1: \documentclass{emulateapj}
2:
3: \citestyle{aa}
4:
5: \begin{document}
6: \title{Correlation between X-ray Lightcurve Shape and Radio Arrival Time in
7: the Vela Pulsar}
8: \author{A. Lommen\altaffilmark{1}, J. Donovan\altaffilmark{1,2},
9: C. Gwinn\altaffilmark{3},
10: Z. Arzoumanian\altaffilmark{4,5}, A. Harding\altaffilmark{4},
11: M. Strickman\altaffilmark{6}, R. Dodson\altaffilmark{7},
12: P. McCulloch\altaffilmark{8}, and D. Moffett\altaffilmark{9}}
13: \altaffiltext{1}{Department of Physics and Astronomy, Franklin and Marshall College, Lancaster, Pennsylvania.}
14: \altaffiltext{2}{Department of Astronomy, Columbia University, New York, New York.}
15: \altaffiltext{3}{Department of Physics, University of California, Santa Barbara, California.}
16: \altaffiltext{4}{NASA Goddard Space Flight Center, Greenbelt, Maryland.}
17: \altaffiltext{5}{Universities Space Research Association, Columbia, Maryland.}
18: \altaffiltext{6}{Code 7651.2, Naval Research Laboratory, Washington, DC.}
19: \altaffiltext{7}{Observatorio Astron\'omico Nacional,
20: Madrid, Espa\~na}
21: \altaffiltext{8}{University of Tasmania, Tasmania, Australia.}
22: \altaffiltext{9}{Furman University, Greenville, South Carolina.}
23:
24: \submitted{Submitted to the Astrophysical Journal, 13 Sep 2006; Revised,
25: 04 Nov 2006; Accepted, 11 Nov 2006}
26:
27: \begin{abstract}
28:
29: We report the results of simultaneous observations of the Vela pulsar
30: in X-rays and radio from the RXTE satellite and the Mount Pleasant
31: Radio Observatory in Tasmania. We sought correlations between the
32: Vela's X-ray emission and radio arrival times on a pulse by pulse
33: basis. At a confidence level of 99.8\% we have found significantly
34: higher flux density in Vela's main X-ray peak during radio pulses that
35: arrived early. This excess flux shifts to the `trough' following the
36: 2nd X-ray peak during radio pulses that arrive later. Our results
37: suggest that the mechanism producing the radio pulses is intimately
38: connected to the mechanism producing X-rays. Current models using
39: resonant absorption of radio emission in the outer magnetosphere as a
40: cause of the X-ray emission are explored as a possible explanation for
41: the correlation.
42: \end{abstract}
43:
44: \keywords{pulsars: X-ray -- pulsars: radio -- pulsars:
45: individual(\objectname {Vela})}
46:
47: \section{Introduction}
48:
49: The Vela pulsar (PSR B0833$-$45) has been well-studied. Much
50: observational work has been done to understand Vela's emission in
51: individual wavelength regions; e.g. the work of
52: \nocite{Krishnamohan83} Krishnamohan \& Downs (1983, hereafter KD83)
53: in the radio regime, observations by \citet{Ogelman93} in the X-ray
54: regime, and studies by \cite{Kanbach94} in the gamma ray regime.
55: Observations of Vela's spectrum allow for the possibility of both
56: polar cap \citep{Daugherty96} and outer-gap \citep{Cheng00} models of
57: emission. Vela's pulse profiles in individual regions have been
58: phase-aligned using the phase of the radio pulse across the optical,
59: X-ray, and gamma ray wavelength bands \nocite{Harding02} (Harding et
60: al. 2002, hereafter H02). This article works to further relate Vela's
61: high energy emission to its low-energy (radio) emission.
62:
63: X-ray observations of Vela are challenging. Though Vela is the
64: strongest gamma ray source in the sky, the pulsar's spectral power
65: drops off in the hard X-ray band, making its X-ray emission very
66: difficult to detect. Additionally, the pulsed emission is overwhelmed
67: by the bright but unpulsed background of the X-ray emission nebula in
68: which the pulsar is embedded \citep{Helfand01}.
69:
70: The single-peaked pulse profile of the Vela pulsar in radio
71: wavelengths is much simpler than its high energy counterparts,
72: although several studies have revealed compound emission. KD83
73: detected peak-intensity dependent changes in the pulse-shape with the
74: strongest pulses arriving earlier than the averaged profile. They
75: conclude that the radio peak is composed of four different components
76: originating at different heights in the emission cone with components
77: lower in the magnetosphere arriving later. In this article one aspect
78: we explore is whether a similar connection persists in the X-ray
79: regime, i.e. whether the X-ray pulses associated with early-arriving
80: radio pulses have a different shape and/or a different flux than
81: others.
82:
83: Related work on other pulsars includes experiments probing the
84: relationship between the Crab pulsar's ``giant" pulsed emission and
85: its gamma ray emission (Lundgren et al., 1995, hereafter
86: Lu95\nocite{Lundgren95}) or its optical emission (Shearer et al.,
87: 2003, hereafter Sh03\nocite{Shearer03}). They reached opposite
88: conclusions. Lu95 observe no correlation within their sensitivity,
89: indicating that variations in radio flux are caused by changes in
90: radio coherence, which only affects the radio intensity. Sh03 observe
91: a significant correlation, and they thus conclude that the increased
92: emission in the optical and radio is caused by an increased pair
93: production efficiency.
94:
95: Patt et al. (1999)\nocite{Patt99} study pulse-to-pulse variability in
96: the X-ray regime for the Crab pulsar, and they find the pulses to be
97: steady to 7\%. Using this result, as well as previous work showing
98: that the Crab exhibits giant radio pulses roughly every two minutes,
99: they conclude that the radio and X-ray emission mechanisms are not
100: closely related, even though it is likely that the optical and X-ray
101: emission regions exist in the same section of the magnetosphere
102: \citep{Patt99}.
103:
104: Additional experiments linking pulsar emission in the radio and X-ray
105: regimes have been performed by Cusumano et
106: al. (2003)\nocite{Cusumano03} and Vivekanand
107: (2001)\nocite{Vivekanand01}. Cusumano et al. show that in PSR B1937+21
108: there is close phase alignment between X-ray pulses and giant radio
109: pulses, suggesting a correlation in their emission regions.
110: Vivekanand, on the other hand, finds that the X-ray flux variations
111: are so much smaller than those at radio wavelengths that they are
112: inconsistent with the existence of any relationship between the
113: charged emitters in the two wavelength regimes.
114:
115: Giant radio pulses have not been shown to exist in Vela, but Johnston
116: et al. (2001, hereafter J01) \nocite{Johnston01} discovered
117: microstructure and `giant micropulses' in the Vela pulsar. The giant
118: micropulses have flux densities no more than ten times the mean flux
119: density and have a typical pulse width of $\sim 400 \mu s$.
120:
121: By doing a pulse-by-pulse analysis of the Vela pulsar in X-ray and
122: radio wavelengths, we will show in this paper that the Vela pulsar's
123: X-ray and radio emission mechanisms must be related. We will discuss
124: the X-ray and radio observations in
125: \S \ref{sec.observations}, our analysis in
126: \S \ref{sec.analysis}, the effects of scintillation in
127: \S \ref{sec.scintillation}, a discussion of interpretations in
128: \S \ref{sec.discussion}, and finally our conclusions and related future work in
129: \S \ref{sec.conclusion}.
130:
131: \section{Observations}
132: \label{sec.observations}
133:
134: Our data consist of 74 hours of simultaneous radio and X-ray
135: observations taken over three months at the Mount Pleasant Radio
136: Observatory in Tasmania and with the RXTE satellite. The radio data
137: were acquired during 12 separate observations between 30 April and 23
138: August, 1998.
139:
140: The radio data were collected as part of the long term monitoring
141: program of ten young pulsars, including Vela \citep{Lewis03}. These
142: data were collected with the 26m antenna at 990.025 MHz using the
143: incoherently de-dispersed single pulse system (full description in
144: Dodson, McCulloch, \& Lewis, 2002\nocite{Dodson02}). All individual
145: pulses from Vela are detectable, and the pulse height, integrated
146: area, and central time of arrival (for the solar-system barycenter)
147: were calculated from cross-correlation with a high signal to noise
148: template in the usual fashion.
149:
150: The X-ray data were taken during the same three months, yielding 265
151: ks of usable simultaneous observation. For the purposes of this
152: project, only top-layer data from RXTE's Proportional Counter Units
153: (PCUs) in Good Xenon mode in the energy range of 2-16 keV were
154: used. Other filtering parameters included were standard RXTE criteria:
155: elevation was greater than 10 degrees, offset was less than 0.02
156: degrees, the data were taken with at least 3 PCUs on, time since SAA
157: was at least 30 minutes, and electron0 was less than 0.105.
158:
159: \section{Analysis}
160: \label{sec.analysis}
161:
162: We filtered the X-ray photon arrival times and transformed them to the
163: Solar System Barycenter (SSB) using the standard FTOOLS
164: \citep{Blackburn95} package. We calculated the pulsar phase at the
165: time of each X-ray photon, using the radio pulsar-timing program
166: TEMPO$^{10}$, and the ephemeris downloadable from Princeton
167: University\footnote{See http://pulsar.princeton.edu/tempo}. We
168: matched each X-ray photon with the radio pulse that arrived at the SSB
169: at the same time. The precise time span associated with each radio
170: pulse was given by our best model for arrival time of the peak of the
171: radio pulse, $\pm 0.5\times$ the instantaneous pulse period calculated
172: via the model. Photons arriving on the borderline were associated
173: with the earlier pulse. We then compared pulse profiles for X-rays
174: segregated according to the arrival time of the radio pulse.
175:
176: Single radio pulses arrive at a range of times around the predicted
177: arrival time, as KD83 found. The histogram of residual arrival times
178: for radio pulses, relative to the prediction of our best model, is
179: shown in Figure \ref{fig.phasedist}. In the figure, and in our
180: analysis, the average residual from each 5-minute segment of data was
181: subtracted from all the data in that segment in order to account for
182: any systematic wandering of the pulse arrival times relative to our
183: model. The distribution of arrival times is slightly skewed, with a
184: tail at late arrival times, so that the mean of the distribution is
185: slightly later than the mode (at the peak of the histogram), as Figure
186: \ref{fig.phasedist} shows. We divided all of the pulses into 10
187: deciles, by the residual phase of the radio pulse, with equal numbers
188: of pulses in each decile. Figure \ref{fig.phasedist} shows our
189: division of the residual phase of the radio pulse into the deciles.
190: The deciles are well mixed in time, i.e. no particular observing epoch
191: dominates any decile. Removing the 5-minute average residual
192: eliminates effects of long-term trends that might appear independently
193: in the radio pulse-timing and X-ray photon counting data.
194:
195: \begin{figure}
196: \plotone{f1.eps}
197: \caption{The number of radio pulses vs phase relative to a predictive
198: long-term timing model over the entire data set. The dotted lines
199: show the average position of boundaries of the 10 bins that were used
200: to make the 10 profiles shown in Figure \ref{fig.hists}.
201: \label{fig.phasedist}
202: }
203: \end{figure}
204:
205: We formed an X-ray pulse profile, integrated from 2$-$16 keV, for each
206: of the deciles of radio-pulse arrival times, from the X-ray photons
207: associated with each. Figure \ref{fig.hists} shows the 10 resulting
208: X-ray profiles. Each profile contains 13 bins in phase with the radio
209: peak being at the left edge of each plot on the border between bins 13
210: and 1. The X-ray profiles are significantly different; the X-ray
211: pulse changes in shape with the arrival of the radio pulse. We denote
212: the ten X-ray profiles by their ``lateness": profile 1 comprises the
213: decile of X-ray photons from the earliest radio-pulse arrivals, and
214: profile 10 the decile from the latest radio-pulse arrivals. In
215: particular it appears that the first X-ray pulse is sharper and
216: stronger in the earlier deciles. Overall 2 distinct X-ray peaks are
217: visible in each of the first 5 profiles, whereas in profiles 6-10 the
218: two peaks are difficult to distinguish. In order to quantify these
219: changes, we performed a number of statistical tests on these 10
220: profiles including a full Monte Carlo simulation described at the end
221: of this section.
222:
223: \begin{figure*}
224: \plotone{f2.eps}
225: \caption{ Full-period X-ray lightcurves for photons detected during
226: radio pulse arrival times falling in the 10 decile bins shown in
227: Figure \ref{fig.phasedist}.
228: \label{fig.hists}
229: }
230: \end{figure*}
231:
232: We observe no significant trend in total X-ray flux with
233: lateness. Figure \ref{fig:total_counts} shows total counts for each of
234: the 10 deciles. From these data, we determine Pearson's correlation
235: coefficient of $r=-.26$ of total X-ray counts with radio-pulse
236: lateness with a confidence interval of 54\%. Nominally, this is not
237: distinguishable from the hypothesis of zero correlation. Note,
238: however, that Pearson's $r$ is not necessarily the best statistic for
239: this comparison, as we discuss below.
240:
241: \begin{figure}[b]
242: \plotone{f3.eps}
243: \caption{
244: Total X-ray counts vs `lateness' decile.
245: \label{fig:total_counts}
246: }
247: \end{figure}
248:
249: We do find that the total counts reported in Figure
250: \ref{fig:total_counts} are inconsistent with the Gaussian distribution
251: expected for this limit of the Poisson distribution produced by shot
252: noise, at 99.7\% confidence, as determined from a $\chi^2$ test. This
253: suggests that there is indeed an evolution of X-ray profile with radio
254: lateness. If we consider the counts in individual bins, we find that
255: they do deviate from the mean more than one would expect from Poission
256: statistics, as well. A simple $\chi^2$ test determines that in bins 2
257: and 3 we can reject the hypothesis of Poisson noise around the mean
258: value with 81\% and 83\% confidence respectively. In other words,
259: bins 2 and 3 show larger changes than we would expect a priori. The
260: changes among the 10 profiles in each of the other 11 lateness bins
261: are consistent with Poisson noise.
262:
263: The $\chi^2$ test, however, is not suited to detecting trends. In
264: order to detect trends we look to Pearson's correlation coefficient,
265: $r$ and its associated confidence interval. Pearson's $r$ has limited
266: validity in our case because it usually assumes that both variables
267: are drawn from random distributions with nearly Gaussian statistics.
268: In our case, lateness is deterministic rather than random and
269: Gaussian. Shot noise in the profile is random and Gaussian, but
270: differences of the X-ray profile among deciles need not be.
271: Nonetheless, this is a useful first step to take to estimate the
272: significance of any trends. Note also that Pearson's $r$ itself,
273: though widely used to measure the strength of a correlation, does not
274: judge the existence of a correlation. We therefore place more stake
275: in the confidence interval, although this involves additional
276: assumptions about the variables correlated \citep{NR14.5}. If our
277: visual impression of the profiles is accurate we would expect that in
278: the trough, made up of bins 12, 13 and 1, the correlations are
279: positive (increasing counts with increasing lateness). All 3 bins do
280: indeed show positive correlations, 0.29, 0.66 and 0.35 respectively
281: with associated confidence intervals of 58\%, 96\% and 68\%. In other
282: words, 2 of the 3 bins show a significant (greater than 1$\sigma =
283: 65$\%) correlation. Likewise we would expect that the first X-ray
284: peak, made up of bins 2 and 3, would show negative correlations
285: (decreasing counts with increasing lateness). Bins 2 and 3 do in fact
286: show negative correlations of $-0.73$ and $-0.48$ with 98\% and 85\%
287: confidence, consistent with our interpretation of the first peak
288: decreasing with lateness. The only other bin that displays a
289: significant correlation is bin 8, which shows a negative correlation
290: of $-0.63$ with 95\% confidence.
291:
292: \begin{figure*}
293: \plottwo{f4a.eps}{f4b.eps}
294: \caption{
295: Left: $Y_a = $bin2 + bin3 $-$ bin12 $-$ bin13 vs the lateness of the radio
296: pulse.
297: Right: $Y_b = $bin2 + bin3 $-$ bin13 $-$ bin1 vs the lateness of the radio
298: pulse.
299: \label{fig:plot_compare_10}
300: }
301: \end{figure*}
302:
303: Next we looked to see if pairs of bins could be combined together to
304: better detect the signal. Based on our results described above we
305: were curious about whether the quantity $Y_{i,a} =
306: (c_{i\,2}+c_{i\,3})-(c_{i\,12}+c_{i\,13})$ or $Y_{i,b} =
307: (c_{i\,2}+c_{i\,3})-(c_{i\,13}+c_{i\,1})$ would show significant
308: trends with lateness. Here $c_{i\,1}, c_{i\,2}, c_{i\,3}, c_{i\,12},
309: c_{i\,13}$ are the measured counts in bins 1, 2, 3, 12, and 13, at
310: lateness $i$. $Y_{i,a}$ and $Y_{i,b}$ effectively measure, for the 10
311: profiles, the height of the peak minus the trough and are shown in
312: Figure \ref{fig:plot_compare_10}. Via a simple $\chi^2$ test we can
313: reject with 97\% confidence the hypothesis that the parent
314: distribution of either of the $Y$'s is a constant at the mean value.
315: More importantly Figure \ref{fig:plot_compare_10} shows a clear
316: systematic trend of $Y$ with lateness. A line fitted to the data and
317: its associated Poisson uncertainty shown in Figure
318: \ref{fig:plot_compare_10} yields a slope of $m=-297$ and $-308$
319: respectively with a formal 1-standard deviation uncertainty of
320: $\sigma=76$ for both, so both represent (not independently) $\sim 4 $
321: standard deviation detections of a trend in these data. For the sake
322: of completeness we tried all possible differences of pairs of adjacent
323: bins. The next highest magnitude slope was 208, but this and all other
324: significant correlations included some subset of bins 12, 13, 1, 2,
325: and 3. The correlation coefficients and associated confidence
326: intervals for $Y_{i,a}$ and $Y_{i,b}$ are $-0.91$ at 99.97\% and
327: $-0.84$ at 99.74\%.
328:
329:
330: Given the limited validity of the Pearson's $r$ coefficient with
331: non-random, non-normally distributed data, we sought a more rigorous
332: estimation of the possibility that such a significant trend would
333: arise amongst 10 such profiles merely by chance, i.e. merely by random
334: fluctuation of each bin around its mean. To answer this question we
335: performed a Monte Carlo simulation with $10^5$ realizations of the
336: data. Each simulated data set consisted of 10 profiles, each of 13
337: bins. The counts in each bin were chosen as a Gaussian deviate of the
338: X-ray pulse profile, averaged over lateness, and with variance set by
339: Poisson statistics. Thus, in the simulated data sets all differences
340: among the 10 profiles arose entirely from counting statistics. In
341: each simulated data set, for every possible pair of summed adjacent
342: bins (\{1,2 \& 3,4\}, \{1,2 \& 4,5\} ...\{j,k \& n,o\}, with bins
343: \{j,k\} adjacent and bins \{n,o\} adjacent: 66 pairs in all) we
344: computed $Y_i$:
345: $$
346: Y_i = c_{ij}+c_{ik} - (c_{in} + c_{io})
347: $$
348: where $i=1$ corresponds to the earliest profile in terms of radio
349: phase, and $c_{ij}$ is the number of counts in the jth bin of the ith
350: profile. For the set of points \{$i$, $Y_i$\} where $m$ and $\sigma$
351: are the slope and uncertainty of the best-fit straight line we
352: computed the following statistic:
353: $$
354: \Gamma = (m/\sigma)^2
355: $$
356: As for the fit to our data described above, we weighted the data by
357: the uncertainties as given by Poisson statistics. We then compared
358: the largest $\Gamma$ for each simulated data set to the largest
359: $\Gamma$ for the actual data (16.4) and found a probability of 0.0024
360: that the correlation we observe could have been obtained by chance.
361:
362: We tried other statistics to see if we could find one more
363: sensitive to the presence of a correlation like that we observed.
364: The most sensitive one included Pearson's
365: $r$ as follows:
366: $$
367: \Gamma^\prime = r^2 + (m/\sigma^\prime)^2
368: $$
369: where $\sigma^\prime$ is the $\sqrt{{\rm mean}(c_{ij}+c_{ik},
370: c_{il}+c_{im})}$. $\Gamma^\prime$ yielded a significance of 0.0007.
371: $\Gamma^\prime$ is more sensitive than $\Gamma$ to the proximity of
372: our data to the fitted line, so it yields a smaller probability of
373: false detection. Regardless, we retain the more straightforward
374: $\Gamma$ as a conservative estimate of the significance of our result.
375:
376: Of use in interpreting our results is knowledge of the character of
377: the radio residuals by which we are binning the X-ray photons. KD83
378: did much work in this area, but one question they do not address
379: directly is the following: if a particular pulse is ``early", what is
380: the chance that the next pulse will also be early? We calculated the
381: autocorrelation function of radio residual, shown in figure
382: \ref{fig:autocorr}. The function is normalized by the autocorrelation
383: at zero lag. The function at a lag of 1 pulse is represented by the
384: left edge of the plotted curve, at a value of 0.066. The figure shows
385: that the pulsar has very little ``memory" of the lateness of the
386: previous pulse, although it is interesting to note there is finite
387: correlation out to 40 pulse periods.
388:
389: \begin{figure}[b]
390: \plotone{f5.eps}
391: \caption{
392: Autocorrelation of radio residual vs lags (pulses).
393: \label{fig:autocorr}
394: }
395: \end{figure}
396:
397: \section{Scintillation}
398: \label{sec.scintillation}
399:
400: In contrast to effects intrinsic to the pulsar, scintillation is
401: unlikely to produce the observed association, because it does not
402: affect X-rays; scintillation might erase such a correlation but it
403: cannot introduce it. Diffractive scintillation for Vela at our
404: frequency has a characteristic bandwidth of 2 kHz (Gupta 1995), far
405: smaller than our observing bandwidth of 6.4 MHz. We therefore average
406: over $\sim$3200 independent scintillation elements. Refractive
407: scintillation modulates flux density over a wide bandwidth and has
408: timescale $\sim$25 days \citep{Desai92}. This was shorter than the
409: total span of the observations, but much longer than the span of any
410: one observation. To combat any effect that this might have on the
411: observed arrival times, we defined the 10 deciles separately for each
412: observation; i.e. the specific values of residual that separated each
413: of the 10 bins were calculated for each individual radio
414: observation. Again, these cutoffs were defined in such a way that an
415: equal number of radio pulses was associated with each decile.
416:
417: We assumed that the pulsar was stationary relative to a refractive
418: scintillation element during each single observation, and further
419: tests to ensure that the length of the observation was not a factor
420: were performed by renormalizing the cutoffs using both one hour and
421: five minute timespans. Reanalyzing the X-ray data in one hour segments
422: did not significantly change our results, and the five minute spans
423: were found to be too short to accurately represent Vela's emission.
424:
425: \section{Discussion}
426: \label{sec.discussion}
427:
428: The early radio emission could result from coherent radiation along a
429: different set of field lines (i.e. more leading) or from radiation at
430: a higher altitude, or both. The work of KD83 suggests both. The early
431: radio emission may also result from stochastic fluctuations in the
432: radio beam intensity which would lead to pulse arrival time changes by
433: changing the shape of the radio beam as it takes a finite time to
434: sweep across our line of sight.
435:
436: Petrova \nocite{Petrova03} (2003, and references therein) offers a
437: physical model that could explain the observed relationship between
438: the radio and X-ray emission. Her model suggests that resonant
439: absorption of radio emission from the outer magnetosphere leads to an
440: increase in the pitch angles and momenta of the secondary pairs, which
441: then leads to optical and higher energy emission by spontaneous
442: synchrotron radiation. How could this model produce a changing
443: X-ray pulse shape? Due to the effects of rotation of the
444: magnetosphere and aberration, the early radio emission can cross a
445: larger number of higher altitude field lines and at larger angles,
446: thereby maximizing the opportunity for absorption by particles on
447: those field lines, and therefore the production of high-energy
448: radiation. Conversely a late radio photon on a path almost directly
449: along the magnetic pole may escape the magnetosphere with many fewer
450: interactions, since it will cross fewer open field lines and at
451: smaller angle. The details of the above are somewhat unimportant as
452: our knowledge of the magnetosphere and the plasma therein is limited,
453: but the point is that as different parts of the radio beam are active,
454: resonant absorption may happen at different rates in different parts
455: of the magnetosphere, causing continuous change in the shape of the
456: observed X-ray emission.
457:
458: More generally our results imply a connection between the radio and
459: X-ray emission mechanisms for Vela that is not consistent with outer
460: gap models. In these models, the high energy emission results from a
461: gap connection to the pole opposite from that producing the radio
462: emission. The pole and outer gap associated with the same set of field
463: lines are not visible to one observer. It is not clear how a
464: correlation could exist between the radio and high energy regimes in
465: these models.
466:
467: The giant micropulse emission observed by J01 would cause a single
468: radio pulse to arrive about 1 ms early, so it is realistic to consider
469: the possibility that the giant micropulse emission is primarily
470: responsible for the early arrival of the radio pulse. However, out of
471: 20,085 pulses that J01 observed at 1413 MHz, 14 of them contained
472: giant micropulses. So the giant micropulses may be influencing the
473: first of our 10 deciles, but cannot contribute to the effect observed
474: in the other 9 deciles. We conclude that Vela's giant micropulse
475: emission cannot be responsible for the effect presented here.
476:
477: \section{Conclusions and Future Work}
478: \label{sec.conclusion}
479:
480: We have evidence that links features of Vela's X-ray emission with
481: features of its radio emission. First, we find that X-ray pulses
482: associated with early radio pulses show stronger emission at the main
483: X-ray peak which is the sharper of the two. Similarly X-ray pulses
484: associated with later radio pulses show stronger emission at the
485: trough following the 2nd X-ray peak, a region in phase near the radio
486: peak. The trend we measure has a 0.2\% probability of appearing in
487: the data by chance. We conclude that there is a close relationship
488: between X-ray and radio emission in the Vela pulsar.
489:
490: We plan to further characterize the relationship between the radio and
491: high energy emissions of pulsars to identify their origins and
492: constrain magnetospheric models. In particular, we will explore the
493: dependence of radio-to-X-ray correlations on the radio frequency and
494: polarization properties of individual Vela pulses, both of which carry
495: information about emission altitudes. Similar observations of other
496: pulsars also promise useful insights as probes of different magnetic
497: field strengths and emission/viewing geometries.
498:
499: \acknowledgments
500:
501: Many thanks to Ben Stappers, Michael Kramer, Russell Edwards, Wim
502: Hermsen, David Helfand, Paul Ray, Simone Migliari and Tiziana Di Salvo
503: for helpful comments. AL is grateful for the hospitality of the
504: Australia Telescope National Facility and for a Research Corporation
505: Grant in support of this research. ZA was supported by NASA grant
506: NRA-99-01-LTSA-070. CG acknowledges support of NSF-AST-9731584. AH
507: acknowledges support from the NASA Astrophysics Theory Program. RD
508: acknowledges support as a Marie-Curie fellow via EU FP6 grant
509: MIF1-CT-2005-021873. This research has made use of data obtained from
510: the High Energy Astrophysics Science Archive Research Center
511: (HEASARC), provided by NASA's Goddard Space Flight Center.
512:
513: \input{ms.bbl}
514:
515: \end{document}
516: