1: %\documentclass[manuscript]{aastex}
2: \documentclass[12pt, preprint]{aastex}
3: %\documentclass{emulateapj}
4: \usepackage{amsmath}
5: %\usepackage{apjfonts}
6:
7: \shorttitle{Cyclic evolution of radio pulsars}
8: \shortauthors{Beskin et al.}
9:
10: \begin{document}
11:
12: \title{Cyclic evolution of radio pulsars on the time scale of hundreds
13: of years}
14:
15: \author{G. Beskin\altaffilmark{1, 3}, A. Biryukov\altaffilmark{2} and
16: S.Karpov\altaffilmark{1, 3} }
17:
18: \altaffiltext{1}{Special Astrophysical Observatory of Russian Academy of
19: Sciences, Russia}
20: \altaffiltext{2}{Sternberg Astronomical Institute of Moscow State University,
21: Moscow, Russia}
22: \altaffiltext{3}{Isaac Newton Institute of Chile, SAO branch, Russia}
23:
24: \begin{abstract}
25: The recent massive measurements of pulsar frequency second derivatives have shown
26: that they are 100-1000 times larger than expected for standard pulsar
27: slowdown low. Moreover, the second derivatives as well as braking indices
28: are even negative for about half of pulsars. We explain these paradoxical results on the
29: basis of the statistical analysis of the rotational parameters
30: $\nu$, $\dot \nu$ and $\ddot \nu$ of the subset of 295 pulsars taken mostly from
31: the ATNF database. We have found strong correlation of $\ddot \nu$ and
32: $\dot \nu$ either for $\ddot\nu > 0$ (correlation coefficient $r\approx0.9$)
33: and $\ddot\nu < 0$ ($r\approx0.85$), and of $\nu$ and $\dot\nu$
34: ($r\approx0.7$). We interpret these dependencies as evolutionary ones due to
35: $\dot\nu$ being nearly proportional to characteristic age $\tau_{ch}$. The derived
36: statistical relations as well as "anomalous" values of $\ddot\nu$ are well
37: explained in the framework of the simple model of cyclic evolution of the
38: rotational frequency of the pulsars. It combines the secular change
39: of $\nu_{tr}(t)$, $\dot\nu_{tr}(t)$ and $\ddot\nu_{tr}(t)$ according to the
40: power law with $n\approx5$ and harmonic oscillations of 100--1000 years
41: period with an amplitude from $10^{-3}$ Hz for young pulsars to $10^{-10}$ Hz
42: for elder ones. The physical nature of these cyclic variations of the
43: rotational frequency may be similar to the well-known red timing noise,
44: however, with much larger characteristic time scale.
45: \end{abstract}
46:
47: \keywords{methods: data analysis --- methods: statistical --- pulsars:
48: general}
49:
50: \section{Introduction}
51:
52: The slowdown of radio pulsars is due to the electromagnetic emission energy
53: losses. According to this "classical" approach, their rotational frequencies
54: $\nu$ evolve following the slowdown law $\dot\nu=-K\nu^n$, where $K$ is the
55: positive constant that depends on the magnetic dipole moment and the moment
56: of inertia of the neutron star, $n$ is the braking index. The latter can be
57: determined observationally from measurements of $\nu$, $\dot \nu$ and $\ddot
58: \nu$ as $n={\nu \ddot \nu}/{\dot \nu^2}$. For pure dipole structure of pulsar
59: magnetosphere $n=3$, the pulsar wind decreases this value down to $n=1$,
60: while for multipole magnetic field it is up to $n=5$ \citep{man77, man85,
61: bla88}. For such law of pulsar evolution the second derivatives of their
62: rotational frequencies (and therefore their braking indices) can't be
63: measured for objects with the characteristic age $\tau_{ch} =
64: -\nu / 2 \dot\nu > 10^6$ years as their
65: % second derivatives
66: $\ddot \nu < 10^{-30}$ s$^{-3}$, that is much lower than the current
67: detection limit \citep{liv05}. At the same time the massive measurements of
68: pulsar frequency second derivatives $\ddot \nu$ have shown that their values
69: are much larger than expected for standard spin-down law and are even
70: negative for about half of all pulsars. The corresponding braking
71: indices are from $-10^6$ to $10^6$ \citep{d'a93, chu03, hob04}.
72:
73: These strange results are usually interpreted as an influence of a red timing
74: noise, which manifests itself as irregular variations of pulse arrival
75: times relative to the secular frequency evolution \citep{gro75, cor80, cor85},
76: described through Taylor series expansion as
77: \begin{equation}
78: \nu(t)=\nu_0+\dot\nu_0(t-t_0)+\frac12\ddot\nu_0(t-t_0)^2+...
79: %\label{eq_1}
80: \end{equation}
81: where $t_0$, $\nu_0$, $\dot \nu_0$ and $\ddot \nu_0$ are initial time moment,
82: frequency and its derivatives at this moment. Historically, the secular
83: evolution of pulsar rotation rate has been considered as the linear trend of
84: the frequency behaviour, while quadratic term of the series (1) (with cubic
85: behaviour in pulse arrival times), measured as $\ddot \nu$, has been treated
86: as the signature of non-monotonic (on time scale of the observations)
87: processes -- red noise, post-glitch recovery, pulsar orbital motion and free
88: precession \citep{gro75, cor80, she96, sta00}. However, recently the increase
89: of the sensitivity of the receivers and time spans of the continuous
90: timing (from several to thirty years) made it possible to estimate the
91: higher-order coefficients of the series (1) \citep{cam00, sco03, hob04,
92: liv05}. The results of these studies show that $\ddot \nu$, while changing on
93: the time scales shorter than observation time spans, {\it on average}
94: characterizes the pulsar secular evolution rather than noise processes. The
95: latter, causing the variations mentioned earlier, worsens the accuracy of the
96: second derivative determination, while can't change its mean value
97: hundreds times \citep{cam00, hob04, liv05}.
98:
99: In this Letter we report the detection of correlation of $\ddot\nu$ and
100: $\dot\nu$ either for $\ddot\nu > 0$ and $\ddot\nu < 0$ and correlation
101: of $\nu$ and $\dot\nu$. We show that these statistical relations as well
102: as "anomalous" values of $\ddot\nu$ can be explained by the combination
103: of secular decrease of pulsar frequencies with $n\approx 5$ and their cyclic
104: variations on the time scale of several hundreds of years.
105:
106: \section{Statistical analysis of the ensemble of pulsars}
107:
108: Our statistical analysis is based on the assumption that numerous
109: measurements of the pulsar frequency second derivatives reflect their secular
110: evolution on the time scale larger than the duration of observations, and
111: uses the parameters of nearly 300 pulsars.
112:
113: \begin{figure}[t]
114: {\centering \resizebox*{1\columnwidth}{!}{\includegraphics[angle=270]
115: {f1.eps}} \par}
116: \caption{The $\ddot \nu - \dot \nu$ diagram. The figure shows the pulsars
117: from the work \citep{hob04} as
118: circles, and the objects measured by other groups as squares. Empty symbols
119: represent the pulsars associated
120: with the supernova remnants, and therefore -- relatively young ones.
121: Analytical fits for both positive and negative
122: branches are shown as solid lines.}
123: %\label{fig_1}
124: \end{figure}
125:
126: From 389 objects of ATNF catalogue \citep{man05} with known $\ddot \nu$ we
127: have compiled a list of "ordinary" radio pulsars with $P>20$ ms
128: and $\dot P
129: >10^{-17}$ s s$^{-1}$, excluding recycled, anomalous and binary pulsars, and
130: with relative accuracy of second derivative measurements better than 75\%.
131: It has been appended with 26 pulsars from other sources \citep{d'a93, chu03}.
132: The parameters of all pulsars have been plotted on the $\ddot \nu - \dot \nu$
133: diagram (Fig. 1). Here the circles are the objects with second derivatives
134: measured at the Jodrell Bank Observatory \citep{hob04}, squares -- the results
135: of other groups \citep{d'a93, chu03}, and empty symbols represent young
136: pulsars associated with supernova remnants.
137:
138: The main result of the statistical analysis of these data is a significant
139: correlation of $\ddot \nu$ and $\dot \nu$, either for 168 objects with $\ddot
140: \nu > 0$ (correlation coefficient $r\approx0.90$) and for 127 objects with
141: $\ddot \nu < 0$ ($r\approx0.85$). Either groups follow the nearly linear laws,
142: however they are not exactly symmetric with respect to $\ddot \nu=0$. Also,
143: the young pulsars with $\dot \nu < -10^{-11}$ s$^{-2}$ and $\ddot \nu < 0$
144: are absent.
145:
146: \begin{figure}[t]
147: {\centering \resizebox*{1\columnwidth}{!}{\includegraphics[angle=270]
148: {f2.eps}} \par}
149: \caption{The correlation of $\dot\nu$ and $\tau_{ch}$.}
150: %\label{fig_2}
151: \end{figure}
152:
153: Figure 2 shows the dependence of $\dot \nu$ on the characteristic age
154: $\tau_{ch}$. Of course, due to practical proportionality of these parameters
155: there is a strong dependence of $\ddot \nu$ on $\tau_{ch}$ either for pulsars
156: with $\ddot \nu >0$ (positive branch, $r \approx 0.85$) and for $\ddot \nu <
157: 0$ (negative branch, $r \approx 0.75$) (Fig. 3a). Also, we have found the
158: significant correlation of $n$ and $\tau_{ch}$ (Fig. 3b). Note some important
159: points of these results.
160:
161: \begin{itemize}
162: \item{
163: Measurements of different observational groups show the same laws, so
164: the correlations discovered are not related to the peculiarities of the
165: receiving devices of a single radio telescope.
166: }
167: \item{
168: Young pulsars confidently associated with supernova remnants are
169: systematically shifted to the left in Figure 1 and Figure 2 (empty symbols).
170: The order of their physical ages roughly corresponds to that of
171: characteristic ages. It means that any dependence on $\dot \nu$ or
172: $\tau_{ch}$ may be interpreted as the dependence on time.
173: }
174: \item{
175: Separate analysis of the whole set of parameters of pulsars with $\ddot \nu > 0$
176: and $\ddot \nu < 0$ (galactic positions, fluxes, luminosities, magnetic fields,
177: dispersion and rotation measures) shows the statistical identity of these
178: sub-groups.
179: }
180: \end{itemize}
181:
182: So, the $\ddot\nu - \dot \nu$ diagram (Fig. 1) may be interpreted as an
183: evolutionary one. In other words, each pulsar during its evolution moves
184: along the branches of this diagram while increasing the value of its
185: $\dot \nu$ (which corresponds to the increase of its characteristic age).
186: However, there is an obvious contradiction, as for $\ddot \nu < 0$ (negative
187: branch), $\dot \nu$, being negative, may only decrease with time, so the
188: motion along the negative branch may be only backward! Certainly, this
189: contradiction is easily solved by assuming non-monotonic behaviour of
190: $\nu(t)$, which may have some cyclic (quasi-periodic) component along with
191: secular one. Figure 4 shows the example of such cyclic behaviour of a pulsar.
192: The characteristic time scale $T$ of such cycles must be much shorter than
193: the pulsar life time and at the same time much larger than the time scale of
194: the observations. Each pulsar during its evolution repeatedly changes the
195: sign of $\ddot \nu$ and spends roughly half of its life time on each branch.
196: The asymmetry of the branches reflects the positive secular behaviour of
197: $\ddot \nu_{tr}(t)$, and so, secular increase of $\dot\nu(t)$. Systematic
198: decrease of branches separation reflects the decrease of the oscillations
199: amplitude and/or the increase of its characteristic time scale. Any
200: non-monotonic variations of $\nu(t)$, like post-glitch recovery, standard red
201: noise or orbital motion, will manifest itself in a similar way on the
202: $\ddot\nu - \dot \nu$ diagram and lead to extremely high values of
203: $\ddot \nu$ \citep{gro75, cor80, she96, sta00}. However, their characteristic
204: time scales vary from weeks to years, and they are detected immediately. In
205: our case we have processes on time scales of hundreds of years, and their
206: study is possible only by means of statistical methods, assuming that the
207: ensemble of pulsars of various age models the behaviour of a single pulsar
208: during its evolution.
209:
210: \begin{figure}[t]
211: {\centering \resizebox*{1\columnwidth}{!}{
212: \includegraphics[angle=270]{f3a.eps}
213: }\par}
214: {\centering \resizebox*{1\columnwidth}{!}{
215: \includegraphics[angle=270]{f3b.eps}
216: }\par}
217: \caption{(a) The correlation of $\ddot\nu$ and $\tau_{ch}$. Empty symbols
218: represent the young pulsars associated with the supernova remnants.
219: It is clear that the pulsar characteristic age roughly corresponds
220: to the real one. (b) The correlation of $n$ and $\tau_{ch}$.}
221: %\label{fig_2}
222: \end{figure}
223:
224: \begin{figure}[t]
225: {\centering \resizebox*{1\columnwidth}{!}{\includegraphics[angle=270]
226: {f4.eps}} \par}
227: \caption{The sample of pulsar cyclic behaviour on the $\ddot \nu - \dot \nu$
228: diagram. The solid line shows the evolution of a pulsar according to the
229: simple harmonic model presented in (2) -- (7) with the oscillation period
230: $T=1000$ years. The dashed line shows the corresponding secular evolution
231: term $\ddot\nu_{tr}(t)$, and the oscillations are around it, symmetric in
232: linear scale. For $\dot\nu<-7\cdot10^{-12}$ s$^{-2}$, the amplitude of the
233: oscillations is smaller than the secular term, thus $\ddot\nu$ is always
234: positive, for larger values it changes its sign repeatedly, spending
235: approximately half of the time in the negative region.}
236: %\label{fig_3}
237: \end{figure}
238:
239: Note that the correlations between $|\ddot P|$ and $\dot P$, similar to these
240: discovered here (but without distinguishing between objects with $\ddot \nu >
241: 0$ and $\ddot \nu < 0$), were discussed earlier \citep{cor85, arz94, lyn99}
242: as an evidence of higher timing noise of young pulsars.
243:
244: \section{The simplest model of pulsar cyclic evolution}
245:
246: Cyclic variations of the pulsar rotational frequency may be complicated --
247: quasi-periodic, multi-harmonic, stochastic. As a rough initial approximation,
248: we use a simple model of harmonic oscillations superimposed on secular
249: evolution. In this case,
250: \begin{equation}
251: \nu(t)=\nu_{tr}(t) + A(t)\sin\left[{\Omega(t)t + \phi_0}\right] ,
252: %\label{eq_2}
253: \end{equation}
254: where $A(t)$ is the amplitude of periodic frequency variations,
255: $\Omega(t)=2\pi / T(t)$ -- its frequency, and $\nu_{tr}(t)$ is the
256: secular evolution term. This leads to
257: \begin{equation}
258: \dot \nu(t)=\dot \nu_{tr}(t) + A(t)\Omega(t)\cos\left[{\Omega(t)t +
259: \phi_0}\right] ,
260: %\label{eq_3}
261: \end{equation}
262: \begin{equation}
263: \ddot \nu(t)=\ddot \nu_{tr}(t) - A(t)\Omega^2(t)\sin\left[{\Omega(t)t +
264: \phi_0}\right] ,
265: %\label{eq_4}
266: \end{equation}
267: where we neglected the terms with $A(t)$ and $\Omega(t)$ derivatives due
268: to its relative smallness. From (3) and (4) it is easy to get the relation
269: between $\ddot \nu(t)$ and $\dot \nu(t)$ in the cycloidal form
270: $[\ddot \nu(t)-\ddot\nu_{tr}(t)]^2=\Omega^4(t)A^4(t)- \Omega^2(t)[\dot\nu(t)-
271: \dot\nu_{tr}(t)]^2$, which is plotted in Figure 4.
272:
273: By marking with "+" and "-" subscripts the values related to the positive
274: and negative branches on the $\ddot\nu - \dot\nu$ diagram (Fig. 1),
275: correspondingly, and pointing to the amplitude values of $\nu$ and $\ddot
276: \nu$ (i.e. assuming $\sin{\left[\Omega(t)t + \phi_0\right]} = \mp 1$), we get
277: \begin{equation}
278: \nu_{\pm}(t) = \nu_{tr}(t) \mp A(t)
279: %\label{eq_5}
280: \end{equation}
281: \begin{equation}
282: \dot \nu_{\pm}(t)=\dot \nu_{tr}(t)
283: %\label{eq_6}
284: \end{equation}
285: \begin{equation}
286: \ddot \nu_{\pm}(t) = \ddot \nu_{tr}(t) \pm A(t)\Omega^2(t).
287: %\label{eq_7}
288: \end{equation}
289: So, due to (6), the branches of $\ddot\nu - \dot\nu$ diagram describe the
290: dependence of pulsar frequency second derivative $\ddot\nu$ on the secular
291: behaviour of its first derivative $\dot \nu_{tr}$, which corresponds to the
292: dependence on time.
293:
294: Relations (2) -- (7) describe the apparent evolution of some "average"
295: pulsar. Of course, in the ensemble of all pulsars the scatter of points on
296: the $\ddot\nu - \dot\nu$ diagram reaches 4 orders of magnitude due to
297: distribution of the neutron star parameters and initial conditions. Due to
298: this fact it is impossible to determine $\ddot \nu_{tr}(t)$ directly as the
299: half-sum of $\ddot \nu_{+}(t)$ and $\ddot \nu_{-}(t)$ from (7) and Fig. 1, as
300: it requires the subtraction of two close values with significant errors.
301:
302: First of all we compute the behaviour of $\nu_{tr}(t)$ (or, $\nu_{tr}(\dot
303: \nu_{tr})$) by plotting the studied pulsar group on the $\dot \nu - \nu$ diagram
304: (Fig. 5). The objects with $\ddot \nu > 0$ and $\ddot \nu < 0$ are marked as
305: filled and empty circles, correspondingly. It is easily seen that the
306: behaviour of these two sub-groups is the same, i.e. $A(t)$ according to (5)
307: is significantly smaller than the intrinsic scatter of $\nu(\dot \nu)$.
308: However, a strong correlation between $\nu_{tr}$ and $\dot\nu_{tr}$
309: ($r\approx0.7)$ is seen, and
310: \begin{equation}
311: \dot\nu_{tr}=-C\nu_{tr}^n,
312: %\label{eq_9}
313: \end{equation}
314: where $C=10^{-15.26\pm1.38}$ and $n=5.13\pm0.34$.
315: So, the secular evolution of the "average" pulsar is according to the
316: "standard" spin-down law with $n\approx5$! This value may suggest the
317: importance of multipole components of pulsar magnetic field, or the deviation
318: of the angle between pulsar rotational and dipole axes from ${\pi}/{2}$
319: \citep{man77, con05}. From (8) we may easily get the relation between
320: $\ddot\nu_{tr}$ and $\dot\nu_{tr}$
321: \begin{equation}
322: \ddot\nu_{tr} (\dot\nu_{tr}) = nC^{\frac1n}(-\dot\nu_{tr})^{2-\frac1n},
323: %\label{eq_10}
324: \end{equation}
325: which is shown in Figure 6 as a thick dashed line. The amplitude of the
326: $\ddot \nu(t)$ oscillations, $A_{\ddot \nu}(t)$, may be easily computed
327: from (7) using fits for $\ddot\nu_{+}$ and $\ddot\nu_{-}$ in Figure 1,
328: \begin{equation}
329: A_{\ddot \nu}(t) = A(t)\Omega^2(t)=\frac12\left[\ddot \nu_{+}(t) - \ddot
330: \nu_{-}(t)\right]
331: %\label{eq_8}
332: \end{equation}
333:
334: \begin{figure}[t]
335: {\centering \resizebox*{1\columnwidth}{!}{\includegraphics[angle=270]
336: {f5.eps}} \par}
337: \caption{The $\dot \nu - \nu$ diagram for the pulsars with the measured
338: second derivative. The filled symbols are objects with positive $\ddot\nu$,
339: the empty ones -- with negative. The behaviour of both subsets is the same.
340: The solid line represents the best fit, corresponding to the $n\approx5$
341: braking index, the dotted ones -- 1-$\sigma$ range.}
342: %\label{fig_4}
343: \end{figure}
344:
345: \begin{figure}[t]
346: {\centering \resizebox*{1\columnwidth}{!}{\includegraphics[angle=270]
347: {f6.eps}} \par}
348: \caption{The $\ddot \nu - \dot \nu$ diagram with the simple harmonic
349: oscillations model. The solid line is the amplitude $A_{\ddot\nu}$
350: of the frequency second derivative variations according to (10),
351: the dashed -- the secular term $\ddot\nu_{tr}$ from (9), and the dot-dashed
352: lines are the envelopes of the oscillations $\ddot\nu_{tr}\pm A_{\ddot\nu}$
353: with 1-$\sigma$ ranges (dotted lines). Each pulsar spends the majority of
354: its life time at or very near the envelopes.}
355: %\label{fig_5}
356: \end{figure}
357:
358: The result is shown in Figure 6 as a solid line. The amplitude $A_{\ddot\nu}$
359: is decreasing almost linearly with the increase of $\dot \nu_{tr}$, i.e. life
360: time. This may be related to the decrease of either $A(t)$ or $\Omega(t)$.
361:
362: From Figure 6 it is seen that our simple model formulated in (2) -- (7)
363: describes the observational data rather well, at least for the pulsars with
364: $\dot\nu>-10^{-11}$ s$^{-2}$, i.e. with $\tau_{ch}>10^4$ years. For
365: $\dot\nu<-7\cdot10^{-12}$ s$^{-2}$, the $A_{\ddot\nu}$ is smaller than
366: $\ddot\nu_{tr}$, and during the oscillations $\ddot\nu>0$ always, that
367: explains the absence of pulsars of the negative branch of $\ddot\nu-\dot\nu$
368: diagram in this region. The situation changes dramatically for
369: $\dot\nu>10^{-11}$ s$^{-2}$, i.e. $\tau_{ch} > 10^4$ years. Here
370: $A_{\ddot\nu}$ becomes larger than $\ddot\nu_{tr}$ and pulsars with
371: $\ddot\nu < 0$ and $n < 0$ appear. In other words, the cyclic behaviour
372: according to the model described by (2) -- (7) begins to manifest itself in
373: the observing quantities, -- $\ddot\nu$ and $n$ begins to deviate
374: significantly from the predictions of the "standard" spin-down law evolution.
375:
376: On the basis of these assumptions it is possible to get a rough estimate
377: of the maximal period of the cyclic behaviour as the several times of the
378: time scale of the vanishing of $\ddot\nu$ for the young pulsars with the
379: measured $\dddot\nu$ as $T_{max}\approx-4{\ddot\nu}/{\dddot\nu}$. For {\it
380: PSR B1509-58} (with $\ddot\nu\approx2\cdot10^{-21}$ s$^{-3}$ and
381: $\ddot\nu\approx-1.3\cdot10^{-31}$ s$^{-4}$) \citep{liv05} $T_{max}\approx
382: 2\cdot10^{3}$ years ($\Omega_{min}\approx10^{-10}$ s$^{-1}$). For {\it Crab}
383: (with $\ddot\nu\approx1.2\cdot10^{-20}$ s$^{-3}$ and $\ddot\nu\approx-
384: 1.4\cdot10^{-30}$ s$^{-4}$) \citep{sco03} $T_{max}\approx 10^{3}$ years
385: ($\Omega_{min}\approx2\cdot10^{-10}$ s$^{-1}$). On the other hand, the
386: estimations of $\Omega$ may be done using the relations (2) -- (4), for the
387: pulsars with several measurements of rotational parameters at different
388: epochs with enough separation. So, the lower limit is $\Omega_{min}\approx
389: \sqrt{\left|{\Delta\ddot\nu}/{\Delta\nu}\right|}$. From the results of
390: observations of the old ($\tau \propto 10^6$ years) pulsars, {\it B0823+26},
391: {\it B1706-16}, {\it B1749-28} and {\it B2021+51} with 7 to 10 years
392: separation, we have obtained the same estimate of $T_{max}\approx10^3 \div
393: 2\cdot 10^3$ years. Finally, the period of cyclic oscillations is likely to
394: be in the range of 100 -- 1000 years ($2\cdot10^{-10} < \Omega < 2\cdot
395: 10^{-9}$ s$^{-1}$). Here we have suggested that the minimal cycle period has
396: to be several times the pulsar observation period -- 30 years.
397:
398: Taking into account that $A_{\ddot \nu} = \Omega^2A$ is changing several
399: orders of magnitude, our initial assumption of relative constancy of $\Omega$
400: and significant decrease of $A$ during the evolution is somehow justified.
401: Assuming $\Omega\approx10^{-9}$ s$^{-1}$, we get for the amplitude of
402: frequency oscillations $A={A_{\ddot \nu}}/{\Omega^2}$ values from $4\cdot
403: 10^{-4}$ s$^{-1}$ for the youngest pulsars till $10^{-10}$ s$^{-1}$ for the
404: eldest.
405:
406: \section{Discussion and conclusions}
407:
408: The physical reasons of the discussed cyclic variations of the pulsar
409: rotational frequency may be similar to the ones of the well-known red timing
410: noise. In the first place, the free precession \citep{sta00} is the periodic
411: effect most suitable for the proposed harmonic model. Note that the
412: interaction of the superfluid core and the crust of the neutron star does not
413: depress its development \citep{alp05}. Several processes had been proposed
414: for the explanations of the red noise \citep{cor80, cor81, che87} -- from the
415: collective effects in the superfluid core till the electric current fluctuations in
416: the pulsar magnetosphere. The possibility of its utilization for the
417: explanation of the frequency variations on the time scale of hundreds of
418: years is yet to be analyzed.
419:
420: The argument towards the similarity between the discussed variations and
421: the red timing noise is the coincidence of the red noise $\ddot\nu$ amplitude
422: extrapolated according to its power spectrum slope \citep{bay99} to the time
423: scale of hundreds of years, with $A_{\ddot\nu}$ in our model for the same
424: $\dot\nu$, i.e. the same ages (see Fig. 6).
425:
426: Finally, note that the real evolution of the pulsar rotational frequencies
427: may be significantly more complicated than the proposed simple model. The
428: variations may be non-harmonic, their amplitudes, phases and frequencies may
429: change randomly. The principal point is that all the pulsars evolve this
430: way, and the time scale of such variations must significantly exceed several
431: tens of years. That explains the anomalous values of the observed $\ddot\nu$
432: and braking indices.
433:
434: \acknowledgments
435:
436: This work has been supported by the Russian Foundation for Basic
437: Research (grant No 04-02-17555), Russian Academy of Sciences (program
438: "Evolution of Stars and Galaxies"), and by the Russian Science Support
439: Foundation.
440:
441: \begin{thebibliography}{1}
442:
443: \bibitem[Alpar(2005)]{alp05}
444: Alpar, M. A. 2005, preprint (astro-ph/0505073)
445:
446: \bibitem[Arzoumanian et al.(1994)]{arz94}
447: Arzoumanian, Z., Nice, D. J., Taylor, J. H., \& Thorsett, S. E. 1994,
448: \apj, 422, 671
449:
450: \bibitem[Baykal et al.(1999)]{bay99}
451: Baykal, A., Alpar, M.A., Boynton, P.E. \& Deeter, J.E. 1999,
452: \mnras, 306, 207
453:
454: \bibitem[Blandford \& Romani(1988)]{bla88}
455: Blandford, R.D \& Romani R.W. 1988, \mnras, 234, 57P
456:
457: \bibitem[Camilo et al.(2000)]{cam00}
458: Camilo, F., Kaspi, V. M., Lyne, A. G., Manchester, R. N., Bell, J. F.,
459: D'Amico, N., McKay, N. P. \& Crawford, F. 2000, \apj, 541, 367
460:
461: \bibitem[Cheng (1987)]{che87}
462: Cheng, K.S. 1987, \apj, 321, 799
463:
464: \bibitem[Chukwude(2003)]{chu03}
465: Chukwude, A.E. 2003, \aap, 406, 667
466:
467: \bibitem[Contopoulus \& Spitkovsky(2005)]{con05}
468: Contopoulus, I., \& Spitkovsky, A. 2005, preprint (astro-ph/0512002)
469:
470: \bibitem[Cordes \& Helfand(1980)]{cor80}
471: Cordes, J. M. \& Helfand, D. J. 1980, \apj, 239, 640
472:
473: \bibitem[Cordes \& Greenstein(1981)]{cor81}
474: Cordes, J. M. \& Greenstein, G. 1981, \apj, 245, 1060
475:
476: \bibitem[Cordes \& Downs(1985)]{cor85}
477: Cordes, J. M. \& Downs, G. S. 1985, \apjs, 59, 343
478:
479: \bibitem[D'Alessandro et al.(1993)]{d'a93}
480: D'Alessandro, F., McCulloch, P. M., King, E. A., Hamilton, P. A. \& McConnell,
481: D. 1993, \mnras, 261, 883
482:
483: \bibitem[Groth(1975)]{gro75}
484: Groth, E.J. 1975, \apj, 529, 431
485:
486: \bibitem[Hobbs et al.(2004)]{hob04}
487: Hobbs, G. B., Lyne, A. G., Kramer, M., Martin, C. E. \& Jordan, C.
488: 2004, \mnras, 353, 1311
489:
490: \bibitem[Livingstone et al.(2005)]{liv05}
491: Livingstone, M. A., Kaspi, V. M. Gavriil, F. P., \& Manchester, R. N. 2005,
492: \apj, 619, 1046
493:
494: \bibitem[Lyne(1999)]{lyn99}
495: Lyne, A. 1999, in Pulsar Timing, General Relativity and the Internal
496: Structure of Neutron Stars, ed. Z. Arzoumanian, F. Van der Hooft, \&
497: E. P. J. van den Heuvel (Amsterdam: Koninklijke Nederlandse Akademie van
498: Wetenschappen), 141
499:
500: \bibitem[Manchester \& Taylor(1977)]{man77}
501: Manchester, R. N. \& Taylor, J. H. 1977, Pulsars (San Francisco: Freeman)
502:
503: \bibitem[Manchester, Durdin \& Newton(1985)]{man85}
504: Manchester, R. N., Durdin, J. M., \& Newton, L. M. 1985, Nature, 313, 374
505:
506: \bibitem[Manchester et al.(2005)]{man05}
507: Manchester, R. N., Hobbs, G. B., Teoh, A. \& Hobbs, M. 2005, \aj, 129,
508: 1993
509:
510: \bibitem[Scott, Finger \& Wilson(2003)]{sco03}
511: Scott, D., Finger, M. \& Wilson, C. 2003, \mnras, 344, 412
512:
513: \bibitem [Shemar \& Lyne(1996)]{she96}
514: Shemar, A. L. \& Lyne, A. G. 1996, \mnras, 282, 677
515:
516: \bibitem[Stairs, Lyne \& Shemar(2000)]{sta00}
517: Stairs, I. H., Lyne, A. G. \& Shemar, A. L. 2000, Nature, 406, 484
518:
519: \end{thebibliography}
520:
521: \end{document}
522:
523: