Radio Observations of PSR B1259-63 through the 2004 periastron passage

0.36 MB·English

by  Simon Johnston

Mon.Not.R.Astron.Soc.000,1–7(2004) Printed2February2008 (MNLATEXstylefilev1.4)
Radio Observations of PSR B1259–63
through the 2004 periastron passage
Simon Johnston1, Lewis Ball2, N. Wang1,3 & R. N. Manchester3
1School of Physics, University of Sydney, NSW 2006, Australia.
2Australia Telescope National Facility, CSIRO, P.O. Box 276, Parkes, NSW2870, Australia
3Australia Telescope National Facility, CSIRO, P.O. Box 76, Epping, NSW1710, Australia
5
0
0 2February2008
2

n
a ABSTRACT
J We report here on extensive radio observations of the binary system containing
1 PSR B1259−63 and the Be star SS 2883, made around the time of the 2004 peri-
3 astron. This is the fourth periastron to have been observed in detail. As in previous
observations, changes in the pulsar’s dispersion and rotation measures are detected

1 overaperiodspanning200days.Weshowthattheeclipseofthepulsedemissionlasts
v from16dayspriortoperiastronto15daysafterperiastronandisconsistentfromone
0 periastron to the next. We demonstrate that the timing solution proposed by Wang
6
et al. (2004) provides a good fit through the 2004 periastron. The light curve of the
6
transient unpulsed radio emission is broadly similar from one periastron to the next.
1
For this periastron, however, the light curve is strongly peaked post-periastron with
0
ratherlittle enhancementpriortoperiastron,incontrasttothe2000periastronwhere
5
0 thepeakfluxdensitiesweremoreequal.Theseobservationsremainconsistentwiththe
/ interpretation that the pulsar passes through the dense circumstellar disk of the Be
h
starjustbeforeandjustafterperiastron.Theobserveddifferencesfromoneperiastron
p
to the next can be ascribed to variations in the local disk density and magnetic field
-
o structure at the time the pulsar enters the disk.
r
t Key words: pulsars:individual: PSR B1259−63
s
a
:
v
i
X
1 INTRODUCTION andfallsoffasapower-lawwithdistancefromthestar.The
r disk is likely to be tilted with respect to the pulsar orbital
a
PSR B1259−63 was discovered in a large-scale high fre- plane, and PSR B1259−63 is eclipsed for about 35 days as
quencysurveyoftheGalacticplane(Johnstonetal.1992a). it goes behind the disk.
Itisuniquebecauseitistheonlyknownradiopulsarinorbit This pulsarhasbeen extensivelyobserved sinceitsdis-
about a massive, main-sequence, B2e star (Johnston et al. covery 15 years ago. Determining the timing properties of
1992b). PSR B1259−63 has a short spin period of ∼48 ms the pulsar has proved a difficult task. Initial timing solu-
and moderate period derivative of 2.28×10−15, implying a tions obtained by Manchester et al. (1995) and Wex et al.
high spin-down energy of 8.2×1035 ergs−1 and a charac- (1998) failedtoaccuratelydescribethedatathroughsubse-
teristic age of only 330 kyr. It has a long orbital period of quent periastron passages. The most recent update on the
∼1237days,andalargeeccentricityof0.87,withaprojected timingofPSRB1259−63byWang,Johnston&Manchester
semi-major axis, asiniof ∼1300 light-s. (2004) provides a comprehensive review of the many diffi-
Thecompanion,SS2883,isa10thmagnitudestarwith cultiesinvolved.Theirmodelshowsthatthepulsarglitched
a mass of about 10 M and a radius of 6 R . Assuming inlate1997andthatthetimingresidualsarebestmodelled
⊙ ⊙
the pulsar mass is 1.4 M implies the inclination of the byincluding a jump in asini at each periastron.
⊙
orbit to the plane of the sky is i ∼ 36◦. Be stars as a class Transientunpulsedradioemissionseenaroundthetime
have a hot, tenuous polar wind and a cooler, high density, of periastron has been discussed by Johnston et al. (1996),
equatorialdisk.TheexpectedmasslossratefromaB2estar Johnstonetal.(1999)andConnorsetal.(2002),withphysi-
is∼10−6 M .Johnstonetal.(1994) observedHαemission calmodelsfortheemissionproposedbyMelatos,Johnston&
⊙
lines from the disk at 20 stellar radii (R ), just inside the Melrose(1995),Balletal.(1999)andConnorsetal.(2002).
∗
pulsar orbit of 24 R at periastron. The density of the disk UnpulsedemissionhasalsobeendetectedintheX-ray(sum-
∗
material is high near the stellar surface (108−1010 cm−3), marisedinHirayamaetal.1996) andinthesoftγ-rayband
(cid:13)c 2004RAS
2 Johnston et al.
byGroveetal.(1995)andmorerecentlyShawetal.(2004),
Table1.DMandRMvariationsandtheinferredmagneticfield
withmodellingoftheemissiondescribedinTavani&Arons forthe2004periastronobservations.TheerrorinDMistypically
(1997). 0.2cm−3 pc.
Kirk, Ball and Skjaeraasen (1999) proposed that the
systemshouldbedetectableathighenergiesthroughinverse Date Day ∆DM RM B
k
Compton scatteringoftheUVphotonsfrom theBestarby (cm−3pc) (radm−2) (mG)
the relativistic pulsar wind. This model has largely been
confirmed through a recent detection of the system at TeV 2004Jan27 −39.8 3.0 −3100±300 −1.3±0.1
2004Jan29 −37.7 2.5 −3800±400 −1.9±0.2
energies byAharonian et al. (2004).
2004Jan31 −35.7 3.4 +1700±200 +0.6±0.1
2004Feb02 −33.7 5.6
2004Feb04 −31.7 4.0 −1500±200 −0.5±0.1
2004Feb06 −29.7 3.9
2 OBSERVATIONS
2004Feb08 −27.7 5.9
2004Feb10 −25.7 4.9 +11500±1100 +2.9±0.3
2.1 Pulsed emission
2004Feb12 −23.7 5.3
Observations of the pulsar were carried out with the 64- 2004Feb14 −21.7 5.5
m radio telescope located in Parkes, NSW, at frequencies 2004Feb17 −18.8
2004Feb18 −17.8 19.5
between 680 and 8640 MHz.
2004Feb20 −15.8
Two independent backend systems were employed; fil-
2004Mar23 16.1 3.2
terbanksystemswhichrecordedtotalpowerandacorrelator
2004Mar25 18.1
systemwhichretainedfullpolarisation information.Thefil-
2004Mar28 21.2 1.0
terbanksystemwasusedat680and1500MHzonly.Atthe 2004Mar29 22.1 1.2
lower frequency it consists of 256 frequency channels each 2004Mar31 24.1 0.8 +6000±600 +10±3
of width 0.250 MHz for a total bandwidth of 64 MHz and 2004Apr02 26.2 0.7 −2500±300 −4.5±1.4
at the higher frequency consists of 512 frequency channels 2004Apr04 28.2 0.5 −280±30 −0.7±0.3
each of width 0.5 MHz for a total bandwidth of 256 MHz. 2004Apr08 32.2 0.4 −500±50 −1.4±0.7
In each case the analogue signal is one-bit digitised every 2004Apr14 38.2 0.4 −1080±100 −3.3±1.6
2004Apr22 46.1 0.2 −360±40 −2±2
250µsandwrittentotapeforoff-lineanalysis.Thisanalysis
involvedde-dispersionandfoldingatthetopocentricperiod
toproduceapulseprofile.Correlatordatawereobtainedat
frequencies around 1370, 3100 and 8640 MHz with a total
bandwidthof256,512and512MHzatthethreefrequencies.
Channelbandwidthswere 1MHzand therewere 512 phase date of the observation and the second column shows the
bins across the pulsar period. The data were folded on-line offset in days from the 2004 periastron epoch (which we
foranintervalof60sandwrittentodisk.Off-lineprocessing denoteas T hereafter).
usedthePsrchivesoftwareapplication(Hotan,vanStraten
& Manchester 2004) specifically written for analysis of pul-
sar data. Processing involved calibration and de-dispersion
and yielded full Stokespulse profiles.
In a typical observation, data were obtained at 680,
2.2 Unpulsed emission
1500, 3100 and 8640 MHz in the space of 3 h. The time-
of-arrival (ToA) of the fiducial point in the pulsar profile Observations were also made with the Australia Telescope
was calculated for each observation by convolving with a Compact Array (ATCA), an east-west synthesis telescope
standard template which is frequency dependent.The tem- located nearNarrabri,NSW,whichconsists ofsix22-m an-
plates at different frequencies were aligned in pulse phase tennas on a 6 km track. ATCA observations can be made
as shown in Figure 1 of Wang et al. (2004). Any offset be- simultaneouslyateither1.4and2.4GHzor4.8and8.4GHz
tweenthearrivaltimesatthedifferentfrequencieswasthen withabandwidthof128MHzateachfrequencysubdivided
attributed to a change in dispersion measure (DM) of the into 32 spectral channels, and full Stokes parameters. The
pulsar and a fit was doneto obtain the offset DM. ATCA is also capable of splitting each correlator cycle into
The polarisation information allowed us to obtain the bins corresponding to different phases of a pulsar’s period,
rotation measure (RM) for each observation. There are two and in our case the pulse period of ∼48 ms was split into
possiblemethodsforobtainingtheRM.Thefirstistomea- 16 phase bins. This allows a measurement of off-pulse and
surethechange in position angle across theband.This can on-pulseflux densities to bemade simultaneously.
bedoneeitherat agiven pulsarphaseorbytimeaveraging Initialdatareductionandanalysiswerecarriedoutwith
across several phase bins. The second method is to max- the Miriad package using standard techniques. After flag-
imisethelinearlypolarisedfluxbychoosingtrialRMs.This gingbaddata,theprimarycalibratorwasusedforfluxden-
method works well for low signal to noise ratios as it uses sity and bandpass calibration and the secondary calibrator
the entire pulse window. In fact, for PSR B1259−63, there wasusedtosolveforantennagains,phasesandpolarisation
is very little change in the position angle across each com- leakageterms.Aftercalibration,thedataconsistof13inde-
ponent,and thetwo methods yield similar results. pendentfrequencychannelseach8MHzwideforeachofthe
Table 1 shows the log of the observations of the pulsar 16 phase bins. Subsequent analysis of the data was carried
made with the Parkes telescope. The first column gives the out as described in Connors et al. (2002).
(cid:13)c 2004RAS,MNRAS000,1–7
Radio Observations of PSR B1259–63 3
Figure1.Post-fittimingresidualsforPSRB1259−63from1991
Januaryto2004Septemberwithamodelincludingjumpsinasini Figure 2. Dispersion Measure variations in PSR B1259−63
ateachperiastron.Thermsresidualis350µs.Dashedlinesmark aroundthe2004periastron.TheDMcontributionfromtheinter-
stellar medium is assumed to be 146.7 cm−3pc. The error bars
theperiastronepochs.
aretypically0.2cm−3pcandaresmallerthanthesymbols.
3 RESULTS
3.1 Pulsed emission 3.1.3 RM variations
3.1.1 Timing Column 4 of Table 1 shows the RM as a function of epoch.
The typical error bars are of order 10 per cent. The RM
Figure 1 shows the complete timing residuals from
changessignificantlybothinmagnitudeandinsignbetween
PSRB1259−63.Therearemorethan1200independenttim-
the observations but there is little evidence of a change in
ing points, with a data span from 1991 January to 2004
RM within the duration of a single observation. On Feb 2
September including five periastron passages. As described (T−34)wewereunabletomeasureanRMevenat8.4GHz,
in earlier papers, the DM of the pulsar changes near peri- onT−30thepolarisation qualityofthedataispoorandon
astron and this extra DM needs to be accounted for in the T−28thepulsarisveryweakandnoRMinformationcould
timingsolution.Forthisperiastron,observationsweremade be extracted. Following the very large RM value on T−26,
at frequencies between 0.64 and 8.4 GHz and the DM was
the pulsar appeared completely depolarised at all frequen-
calculated by measuring the time offset between the TOAs
cies. After the pulsar re-emerged from the eclipse, no RMs
at thedifferent frequencies (see e.g. Wex et al. 1998).
could initially be measured. A high, positive RM was mea-
In Wang et al. (2004) we determined that the best fit suredonT+24,subsequentvalueswerenegativethereafter.
solution involved adding jumps in the valueof asini at pe-
It is noticeable that the RM varies by more than a factor
riastron and that a small glitch occured at MJD 50691. We
of 10 in the post-periastron observrations and yet the DM
havenowextendedthat modeltocoverthe2004 periastron
variesonlybyafactorof2.Therearethreepotentialexpla-
passage,obtainingsuccessivejumpsinasiniof60.7,−26.3,
nationsforthedepolarisationofthesignal.Itispossiblethat
2.8, 4.2 and −7.8 ms. The fitted jumps for the first 4 peri-
theRMissohighthatevenacrossasinglefrequencychannel
astronsarewithinafewpercentofthoseobtainedbyWang
the position angle varies significantly, thus depolarising the
et al. (2004). signal. This would imply an RM greater than 105 rad m−2
for the 8.4 GHz observations. Secondly, the RM could be
highly variable on the timescale of a few minutes, causing
3.1.2 DM variations
depolarisation when time averaging occurs. Finally, there
Column 3 of Table 1 lists the DM variations and these are may be different values of RM along different ray paths,
displayedinFigure2.ThetypicalerrorintheDMvaluesare especially when the pulsar becomes scatter broadened. We
0.2cm−3pc.Althoughthepulsarwasundetectedat1.4,3.1 believe it is likely that all three occur close to the point of
and8.4GHzonT−18.8,thelastdetectionbeforeperiastron eclipse of thepulsar.
occurredonT−17.8whentheDMchangewas19.5cm−3pc. TheRMvaluesmeasuredaresignificantlydifferentfrom
SubsequentobservationsonT−15.8failedtodetectthepul- those obtained by Connors et al. (2002) for the 2000 pe-
sar. The eclipse lasted until T+16.1 although on this date riastron. In 2000 the pulsar was depolarised from T−46
the pulsar was not detected at 1.4 GHz. Two days later, onwards, whereas in 2004 polarisation was detected up to
there is a marginal detection of the pulsar at 3.1 GHz but T−27. Large and variablepolarisation was detected follow-
not at higher or lower frequencies. After the exit from the ing periastron in both 2000 and 2004, although the large
eclipsetheDMchangewasonly3.2anddecayedoveranin- valueof−7700radm−2 measuredin2000wasnotrepeated
tervalof∼20days.Thisisbroadlyinlinewithchangesseen in 2004.
during previous periastron passages (see Figure 2 in Wang ArmedwithboththeDMandtheRMwecancompute
et al. 2004). themagneticfieldparalleltothelineofsight,B inmGvia
k
(cid:13)c 2004RAS,MNRAS000,1–7
4 Johnston et al.
Table 2. Flux densities for pulsed emission and total emission (pulsed plus unpulsed) for all 15 observations made with the ATCA
in 2004. Total flux densities and errors were obtained from the miriad routine imfit. Pulsed flux densities and errors were obtained
using the routine uvflux. Pulsar flux densities are omitted when no significant pulses could be seen. The last two observations used
ultra-compactarraysandtheimagingwastoopoortoobtainreliablefluxdensityestimates.
ObservingFrequency
Date Day 1384MHz 2496MHz 4800MHz 8400MHz
Pulsar Total Pulsar Total Pulsar Total Pulsar Total
(mJy) (mJy) (mJy) (mJy) (mJy) (mJy) (mJy) (mJy)
Feb14 −21.5 2.4±0.2 4.7±0.5 3.3±0.2 4.8±0.5 2.5±0.2 2.9±0.5 0.8±0.2 0.8±0.3
Feb21 −14.7 — 12.4±2.3 — 12.2±0.5 — 7.8±0.6 — 5.0±0.5
Feb27 −8.7 — 14.7±2.2 — 13.6±0.5 — 10.4±0.6 — 7.3±0.7
Mar04 −2.8 — 16.9±2.3 — 12.0±0.6 — 9.9±0.7 — 5.8±0.8
Mar10 +3.2 — 13.2±1.8 — 14.2±0.5 — 10.5±0.8 — 7.6±0.6
Mar15 +8.2 — 12.5±0.9 — 9.8±0.7 — 4.8±1.4 — 1.5±0.7
Mar23 +16.1 — 20.0±1.0 0.6±0.2 17.8±0.5 0.6±0.2 10.2±1.8 0.2±0.2 4.3±1.0
Mar28 +21.1 3.2±0.2 51.2±2.9 3.6±0.2 44.3±2.0 2.4±0.2 27.5±2.9 1.9±0.2 14.9±3.2
Apr04 +28.1 5.1±0.2 27.0±1.4 3.6±0.2 22.0±0.7 1.6±0.2 13.0±1.1 1.1±0.2 7.5±1.2
Apr09 +33.5 4.6±0.2 27.6±3.6 3.1±0.2 9.7±2.0 1.5±0.2 8.5±2.0 0.8±0.2 4.0±1.5
Apr14 +38.5 5.9±0.2 20.5±3.8 5.7±0.2 19.4±2.6 2.7±0.2 7.5±1.8 1.8±0.2 3.8±0.7
May10 +64.2 2.2±0.2 10.4±0.7 2.5±0.2 10.7±0.5 1.1±0.2 5.8±0.9 0.5±0.2 3.7±0.7
Jun09 +94.0 3.2±0.2 11.1±1.8 2.8±0.2 4.0±0.6 1.4±0.2 1.3±0.3 0.6±0.2 0.2±0.2
Aug04 +150.1 6.7±0.2 — 4.6±0.2 — 2.2±0.2 — 0.3±0.2 —
Sep09 +186.1 4.8±0.2 — 2.8±0.2 — 2.4±0.2 — 1.6±0.2 —
B =1.232×10−3 RM (1) it subsequently decreased more slowly. The unpulsed emis-
k ∆DM sionwasstilldetectableatlevelsofafewmJyatthetimeof
ValuesofB areshownincolumn5ofTable1.Theerrorbar thefinalobservationaroundT+65.Thereissomesuggestion
k
in B is dependent on the RM error for the pre-periastron of absorption at frequencies below 2.4 GHz in the emission
k
data and theDM error for thepost-periastron data. observed at the earliest times, T−22 and T−15. At other
epochs thespectrum is well fitted bya simple power law of
theform Sν =Cνα with α∼−0.6.
3.2 Unpulsed emission
Table2liststhepulsedfluxdensityandthetotalfluxdensity
4 COMPARISON OF PERIASTRON
detected at the four frequencies observed with the ATCA.
PASSAGES
Dashes indicate that the pulsar was not detected at that
epoch. The dates of these observations were carefully se- The top panel of Figure 3 shows the light curve for each of
lected to provide good sampling of the transient unpulsed the four frequencies from the 2004 periastron. Below that
emissionbasedontheobservationsofthepreviousthreepe- panel we show the data from the three preceding perias-
riastronpassages.ThenondetectionofthepulsaratT−14.7 tron passages in 1994, 1997 and 2000. Figure 4 shows the
is consistent with the lack of pulsed emission observed at unpulsed emission at 1384 MHz from all four periastron
Parkes one day earlier. The re-appearance of the pulsar at passages superposed, to allow for easy comparison of the
T+16.1at thehigherfrequenciesisalsoconsistentwith the observed light curves.
Parkes observations. The similarities between the four periastron passages
Figure3showsthefluxdensityofthetransientunpulsed are striking. In the three cases where pre-periastron obser-
emission detectedfrom thePSR B1259−63 system through vationswereobtained,theunpulsedemissionwasdetectable
the 2004 periastron passage (top panel), together with the by T−20, subsequently increased to a maximum around
corresponding data from the previous three periastron pas- T−7 and then decayed slowly to a local minimum around
sages(lowerpanels).Althoughthesamplingratewasnotas T+10. In all four cases the flux density after periastron in-
highin2004asineither1997or2000,itisunlikelythatany creasedfromalocalminimumaroundT+10toamaximum
significant feature of thelight curvehas been missed. closetoT+20,thendecreasedrelativelyrapidlyuntilT+30,
Unpulsed emission was detected at a low level on and subsequentlydeclined more slowly.
the first day that observations were made, at T−22. The The general behaviour of the unpulsed emission has
flux density at all four frequencies observed then increased been interpreted as resulting from the interaction of the
steadily, reaching a maximum (∼17 mJy at 1.4 GHz) be- pulsarandits wind with thedensedisk surroundingtheBe
tweenT−10andT−3,andsubsequentlydecreasedtoamin- starcompanion(Balletal.1999).Thepulsarpassesthrough
imum around T+7. The flux density at all four frequencies the disk twice, once a few days before periastron and then
then increased quite rapidly peaking around T+22 (nearly againafewdaysafterward.Theunpulsedemissionhasbeen
50 mJy at 1.4 GHz), significantly higher than the level at- broadly interpreted as resulting from two emission regions
tained prior to periastron. Just 5 days later, at T+27, the associated with these interactions.
fluxdensityhaddroppedtoaroundhalfitspeakvalue,and However, there are also significant differences between
(cid:13)c 2004RAS,MNRAS000,1–7
Radio Observations of PSR B1259–63 5
Table 3. Observations bracketting the disappearance and reap-
pearanceofpulsedemissionatthefourperiastronepochs.Aster-
isksindicatedaysonwhichthesourcewasbothdetectedandnot
detected inseparateobservations.
periastron daynumber
epoch last firstnon- lastnon- first
detection detection detection detection
1994Jan9 −20.1 − 13.9 23.9
1997May29 −18.9 −17.5 13.2 15.2*
2000Oct17 −18.4 −16.5 13.3 18.5
2004Mar7 −18.3 −15.3 13.7 15.7*
the light curves at the four periastron passages which ap-
pear to be robust despite the different data sampling inter-
vals. In 1997 and 2000 the flux density rises rapidly to its
pre-periastron value over ∼ 10 days. In 2004 the increase
apparentlyoccurs more slowly, at least at 1.4 and 2.4 GHz,
but this could be due to decreasing optical depth. In 1997
and2004thefirstmaximumisnotwelldefined,butin2000
it appears as a distinct peak between T−10 and T−8. In
1997and2004thepeakfluxdetectedafterperiastronisap-
proximately twice that seen before periastron, but in 2000
the peak before periastron is comparable to that after pe-
riastron. In 1994, 2000 and 2004 the peak after peristron
occurs between 20 and 25 days after periastron, while in
1997 the post-peristron peak is significantly earlier, occur-
Figure3.Transientradioemissionaroundthetimeofperiastron ring between T+15 and T+20.
forfourdifferentperiastronpassages.Eachplotdisplaysfourdif- These differences in the unpulsed emission most likely
ferentfrequencies,1384,2496,4800and8640MHz,inorderfrom reflect differencesin theproperties of thedisk at thesiteof
top down. The1994 and1997 data werepreviouslypublishedin the pulsar crossings. The behaviour of the pulsed emission
Johnston et al. (1999) and the 2000 data published by Connors providesacoarseprobeofthediskgeometry,sincethedisap-
etal.(2002). pearanceofpulsedemissionpre-periastron,anditsreappear-
ance post-periastron, are interpreted as marking the times
when the pulsar passes into the disk and then re-emerges
frombehindit.Theepochsoftheobservationswhichbracket
the start and end of the eclipse of the pulsed emission are
summarised in Table 3.
The best constrained are the epochs of re-appearance
after peristron in 1997 and 2004. On both these occasions
observations with the Parkes telescope on T+15 included
both detections and non-detections of the pulsed emission
—consistentwiththatbeingtheactualdayofre-emergence.
The epoch of the pulsar disappearance before periastron in
1997isalsowellconstrained.ParkesobservationsonT−18.9
detectedthepulsarand ATCAobservations onT−17.5 did
not.Unfortunatelytheothereclipseepochsarelesswellde-
termined, but all are consistent with an eclipse starting on
T−18 and ending between T+14 and T+15.
Comparisonoftheepochsofthepulsareclipsewiththe
light curve of the unpulsed emission suggests the following.
Just before the pulsar goes into the eclipse prior to peri-
Figure 4. Light curves of the unpulsed emission at 1384 MHz astron, and at the time when the dispersion measure has
from the four periastron passages. Solid line denotes 2004 data, changed dramatically, the unpulsed emission starts to rise.
dashed line denotes 2000 data, dash-dot line is 1997 data and
Theimplicationsarethattheonsetoftheunpulsedemission
dottedlineis1994data.
occursasthepulsarenterstheBestardisk.There-detection
ofthepulsedemissionoccursjustbeforethepost-periastron
peak of the unpulsed emission. This is consistent with the
peakoftheunpulsedemissioncoincidingwiththeendofthe
interaction between the pulsar and the disk, and hence the
(cid:13)c 2004RAS,MNRAS000,1–7
6 Johnston et al.
through adiabatic losses as the emitting region is advected
outward in theexpandingradial flow of theBe star disk.
The observed epochs of the eclipse of the pulsed emis-
sionprovidelittleevidenceofgrossdifferencesinthediskge-
ometry between the different periastron passages, although
the sparse sampling of the observations does not rule out
some variation between the periastron eclipse epochs. On
the other hand, differences in the DM and RM variations
between periastron passages as the pulsar enters and re-
emerges from eclipse indicate that there are significant dif-
ferencesin thelocal propertiesoftheBestarwind/disk en-
counteredbythepulsar.Theselocalproperties,particularly
the number density of the wind and disk, will have a di-
rect effect on determiningtheproperties of thepulsarwind
bubbleandhencetheobservedradioemission fromthepre-
periastron and post-periastron encounter.
If synchrotron losses dominate, the initial flux density
decay following each peak is linear, with the decay time of
Figure 5. DM variations as a function of days from periastron
thepre-andpost-periastronsourcesdeterminedbythemag-
after combining all data from 1990 January to 2004 September.
The errorbars are smallerthan the symbols.The DM contribu- neticfieldstrengthwithintherelevantsourceregion.Ballet
tionfromtheinterstellarmediumisassumedtobe146.7cm−3pc. al. (1999) proposed that the1997 observations were consis-
tentwithadecaytimeofapproximately72daysforemission
region associated with thepre-periastron disk crossing, and
ceasing of the supply of relativistic electrons to the pulsar just9daysforthepost-periastronsource.TakingT−18and
wind – Be star disk bubble. T+13 as the epochs of the pre- and post-periastron cross-
Figure5combinesalltheDMvariationsmeasuredsince ings, the corresponding orbital separations are 40 R and
∗
1990 onto a single plot. Again, the results are largely con- 33R .IfthediskoftheBestarwashomogeneousthiswould
∗
sistent from one periastron to the next (although the sam- imply a stronger magnetic field – and hence shorter decay
plingisfarfromuniform).Theeclipseofthepulsedemission time–forthepost-periastronsourcethanthepre-periastron
is clearly delineated. The asymmetry between the pre- and source, because the smaller pulsar/Be star separation will
post-periastron data is striking. This is largely because of result in a smaller pulsar wind bubble. After theinitial lin-
the geometry of the system; the pulsar is on the far side of ear decay, synchrotron losses produce a distinct break in
theBe star with respect to theobserver prior to periastron the light curve once losses become significant at the energy
and on the near side following periastron. The path length of the electrons responsible for the emission at the observ-
is therefore correspondingly larger resulting in a larger DM ing frequency. This break is followed by a rapid frequency-
variations pre-periastron. dependent decay in the radio flux. No obvious evidence for
such a break has been detected.
Connorsetal.(2002)suggestedthattheunpulsedemis-
sion associated with the 2000 periastron fitted better with
5 DISCUSSION
adiabaticlossesasthedominantmechanismforthedecay.In
The differences between the light curves of the unpulsed thiscasethefluxdensitydecayfollowsapowerlaw,withthe
emission observed through the 1997, 2000 and 2004 perias- differencebetweenthedecayofthepre-andpost-periastron
tronpassagesaredifficulttoreconcilewiththesimplemod- sources determined by the relevant pulsar Be star separa-
els so far proposed. Melatos, Johnston & Melrose (1995) tion. This provides a good qualitative match to the 2000
establishedthatfree-freeabsorptioninaninclineddiskpro- periastron light curvesalthough it isdifficulttoaccountfor
vided a good fit to the observed variations in DM and RM theemissionplateaubetweenT andT+15.Thismodeldoes
through periastron. not providea good fit to theemission observed in 1997.
Ball et al. (1999) proposed that the unpulsed emission The 2004 light curves are difficult to fit with either
wassynchrotronradiationfromtwopopulationsofelectrons of these models for two reasons. The first phase of emis-
impulsivelyacceleratedasthepulsarpassedthroughtheBe sion prior to T+5, with its apparently slower increase and
stardisk before andafter periastron. They showed that the poorly-defined peak, is qualitatively different to that seen
qualitative behaviour of the unpulsed emission was consis- in either 1997 or 2000. The combination of a rapid initial
tentwithasituationwheretwosuchsourcescooledprimar- post-periastrondecreasefollowedbyamuchslowerdecrease
ily as a result of synchrotron losses. The rise times to the issimilartothebehaviourin1997and1994,butatthetime
two emission peaks were taken to be the same, around 10 of the transition the flux is much higher than the extrapo-
days, corresponding the the interaction time between the lated pre-periastron decrease.
pulsar wind and the Be star disk, roughly twice the time ThesystemhasrecentlybeendetectedatTeVenergies
takenforthepulsaritself topassthroughthedisk.Theun- by Aharonian et al. (2004) in observations using the HESS
pulsedemissionobservedduringthe2000periastronshowed telescopes. An initial detection was made 10 days prior to
a somewhat different behaviour, and Connors et al. (2002) the2004periastron.Afteragapintheobservingbecauseof
presentedamodelbasedonthatfromBalletal.(1999),but thefullmoon,observationsstarted again at T+10,afteran
withthetwosynchrotronemittingregionscoolingprimarily initialnon-detectionthefluxrosesharply,peakingatT+20.
(cid:13)c 2004RAS,MNRAS000,1–7
Radio Observations of PSR B1259–63 7
It then began a slow decay with some detectable flux still REFERENCES
present100daysafterperiastron.Themaximumfluxabove
AharonianF.etal.,2004,A&A,Submitted
230GeVwas7percentofthatoftheCrabNebula,andthe
Ball L. T., Melatos A., Johnston S., Skjaeraasen O., 1999, ApJ,
spectrum between 0.4 and 3 TeV is well fitted by a power
514,L39
law with photon index ∼2.7. Both the amplitude and the Connors T. W., Johnston S., Manchester R. N., McConnell D.,
spectrum were close to that predicted by the Kirk et al. 2002,MNRAS,336,1201
(1999) model in which the Be star photons are scattered Grove E., Tavani M., Purcell W. R., Kurfess J. D., Strick-
by electrons in the shocked region of the pulsar wind. The manM.S.,AronsJ.,1995,ApJ,447,L113
observed TeV light curve is poorly sampled, but appears Hirayama M., Nagase F., Tavani M., Kaspi V. M., Kawia N.,
remarkablysimilartotheradiolightcurve.Aharonianetal. AronsJ.,1996,Proc.Astr.Soc.Jap.,48,833
Hotan A. W., van Straten W., Manchester R. N., 2004,
(2004) show that the TeV peak following periastron occurs
Proc.Astr.Soc.Aust.,21,302
at approximately thesame epoch as thesecond radio peak,
Johnston S., Lyne A. G., Manchester R. N., Kniffen D. A.,
and liketheradio fluxtheTeV gamma ray emission decays
D’AmicoN.,LimJ.,AshworthM.,1992a,MNRAS,255,401
over100 days or so.
Johnston S., Manchester R. N., Lyne A. G., Bailes M.,
KaspiV.M.,QiaoG.,D’AmicoN.,1992b,ApJ,387,L37
Johnston S., Manchester R. N., Lyne A. G., Nicastro L., Spy-
romilioJ.,1994,MNRAS,268,430
6 CONCLUSIONS Johnston S., Manchester R. N., Lyne A. G., D’Amico N.,
Bailes M., Gaensler B. M., Nicastro L., 1996, MNRAS, 279,
Itisnowapparentthatwhilethegrossfeaturesoftheradio 1026
emission from the PSR B1259−63/SS2883 system are re- Johnston S., Manchester R. N., McConnell D., Campbell-
peated through each periastron passage, thedetails of both WilsonD.,1999,MNRAS,302,277
the pulsed and the transient unpulsed emission differ con- KirkJ. G., Ball L., Skjaeraasen O., 1999, Astroparticle Physics,
siderably from one periastron to thenext. 10,31
Thedatafromthefourperiastronencounterssofarob- Manchester R. N., Johnston S., Lyne A. G., D’Amico N.,
BailesM.,NicastroN.,1995,ApJ,445,L137
servedatradiofrequenciesremainconsistentwiththeinter-
MelatosA.,JohnstonS.,MelroseD.B.,1995,MNRAS,275,381
pretationthatthepulsarpassesthroughadensediskaround
Shaw S. E., Chernyakova M., Rodriguez J., Walter R.,
its Be star companion just before peristron and is obscured
KretschmarP.,Mereghetti S.,2004,A&A,426,L33
for some 33 days before re-emerging. The data imply that
TavaniM.,AronsJ.,1997,ApJ,477,439
there is little difference between the geometry of the en- Wang N., Johnston S., Manchester R. N., 2004, MNRAS, 351,
counter between the pulsar and the Be star disk from one 599
pass tothenext.Unpulsedemission appears neartheonset Wex N., Johnston S., Manchester R. N., Lyne A. G., Stap-
oftheencounterbetweenthepulsarandtheBestardiskand persB.W.,BailesM.,1998,MNRAS,298,997
the proposal that it originates from the two regions where
the pulsar passes through the disk remains consistent with
the observations. However, there are significant differences
between the DM and RM of the pulsar and the transient
unpulsed emission from one periastron to the next. These
differences all suggest that the local properties of the Be
star disk encountered by the pulsar, such as the number
densityand magnetic field,vary considerably between peri-
astronpassages.Modellingthedetaileddependenceofemis-
sionobservedthrougheachindividualperiastronistherefore
unlikely to beproductive.
Thesimilarity betweenthelightcurvesoftheunpulsed
radio emission and the TeV emission is intriguing. It raises
anapparentcontradiction,inthattheradioemissionismost
likely associated with the pulsar–disk interaction while the
propertiesoftheTeVemission areverywellfittedbymodel
predictions that do not involve the disk. This is certain to
become clearer as more detailed comparisons of the radio,
X-ray and gamma-ray observations and the models of the
system are completed.
ACKNOWLEDGMENTS
The Australia Telescope is funded by the Commonwealth
of Australia for operation as a National Facility managed
by the CSIRO. We are grateful to Willem van Straten and
AidanHotan forconstant software supportthroughoutthis
project.
(cid:13)c 2004RAS,MNRAS000,1–7