: 0.3-0.4 & 0.0035 & 0.0076 & 0.0011 & 0.0003 & 0.0028 & 0.0004 \\
514: 0.4-0.5 & -0.0011 & 0.0130 & 0.0032 & -0.0008 & 0.0067 & 0.0013 \\
515: 0.5-0.6 & -0.0052 & 0.0116 & 0.0016 & 0.0031 & 0.0083 & 0.0010 \\
516: 0.6-0.7 & 0.0039 & 0.0152 & 0.0028 & -0.0008 & 0.0135 & 0.0025 \\
517: 0.7-0.8 & 0.0015 & 0.0092 & 0.0013 & -0.0001 & 0.0077 & 0.0010 \\
518: 0.8-0.9 & 0.0026 & 0.0104 & 0.0013 & -0.0018 & 0.0079 & 0.0010 \\
519: 0.9-1.0 & -0.0035 & 0.0088 & 0.0013 & 0.0007 & 0.0063 & 0.0008 \\
520: \enddata
521: \end{deluxetable}
522: 
523: 
524: 
525: 
526: \begin{figure}
527: \plotone{f1.eps}
528: \caption{\it Typical spectrum of SS433 taken from the
529: Hartung-Boothroyd Observatory 24-inch telescope.  Note the clear
530: appearance of the moving Balmer lines from the jets.}
531: \end{figure}
532: 
533: \begin{figure}
534: \epsscale{0.8}
535: \plotone{f2a.eps}
536: \plotone{f2b.eps}
537: \caption{\it Doppler shift data, model fit, and residuals versus time
538: based on the best-fit parameters in Table 1 for $z_1$, the ``receding
539: jet''.  The vertical extent of the plotting symbols shows the typical
540: $\pm 1 \sigma$ uncertainty in the Doppler shift.}
541: \end{figure}
542: 
543: 
544: \begin{figure}
545: \epsscale{0.8}
546: \plotone{f3a.eps}
547: \plotone{f3b.eps}
548: \caption{\it Doppler shift data, model fit, and residuals versus time
549: based on the best-fit parameters in Table 1 for $z_2$, the
550: ``approaching jet''.  The vertical extent of the plotting symbols
551: shows the typical $\pm 1 \sigma$ uncertainty in the Doppler shift.}
552: \end{figure}
553: 
554: 
555: \begin{figure}
556: \epsscale{0.8}
557: \plotone{f4a.eps}
558: \plotone{f4b.eps}
559: \caption{\it Doppler shift data, model fit, and residuals versus
560: precessional phase based on the best-fit parameters in Table 1 for:
561: (top) $z_1$, the ``receding jet''; (bottom) $z_2$, the ``approaching''
562: jet.  The vertical extent of the plotting symbols shows the typical
563: $\pm 1 \sigma$ uncertainty in the Doppler shifts.}
564: \end{figure}
565: 
566: \begin{figure}
567: \plotone{f5.eps}
568: \caption{\it Doppler shift residuals from the kinematic model.  Note the clear anti-correlation between $z_1$ and $z_2$ residuals.}
569: \end{figure}
570: 
571: \begin{figure}
572: \epsscale{0.8}
573: \plotone{f6a.eps}
574: \plotone{f6b.eps}
575: \caption{\it Phase residuals deduced from velocity residuals to the kinematic model, averaged over 10-day intervals.}
576: \end{figure}
577: 
578: \begin{figure}
579: \epsscale{0.8}
580: \plotone{f7a.eps}
581: \plotone{f7b.eps}
582: \caption{\it Phase residuals deduced from velocity residuals to the kinematic model, averaged over 30-day intervals.}
583: \end{figure}
584: 
585: \begin{figure}
586: \epsscale{0.8}
587: \plotone{f8a.eps}
588: \plotone{f8b.eps}
589: \caption{\it Jet velocity amplitude required to match the observed
590: Doppler shifts as a function of time.}
591: \end{figure}
592: 
593: \begin{figure}
594: \plotone{f9.eps}
595: \caption{\it Jet velocity amplitude required to match the observed
596: Doppler shifts as a function of precessional phase.}
597: \end{figure}
598: 
599: 
600: 
601: \end{document}
602: 
603: 
604: 
605: 
606: 
607: 	
608: 
609: