1: \documentclass[useAMS,usenatbib]{mn2e}

2: \usepackage{graphics, graphicx}

3: % If your system has the AMS fonts version 2.0 installed, MN.sty can be

4: % made to use them by uncommenting the line: %\AMStwofontstrue

5: %

6: % By doing this, you will be able to obtain upright Greek characters.

7: % e.g. \umu, \upi etc.  See the section on "Upright Greek characters" in

8: % this guide for further information.

9: %

10: % If you are using AMS 2.0 fonts, bold math letters/symbols are available

11: % at a larger range of sizes for NFSS release 1 and 2 (using \boldmath or

12: % preferably \bmath).

13: 

14: \newcommand{\NrHighEdotGlast}{61}

15: \newcommand{\TotNrPulsars}{352}

16: \newcommand{\SlopeWPrelation}{$-0.30\pm0.05$}

17: \newcommand{\SlopeWPrelationwithouterror}{$-0.30$}

18: 

19: \newif\ifAMStwofonts

20: %\AMStwofontstrue

21: 

22: %%%%% AUTHORS - PLACE YOUR OWN MACROS HERE %%%%%

23: \def\lapp{\ifmmode\stackrel{<}{_{\sim}}\else$\stackrel{<}{_{\sim}}$\fi}

24: \def\gapp{\ifmmode\stackrel{>}{_{\sim}}\else$\stackrel{>}{_{\sim}}$\fi}

25: \def\arcdeg{\hbox{$^\circ$}}

26: \newcommand{\prog}[1]{\textsc{#1}}

27: \newcommand{\degrees}[1]{\ensuremath{#1^\circ}}

28: 

29: \input{figures.tex}

30: 

31: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

32: 

33: \title[Characteristics of energetic pulsars]

34: {Profile and polarization characteristics of energetic pulsars}

35: \author[Weltevrede \& Johnston]

36: {Patrick Weltevrede and Simon Johnston \\

37: Australia Telescope National Facility, CSIRO, P.O. Box 76, 

38: Epping, NSW 1710, Australia.

39: }

40: \date{\today}

41: \pagerange{\pageref{firstpage}--\pageref{lastpage}}

42: \pubyear{2008}

43: \begin{document}

44: \maketitle

45: \label{firstpage}

46: 

47: \begin{abstract}

48: In this paper we compare the characteristics of pulsars with a high

49: spin-down energy loss rate ($\dot{E}$) against those with a low

50: $\dot{E}$.  We show that the differences in the total intensity pulse

51: morphology between the two classes are in general rather subtle.  A

52: much more significant difference is the fractional polarization which

53: is very high for high $\dot{E}$ pulsars and low for low $\dot{E}$

54: pulsars.  The $\dot{E}$ at the transition is very similar to the death

55: line predicted for curvature radiation. This suggests a possible link

56: between high energy and radio emission in pulsars and could imply that

57: $\gamma$-ray efficiency is correlated with the degree of linear

58: polarization in the radio band.  The degree of circular polarization

59: is in general higher in the second component of doubles, which is

60: possibly caused by the effect of co-rotation on the curvature of the

61: field lines in the inertial observer frame.

62: 

63: The most direct link between the high energy emission and the radio

64: emission could be the sub-group of pulsars which we call the energetic

65: wide beam pulsars. These young pulsars have very wide profiles with

66: steep edges and are likely to be emitted from a single magnetic

67: pole. The similarities with the high energy profiles suggest that both

68: types of emission are produced at the same extended height range in

69: the magnetosphere.  Alternatively, the beams of the energetic wide

70: beam pulsars could be magnified by propagation effects in the

71: magnetosphere. This would naturally lead to decoupling of the wave

72: modes, which could explain the high degree of linear polarization. As

73: part of this study, we have discovered three previous unknown

74: interpulse pulsars (and we detected one for the first time at

75: 20~cm). We also obtained rotation measures for 18 pulsars whose values

76: had not previously been measured.

77: 

78: \end{abstract}

79: 

80: \begin{keywords}

81: polarization --- pulsars:general --- pulsars: individual PSRs J0905--5127, J1126--6054, J1611--5209, J1637--4553 --- radiation mechanisms: non-thermal 

82: \end{keywords}

83: 

84: \section{Introduction}

85: 

86: Pulsars are observed to be spinning down with time.  The spin-down

87: energy loss rate $\dot{E}$, which is the loss of kinetic energy, is

88: given by

89: \begin{equation}

90: \dot{E} = 4\pi^2I\dot{P}P^{-3}

91: \end{equation}

92: where $I$ is the moment of inertia of the star (generally taken to be

93: $10^{45}$~g$\,$cm$^{2}$), $P$ its spin period and $\dot{P}$ its spin

94: down rate. Some of the loss of spin-down energy emerges as radiation

95: across the entire electromagnetic spectrum from radio to

96: $\gamma$-rays. The radio emission accounts for only $\sim10^{-6}$ of the

97: energy budget (e.g. \citealt{lk05}) whereas up to a few percent is

98: emitted in the $\gamma$-ray band (e.g. \citealt{tho04}), with the rest

99: converted to magnetic dipole radiation and some form of pulsar wind.

100: 

101: It has been evident for more than a decade that pulsars with high

102: $\dot{E}$ have different polarization characteristics to those with

103: lower $\dot{E}$. Many high $\dot{E}$ pulsars are highly linearly

104: polarized (e.g. \citealt{qmlg95,hlk98,cmk01}). The pulse profiles of

105: high $\dot{E}$ pulsars are believed to be generally simple, consisting

106: of either one or two prominent components

107: (e.g. \citealt{hmt71,ran83}). \cite{jw06} found that, in the high

108: $\dot{E}$ pulsars with double profiles, the total power and the

109: circular polarization usually dominates in the trailing component and

110: that the swing of position angle (PA) of the linearly polarized

111: radiation is steeper under the trailing component. They interpreted

112: these results as showing that the beam of high $\dot{E}$ pulsars

113: consisted of a single conal ring at a relatively high

114: height. \cite{kj07} incorporated these results into their pulsar beam

115: model. In their model, there is a sharp distinction between pulsars

116: with $\dot{E}>10^{35}$ erg~s$^{-1}$ and those with smaller $\dot{E}$.

117: 

118: High $\dot{E}$ pulsars are not only interesting because of their

119: distinct properties in the radio band, but also because a subset of

120: them emit pulsed high energy emission. There are three different

121: families of high energy emission models in the literature which places

122: the emitting regions at different locations in the pulsar

123: magnetosphere. In the polar cap models (e.g. \citealt{dh96}) the

124: emitting region is close to the neutron star surface, while outer gap

125: models (e.g. \citealt{chr86a}) place the emitting region near the

126: light cylinder. Finally, in slot gap models (e.g. \citealt{mh04}) the

127: particle acceleration occurs in a region bordering the last open field

128: lines. In the polar cap models the young pulsars are thought to

129: produce pairs through curvature radiation (e.g. \citealt{hm01a}),

130: while older pulsars produce pairs only through inverse-Compton

131: scattering (e.g. \citealt{mh04b}). In the outer gap model pairs are

132: formed by the interaction of thermal X-ray photons from the neutron

133: star surface with $\gamma$-ray photons (e.g. \citealt{rom96a}). All

134: models have in common that high $\dot{E}$ pulsars should be brighter

135: $\gamma$-ray sources than low $\dot{E}$ pulsars, something which

136: is confirmed by EGRET (\citealt{tho04}).

137: 

138: We have recently embarked on a long-term timing campaign to monitor a

139: large sample of young, high $\dot{E}$ pulsars. The ephemerides

140: obtained from timing will be used to provide accurate phase tagging of

141: $\gamma$-ray photons obtained from the Fermi Gamma-Ray Space Telescope

142: (formerly known as GLAST \citealt{sgc+08}) and AGILE

143: (\citealt{ppp+08}) satellites, with the expectation that the number of

144: $\gamma$-ray pulsars will increase from the current 7 to over 100

145: (\citealt{gsc+07}). Of the $\sim80$ non millisecond pulsars in the

146: pulsar catalogue maintained by the

147: ATNF\footnote{http://www.atnf.csiro.au/research/pulsar/psrcat}

148: \citep{mhth05} with $\dot{E}>10^{35}$ erg~s$^{-1}$, we have obtained

149: polarization profiles at 1.4~GHz for \NrHighEdotGlast, a substantial

150: increase in the number available to previous studies. In this paper,

151: therefore, we examine the differences between high and low $\dot{E}$

152: pulsars. In total, we use pulse profiles from {\TotNrPulsars} pulsars,

153: which includes the {\NrHighEdotGlast} energetic pulsars and a

154: comparison sample of intermediate and low $\dot{E}$ pulsars in order

155: to draw general conclusions about the pulsar population.

156: 

157: The paper is organized as follows. We start with explaining the

158: details of the observations and the data analysis. In section 3 we

159: then describe the polarization profiles of four pulsars for which we

160: found an interpulse at 20~cm and present new rotation measures. In

161: section 4 the the total intensity profiles of the pulsars are

162: discussed, followed by a discussion of the polarization

163: properties. Finally we will discuss the results in section 6, followed

164: by the conclusions. The polarization profiles of all the pulsars can

165: be found in appendix A (those for which we have a 20~cm and a 10~cm

166: profile) and appendix B (those for which we only have a 20~cm

167: profile). The plots of the pulse profiles can also be found on the

168: internet\footnote{http://www.atnf.csiro.au/people/joh414/ppdata/index.html}.

169: Finally, a table with derived properties from the pulse profiles can

170: be found in appendix C\footnote{This table is also available in

171: electronic form at the CDS via anonymous ftp to {\tt

172: cdsarc.u-strasbg.fr (130.79.128.5)} or via {\tt

173: http://cdsweb.u-strasbg.fr/cgi-bin/qcat?J/MNRAS/}}. The appendices are

174: only available in the on-line version of this publication.

175: 

176: \section{Observations and data analysis}

177: 

178: The procedure to generate pulse profiles for the pulsars which are

179: timed for the Fermi and AGILE satellites is complicated by the

180: fact that the pulse profiles of individual (short) observations have

181: typically a low signal to noise ratio ($S/N$). It is therefore

182: required to sum all the available observations in order to obtain a

183: template profile with a higher $S/N$. This procedure is described in

184: some detail in this section.

185: 

186: \subsection{Observations}  

187: 

188: 

189: All the observations were made at the Parkes telescope in

190: Australia using the centre beam of the 20~cm multibeam receiver (which

191: has a bandwidth of 256~MHz and has a noise equivalent flux density of

192: $\sim$35 Jy on a cold sky) and the 10/50~cm receiver (which has at

193: 10~cm a bandwidth of 1024~MHz and has a noise equivalent flux density of

194: $\sim$49 Jy on a cold sky). This paper will focus mainly on the 20~cm

195: data, because that is wavelength at which the majority of observations

196: were made. However for some highly scattered pulsars it is also useful

197: to consider the 10~cm data. The 50~cm data is not used, because the

198: profiles are scattered at that frequency in many cases. The timing

199: program started in April 2007 and each pulsar is typically observed

200: once per month at 20~cm and twice per year at 10 and 50~cm. The two

201: polarization channels of the linear feeds of the receiver were

202: converted into Stokes parameters, resampled and folded at the pulse

203: period by a digital filterbank. In our case a pulse profile with 1024

204: bins and 1024 frequency channels was dumped every 30 seconds on hard

205: disk. Before each observation a calibration signal, injected into the

206: feed at a \degrees{45} angle to the probes, was recorded which is then

207: used to determine the phase delay and relative gain between the two

208: polarization channels.

209: 

210: The data were processed using the {\sl PSRCHIVE} package

211: (\citealt{hvm04}). The data of each observing session were first

212: checked for narrow band radio frequency interference (RFI). An

213: automatic procedure using the median smoothed difference of the

214: bandpass, was used to identify the affected frequency channels in the

215: calibration observations. The flagged channels were left out of all

216: the observations of a particular observing day, making the automatic

217: procedure more robust in finding weaker RFI which is not always

218: identified. The remaining frequency channels were added together and

219: the resulting sequence of profiles was then visually inspected for

220: impulsive RFI. The sub-integrations in where RFI was particular strong

221: were left out of further data processing.

222: 

223: The 20~cm multibeam receiver has a significant cross-coupling between

224: the two dipoles affecting the polarization of the pulsar signal. For

225: instance, a highly linearly polarized signal induces artificial

226: circular polarization. These effects are measured as a function of

227: parallactic angle for PSR J0437--4715 for the Parkes Pulsar Timing

228: Array project (Manchester et al. 2008, in preparation), which allows

229: the construction of a polarimetric calibration model

230: (\citealt{van04}). We have applied this model to all the observations

231: using the 20~cm multibeam receiver, which reduces the artifacts in the

232: Stokes parameters considerably.

233: 

234: \subsection{\label{SctSumming}Summing of the individual observations}

235: 

236: For some pulsars the timing noise is so severe that the pulse period

237: predicted by the timing solution in the pulsar catalogue is not

238: accurate enough to fold the data. In such a case the pulsar appears to

239: drift in longitude in successive sub-integrations. We therefore

240: applied the updated timing solutions to align the sub-integrations

241: within individual observations.

242: 

243: To produce high $S/N$ profiles the individual observations must be

244: added together. Because many pulsars involved in this timing program

245: have severe timing noise and show glitches, it is difficult to use the

246: timing solution to add the observations together. Instead a scheme was

247: followed in which the observations are correlated with each other in

248: order to find the offsets in pulse longitude between the

249: profiles. These offsets were applied directly to the individual

250: observations using custom software in order to sum the profiles. The

251: sum of the profiles (i.e. the standard or template) has a higher $S/N$

252: than the individual profiles and can then be correlated with the

253: individual observations to determine the offsets in pulse longitude

254: with higher precision, hence making a more accurate standard. This

255: procedure is repeated one more time to make the final pulse profile.

256: 

257: \subsection{Faraday de-rotation}

258: 

259: The interstellar medium interacts with the radio waves of pulsars,

260: causing a number of frequency and time dependent effects. One of these

261: effects is Faraday rotation, where the interstellar magnetic field

262: component parallel to the line of sight causes a difference in the

263: propagation speeds of the left- and right-hand circular polarization

264: signal components. This effect causes the polarization vector to

265: rotate in the Stokes Q and U plane and the angle is a function of

266: frequency and the rotation measure (RM). It is therefore necessary to

267: de-rotate Stokes Q and U before summing the frequency channels in an

268: observation.

269: 

270: A similar procedure has to be followed when the profiles of different

271: observations are summed together, because different frequency channels

272: were flagged and deleted in different observations. This means that

273: although the centre frequencies are identical for the different

274: observations, their weighted mid-frequencies are slightly

275: different. The {\sl PSRCHIVE} package de-rotates Stokes Q and U with

276: respect to this weighted mid-frequency of the band and therefore it is

277: necessary to take the RM into account when profiles of different

278: observations are summed together. This is done by rotating Stokes Q

279: and U of each observation with respect to infinite frequency using

280: custom software before adding individual observations together.

281: 

282: \subsection{Making frequency standards}

283: 

284: In order to be able to measure the RM for pulsars for which no

285: sufficiently accurate values were available one needs to keep

286: frequency resolution. This is done by summing the observations

287: together using custom software which takes into account the pulse

288: longitude offsets found by correlating the profiles of the individual

289: observations (as described section \ref{SctSumming}). A complication

290: is that {\sl PSRCHIVE} de-disperses the data with respect to the

291: non-weighted centre frequency of the band, while the pulse longitude

292: offsets are determined using de-dispersed profiles with respect to the

293: weighted mid-frequency. It is therefore necessary to include a

294: dispersion time delay corresponding with the difference in the

295: weighted and non-weighted mid-frequency when the observations are

296: added together.

297: 

298: 

299: 

300: \section{Results on individual pulsars}

301: 

302: \subsection{Newly discovered interpulses}

303: 

304: \begin{figure}

305: \begin{center}

306: \includegraphics[height=0.99\hsize,angle=270]{J0905-5127.20.paperplot.ps}\\

307: \vspace*{5mm}

308: \includegraphics[height=0.99\hsize,angle=270]{J1126-6054.20.paperplot.ps}\\

309: \vspace*{5mm}

310: \includegraphics[height=0.99\hsize,angle=270]{J1611-5209.20.paperplot.ps}\\

311: \vspace*{5mm}

312: \includegraphics[height=0.99\hsize,angle=270]{J1637-4553.20.paperplot.ps}

313: \caption{\label{Fig_newIP}The pulse profile of the four pulsars for

314: which we report an interpulse at an observing wavelength of 20~cm.

315: The top panels show the total intensity profile (solid line), linear

316: polarization (dashed line) and circular polarization (dotted

317: line). The peak intensity of the profiles are normalized to one. The bottom

318: panels show the PA of the linear polarization (for the pulse longitude

319: bins in where the linear polarization was detected above $2 \sigma$).}

320: \end{center}

321: \end{figure}

322: 

323: While analysing the data described in this paper we discovered four

324: interpulses which have not been previously reported at 20~cm in the

325: literature. The polarization profiles of these pulsars are shown in

326: Fig. \ref{Fig_newIP} and discussed in some detail below. In all

327: cases there is no evidence that the interpulse only appears

328: sporadically rather than be weakly present in all observations.

329: 

330: \subsubsection{PSR J0905--5127}

331: 

332: Profiles of this pulsar were presented first in \cite{dsb+98}. In

333: their figure, there appears to be little sign of the interpulse at

334: 20~cm, with perhaps a hint at 70~cm.  In our observations at 20~cm,

335: the interpulse is very weak in comparison to the main pulse with an

336: intensity ratio of $\sim$17. The separation between the centroid of

337: the main and interpulses is 175\degr. The main pulse is a clear double

338: with a total width of $\sim\degrees{20}$ but not much structure can be

339: discerned in the interpulse because of the low $S/N$, although its

340: width appears to be narrower than that of the main pulse. The

341: interpulse is separated by \degrees{180} from the trailing component

342: of the main pulse, suggesting that the interpulse could be the

343: trailing component of a double. The polarization swing across both the

344: main and interpulses is rather flat, and we cannot attempt a rotating

345: vector model fit (RVM; \citealt{rc69a}). We do not have sufficient

346: $S/N$ to make any claims about the interpulse at either 10~cm or 50~cm

347: from our data.

348: 

349: 

350: \subsubsection{PSR J1126--6054}

351: 

352: The interpulse of PSR J1126--6054 is too weak to be seen in the

353: profile presented in \cite{jlm+92}. However, it is just visible in our

354: 20~cm data with a peak amplitude about one tenth of that of the main

355: pulse. The peak-to-peak separation between the main and interpulses is

356: $\sim$174\degr. This low $\dot{E}$ pulsar has a low degree of linear

357: polarization in its main pulse, which explains the absence of

358: significant linear polarization in the much weaker interpulse.

359: The interpulse appears to be significantly narrower, though the low

360: $S/N$ makes the width difficult to measure. At 50~cm, the interpulse

361: is marginally stronger with respect to the main pulse whereas at 10~cm

362: it is not detected.

363: 

364: \subsubsection{PSR J1611--5209}

365: 

366: There is no obvious interpulse at 20~cm in the profile presented in

367: \cite{jlm+92}. However, \cite{kjm05} reported a low-amplitude

368: interpulse in their 10~cm data.  In our 20~cm data we clearly see the

369: interpulse which has a peak amplitude less than 0.1 that of the main

370: pulse. The separation between the main and interpulse is

371: $\sim$177\degr. The main pulse has a total width of $\sim$10\degr\ and

372: consists of at least two components with a low fractional

373: polarization. The low $S/N$ in the interpulse precludes any

374: measurement of the polarization, but the overall width seems similar

375: to main pulse.

376: 

377: 

378: \subsubsection{PSR J1637--4553}

379: 

380: This pulsar has a very weak interpulse (about one tenth in amplitude

381: compared to the main pulse), which is perhaps just visible in the

382: existing literature (\citealt{jlm+92}). The separation between the

383: main and interpulses is $\sim\degrees{173}$, and although weak, the

384: interpulse seems to be the same width as the $\sim\degrees{20}$ of the

385: main pulse. The polarization of the interpulse is hard to determine,

386: although the main pulse is virtually 100\% polarized. At 50~cm, the

387: interpulse has the same separation from the main pulse and roughly the

388: same relative amplitude as at 20~cm. Our low $S/N$ at 10~cm makes the

389: interpulse undetectable.

390: 

391: 

392: \subsection{New rotation measures}

393: 

394:  

395: \begin{table}

396: \begin{tabular}{crrrlr}

397: \hline

398: Name & \multicolumn{1}{c}{RM} & \multicolumn{1}{c}{$l$} & \multicolumn{1}{c}{$b$} & \multicolumn{1}{c}{DM} & \multicolumn{1}{c}{$d$}\\

399:  & \multicolumn{1}{c}{[rad m$^{-2}$]} & \multicolumn{1}{c}{[deg]} & \multicolumn{1}{c}{[deg]} & \multicolumn{1}{c}{[pc cm$^{-3}$]} & \multicolumn{1}{c}{[kpc]}\\

400: \hline

401: J1052--5954 & $-280\pm24$      & 288.55 & -0.40 &  \hspace*{3mm}491    &   13.55  \\

402: J1115--6052 & $257\pm18$       & 291.56 & -0.13 &  \hspace*{3mm}228.2  &   6.76   \\

403: J1156--5707 & $238\pm19$       & 295.45 & 4.95  &  \hspace*{3mm}243.5  &   20.40  \\

404: J1524--5625 & $180\pm20$       & 323.00 & 0.35  &  \hspace*{3mm}152.7  &   3.84   \\

405: J1524--5706 & $-470\pm20$      & 322.57 & -0.19 &  \hspace*{3mm}833    &   21.59  \\

406: J1638--4417 & $160\pm25$       & 339.77 & 1.73  &  \hspace*{3mm}436.0  &   8.46   \\

407: J1702--4128 & $-160\pm20$      & 344.74 & 0.12  &  \hspace*{3mm}367.1  &   5.18   \\

408: J1705--3950 & $-106\pm14$      & 346.34 & 0.72  &  \hspace*{3mm}207.1  &   3.86   \\

409: J1737--3137 & $448\pm17$       & 356.74 & 0.15  &  \hspace*{3mm}488.2  &   5.88   \\

410: J1738--2955 & $-200\pm20$      & 358.38 & 0.72  &  \hspace*{3mm}223.4  &   3.91   \\

411: J1801--2154 & $160\pm40$       & 7.83   & 0.55  &  \hspace*{3mm}387.9  &   5.15   \\

412: J1809--1917 & $41\pm17$        & 11.09  & 0.08  &  \hspace*{3mm}197.1  &   3.71   \\

413: J1815--1738 & $175\pm20$       & 13.18  & -0.27 &  \hspace*{3mm}728    &   9.06   \\

414: J1828--1101 & $45\pm20$        & 20.49  & 0.04  &  \hspace*{3mm}607.4  &   7.26   \\

415: J1837--0604 & $450\pm25$       & 25.96  & 0.26  &  \hspace*{3mm}462    &   6.19   \\

416: J1841--0345 & $447\pm15$       & 28.42  & 0.44  &  \hspace*{3mm}194.32 &   4.15   \\

417: J1845--0743 & $440\pm12$       & 25.43  & -2.30 &  \hspace*{3mm}281.0  &   5.85   \\

418: J1853--0004 & $647\pm16$       & 33.09  & -0.47 &  \hspace*{3mm}438.2  &   6.58   \\

419: \hline

420: \end{tabular}

421: \caption{\label{newRMs}The pulsars for which new values of the RM were

422: measured. From left to right the columns are the pulsar name, the

423: measured rotation measures, the galactic longitude and latitude, the

424: dispersion measure and the best available distance estimate.}

425: \end{table}

426: 

427: As mentioned in section 2, the interstellar magnetic field parallel to

428: the line of sight causes the polarization vector to rotate in the

429: Stokes Q and U plane. In order to derive the degree of linear

430: polarization it is therefore necessary to correct for this rotation

431: before summing the frequency channels across the frequency band. The

432: amount of rotation of the PA depend on the rotation measure RM and

433: values for the RM were obtained from the pulsar catalogue. However,

434: not all the pulsars have a published value for its RM (or one with

435: sufficient accuracy). We therefore measured the RM for a number of

436: objects in our sample.

437: 

438: The RM can be measured by fitting the change of the PA ($\psi$)

439: across the frequency band with the Faraday rotation formula

440: \begin{equation}

441: \label{EqRM}

442: \psi\left(\lambda\right)=\psi_\infty+RM\,\lambda^2,

443: \end{equation}

444: where $\lambda$ is the observing wavelength of the considered

445: frequency channel and $\psi_\infty$ is the PA at infinite

446: frequency. When different pulse longitude bins of the pulse profile

447: show a similar frequency dependence of $\psi$ one can be confident in

448: the measured RM. The RM is obtained by calculating the weighted

449: average of the fits of equation (\ref{EqRM}) for different bins in

450: where there is enough linear polarization present. The new RM values

451: are listed in Table \ref{newRMs}. Only PSR J1809--1917 has a

452: previously published RM which we include in the table because our

453: value of 41 rad m$^{-2}$ differs significantly from the 130 rad

454: m$^{-2}$ quoted by Han et al. (2006).

455: 

456: 

457: \section{Total intensity pulse profiles}

458: 

459: In this and the following section we investigate if, and how, the

460: beams of high $\dot{E}$ pulsars differ from those of low $\dot{E}$

461: pulsars. As we are going to investigate basic pulse profile properties

462: in a statistical way, it is important to consider the effects of a low

463: $S/N$ and interstellar scattering. Because the high $\dot{E}$ pulsars

464: tend to be younger they have on average lower galactic latitude than

465: older pulsars, hence they tend to be more affected by interstellar

466: scattering. Therefore low $S/N$ observations and profiles which are

467: clearly affected by interstellar scattering were excluded from the

468: statistics.

469: 

470: 

471: 

472: \subsection{Pulse profile morphology}

473: 

474: It has been pointed out by several authors

475: (e.g. \citealt{hmt71,ran83,jw06,kj07}) that the profiles of high

476: $\dot{E}$ pulsars are relatively simple. A problem with measuring

477: ``profile complexity'' is that it is not a well defined quantity,

478: hence it is highly subjective. In order to make the results objective

479: and better reproducible one should quantify the amount of complexity

480: in a mathematical way. We will therefore explore ways to quantify

481: different aspects of profile complexity, because it is difficult to

482: come up with a definition which covers all facets of profile

483: complexity.

484: Only the total intensity (Stokes parameter I) profiles are

485: considered in this section, while the polarization properties are

486: investigated in the next section.

487: 

488: 

489: \subsubsection{Profile classification}

490: 

491: \begin{table}

492: \begin{tabular}{c@{\,} c@{} c c c c c}

493: \hline

494: \multicolumn{3}{c}{$\dot{E}$ [erg~s$^{-1}$]} & Single & Double & Multiple & Total\\

495: \hline

496: $10^{35}$ & -- & $10^{38}$ & 27 (53\%) & 17 (33\%) & 7 (14\%) & 51\\

497: $10^{33}$ & -- & $10^{35}$ & 53 (47\%) & 43 (38\%) & 16 (14\%) & 112\\

498: $10^{28}$ & -- & $10^{33}$ & 52 (46\%) & 46 (40\%) & 16 (14\%) & 114\\

499: \hline

500: \end{tabular}

501: \caption{\label{classification}The classification of the profiles for

502: different $\dot{E}$ bins. Pulsars with a $S/N < 30$ were excluded as

503: well as the profiles marked to show substantial scattering.}

504: \end{table}

505: 

506: Pulse profiles are often described in terms of ``components'', which

507: are attributed to structure in the pulsar beam. There are different

508: models in the literature describing the structure of the radio beam of

509: pulsars. The beam could be composed out of a core and one or more

510: cones (\citealt{ran83}), randomly distributed patches (\citealt{lm88})

511: or patchy cones (\citealt{kj07}). In these models each component of

512: the pulse profile originates from a different physical location in the

513: magnetosphere. Because the components overlap in many cases and

514: because their shapes are not uniquely defined, it is difficult to

515: objectively classify profiles. Following \cite{kj07}, we have

516: classified the profiles by eye into three classes depending on the

517: number of distinct (possibly overlapping) peaks in the pulse

518: profile. These classes are named ``single'', ``double'' and

519: ``multiple'', depending on if one, two or more peaks were

520: identified. Although this classification is subjective, it should be

521: considered as a rough measure for the complexity.

522: 

523: Table \ref{classification} shows the percentage of pulsars in each

524: class for three different $\dot{E}$ bins.  For the pulsars which are

525: significantly scattered at 20~cm we used 10~cm data (when available)

526: and we omitted profiles with $S/N<30$ to improve

527: significance. Compared with \cite{kj07} we find relatively more

528: singles and less multiples and also the difference between high and

529: low $\dot{E}$ pulsars is less pronounced. This might partially reflect

530: the subjectivity of profile classification, but it may also be related

531: to the fact that the classification of \cite{kj07} was based on

532: polarization properties. For instance, rapid changes in the PA-swing

533: are often found in between components and can therefore be interpreted

534: as an indication for the presence of multiple components.  The

535: polarization properties are discussed in a separate section in this

536: paper.

537: 

538: \subsubsection{\label{SctMathDecomp}Mathematical decomposition of the profiles}

539: 

540: Because profile components can overlap and can have various shapes, it

541: is in many cases not clear how many separate emission components there

542: are. Also, because the classification is done by eye it is highly

543: subjective at which level of detail the profile is separated into

544: components. A more objective way to decompose the profile into

545: components is to describe the profiles as linear combinations of basis

546: functions. The number of required functions to fit the profile

547: is then a measure for the complexity of the pulse profiles.

548: 

549: \begin{figure}

550: \begin{center}

551: \includegraphics[height=0.99\hsize,angle=270]{vonMises.ps}

552: \caption{\label{Fig_vonmises}The decomposition of the pulse profile of

553: PSR J1803--2137 at 20~cm into five von Mises functions (the dotted

554: curves). The sum of these functions (the thick solid line) is a good

555: representation of the observed pulse profile (thin solid line).}

556: \end{center}

557: \end{figure}

558: 

559: Gaussian functions are often used to decompose profiles

560: (e.g. \citealt{kwj+94}), but we have chosen to use von Mises functions

561: (\citealt{von18}), which are defined as

562: \begin{equation}

563: \label{Eqvonmises}

564: I(x) = I_\mathrm{peak}e^{\kappa\cos\left(x-\mu\right) - \kappa}.

565: \end{equation}

566: Here $\mu$ is the location of the peak (in radians), $I_\mathrm{peak}$

567: is the peak intensity and $\kappa$ is the concentration (which

568: determines the width of the peak). The shape of these functions is

569: very similar to Gaussians (see Fig. \ref{Fig_vonmises}), but they can

570: often fit the edges of components slightly better. The main difference

571: is that von Mises functions are circular, hence they are also known as

572: circular normal distributions. A fitting routine for von Mises

573: functions is part of the {\sl PSRCHIVE} software package. 

574: 

575: There is a subtle difference between the required number of fit

576: functions and the number of components in the pulse profile. The first

577: is just a mathematical measure of complexity, while the latter is the

578: number of distinct physical emission locations in the pulsar

579: magnetosphere which are visible in the line of sight. These numbers

580: can be different, because there is no a priori reason to believe that

581: the shape of a profile component can be described by a single, simple,

582: mathematical basis function which is the same for all pulsars.  For

583: instance, a profile which shows a tail because of interstellar

584: scattering can have one component (``single profile''), but it can

585: only be fitted by a number of von Mises functions. Another example can

586: be seen in the decomposition as shown in

587: Fig. \ref{Fig_vonmises}. Although the component between pulse

588: longitude \degrees{70} and \degrees{120} is fit by two von Mises

589: functions, the smooth shape does suggest that it is a single

590: asymmetric emission component. By using more complex asymmetric

591: mathematical functions it might be possible to decompose some profiles

592: in a smaller number of fit functions. However, in effect this is the

593: same as to fit a larger number overlapping more simple symmetric

594: functions which have less fit-parameters per function. 

595: 

596: There is not always one unique solution for the decomposition of a

597: profile and therefore the decomposition does not necessarily give

598: additional insight in how profiles are composed out of distinct

599: physical components. Nevertheless a noise free mathematical

600: description of a pulse profiles can be used as a measure for its

601: complexity.  Moreover it is a very useful technique which makes it

602: easier to measure profile properties such as pulse widths. An

603: additional advantage of a mathematical description of the profile is

604: that one can more accurately determine the component widths for

605: pulsars which have overlapping components.

606: 

607: 

608: \begin{figure}

609: \begin{center}

610: \includegraphics[height=0.95\hsize,angle=270]{avnrcomponents.ps}

611: \caption{\label{Fig_ppdot_complexity}The histogram of the average

612: number of von Mises functions required to fit the profiles when the

613: $S/N$ is scaled down to 30 for different $\dot{E}$ bins (solid

614: line). The dashed histogram shows the number of pulsars contributing

615: in each bin. Only profiles with a $S/N\geq100$ are included.

616: }

617: \end{center}

618: \end{figure}

619: 

620: When using the number of mathematical fit functions as a measure

621: of complexity it is important to take into account the $S/N$ ratio

622: of the profiles. A higher $S/N$ profile will require a larger number

623: of mathematical basis functions to fit its shape, even though the

624: profile is not necessarily more complex. In order to avoid this

625: effect, we determined how many of the fit functions would have a

626: significant contribution to the total integrated intensity of the

627: pulse profile when the $S/N$ would have been 30.  We only considered

628: profiles with a $S/N\ge100$ to ensure that all weak components which

629: are just significant when the $S/N$ would have been 30 are spotted by

630: eye.

631: 

632: Fig. \ref{Fig_ppdot_complexity} shows the average number of von Mises

633: functions required to fit the profiles when the $S/N$ is scaled down

634: to 30 for different $\dot{E}$ bins.  There is not much evidence that

635: the profile complexity is very different for high and low $\dot{E}$

636: pulsars. The absence of a significant correlation between $\dot{E}$

637: and the number of fit functions is confirmed by calculating the

638: Spearman rank-order correlation coefficient (\citealt{ptvf92}) of the

639: unbinned data, which is a non-parametric measure of correlation. Among

640: the most complex profiles, according to this classification scheme,

641: are those of PSRs J1034--3224 and J1745--3040, which indeed have

642: complex looking profile shapes.  An other pulsar which is ranked at

643: the same level of complexity is PSR J1302--6350, which has to the eye

644: a relatively simple double peaked profile, but its highly asymmetric

645: components require relatively many mathematical functions to fit.  We

646: will therefore try a different method to define profile complexity

647: below.

648: 

649: 

650: \subsubsection{The dimensionless double separation}

651: 

652: \begin{figure}

653: \begin{center}

654: \includegraphics[height=0.95\hsize,angle=270]{double_separation.ps}

655: \caption{\label{Fig_double_sep}The dimensionless double separation

656: (the ratio of the separation between the components and the average of

657: their full width half maxima) versus the spin-down energy loss rate

658: for all the observed pulsars at 20~cm which are classified to be

659: doubles with a $S/N\geq30$. }

660: \end{center}

661: \end{figure}

662: 

663: As described in section \ref{SctMathDecomp}, the profiles of the high

664: $\dot{E}$ pulsars J1015--5719 and J1302--6350 were ranked as highly

665: complex, while they appear to be ``simple'' to the eye. One property

666: of these profiles which makes them look simple is that they are

667: doubles with well separated components. We therefore tested the

668: hypothesis that the doubles of high $\dot{E}$ pulsars have more

669: clearly separated components than those of the low $\dot{E}$

670: pulsars. How clearly the components of doubles are separated can be

671: quantified by calculating a quality factor, which we define to be

672: \begin{equation}

673: Q_\mathrm{sep} = \frac{\mathrm{\Delta\phi_\mathrm{sep}}}{\frac{1}{2}\left(\mathrm{FWHM}_1+\mathrm{FWHM}_2\right)}.

674: \end{equation}

675: This dimensionless double separation is the ratio of the separation

676: between the components $\Delta\phi_\mathrm{sep}$ and the average of

677: the full width half maxima of the components $\mathrm{FWHM}_1$ and

678: $\mathrm{FWHM}_2$. Higher values of $Q_\mathrm{sep}$ imply that the

679: components are separated more compared with the width of the

680: components.

681: 

682: Fig. \ref{Fig_double_sep} shows $Q_\mathrm{sep}$ versus $\dot{E}$ for

683: all profiles at 20~cm which were classified to be doubles and have a

684: $S/N\geq30$. There is no evidence that the components of doubles of

685: low $\dot{E}$ pulsars are more likely to be overlapping, which is

686: confirmed by calculating the Spearman rank-order correlation

687: coefficient. According to this measure the most clearly separated

688: doubles are PSRs J1302--6350, J1733--3716, J1901--0906 and

689: J2346--0609.

690: 

691: \subsubsection{Profile symmetry}

692: 

693: A factor which was not taken into account in the previous sections is

694: the amount of symmetry in the profile. For instance, PSR J1302--6350

695: has highly asymmetric profile components, but the profile as a whole

696: appears symmetric and could therefore be regarded as ``simple''. It

697: is therefore interesting to consider the degree of symmetry of the

698: profiles, which can be

699: measured by cross-correlating the profile with its mirror-image. We

700: define the degree of profile symmetry to be the ratio of the maximum

701: value of the cross-correlation function between the profile and the

702: time reversed profile, and the maximum value of auto-correlation

703: function of the profile. The degree of symmetry is therefore

704: normalized to 1 for completely symmetric profiles and it decreases for

705: more asymmetric profiles.

706: 

707: \begin{figure}

708: \includegraphics[height=0.95\hsize,angle=270]{symmetry.ps}

709: \caption{\label{Fig_symmetry}The degree of profile symmetry versus

710: $\dot{E}$. Pulsars with a $S/N < 30$ and profiles with substantial

711: scattering were excluded and only 20~cm data was considered. There is

712: no evidence that high $\dot{E}$ pulsars are more symmetric.}

713: \end{figure}

714: 

715: The degree of symmetry versus $\dot{E}$ is shown in

716: Fig. \ref{Fig_symmetry}. The pulsar with the lowest measured degree of

717: symmetry is PSR B1747--31, which has a relatively narrow and bright

718: leading component and a much broader and weaker trailing

719: component. Also the complex main pulse of PSR B1055--52 can be found

720: at the lower end of this figure. There is no indication for any

721: correlation, which is confirmed by the Spearman rank-order correlation

722: coefficient. Like for the other measures of complexity, it is hard to

723: quantify that pulse profiles of high $\dot{E}$ pulsars are more simple

724: than those of the low $\dot{E}$ pulsars.

725: 

726: 

727: 

728: \subsection{\label{Sectbeamwidths}Pulse widths versus $P$}

729: 

730: A basic property of the emission beam of a pulsar is its half opening

731: angle $\rho$. It is found that the opening angle is proportional to

732: $P^{-1/2}$ (e.g. \citealt{big90b,kxl+98,gks93,ran93}),

733: which is expected if the edge of the active area of the polar cap is

734: set by the last open field lines. In order to derive the opening angle

735: from the measured profile width one needs to know how the emission

736: beam intersects the line of sight. Because the orientation of the line

737: of sight with respect to the pulsar beam is for most pulsars at best

738: only poorly constrained, it is difficult to obtain accurate opening

739: angles. For a large sample of pulsars the unknown geometrical factors

740: should average out and therefore the profile width and $\rho$ should

741: have the same $P$ dependence. The unknown geometry will cause

742: additional scattering around the correlation between the pulse width

743: and $P$.

744: 

745: \begin{figure}

746: \includegraphics[height=0.95\hsize,angle=270]{width_p.ps}

747: \caption{\label{Fig_pulsewidths_p}The measured profile 10\% widths

748: versus $P$. The solid line is the power law fit through the data,

749: which has a slope of \SlopeWPrelationwithouterror. The dashed line

750: indices the fit of a second order polynomial through the data points

751: (which is statistically not better than the power law fit).  The

752: filled points are pulsars with an $\dot{E} \ge 10^{34}$ erg~s$^{-1}$

753: and the open points have a lower $\dot{E}$.  Pulsars with a $S/N < 30$

754: and profiles with substantial scattering were excluded. All the shown

755: observations were done at a wavelength of 20~cm.}

756: \end{figure}

757: 

758: The measured pulse widths at 10\% of the peak intensity ($W_{10}$)

759: indeed show a slight anti-correlation with $P$

760: (Fig. \ref{Fig_pulsewidths_p}), while there is no indication for a

761: dependence with $\dot{P}$ (not shown). The slope is measured by

762: reduced $\chi^2$ fitting (the data points are weighted equally), which

763: results in a slope of \SlopeWPrelation\footnote{The correlation

764: is slightly different, although within the error identical to the

765: value quoted in \cite{wj08a}. This is because since the publication of

766: that paper more timing observations have been added.}, comparable

767: with the fit obtained from the data of \cite{gl98} by

768: \cite{wj08a}. The slope of the correlation is therefore slightly less

769: than what is expected from theory. This conclusion, in combination

770: with the period distribution of pulsars with interpulses, provides

771: convincing evidence in favour for the evolution of the pulsar beam

772: towards alignment with the rotation axis (\citealt{wj08a}).

773: 

774: If there is any deviation from a power law relationship between

775: $W_{10}$ and $P$, then it would be that the slope of the correlation

776: is steeper for faster rotating pulsars. Although the fit of a

777: second order polynomial through the data-points indeed show this

778: trend, it is statistically not much better than the first order fit.

779: High $\dot{E}$ pulsars are in general spinning faster then low

780: $\dot{E}$ pulsars, and therefore one could conclude that the pulse

781: widths of high $\dot{E}$ pulsars have a stronger dependence on $P$

782: than the low $\dot{E}$ pulsars. To illustrate this the pulsars with

783: high and low values of $\dot{E}$ are marked differently in

784: Fig. \ref{Fig_pulsewidths_p}. 

785: One could argue that there is not much evidence for a correlation for

786: the low $\dot{E}$ pulsars, while this is clearer for the high

787: $\dot{E}$ pulsars. But, as the fit second order polynomial was

788: statistically not much better than the fit of a power law, this

789: conclusion is also not significant. If this correlation exist, then

790: it would suggest that the profile widths of the high $\dot{E}$

791: pulsars follow the theoretical prediction more closely than those of

792: the low $\dot{E}$ pulsars, which could indicate that the emission

793: geometry for high $\dot{E}$ pulsars is more simple.

794: 

795: 

796: 

797: \subsection{Pulse widths versus $\dot{E}$}

798: 

799: \begin{figure}

800: \includegraphics[height=0.95\hsize,angle=270]{profile_widths_edot.ps}

801: \caption{\label{Fig_pulsewidths_edot}The measured profile 10\% widths

802: versus $\dot{E}$. The triangles indicate the pulsars which are

803: classified to have interpulses. Pulsars with a $S/N < 30$ and

804: profiles with substantial scattering were excluded. All the shown

805: observations were done at a wavelength of 20~cm. Three clusterings of

806: pulsars are highlighted by ellipses, which are discussed in the text.}

807: \end{figure}

808: 

809: In the previous subsection we found that $W_{10}$ is correlated with

810: $P$. One can expect that the correlation with $\dot{E}$ is weaker,

811: because $W_{10}$ was found to be uncorrelated with $\dot{P}$. Indeed,

812: Fig. \ref{Fig_pulsewidths_edot} shows that for most pulsars $W_{10}$

813: is as good as uncorrelated with $\dot{E}$. But remarkably, unlike in

814: Fig. \ref{Fig_pulsewidths_p} there are a number of outliers which are

815: clustered in relatively well defined regions in $\dot{E}$-space. These

816: outliers are indicated by the ellipses and each group will be

817: discussed separately below.

818: 

819: The first group of outliers are the pulsars in the ellipse at the left

820: hand side of Fig. \ref{Fig_pulsewidths_edot}. Although the profiles

821: are clearly wider than most pulse profiles, they form a continuous

822: distribution with the narrower profiles. These low $\dot{E}$ pulsars

823: are PSRs J1034--3224, J1655--3048 and J2006--0807 (which have complex

824: looking profiles) and PSRs J1133--6250 and J1137--6700 (which are

825: doubles with a clear saddle between the components).  The profiles of

826: these pulsars are most likely broad because their beam is close to

827: alignment with the rotation axis, making the beam intersect the line

828: of sight for a relatively long fraction of the rotation period. There

829: is evidence that the beam evolves to alignment with the rotation axis

830: over time (e.g. \citealt{wj08a}), so it is not surprising that these

831: aligned pulsars are old pulsars with low $\dot{E}$ values. For

832: two of these pulsars estimates for the angle between the magnetic axis

833: and the rotation axis can be found in the literature. These

834: polarization studies indeed suggest that the beam of PSRs J1034--3224

835: \citep{mhq98} and J2006--0807 \citep{ran93b,lm88} are close to

836: alignment.

837: 

838: The second group of pulsars with wide profiles are the pulsars with

839: interpulses, which are marked with triangles in

840: Fig. \ref{Fig_pulsewidths_edot}. These are PSRs J0834--4159,

841: J0905--5127, B0906--49, B1124--60, J1549--4848, B1607--52, B1634--45,

842: B1702--19, B1719--37, B1736--29, J1828--1101 and J1843--0702. The

843: profiles of these pulsars are characterised by having an interpulse

844: which is separated by approximately \degrees{180} in pulse longitude

845: from the main pulse. This separation is much larger than the widths of

846: the main- and interpulse. The most natural explanation for these

847: interpulses is that the emission of the main- and interpulse

848: originates from opposite magnetic poles. These pulsars are

849: concentrated to high values for $\dot{E}$. This is partially a

850: selection effect in the sample of pulsars which are included in the

851: Fermi timing program, but it has also shown by \cite{wj08a} that

852: interpulses are more likely to be detected in young (high $\dot{E}$)

853: pulsars.

854: 

855: The third group of pulsars with wide profiles can also be found at the

856: high $\dot{E}$ end of Fig. \ref{Fig_pulsewidths_edot}. These are PSRs

857: J1015--5719, B1259--63, J1803--2137, J1809--1917 and J1826--1334. Like

858: the group of pulsars with wide profiles at the low $\dot{E}$ end of

859: the figure, this group appears to form a continuum with the pulsars

860: with narrow profiles. We will refer to this group as the {\em

861: energetic wide beam pulsars}. Their profiles show a double structure

862: and they are exceptionally wide, but they are not separated by exactly

863: \degrees{180} in pulse longitude. In contrast to the group of

864: interpulses, this separation is not much larger than the width of the

865: individual components. The two components are often highly asymmetric

866: with steep edges at opposite sides, making the profile as a whole to

867: have a high degree of mirror symmetry. For some of these pulsars a

868: weak bump is detected in between the components, which disappears at

869: higher frequencies. The dependence of the PA on pulse longitude is

870: usually simple and straight.

871: 

872: It is not clear if PSR B1055--52 should be classified as an energetic

873: wide beam pulsar or a pulsar with an interpulse. On the one hand the

874: separation between the main- and interpulse is larger than the width

875: of the individual components, but on the other hand the components are

876: very wide and the interpulse is not exactly \degrees{180} away from

877: the main pulse. The location of PSR B1055--52 in

878: Fig. \ref{Fig_pulsewidths_edot} (the lowest triangle at

879: $\dot{E}=3.0\times10^{34}$ erg~s$^{-1}$) suggest that it is well

880: separated from the other energetic wide beam pulsars, although the

881: group of interpulse pulsars appear to have an overlap with the group

882: of energetic wide beam pulsars. Especially the location of PSRs

883: B0906--49 ($\dot{E}=4.9\times10^{35}$ erg~s$^{-1}$) and J1828--1101

884: ($\dot{E}=1.6\times10^{36}$ erg~s$^{-1}$) in the figure are consistent

885: with both groups. Both these pulsars have interpulses at

886: $\sim$\degrees{180} away from the main pulse and this separation is

887: much larger than the component widths (the broad components of

888: J1828--1101 at 20~cm are because of scatter broadening), which is good

889: evidence that both interpulses are emitted from the opposite pole. For

890: PSR B0906--49 the PA-swing is shown to be inconsistent with a wide

891: cone interpretation (\citealt{kj08}).

892: 

893: 

894: The energetic wide beam pulsars are among the pulsars with the highest

895: $\dot{E}$ values ($\dot{E}>5\times10^{35}$ erg~s$^{-1}$), although it

896: is not true that all pulsars with high $\dot{E}$ values are also

897: energetic wide beam pulsars. This is first of all shown by the overlap

898: between the group of pulsars with interpulses and the energetic wide

899: beam pulsars group. Secondly, PSR J1513--5908 which has the highest

900: $\dot{E}$ value in our sample, does not show any evidence of a double

901: structure. Finally, PSR J1028--5819 is an extremely narrow double

902: (\citealt{kjk+08}, point in the bottom left corner of

903: Fig. \ref{Fig_pulsewidths_p}). A high $\dot{E}$ therefore appears to

904: be an important parameter which allows an energetic pulsar to form a

905: wide beam, but there must be more factors involved.

906: 

907: 

908: 

909: 

910: 

911: 

912: 

913: 

914: \subsection{The intensity ratio of the components of high $\dot{E}$ pulsars with double profiles}

915: 

916: \cite{jw06} noted that the trailing component of well separated

917: double profiles of high $\dot{E}$ pulsars tend to dominate in total

918: power (and in circular polarization as we will discuss below). This

919: curious effect seems to be strongest for pulsars with

920: $\dot{E}>10^{35}$ erg~s$^{-1}$ (see for example PSR J1420--6048,

921: Fig. \ref{J1420-6048}). The only exceptions are the Vela pulsar (which

922: has no well separated double profile), PSR J1302--6350 (an energetic

923: wide beam pulsars for which it is not clear which component is the

924: trailing component) and PSR J1831--0952. Nevertheless, in the majority

925: of the cases this correlation holds.

926: 

927: 

928: \section{Polarization}

929: 

930: \subsection{Linear polarization}

931: 

932: 

933: \begin{figure}

934: \includegraphics[height=0.95\hsize,angle=270]{linearpol.ps}

935: \caption{\label{Fig_linearpol}The degree of linear polarization versus

936: $\dot{E}$ of all pulsars observed at 20~cm for which a significant

937: degree of linear polarization was measured. Pulsars which show

938: evidence for scatter broadening were excluded. There are two

939: relatively well defined regions which are almost empty in this

940: diagram. The dashed line shows the linear fit and the solid curve

941: the fit of an arctan function illustrating the step in the degree of

942: linear polarization.}

943: \end{figure}

944: 

945: \begin{figure}

946: \begin{center}

947: \includegraphics[width=0.95\hsize,angle=270]{ppdot.ps}\\

948: \includegraphics[width=0.95\hsize,angle=270]{opm_ppdot.ps}

949: \caption{\label{Fig_ppdot}{\em Top:} The $P-\dot{P}$ diagram of all

950: the observed pulsars at 20~cm for which a significant degree of linear

951: polarization could be measured. The filled and open circles are the

952: pulsars which are respectively more or less than 50\% linearly

953: polarized. {\em Bottom:} The $P-\dot{P}$ diagram of the pulsars which

954: show sudden jumps in their PA-swing (open circles) and those which

955: have a smooth PA-swing (filled circles). }

956: \end{center}

957: \end{figure}

958: 

959: It has been pointed out by several authors that degree of linear

960: polarization is high for high $\dot{E}$ pulsars

961: (e.g. \citealt{qmlg95,hlk98,cmk01,jw06}). This correlation is clearly

962: confirmed, as can be seen in Fig. \ref{Fig_linearpol}. There is a

963: transition from a low to a high degree of linear polarization which

964: happens around $\dot{E}\sim10^{34}-10^{35}$ erg~s$^{-1}$.  Virtually

965: all pulsars with $\dot{E} < 5\times10^{33}$ erg~s$^{-1}$ have less

966: than 50\% linear polarization and for almost all pulsars with with

967: $\dot{E} > 2\times10^{35}$ erg~s$^{-1}$ this percentage is above

968: 50\%. There appears to be a transition region in between where

969: pulsars can both have low and high degrees of polarization, although

970: the transition is remarkably sharp and there are well defined spaces

971: in the figure which are almost empty. The non-linearity of the

972: degree of linear polarization versus $\dot{E}$ is confirmed by fitting

973: an arctan function through the data (solid curve in

974: Fig. \ref{Fig_linearpol}). This is done by minimizing the $\chi^2$

975: using the Levenberg-Marquardt algorithm \citep{mar63} as implemented

976: in \cite{ptvf92} (the data points are weighted equally). The total

977: $\chi^2$ is reduced by 20\% compared with a linear fit, which shows

978: that the step in the degree of linear polarization is important to

979: consider. Adding higher order polynomial terms does not reduce the

980: $\chi^2$ further, suggesting that the step is the most dominant

981: deviation from non-linearity. The position of the steepest point in

982: the fitted function occurs at $\log_{10}{\dot{E}} = 34.50\pm0.08$. 

983: 

984: 

985: The emission of pulsars is thought to be a combination of two

986: orthogonally polarized modes (OPM, e.g. \citealt{mth75}). This aspect

987: of the emission can manifest itself in sharp $\sim\degrees{90}$ jumps

988: in the position angle (PA) over a small pulse longitude range. These

989: jumps are thought to be sudden transitions from the domination of one

990: mode to the other. Jumps in the PA-swing therefore indicates that both

991: modes are present in the emission. The mixing of both modes at a

992: certain longitude will lead to depolarization, so the presence of

993: jumps in the PA-swing could be anti-correlated with the degree of

994: polarization. By comparing the top and bottom panel of

995: Fig. \ref{Fig_ppdot} one can see that the $\dot{E}$ value at which the

996: transition from a low to a high degree of polarization takes place

997: coincides with the $\dot{E}$ value after which pulsars do not have

998: jumps in their PA. This is therefore important evidence that the

999: increase in the degree of linear polarization with $\dot{E}$ is caused

1000: by one OPM dominating the emission. Most high $\dot{E}$ pulsars do not

1001: show OPM jumps, but the reverse is not always true. Low $\dot{E}$

1002: pulsars can have a low degree of linear polarization without evidence

1003: for OPM jumps.

1004: 

1005: There are three curious exceptions in Fig. \ref{Fig_linearpol}

1006: which do not follow the general trend. First of all PSRs

1007: J1509--5850 and J1833--0827 have a low degree of linear polarization

1008: while they have a high $\dot{E}$ ($5.2\times10^{35}$ and

1009: $5.8\times10^{35}$ erg~s$^{-1}$ respectively). However it must be

1010: noted that the leading and trailing components of PSR J1833--0827 are

1011: highly polarized at 10~cm. The degree of linear polarization of

1012: this pulsar shows a drop to zero in the middle of central component,

1013: which could indicate that there is a transition in the dominating

1014: OPM. The other exception is PSR J0108--1431, which has a low

1015: $\dot{E}$ but is nevertheless highly polarized. This could suggest

1016: that this pulsar has some similarities with high energy pulsars.

1017: 

1018: All pulsars without a significant amount of measured degree of linear

1019: polarization fall below the $\dot{E} < 5\times10^{33}$ erg s$^{-1}$

1020: line. The only exception is PSR J1055--6032, which appears to have a

1021: very low degree of polarization. The rule that high $\dot{E}$ pulsars

1022: are highly polarized therefore is confirmed in the majority of all

1023: pulsars.

1024: 

1025: 

1026: 

1027: \subsection{Emission heights}

1028: 

1029: \subsubsection{The emission height derived from the pulse width}

1030: 

1031: The wider pulse profiles of high $\dot{E}$ pulsars are often

1032: attributed to a larger emission height for those pulsars

1033: (e.g. \citealt{man96,kj07}). The divergence of the magnetic (dipole)

1034: field lines away from the magnetic axis makes the half opening angle

1035: $\rho$ of the beam scale with the square root of the emission

1036: height. Under the assumption that the beam of the pulsar is confined

1037: by the last open field lines it follows that

1038: \begin{equation}

1039: \label{EqH}

1040: \rho = \sqrt{\frac{9\pi \,\, h_{\rm em}}{2\,\,\, P\,\, c}}

1041: \end{equation}

1042: (e.g \citealt{lk05}), were $h_{\rm em}$ is the emission height and $c$

1043: the speed of light.

1044: 

1045: \begin{table}

1046: \begin{center}

1047: \begin{tabular}{l c r r r r}

1048: \hline

1049: \multicolumn{1}{c}{Name} & $\dot{E}$ & $h_\mathrm{PA}$ & $h_{90}$ & \multicolumn{1}{c}{$\Delta\phi$} & $R_\mathrm{LC}$\\

1050:  & [erg~s$^{-1}$] & [km] &  [km] & [deg] & [km]\\

1051: \hline

1052: J1015--5719 & $8.27\times10^{35}$ & --   & 5160 & -- & 6674\\

1053: J1302--6350 & $8.25\times10^{35}$ & --   & 3476 & -- & 2279\\

1054: J1803--2137 & $2.22\times10^{36}$ &   461 &  2967 &    8.3 &     6375 \\

1055: J1809--1917 & $1.78\times10^{36}$ &   130 &  1425 &    3.8 &     3948 \\

1056: J1826--1334 & $2.84\times10^{36}$ &   191 &  2522 &    4.5 &     4841 \\

1057: \hline

1058: J0304+1932 & $1.91\times10^{31}$ &    75 &   562 &    0.1 &    66206 \\

1059: J0536--7543 & $1.15\times10^{31}$ &   -21 &  1472 &   -0.0 &    59444 \\

1060: J0614+2229 & $6.24\times10^{34}$ &  1051 &    99 &    7.5 &    15982 \\

1061: J0630--2834 & $1.46\times10^{32}$ &   118 &  2459 &    0.2 &    59376 \\

1062: J0631+1036 & $1.73\times10^{35}$ &  1087 &   278 &    9.1 &    13731 \\

1063: J0729--1448 & $2.81\times10^{35}$ &  1143 &   291 &   10.9 &    12008 \\

1064: J0742--2822 & $1.43\times10^{35}$ &   338 &    68 &    4.9 &     7957 \\

1065: J0835--4510 & $6.91\times10^{36}$ &    32 &    30 &    0.9 &     4263 \\

1066: J0908--4913 & $4.92\times10^{35}$ &   320 &    24 &    7.2 &     5094 \\

1067: J1048--5832 & $2.01\times10^{36}$ &   -24 &   141 &   -0.5 &     5901 \\

1068: J1105--6107 & $2.48\times10^{36}$ &   243 &    52 &    9.2 &     3015 \\

1069: J1119--6127 & $2.34\times10^{36}$ &  2082 &  1406 &   12.3 &    19455 \\

1070: J1123--4844 & $1.76\times10^{32}$ &   540 &   152 &    5.3 &    11682 \\

1071: J1253--5820 & $4.97\times10^{33}$ &   690 &   116 &    6.5 &    12191 \\

1072: J1320--5359 & $1.67\times10^{34}$ &   673 &   158 &    5.8 &    13347 \\

1073: J1359--6038 & $1.21\times10^{35}$ &   544 &    41 &   10.2 &     6084 \\

1074: J1420--6048 & $1.04\times10^{37}$ &   106 &   442 &    3.7 &     3253 \\

1075: J1531--5610 & $9.09\times10^{35}$ &    81 &   136 &    2.3 &     4018 \\

1076: J1535--4114 & $1.98\times10^{33}$ &   474 &   235 &    2.6 &    20654 \\

1077: J1637--4553 & $7.51\times10^{34}$ &   392 &    62 &    7.9 &     5667 \\

1078: J1701--3726 & $2.97\times10^{31}$ &   184 &   984 &    0.2 &   117118 \\

1079: J1705--3950 & $7.37\times10^{34}$ &  -120 &   690 &   -0.9 &    15218 \\

1080: J1709--4429 & $3.41\times10^{36}$ &   563 &   326 &   13.2 &     4889 \\

1081: J1733--3716 & $1.54\times10^{34}$ &   932 &  1969 &    6.6 &    16107 \\

1082: J1740--3015 & $8.24\times10^{34}$ &   818 &    31 &    3.2 &    28952 \\

1083: J1835--1106 & $1.78\times10^{35}$ &   445 &    93 &    6.4 &     7916 \\

1084: J1841--0345 & $2.69\times10^{35}$ &   353 &   415 &    4.2 &     9737 \\

1085: \hline

1086: \end{tabular}

1087: \end{center}

1088: \caption{\label{emissionheights90}The emission height $h_\mathrm{PA}$

1089: is derived from the offset $\Delta\phi$ between the inflection point

1090: of the PA-swing and the centre of the pulse profile. The emission

1091: height $h_{90}$ is derived from the pulse width assuming an orthogonal

1092: rotator ($\alpha=\degrees{90}$) and a line of sight which makes a

1093: central cut through the emission beam ($\beta=\degrees{0}$), which

1094: is the emission height for a typical random geometry. The last column

1095: is the light cylinder radius. The first five pulsars are the pulsars

1096: which we have classified as energetic wide beam pulsars.}

1097: \end{table}

1098: 

1099: Wider beams are more likely to produce wide profiles, although

1100: the observed pulse width also depend on the orientation of the

1101: magnetic axis and the line of sight with respect to the rotation

1102: axis. The relevant parameters are the angle $\alpha$ between the

1103: magnetic axis and the rotation axis and the angle $\zeta$ between the

1104: line of sight and the rotation axis. A related angle is the impact

1105: parameter $\beta = \zeta-\alpha$, which is the angle between the line

1106: of sight and the magnetic axis at its closest approach. For most

1107: pulsars it is extremely difficult to obtain reliable values for these

1108: angles, which makes it hard to derive the emission height from

1109: $W_{10}$.

1110: 

1111: For a sample of pulsars with a random orientation of the magnetic axis

1112: and the line of sight both the $\alpha$ and $\zeta$ distribution are

1113: sinusoidal. Simulations using the model described in \cite{wj08a},

1114: show that the pulse width distribution for such a sample pulsars peaks

1115: at $2\rho$. Some pulsars will have wider profiles because the pulsar

1116: beam is more aligned with the rotation axis, while others will have

1117: narrower profiles because the line of sight grazes the beam. This

1118: implies that the typical pulse width of a large sample of pulsars

1119: which have random orientations of their spin and magnetic axis and

1120: have similar opening angles $\rho$ should be equal to $2\rho$. In

1121: other words, a typical profile width is equal to that which is

1122: expected for an orthogonal rotator ($\alpha=\degrees{90}$) and a line

1123: of sight which makes a central cut through the emission beam

1124: ($\beta=\degrees{0}$). For such geometry Eq. \ref{EqH} can be

1125: rewritten to

1126: \begin{equation}

1127: \label{EqH90}

1128: h_{90} = \frac{cP\left(W_{10}\right)^2}{18\pi},

1129: \end{equation}

1130: which is the emission height for a typical random geometry assuming a

1131: magnetic dipole field and an active area of the polar cap which is set

1132: by the last open field lines.

1133: 

1134: \subsubsection{The emission height derived from the PA-swing}

1135: 

1136: An independent way to estimate the emission height is by measuring the

1137: shift of the PA-swing caused by the co-rotation of the emission region

1138: with the neutron star. In this method it is assumed that the PA-swing

1139: is described by the rotating vector model (RVM; \citealt{rc69a}). The

1140: position angle $\psi$ is then predicted to depend on the pulse

1141: longitude $\phi$ as

1142: \begin{equation}

1143: \tan\left(\psi-\psi_0\right)=\frac{\sin\alpha\;\sin\left(\phi-\phi_0\right)}{\sin\zeta\;\cos\alpha-\cos\zeta\;\sin\alpha\;\cos\left(\phi-\phi_0\right)},

1144: \end{equation}

1145: where $\psi_0$ and $\phi_0$ are the PA and the pulse longitude

1146: corresponding to the intersection of the line of sight with the

1147: fiducial plane (the plane containing the rotation and magnetic axis).

1148: The PA-swing is a S-shaped curve and its inflection point occurs at

1149: $\phi_0$. The RVM fit is shown in the figures of Appendix A and B for

1150: the pulsars which have a roughly S-shaped PA-swing.

1151: 

1152: 

1153: 

1154: If the emission profile is symmetric around the magnetic axis, then one

1155: could expect the inflection point to coincide with the middle of the

1156: pulse profile. However, co-rotation causes the inflection point to be

1157: delayed with respect to the pulse profile. The pulse longitude

1158: difference $\Delta\phi$ between the middle of the profile and the

1159: inflection point of the PA-swing can be used to derive the emission

1160: height (\citealt{bcw91})

1161: \begin{equation}

1162: \label{heighBCW92}

1163: h_\mathrm{PA} = \frac{P\,c\,\Delta\phi}{8\pi }.

1164: \end{equation}

1165: The relative shift of the PA-swing with respect to the profile is

1166: independent of $\alpha$ and $\zeta$ \citep{drh04}. If the

1167: emission height is too large it could be difficult to measure

1168: $\Delta\phi$ because the inflection point of the PA-swing is shifted

1169: beyond the edge of the pulse profile.

1170: 

1171: \subsubsection{The derived emission heights}

1172: 

1173: \begin{figure}

1174: \includegraphics[height=0.95\hsize,angle=270]{heightheight.ps}

1175: \caption{\label{Fig_height}The emission height as derived using two

1176: independent methods, one using the pulse widths ($h_{90}$) and the

1177: other using the shift of the PA-swing with respect to the profile

1178: ($h_\mathrm{PA}$). The solid circles indicate profiles with a clear

1179: double structure. The points should lie on the line if the emission

1180: heights are consistent with each other.}

1181: \end{figure}

1182: 

1183: The emission height derived using the PA-swing ($h_\mathrm{PA}$) and

1184: using the pulse width ($h_\mathrm{90}$) are both listed in Table

1185: \ref{emissionheights90}. This only includes the pulsars which have a

1186: clear S-shaped PA-swing at 20~cm. The typical emission height is a few

1187: hundred km, which is similar to the emission height found by other

1188: authors (e.g. \citealt{bcw91,mr01}).  One can see that for some

1189: pulsars $h_\mathrm{PA}$ is negative, which is obviously

1190: impossible. This can be considered to be a clear warning that the

1191: emission heights for an individual source could be completely wrong,

1192: but one can nevertheless hope that they are meaningful in a

1193: statistical sense. In order to test this we calculated the Spearman

1194: rank-order correlation coefficient between $h_\mathrm{PA}$ and

1195: $h_\mathrm{90}$, which shows that there is no evidence for any

1196: correlation between these parameters. This is also evident from

1197: Fig. \ref{Fig_height}, where these quantities are plotted against each

1198: other. We are therefore forced to accept that even in a statistical

1199: sense the calculated emission heights are inconsistent, supporting

1200: the same conclusion reached by \cite{ml04} based on six pulsars.

1201: 

1202: 

1203: There are a number of reasons why the heights derived using the

1204: two methods could be inconsistent. If the beams are significant

1205: patchy, then the centroid of the profile is not related to the

1206: position of magnetic axis and both methods to derive emission heights

1207: will fail. We therefore made a distinction in Fig. \ref{Fig_height}

1208: between the profiles which are clear doubles and other profiles,

1209: because the double structure could indicate that the pulsar beam is

1210: roughly symmetric around the magnetic axis. As one can see there is no

1211: noticeable difference in the distributions. Another effect that could

1212: be important is the effect of sweepback of the magnetic field

1213: lines. \cite{dh04} derived that the effect of sweepback can dominate

1214: over other effects of co-rotation at low altitudes, making it possible

1215: for the inflection point of the PA-curve to precede the profile

1216: centre. The PA-curve can also precede the profile in case of inward

1217: directed emission (\citealt{dfs+05}).

1218: 

1219: Despite the inconsistency between the derived emission heights using

1220: both methods, it is not true that the emission heights are entirely

1221: random. Most pulsars show a positive emission height $h_\mathrm{PA}$,

1222: indicating that the steepest slope of the PA-swing trails the centroid

1223: of the profile in most cases. In fact, Fig. \ref{Fig_height}

1224: appears to show evidence that it is unlikely that both $h_{90}$ and

1225: $h_\mathrm{PA}$ are large. In Table \ref{emissionheights90} one can

1226: see that the emission of the energetic wide beam pulsars should come

1227: from near the light cylinder in order to explain the width of the

1228: pulse profiles ($h_\mathrm{90}\sim R_\mathrm{LC}$). However, the

1229: derived emission heights from the the PA-swing fits are not unusually

1230: large. In this list one could add the emission height of PSR

1231: J1015--5719, which is estimated by \cite{jw06} to be 380 km. 

1232: 

1233: All the energetic wide beam pulsars with a derived emission height

1234: from the PA-swing can be found below the solid line in

1235: Fig. \ref{Fig_height}, as well as PSRs J1705--3950 and J1733--3716

1236: which have similar profile shapes. It seems unlikely that they all

1237: have beams which are close to alignment with the rotation axis,

1238: which suggest a different reason for the large widths of the profiles

1239: of the energetic wide beam pulsars.  Apparently the emission heights

1240: which are derived from the PA-swings of the energetic wide beam

1241: pulsars are systematically underestimated, or the heights derived from

1242: the profile widths are over estimated. The first case could be

1243: explained by magnetic field line sweepback when the emission height is

1244: low (\citealt{dh04}). The second case implies that the beams of these

1245: pulsars are wider than could be expected from the divergence of the

1246: dipole field lines. The widening of the pulsar beam could, at least in

1247: principle, be caused by propagation effects in the magnetosphere.

1248: 

1249: Another explanation for the deviation of the energetic wide beam

1250: pulsars from the line in Fig. \ref{Fig_height} could be that the two

1251: methods estimate the emission heights at different locations in the

1252: magnetosphere. The method based on the profile width estimates the

1253: emission height at the edge of the of the beam, while the method using

1254: the PA-swing estimates the emission height of the more central

1255: regions of the beam. For the energetic wide beam pulsars $h_{90}$ was

1256: found to be systematically larger than $h_\mathrm{PA}$, which can be

1257: interpreted as evidence for an increase in the emission height at the

1258: edge of the beam. This interpretation will be discussed in more detail

1259: in the following section.

1260: 

1261: Fig. \ref{Fig_height} shows that besides the group of pulsars which

1262: have relatively large $h_{90}$ compared to $h_\mathrm{PA}$, there is

1263: also a group in where the opposite is seen. An explanation could be

1264: that for those pulsars only a fraction of the polar cap is active 

1265: (e.g. \citealt{kg97}). Support for this interpretation is that the

1266: profiles of a number of pulsars in this group are argued to be

1267: produced by partial cones, including PSR J0543+2329

1268: (\citealt{wck+04}), J0614+2229 (\citealt{jkk+07}) and J0659+1414

1269: (\citealt{ew01}).

1270: 

1271: 

1272: \subsection{Circular polarization}

1273: 

1274: \begin{figure}

1275: \includegraphics[height=0.95\hsize,angle=270]{circularpol.ps}

1276: \caption{\label{Fig_circularpol}The degree of circular polarization

1277: (the absolute value of Stokes V) versus $\dot{E}$ of all pulsars

1278: observed at 20~cm for which a significant degree of circular

1279: polarization was measured. Pulsars which show evidence for scatter

1280: broadening were excluded. }

1281: \end{figure}

1282: 

1283: Unlike the degree of linear polarization the degree of circular

1284: polarization appears to be unaffected by $\dot{E}$. Also the fraction

1285: of pulsar which are left- and right-hand circularly polarized is about

1286: 50 per-cent for both the high and low $\dot{E}$ pulsars.

1287: 

1288: \cite{jw06} noted that, besides the total intensity, also the

1289: degree of circular polarization usually dominates in the trailing

1290: components of high $\dot{E}$ pulsars with well separated double

1291: profiles. This correlation is also clearly confirmed in our data for

1292: all pulsars with an $\dot{E}>10^{34}$ erg s$^{-1}$. The only clear

1293: counter example in our data-set could be PSR J1705--1906

1294: ($\dot{E}=7\times10^{34}$ erg s$^{-1}$), which has a high degree of

1295: circular polarization in the leading half of the profile. However, it

1296: must be noted that the single pulse modulation properties of this

1297: pulsar suggest that the leading component is not the leading component

1298: of a double, but rather a precursor to a blended double which forms

1299: the trailing half of the profile (\citealt{wws07}).

1300: 

1301: Remarkable correlations have been reported between the sign of the

1302: circular polarization and the sign of the slope of the

1303: PA-swing. According to \cite{rr90} the sign of the slope of the

1304: PA-swing is correlated with the sign of the circular polarization for

1305: pulsars which are cone dominated and for which the sign of the

1306: circular polarization is the opposite in the two components. But

1307: \cite{hmxq98} did not confirm this correlation. Instead they propose

1308: that there is a correlation between the sign of the circular

1309: polarization and the sign of the slope of the PA-swing for cone

1310: dominated pulsars which have the same sign of circular

1311: polarization. Our data does not show much evidence for either

1312: correlation. Compare for instance the plots for PSR J1826--1334

1313: (positive circular polarization, decreasing PA-swing) with PSR

1314: J1733--3716 (negative circular polarization, decreasing PA-swing).

1315: 

1316: \section{Discussion}

1317: 

1318: 

1319: \subsection{Extended radio emission regions?}

1320: 

1321: We concluded that the differences in the pulse profile morphology of

1322: the high and low $\dot{E}$ pulsars is in general rather subtle,

1323: without an objectively measurable discriminator between them. An

1324: exception is what we call the group of energetic wide beam pulsars,

1325: which do have distinct profile properties and which will be discussed

1326: separately below.  The measured slope of the $W_{10}-P$ correlation

1327: appears to be flatter than the theoretical $P^{-1/2}$ slope.  It is

1328: far from straight-forward to link the deviation of the slope to a

1329: physical mechanism. For example, as explained in \cite{wj08a}, the

1330: measured slope depends on the details of the evolution of the pulsar

1331: spin-down and the alignment of the magnetic axis with the rotation

1332: axis. If the active area of the polar cap is influenced by other

1333: factors than just the opening angle of the last open field lines, or

1334: if the emission height varies from pulsar to pulsar, one can expect to

1335: observe its effects in the $W_{10}-P$ correlation. There is some

1336: marginal evidence that the slope of the correlation is steeper for

1337: high $\dot{E}$ pulsars, suggesting that for high $\dot{E}$ pulsars

1338: these other factors are less important.

1339: 

1340: 

1341: If one believes that the emission geometry is simpler for high

1342: $\dot{E}$ pulsars, one can ask the question of what is causing

1343: this. One factor that could affect the complexity of profiles is the

1344: emission height. Complexity could arise because of multiple distinct

1345: emission heights within the beam (\citealt{kj07}). In their model high

1346: $\dot{E}$ pulsars have only one emission height which is similar for

1347: different pulsars. However, one could also make the argument for an

1348: opposite effect. Maybe the emission of high $\dot{E}$ pulsars is not

1349: emitted from a well defined height, but rather from an extended height

1350: range. This would mean that the observed profiles of high $\dot{E}$

1351: pulsars are a superposition of profiles emitted from a continuum of

1352: heights. The observed sum of those profiles (shifted with respect to

1353: each other by aberration and retardation) will have less complexity

1354: because they are blurred out. Not only are the profiles expected to be

1355: less complex in this scenario, but there is also not much room to vary

1356: the emission height from pulsar to pulsar if the height range is

1357: large. This would make the $W_{10}-P$ correlation follow the

1358: prediction more closely. Large emission height ranges are typical for

1359: high energy models, such as the slot gap models (e.g. \citealt{mh04})

1360: or the two pole caustic models (\citealt{dr03}), hence there could be

1361: parallels with the radio emission for high $\dot{E}$ pulsars. These

1362: parallels could be even more relevant for the energetic wide beam

1363: pulsars.

1364: 

1365: 

1366: \subsection{The energetic wide beam pulsars}

1367: 

1368: As discussed by e.g. \cite{man96}, there is a group of young pulsars

1369: which can be found among the highest $\dot{E}$ pulsars which have very

1370: wide profiles with often steep edges. The profiles are clearly mirror

1371: symmetric, suggesting that the components are the two sides of a

1372: single beam rather than two beams from opposing magnetic poles. This

1373: interpretation is also suggested by the frequency evolution of PSR

1374: B1259--63 (\citealt{mj95}). Because these objects are young, the

1375: typical orientation of the magnetic axis is not expected to be very

1376: different for the highest $\dot{E}$ pulsars and those with

1377: intermediate values, which suggests that some pulsars with high

1378: $\dot{E}$ values can have very different beams compared with other

1379: pulsars.

1380: 

1381: 

1382: 

1383: An interesting analogue can be drawn between the radio profiles of

1384: energetic wide beam pulsars and the high energy profiles of

1385: pulsars. High energy profiles can also be wide doubles which often

1386: have sharp edges (e.g. \citealt{tho04}).  The pulsars which produce

1387: high energy emission are the pulsars with high $\dot{E}$ values, so

1388: there could be a direct link between the high energy pulsars and the

1389: energetic wide beam pulsars. Maybe the radio emission and the high

1390: energy emission are produced at the same location in the

1391: magnetosphere. The sharp edges of the high energy profiles are often

1392: explained by caustics which form because of the combined effect of

1393: field line curvature, aberration and retardation

1394: (e.g. \citealt{mor83}). These caustics occur when the emission is

1395: produced high in the magnetosphere over a large altitude range and if

1396: the magnetic axis is not aligned to the rotation axis

1397: (e.g. \citealt{dr03}). If the radio emission and $\gamma$-ray emission

1398: would come from similar locations, one would expect the radio and

1399: $\gamma$-ray profiles to look alike. Hopefully the Fermi satellite

1400: will find high energy counterparts for these pulsars which allows a

1401: test of this hypothesis.

1402: 

1403: Another way to produce profiles with sharp edges could be the

1404: combination of refraction of radio waves in pulsar magnetospheres in

1405: combination with an emission height which is different for different

1406: field lines (\citealt{wsv+03}). Only the ordinary wave mode is

1407: refracted (e.g. \citealt{ba86}) or scattered (\citealt{pet08a}) in the

1408: magnetosphere. The profiles of the energetic wide beam pulsars can

1409: therefore be expected to be dominated by one polarization mode, which

1410: could potentially also explain their high degree of linear

1411: polarization. The unpolarized bump which is observed in the middle of

1412: some of these profiles could be the un-refracted part of the beam,

1413: which is depolarized because the presence of the extra-ordinary mode.

1414: These central components are strongest at lower frequencies,

1415: consistent with the steeper spectral index which is often observed for

1416: the central components of pulse profiles (e.g. \citealt{ran83}). 

1417: However, it remains to be seen if propagation effects can be strong

1418: enough to explain the extreme pulse widths which are observed.

1419: 

1420: 

1421: The emission geometry appears to be different for high $\dot{E}$

1422: pulsars and is possibly more similar to that of the high energy

1423: emission. However, not all pulsars with high $\dot{E}$ values produce

1424: these extremely wide profiles. Apparently only a subset of the high

1425: $\dot{E}$ pulsars have emission geometries which are very different

1426: from normal radio pulsars. A high $\dot{E}$ is therefore an important

1427: parameter required for the energetic wide beam pulsars, but not the

1428: only one. For instance, maybe only certain configurations of the

1429: plasma distributions enlarge the beam via propagation effects or maybe

1430: not all pulsars have a slot gap which produces radio emission. 

1431: It must also be noted that, like the high $\dot{E}$ radio pulsars, not

1432: all the high energy pulse profiles of pulsars are doubles (e.g. PSR

1433: B1706--44).

1434: 

1435: 

1436: \subsection{Emission heights}

1437: 

1438: There is no evidence that pulsars with large emission heights (derived

1439: from their PA-swing) have wider profiles. It is therefore not clear

1440: what the physical meaning of these emission heights is. There are many

1441: reasons why the derived emission heights could be wrong, including

1442: asymmetric beams, partially active polar caps or sweepback of the

1443: magnetic field lines.  Also if the PA-swing of emission which is

1444: emitted far out in the magnetosphere the PA can be expected to deviate

1445: from the rotating vector model. For instance, the PA-swing for the

1446: outer gap model is predicted to have the steepest slope near the edges

1447: of the profile (\citealt{ry95}), rather then at the pulse longitude

1448: corresponding to the location of the magnetic axis. This would

1449: complicate the calculation of emission heights from the observed

1450: PA-swing considerably. Nevertheless, the fact that most PA-swings

1451: trail the centroid of the profile suggest that the derived emission

1452: heights do carry some information.

1453: 

1454: The emission of the energetic wide beam pulsars should come

1455: from near the light cylinder in order to explain the width of the

1456: pulse profiles. However, the derived emission heights from the

1457: PA-swing fits seem to suggest that the emission heights are not

1458: unusually large. This can be seen as support that the beams of

1459: energetic wide beam pulsars are wide because of propagation effects

1460: instead of caused by a large emission height. An alternative

1461: interpretation is that the emission height at the edge of the beam is

1462: much larger than in the centre of the beam. This would fit in nicely

1463: with the result of \cite{gg03}, who concluded that the outer

1464: components of PSR B0329+54 are emitted from higher in the

1465: magnetosphere. It also fits in nicely with the hypothesis that the

1466: emission of the energetic wide beam pulsars comes from an extended

1467: emission height range, making the emission geometry very similar to

1468: the slot gap model.

1469: 

1470: \subsection{Interpulse problem?}

1471: 

1472: The conclusion that the beams of energetic wide beam pulsars are large

1473: appears to be unavoidable. If this is the case, than one would expect

1474: that it is very likely for the line of sight to intersect the beams of

1475: both poles of the pulsar. Using the model described by \cite{wj08a}

1476: the probability for the line of sight to intersect both beams is

1477: predicted to be 64\%, assuming $\rho=\degrees{75}$ and a random

1478: orientation of the magnetic axis and the line of sight. However, there

1479: is no clear example of an energetic wide beam pulsar which has a

1480: (double peaked) interpulse.

1481: The ``interpulse problem'' is then why we do not observe the

1482: interpulses of the energetic wide beam pulsars.

1483: 

1484: It is argued by \cite{man96} that the steep edges form the outer edge

1485: of an extremely wide beam, which would make the peak-to-peak

1486: separation of the profiles wider than \degrees{180}. In that case the

1487: weak bumps observed for some of these pulsars are then separated by

1488: half a rotational phase from the centre of the profile, which would

1489: make them interpulses. However, because the bumps fill in the region

1490: in between the sharp edges, it seems more likely that the sharp edges

1491: form the inner edges of a wide beam. In that case the profiles are

1492: less wide and the bump forms the centre of the same wide beam.

1493: 

1494: The interpulse problem suggests that the beam sizes are different

1495: for the magnetic poles of the energetic wide beam pulsars. As

1496: discussed above, not all profiles of high $\dot{E}$ pulsars have wide

1497: components. This implies that other criteria have to be met in order

1498: to make the beams wide. These criteria are not necessarily met

1499: simultaneously for both poles, which would reduce the fraction of

1500: pulsars with interpulses. The very different shape of the main- and

1501: interpulse of PSR B1055--52 show that interpulse beams can have very

1502: different shapes, hence possibly also very different sizes. Only 5 of

1503: the 26 pulsars with an $\dot{E}>5\times10^{35}$ erg~s$^{-1}$ in

1504: Fig. \ref{Fig_pulsewidths_edot} are classified to be energetic wide

1505: beam pulsars. Therefore the chance that both poles produces a wide

1506: beam is expected to be only $\sim4\%$ if the chance of producing a

1507: wide beam is independent for each pole.

1508: 

1509: A more extreme point of view to solve the interpulse problem is put

1510: forward by \cite{man96} who argues, following \cite{ml77}, that all

1511: pulsars only have one active wide beam. Although this would trivially

1512: solve the interpulse problem, it does not explain the concentration of

1513: main- interpulse separations near \degrees{180} (see

1514: Fig. \ref{Fig_pulsewidths_edot}).

1515: 

1516: \subsection{Polarization}

1517: 

1518: The degree of linear polarization is found in several studies to

1519: increase with $\dot{E}$. Such behaviour is predicted for the natural

1520: wave modes in the cold plasma approximation (\citealt{hlk98}).  One of

1521: the most surprising results of this paper is the sudden increase in

1522: the degree of linear polarization with $\dot{E}$. This suggest that

1523: pulsars can be separated into two groups which have distinct physical

1524: properties. This could either be in the structure of the magnetosphere

1525: or the physics of the emission mechanism itself. It is remarkable that

1526: over 7 orders of magnitude in $\dot{E}$ the degree in linear

1527: polarization is the only thing that is clearly changing.

1528: 

1529: It has been shown that the degree of polarization is clearly related

1530: to the presence of OPM transitions in the PA-swing. The two plasma

1531: modes (X-mode and O-mode) can be expected to be separated more in

1532: pulse longitude for high $\dot{E}$ pulsars, because the difference in

1533: their refractive indices is larger (e.g. \citealt{hlk98}). This could

1534: prevent the modes from mixing, and therefore prevent

1535: depolarization. However, the fact that high $\dot{E}$ pulsars are less

1536: likely to show jumps in their PA-swing suggests that they only

1537: effectively generate one of the modes. \cite{jhv+05} found that the

1538: velocity vectors of most pulsars make an angle close to either

1539: \degrees{0} or \degrees{90} with the PA of the linear polarization

1540: (measured at the inflection point). This is interpreted as evidence

1541: for alignment of the rotation axis of the star with its proper motion

1542: vector and the bimodal nature of the distribution of angles is

1543: interpreted to be due to the domination of different plasma modes for

1544: different stars. This result therefore suggest that if the emission of

1545: high $\dot{E}$ pulsars is dominated by one mode, it could be either of

1546: the two for different pulsars.

1547: If the profiles of the energetic wide beam pulsars are widened by

1548: refraction, than their emission should be dominated by the O-mode

1549: which can be refracted in the pulsar magnetosphere. 

1550: 

1551: 

1552: A very different interpretation of the sudden increase of the degree

1553: of linear polarization with $\dot{E}$ is based on the fact that the

1554: $\dot{E}$ at the transition is very similar to the death line for

1555: curvature radiation (\citealt{hm02}). This death line could

1556: potentially cause a sudden change in for instance the plasma

1557: distribution in the magnetosphere (which is responsible for the

1558: refraction of the plasma waves), or it could possibly change the

1559: emission mechanism which is responsible for the radio emission. The

1560: possible link between the degree of linear polarization of the radio

1561: emission and the mechanism for the production of the high energy

1562: emission could therefore suggest that the $\gamma$-ray efficiency is

1563: correlated with the degree of linear polarization in the radio band.

1564: It would therefore be extremely interesting to find out if a pulsar

1565: like PSR J0108--1431, which is highly polarized with a low $\dot{E}$,

1566: can be detected by the Fermi satellite.

1567: 

1568: \subsection{Circular polarization}

1569: 

1570: The trend noted by \cite{jw06} that the degree of circular

1571: polarization is usually higher in the second component of doubles is

1572: clearly present in our data as well. It is a possibility that this is

1573: a result of the co-rotation velocity of the emission region. As shown

1574: by for instance \cite{dyk08}, particles travelling along the magnetic

1575: field lines of a rotating dipole will follow stronger curved paths (in

1576: the inertial observer frame) at the leading half of the pulse profiles

1577: compared to the trailing half of the pulse profile. This is the reason

1578: why the observed PA-swing appears to be shifted with respect to the

1579: pulse profile (equation \ref{heighBCW92}). The degree of circular

1580: polarization is in general highest in the central parts of the pulse

1581: profile, there where the curvature of the field lines is weakest. This

1582: could therefore suggest that the location of the highest degree of

1583: circular polarization is, like the PA-swing, shifted to later times by

1584: co-rotation.

1585: 

1586: 

1587: 

1588: 

1589: 

1590: 

1591: \section{Conclusions}

1592: 

1593: In this paper we present and discuss the polarization profiles of a

1594: large sample of young, highly energetic pulsars which are regularly

1595: observed with the Parkes telescope. This sample is compared with a

1596: sample of a similar number of low $\dot{E}$ objects in order to draw

1597: general conclusions about their differences.

1598: 

1599: There is some evidence that the total intensity profiles of high

1600: $\dot{E}$ pulsars are slightly simpler based on a classification by

1601: eye. However, there is no difference in the complexity of the

1602: mathematical decomposition of the profiles, the amount of overlap

1603: between the components of doubles or the degree of profile symmetry.

1604: We therefore conclude that differences in the total intensity pulse

1605: morphology between high and low $\dot{E}$ pulsars are in general

1606: rather subtle. High $\dot{E}$ pulsars appear to show a stronger

1607: $W_{10}-P$ correlation which is closer to the theoretical expectation,

1608: suggesting that for high $\dot{E}$ pulsars there are less complicating

1609: factors in the emission geometry.

1610: 

1611: A much more pronounced difference between high and low $\dot{E}$

1612: pulsars is the degree of polarization. The degree of polarization was

1613: already known to increase with $\dot{E}$, but our data shows there is

1614: a rapid transition between relatively unpolarized low $\dot{E}$

1615: pulsars and highly polarized high $\dot{E}$ pulsars. The increase in

1616: the degree of polarization is related to the absence of OPM

1617: jumps. Refraction of the radio emission is expected to be more

1618: effective in the magnetosphere of high $\dot{E}$ pulsars, which could

1619: prevent depolarization because of mixing of the plasma modes. The

1620: absence of OPM jumps suggest that one of the two modes (not

1621: necessarily the same for different pulsars) dominates over the other.

1622: The $\dot{E}$ of the transition is very similar to the death line for

1623: curvature radiation, which could be the reason why the transition is

1624: relatively sharp. This potential link between the high energy

1625: radiation and the radio emission could mean that the $\gamma$-ray

1626: efficiency is correlated with the degree of linear polarization in the

1627: radio band.

1628: 

1629: The degree of circular polarization is in general higher in the second

1630: component of doubles. This remarkable correlation is possibly caused

1631: by the effect of co-rotation on the curvature of the field lines in

1632: the inertial observer frame, making this effect very similar to the

1633: shift of the PA-swing predicted for a finite emission height. In

1634: addition, the trailing component usually dominates in total power.

1635: 

1636: 

1637: The $W_{10}-\dot{E}$ distribution clearly shows sub-groups which are

1638: not visible in the pulse $W_{10}-P$ distribution, suggesting that

1639: $\dot{E}$ is an important physical parameter for pulsar

1640: magnetospheres. Besides a group of pulsars which probably have beams

1641: aligned with the rotation axis and a group of pulsars with interpulses

1642: which are probably orthogonal rotators there is a group of energetic

1643: wide beam pulsars. These young pulsars have very wide profiles with

1644: often steep edges which are likely to be emitted from a single pole.

1645: 

1646: 

1647: The profile properties of the energetic wide beam pulsars are similar

1648: to those of the high energy profiles, suggesting another possible link

1649: with the high energy emission. We therefore propose that the emission

1650: of these pulsars could come, like the high energies, from extended

1651: parts of the magnetosphere. 

1652: The extended height range from where the emission is emitted will

1653: smear out the complex features of the profiles.

1654: A large height range could also prevent the emission height to vary

1655: much from pulsar to pulsar, which would result in a stronger

1656: $W_{10}-P$ correlation for high $\dot{E}$ pulsars, as is indeed

1657: observed. If the radio emission and $\gamma$-ray emission of these

1658: pulsars indeed come from similar locations in the magnetosphere, one

1659: would expect the radio and $\gamma$-ray profiles to look alike,

1660: something that potentially can be tested by the Fermi satellite.

1661: 

1662: An alternative mechanism to produce the profiles of the energetic wide

1663: beam pulsars could be the combination of refraction (or scattering) of

1664: radio waves in pulsar magnetospheres with an emission height which is

1665: different for different field lines (\citealt{wsv+03}). Refraction

1666: (and scattering) is most severe for the ordinary wave mode, suggesting

1667: that these profiles are dominated by one polarization mode. This would

1668: be consistent with the high degree of linear polarization observed for

1669: these pulsars. The unpolarized bump in the middle of the profiles of

1670: the energetic wide beam pulsars could be the un-refracted part of the

1671: beam, which is depolarized because the mixing of the plasma modes. 

1672: 

1673: 

1674: Measurements of the emission height could potentially discriminate

1675: between the refraction model and the extended emission height model

1676: for the energetic wide beam pulsars. There is no evidence that pulsars

1677: with large emission heights (derived from their PA-swing) have wider

1678: profiles. It is therefore not clear what the physical meaning of these

1679: emission heights are. It could supports the idea that the beams of

1680: energetic wide beam pulsars are wide because of refraction instead of

1681: caused by a large emission heights. However, it could also mean that

1682: the emission height of the outer parts of the beam is much larger than

1683: for the central parts, making the emission geometry similar to that of

1684: a slot gap.

1685: 

1686: 

1687: 

1688: 

1689: 

1690: 

1691: 

1692: 

1693: \section*{Acknowledgments}

1694: The authors would like to thank the referee, Axel Jessner, for his

1695: useful comments on this paper. The Australia Telescope is funded by

1696: the Commonwealth of Australia for operation as a National Facility

1697: managed by the CSIRO.

1698: 

1699: \bibliographystyle{mn2e}

1700: \bibliography{journals_apj,modrefs,psrrefs} 

1701: \label{lastpage} 

1702: 

1703: \appendix

1704: \section{Polarization figures at both 10 and 20~cm}

1705: {\bf The astro-ph version is missing 528 figures due to file size

1706: restrictions. Please download the paper including the appendices from

1707: http://www.atnf.csiro.au/people/pulsar/wj08b.pdf.}

1708: %\input{figures10.tex}

1709: %\clearpage

1710: \section{Polarization figures only at 20~cm}

1711: {\bf The astro-ph version is missing 528 figures due to file size

1712: restrictions. Please download the paper including the appendices from

1713: http://www.atnf.csiro.au/people/pulsar/wj08b.pdf.}

1714: %\input{figures20.tex}

1715: \section{Table with the profile properties}

1716: \onecolumn

1717: \input{table.tex}

1718: 

1719: \clearpage

1720: \end{document}

1721: 

1722: