Astronomy & Astrophysics manuscript no.

ms ©ESO 2025
March 3, 2025

Reconstruction of spider system’s observables from orbital period

modulations via the Applegate mechanism
Vittorio De Falco1, 2⋆ , Amodio Carleo3⋆⋆ , Alessandro Ridolfi3 ⋆⋆⋆ , Alessandro Corongiu3⋆⋆⋆⋆

1
Scuola Superiore Meridionale, Largo San Marcellino 10, 80138 Napoli, Italy,
2
Istituto Nazionale di Fisica Nucleare (INFN), sez. di Napoli, Via Cinthia 9, I-80126 Napoli, Italy
3
INAF – Osservatorio Astronomico di Cagliari, Via della Scienza 5, I-09047 Selargius (CA), Italy

Received March 3, 2025; accepted XXX

arXiv:2502.21283v1 [[Link]] 28 Feb 2025

ABSTRACT

Redback and black widow pulsars are two classes of peculiar binary systems characterized by very short orbital periods, very low
mass companions, and, in several cases, regular eclipses in their pulsed radio signal. Long-term timing revealed systematic but
unpredictable variations in the orbital period, which can most likely be explained by the so-called Applegate mechanism. This relies
on the magnetic dynamo activity generated inside the companion star and triggered by the pulsar wind, which induces a modification
of the star’s oblateness (or quadrupole variation). This, in turn, couples with the orbit by gravity, causing a consequent change
in the orbital period. The Applegate description limits to provide estimates of physical quantities by highlighting their orders of
magnitude. Therefore, we derive the time-evolution differential equations underlying the Applegate model, that is, we track such
physical quantities in terms of time. Our strategy is to employ the orbital period modulations, measured by fitting the observational
data, and implementing a highly accurate approximation scheme to finally reconstruct the dynamics of the spider system under study
and the relative observables. Among the latter is the magnetic field activity inside the companion star, which is still a matter of debate
for its complex theoretical modeling and the ensuing expensive numerical simulations. As an application, we exploit our methodology
to examine two spider sources: 47 Tuc W (redback) and 47 Tuc O (black widow). The results obtained are analyzed and then discussed
with the literature.
Key words. Stars: binaries: eclipsing, Stars: magnetic field, Binary Systems

1 1. Introduction by energetic particles and/or γ-rays produced by the pulsar wind 21

1
(Kluzniak et al. 1988; van den Heuvel & van Paradijs 1988; 22
2 Among the more than ∼ 3600 radio pulsars currently known , Ruderman et al. 1989). This led astronomers to propose the so- 23
3 about 14% are millisecond pulsars (MSPs). These are old neu- called evaporation scenario, accordong to which the secondary 24
4 tron stars (NSs) endowed with relatively weak magnetic fields star gets ablated until it fully disappears, thus leaving an iso- 25
5 (B ≃ 107 − 109 G) and very short rotational periods (∼ 1 − lated MSP. However, successive estimates of the mass-loss rate 26
6 10 ms). MSPs represent the progeny of low-mass X-ray binaries showed that the evaporation time scale is likely much longer than 27
7 (LMXBs), where the NS gets spun up by the accreted matter the Hubble time, casting doubts on the effective occurrence of 28
8 (coming from the secondary star via Roche lobe overflow) and this phenomenon (Stappers et al. 1996a,b, 2001). 29
9 finally attains extreme rotation rates. This is known and widely Notwithstanding, eclipsing binary pulsars became increas- 30
10 accepted in the literature as the recycling scenario (Bisnovatyi- ingly important for stellar and binary evolution studies. The re- 31
11 Kogan & Komberg 1974; Alpar et al. 1982; Radhakrishnan & cycling model, initially supported by the observation of accret- 32
12 Srinivasan 1982; Bhattacharya & van den Heuvel 1991; Papitto ing millisecond X-ray pulsars (AMXPs; see e.g., Wijnands & 33
13 et al. 2013). van der Klis 1998; Falanga et al. 2005), was ultimately con- 34
14 Although MSPs are formed in binary systems, about 20% firmed by the discovery of three “transitional” pulsars, (PSR 35
15 of them are isolated (Belczynski et al. 2010). The reason for J1023−0038, J1824−2452I, and J1227−4853), i.e. systems that 36
16 this is still unclear and is matter of debate (van den Heuvel & have been observed swinging between radio-MSP and X-ray bi- 37
17 van Paradijs 1988; Rasio et al. 1989; Bhattacharya & van den nary states (e.g. Papitto et al. 2013; Stappers et al. 2014). 38
18 Heuvel 1991). A possible explanation arose after the discovery Spider pulsars are a subclass of binary MSPs, character- 39
19 of the first eclipsing binary pulsar, PSR B1957+20 (Fruchter ized by tight (orbital period ≲ 1 d) and circular (eccentricities 40
20 et al. 1988), in which the companion star is constantly ablated ≃ 10−3 −10−4 ) orbits (see e.g., Romani et al. 2012; Pallanca et al. 41
⋆
2012; Kaplan et al. 2013), and a light-weight companion. Most 42
e-mail: [Link]@[Link] (but not all) of them are also eclipsing binary pulsars. Depend- 43
⋆⋆
e-mail: [Link]@[Link]
⋆⋆⋆ ing on the mass mc of the companion, they can be further divided 44
e-mail: [Link]@[Link]
⋆⋆⋆⋆
e-mail: [Link]@[Link] into two distinct classes (Roberts 2013): black widows have de- 45
1
See [Link] generate companions with mc ≲ 0.1M⊙ , whereas redbacks have 46
for more details. However, the situation is as follows: within the galac- semi-degenerate companions with mc ≃ 0.1 − 0.4M⊙ . The evo- 47
tic plane, 135 are isolated and 209 are binaries; whereas in globular lutionary scenario of these two types of binary systems has long 48
clusters, 87 are isolated and 91 are binaries. been discussed, and it is now widely accepted to occur through 49
Article number, page 1 of 13
 A&A proofs: manuscript no. ms

50 irradiation processes (Podsiadlowski 1991; D’Antona & Ergma The necessity in having solid theoretical assessments to de- 113
51 1993; Bogovalov et al. 2008, 2012; Chen et al. 2013; Smedley scribe spider systems represents a powerful means to: (i) bet- 114
52 et al. 2015, and references therein). ter understand the stellar magnetic activity, (ii) get insights into 115
53 The long-term timing of several spiders has revealed that the dynamo processes, (iii) obtain more information on the com- 116
54 they often show significant modulations of their orbital periods, panion stars’ equation of state. In particular, the structure and 117
55 and sometimes also of their projected semi-major axis (see e.g., generation of the magnetic field in low-mass stars are still not 118
56 Shaifullah et al. 2016; Ng et al. 2018). These variations generally clear and need to be investigated. The surface magnetic field is 119
57 manifest themselves as recurrent, but not strictly periodic, cy- thought to be of the order of several kG and can be directly ob- 120
58 cles. These observational clues lead to the exclusion, as dynami- served (see Fig. 1 in Han et al. 2023). Instead, the interior fields 121
59 cal explanations, of apsidal motions (Sterne 1939) and presence are based on contrasting theoretical analyses strongly depending 122
60 of third bodies (van Buren 1986). Instead, a plausible reason can on the considered model (see e.g., Yadav et al. 2015; Feiden & 123
61 be the magnetic activity inside the companion star (Hall 1990), Chaboyer 2014; MacDonald & Mullan 2017). 124
62 closely linked to the dynamo action due to the presence of dif- In this work, we employ the Applegate mechanism, which 125
63 ferential rotation and convective zones. besides describing the phenomenology behind the spider pulsar, 126
64 Spider systems are usually found in quasi-tidally-locked con- it also provides an estimate of the order of magnitude of some 127
65 figurations, occurring when there is no relevant angular momen- related physical variables (e.g., luminosity, differential rotations, 128
66 tum transfer between the companion star and its orbit around quadrupole moment etc.). We propose a dynamical formulation 129
67 the pulsar. This is due to the tidal force acting between the co- of the Applegate model, where the gravitational source and phys- 130
68 orbiting bodies through the pulsar irradiation-driven winds (see ical quantities’ dynamics are tracked point by point during their 131
69 e.g., Applegate & Shaham 1994; Bogdanov et al. 2005), entail- time evolution. In addition, we exploit an inverse approach, be- 132
70 ing tidal dissipation of the pulsar on the companion (known ing counter-current with respect to the strategies followed in the 133
71 as tidally-powered star; see e.g., Balbus & Brecher 1976; latest papers (see e.g., Voisin et al. 2020a,b). Indeed, rather than 134
72 Kochanek 1992; Zahn 2008) and synchronous rotation (one finding a physical justification for the ∆P variations, we take its 135
73 hemisphere of a revolving body constantly faces its partner). profile from the long-term observations to reconstruct the gravi- 136
tational source and related observables’ dynamics. The approach 137
74 A possible explanation for the orbital period variations re- we adopt in this paper follows the BTX phenomenological model 138
75 lies on the Applegate mechanism (Applegate & Patterson 1987; (extension of Blandford & Teukolsky (BT) model, Blandford & 139
76 Applegate 1992b,a), where magnetic cycles induce deformations Teukolsky 1976; Bochenek et al. 2015), which allows one to 140
77 on the companion star shape, thus altering its quadrupole mo- fit the observational data within a precise timing baseline. This 141
78 ment, consequentially causing gravitational acceleration and or- scheme gives rise to a set of coupled ordinary differential equa- 142
79 bital period modulations. This phenomenon is triggered by the tions with respect to time, involving the orbital separation and 143
80 irradiation-driven winds from the pulsar, which generates a spin quadrupole moment. However, the ensuing dynamical system is 144
81 torque on the companion. This in turn induces tidal dissipation still difficult to solve analytically. Therefore, we also develop a 145
82 and energy flow, which powers the magnetic dynamo (Applegate mathematical procedure that provides highly accurate approxi- 146
83 & Shaham 1994). mate analytical solutions. This methodology allows for an easy 147
84 The observed orbital period and projected semi-major axis accomplishment of the proposed goals. 148
85 modulations can be influenced by other effects that are intrinsic The paper is structured as follows: in Sec. 2, we describe 149
86 to the system or caused by kinematic reactions relative to the the features of our dynamical model and derive the equations 150
87 observer motion (e.g., emission of gravitational waves, Doppler of motion; in Sec. 3, we propose a reasonable approximation 151
88 corrections, mass-loss of the binary, and tidal bulge forces), pattern to infer an analytical solution; in Sec. 4 our achievements 152
89 which Lazaridis et al. (2011) estimated to be orders of magni- are applied to black widow and redback systems; finally, in Sec. 153
90 tudes smaller than those caused by the gravitational quadrupole 5 we conclude with some discussions and future perspectives. 154
91 moment activity. On the other hand, Lanza et al. (1998); Lanza
92 & Rodonò (1999) proposed an alternative explanation to the
93 gravitational quadrupole moment variations. Their model ap- 2. The model 155
94 plies the tensor virial theorem (Chandrasekhar 1961) to a gen- We present the dynamical version of the Applegate mechanism, 156
95 eral magnetic field geometry to formalize, through an integral which reports the order of magnitude of the physical observables 157
96 approach, the variations in oblateness. This implies a distributed underlying the dynamics of black widow and redback binary sys- 158
97 non-linear dynamo in the convective envelopes of the compan- tems. This model is further enhanced by incorporating orbital pe- 159
98 ion star, which affects not only the quadrupole moment, but also riod modulations’ profile, derived from observations. This gives 160
99 the differential rotation. rise to a set of coupled and non-trivial ordinary differential equa- 161
100 The Applegate mechanism is still the most quoted explana- tions. We also introduce the explicit formulae of some physical 162
101 tion for its good agreement with the observations, whose mea- variables, whose plotted profiles are the goals of our work. 163
102 sured amplitudes of period modulations are ∆P/P ∼ 10−5 (over We deal with binary sources composed of a pulsar of mass 164
103 timescales of decades or longer), companion star’s variable lu- m p and a companion star of mass mc and radius Rc . The two 165
104 minosity over the ∆L/L ∼ 0.1 level, and differential rotations bodies are both treated as test particles, even if the companion 166
105 at the ∆Ω/Ω ∼ 0.01 order. These values are common in spider star should be considered as extended to account for changes in 167
106 systems (see, e.g. Ridolfi et al. 2016; Freire et al. 2017). shape. However, to simplify the mathematical treatment, we still 168
107 More recently, Voisin et al. (2020a,b) improved the Apple- regard this object as a test body and we entrust the quadrupole 169
108 gate picture, also taking into account relativistic corrections. variable to characterize the variations of matter distribution 170
109 They proposed a detailed model for describing the motion of within the star. In Fig. 1 we report a cartoon sketching the ge- 171
110 spider binary systems that allows us to accurately estimate the ometry of the problem under investigation. 172
111 observed ∆P variations with the final objective of improving the The definitions of some of the above parameters require ad- 173
112 timing solution of these gravitational sources. ditional clarifications. The companion star’s mass, mc , is not 174
Article number, page 2 of 13
 De Falco et al. (2024): Reconstruction of spider system’s observables

175 strictly constant over time, as it gradually decreases due to the of the sunspot cycles (connected to the solar subsurface mag- 227
176 mass loss driven by the pulsar wind. However, the fraction of netic activity) is about 11 yr (Baliunas & Vaughan 1985), but 228
177 matter lost over the observational period is so small (around this value can be of much longer duration in spider systems. 229
178 10−10 M⊙ /yr)2 that it is reasonable to approximate mc through From this consideration and since various cycles can have differ- 230
179 a constant value. Similarly, the radius of the companion star, ent durations, we can conclude that the magnetic activity leads 231
180 Rc , is not fixed; it varies as the shape of the star changes due to systematic, but not strictly periodic, changes in the active star, 232
181 to quadrupole variations. Therefore, the definition of Rc refers subsequently causing the observed orbital period variations. 233
182 to the companion star’s radius at rest. These fluctuations in the
183 size dimension can be quantified, and their relative magnitude
184 is estimated to be about 5 − 7% with respect to Rc (Applegate 2.2. Reference frames 234
185 1992a)3 .
Spider systems are fairly tight, implying highly-circular orbits. 235
186 This section opens by briefly recalling the Applegate mech-
Furthermore, the tidal friction predominately acts, entailing a 236
187 anism in Sec. 2.1, representing the core of our work. In the con-
synchronization of spin and orbit of the binary system, i.e., ro- 237
188 struction of the model, we will make use of two reference frames
tational and orbital angular momentum vectors are aligned and 238
189 (RFs) in Sec. 2.2, which allow us to conveniently deal with the
also have the same module. The dynamics occurs in the plane 239
190 ensuing dynamics. The equations of motion are presented in Sec.
P orthogonal to the direction of the orbital angular momentum. 240
191 2.3. We conclude by deriving the formulae of some fundamental
We neglect any additional perturbing effects that could poten- 241
192 physical observables involved in this scenario in Sec. 2.4.
tially lead to three-dimensional motions outside the P plane. 242
The dynamics can be described in two RFs, having both the 243
193 2.1. Applegate mechanism and dynamo action origin in the binary system’s center of mass (CM; see Fig. 2): 244
194 Applegate (1992a) proposed a mechanism to explain the orbital
195 period modulations in eclipsing binary systems as a consequence – orthonormal corotating RF, RCO = {x, y, z}: the z-axis is or- 245
196 of the magnetic activity inside the secondary star, triggered by thogonal to P, where the x- and y-axes are placed. The x-axis 246
197 the pulsar wind. For spider binary systems (i.e., black widows is always directed along the line connecting the two bodies 247
198 and redbacks) the lighter body plays the role of the active star. and pointing towards the companion star. This RF co-rotates 248
199 The main idea underlying this approach is based on the magnetic with the binary system, so the motion in it results to be al- 249
200 activity cycle, which represents the engine that causes a redistri- ways static. The only variations occur along the x-axis, re- 250
201 bution of the angular momentum inside the star, thus modifying ducing thus the whole dynamics to just one dimension; 251
202 its oblateness (see e.g., Warner 1988; Lanza & Rodonò 2002; – orthonormal static RF, RS = {xS , yS , zS }: zS ≡ z, where the 252
203 Donati et al. 2003; Lanza 2006; Bours et al. 2016). This induces xS -axis is directed towards the position of a static and non- 253
204 a variation in the radial component of the gravitational accelera- rotating observer at infinity O∞ , namely from CM to O′∞ (be- 254
205 tion via the gravitational quadrupole-orbit coupling, thus entail- ing the projection of O∞ on P). This RF is fixed in space and 255
206 ing the orbital period modulations, which we eventually detect. all quantities measured in it are labelled by a subscript S. 256
207 This magnetic activity seems to be powered by a dynamo ac-
208 tion, i.e., a process of magnetic field generation through the in-
209 ductive response of a highly-conductive fluid. Indeed, there is a These RFs are related by the map T : RCO → RS . It is deter- 257
210 conversion of mechanical energy into a magnetic one by stretch- mined by employing polar coordinates, the radius rS (t) (coinci- 258
211 ing and twisting the magnetic field lines (Parker 1955). All of dent with the coordinate x) and polar angle θS (t), where t is the 259
212 the aforementioned effects are driven by the subsurface magnetic time. More explicitly, this transformation reads as 260
213 field, located within the star’s convective zones.
214 The dynamo action is the alternation of two phenomena in- (
xS (t) = rS (t) cos θS (t),
215 side the star: (1) the sheared differential rotation at different lat- ⇔ rS (t) = x. (1)
216 itudes contributes to the trasformation of an initially poloidal yS (t) = rS (t) sin θS (t),
217 magnetic field into an enhanced toroidal one through the Alfvén
218 theorem (also known as Ω-effect; see e.g., Parker 1955, 1979;
2.3. Equations of motion 261
219 Browning et al. 2006); (2) the combined action of cyclonic con-
220 vection, buoyancy, and Coriolis forces turns the toroidal mag- Our model is governed by the gravitational quadrupole-orbit dy- 262
221 netic field back to the poloidal one, completing thus the cycle namics (see Sec. 2.3.1) and the time-variation of the quadrupole 263
222 (also known as α-effect; see e.g., Parker 1955, 1979; Choudhuri moment (see Sec. 2.3.2). This differential problem can be solved 264
223 et al. 1995; Charbonneau & Dikpati 2000; Browning et al. 2006). if it is accompanied by the appropriate initial conditions (see Sec. 265
224 The dynamo model is usually gauged on the solar magnetic 2.3.3). We stress again that, even though our model is fully based 266
225 activity, as it shares profound similarities with the problem under on the Applegate mechanism, the related dynamical equations of 267
226 investigation. We know from Sun’s observations that the period motion have never been written in the literature. 268
2
In this investigation, three aspects must be considered: (1) the mass
loss from the companion amounting to 10−10 M⊙ /yr (Pan et al. 2023); 2.3.1. Gravitational quadrupole-orbit coupling dynamics 269
(2) if the Roche lobe is less than the radius of the companion star, there
is mass transfer; (3) the pulsar wind entails a mass loss of 10−12 M⊙ /yr The two bodies are influenced by their mutual gravitational at- 270
(Guerra et al. 2024). We conclude that we can consider constant mass. traction and the companion star’s gravitational quadrupole-orbit 271
3
The companion star’s variation are due to the combination of two
coupling. The problem is first framed in RCO , where the pulsar 272
effects: quadrupole changes amounting to 1 − 5% (Applegate 1992a;
Harvey et al. 1995) and magnetic field dynamo activity causing fluctu- has coordinates (−x p , 0, 0), while the companion star (xc , 0, 0). 273
ations of 2 − 6% (Rappaport et al. 1983; MacDonald & Mullan 2009). The force acting on the pulsar is F p , being the sum of 274
The combined effects can thus range in the interval 5 − 7%. the gravitational force and the quadrupole moment contribution 275
Article number, page 3 of 13
 A&A proofs: manuscript no. ms

Fig. 1. Illustration of a typical spider binary system. A MSP and a companion star are in a tidally locked configuration, moving in synchronous
rotations on quasi-circular and tight orbits. They are separated by a distance rs (t), in a plane orthogonal to the orbital angular momentum vector L,
which is conserved during the binary system’s motion. The irradiation-driven wind from the MSP heats through a tidal dissipation the companion
star, which starts to evaporate, loosing thus mass. This is the eclipsing material, which obstacles the radio signal detected by a telescope located far
from the binary system. In the dashed red box we sketch the companion star interior. This zone is characterized by a convective envelope, where
the fluid in it is highly conductive. Furthermore, the pulsar wind triggers the differential rotation of the companion star, which induces a magnetic
activity through a dynamo action. The subsurface magnetic field is responsible to break the hydrosthatic equilibrium inside the star, inducing
quadrupole moment changes. Notice that the external magnetic field is expected to be locally poloidal, while the subsurface field is toroidal.

the derivative with respect to the time t), together with the total 281
MTOT = m p + mc and reduced µ = m p mc /MTOT masses. After 282
appropriately manipulating Eqs. (2) and (3), we obtain 283
GMTOT 9 GQ(t)
ẍ = − − . (4)
x2 2 x4
Changing RF through the transformation T , we rewrite the 284
above dynamics in polar coordinates in RS . Considering the an- 285
gular component, we obtain the following equation of motion 286
287
1 d h i
µ rS (t)2 θ̇S (t) = 0, (5)
rS (t) dt

Fig. 2. We display the two RFs: RCO (black) and RS (red).

which immediately entails µrS (t)2 θ˙S (t) = L, where L is the mod- 288
ule of the conserved angular momentum L of the system along 289
the zS -axis, which can be calculated using the initial conditions. 290
276 Q(t)4 from the companion star, namely (Applegate 1992a): The problem is defined in the timeframe [t0 , t1 ]. However, 291
! since the setting is invariant under time shifts, we can con- 292
Gm p mc 3 GQ(t)m p sider, without loss of generality, the following normalized in- 293
m p ẍ p = F p ≡ −∂ x + . (2)
x 2 x3 terval [0, 1]5 . The map connecting [t0 , t1 ] with [0, 1] is given by 294
295
277 Instead, the force acting on the companion star is Fc , simply
278 given by the gravitational force from the pulsar: t − t0
t ∈ [t0 , t1 ] → ∈ [0, 1]. (6)
! t1 − t0
Gm p mc
mc ẍc = Fc ≡ ∂ x . (3) We stress that T 0 is the reference time, usually coincident with 296
x
the transition to the orbit periastron6 . This is the moment where 297
279 We define the relative coordinate system x = xc − x p and we extract the parameters and would correspond to the initial 298
280 the relative acceleration ẍ = ẍc − ẍ p (the over dot stands for time T 0∗ = (T 0 − t0 )/(t1 − t0 ). However, the observations begin at 299
4 5
The quadrupole moment is a tensor written in terms of the companion We prefer to work in normalized units for developing the calcula-
star’s inertial tensor. In our hypotheses, we have Q = Q xx (see discus- tions, since this is advantageous during the fitting procedure.
6
sion under Eq. (3) in Applegate 1992a, for details). For circular orbits, T 0 is the ascending node’s epoch passage.

Article number, page 4 of 13

 De Falco et al. (2024): Reconstruction of spider system’s observables

300 t0 < T 0 . Therefore, it is reasonable to set the initial conditions at where Ω is the angular velocity of the outer layers, which can be 341
301 T 0 and extend our solutions back to the earlier time, t0 . reasonably calculated through the Keplerian angular velocity 342
302 The angle θS (t) can be calculated through the formula θS (t) = s
303 ω(t)(t − T 0∗ ), where ω(t) is the angular frequency. Since the sys- Gmc
Ω= . (12)
304 tem is very tight, the bodies move on quasi-circular orbits with Rc 3
305 relative angular velocity (see Eq. (5) in Applegate 1992a)
The variation of J is caused by the spin torque, because the ir- 343
radiation driven wind from the pulsar generates a ram pressure 344
s  
GMTOT  9 Q(t) 
vθ (t) = 1 +
  . (7) contributing to the spin-up of the companion star. Employing Eq. 345
rS (t) 2 mc rS2 (t) (27) in Applegate (1992a), the time-variation of J is8 346
306 Therefore, the angular frequency ω(t) can be estimated applying Gm2c rS (t) 2 d∆P
!
307 the formula for (quasi-)circular motion (Applegate 1992a) J˙ = − , (13)
6πRc Rc dt
vθ (t) where ∆P(t) is also known as orbital period modulations. There- 347
ω(t) = . (8) fore, the resulting differential equation ruling the dynamics of 348
rS (t)
Q(t) is obtained by substituting Eq. (13) into Eq. (11), namely 349
308 It is important to note that θ̇S (t) , ω(t), since mc ΩrS2 (t) d∆P
Q̇(t) = − . (14)
θ̇S (t) = ω(t) + (t − T 0∗ )ω̇(t). (9) 18π dt

309 However, at the beginning (i.e., for t = T 0∗ ) we have θ̇S (T 0∗ ) = 2.3.3. Initial conditions 350
310 ω(T 0∗ ).
311 Instead, for the radial component, we obtain Our model is governed by a system of ordinary differential equa- 351
tions (10) and (14). It must be complemented by the initial con- 352
L2 Gm p mc 9 GQ(t)m p ditions at the time t = T 0∗ to find a unique solution, which are 353
µr̈S (t) = − 2 − , (10) 354
µrS (t)
3
rS (t) 2 rS4 (t)
∗
2 Ṗ(T 0 )
312 where the change of RF can be seen by the appearance of the ṙS (T 0∗ ) = a, rS (T 0∗ ) = a, θS (T 0∗ ) = 0, Q(T 0∗ ) = Q0 , (15)
3 P(T 0∗ )
313 centrifugal force (first term on the right hand side). This dynam-
314 ical system is composed by a second-order ordinary differential where a is the initial separation between pulsar and companion 355
315 equation (10), which should be complemented by the dynamics star and Q0 = 0.1mc R2c /3 the initial quadrupole (cf. Eq. (25) in 356
316 of Q(t)7 , which is disclosed in the next section. Applegate 1992a). The radial velocity is determined employing 357
Kepler’s third law in its differential form, as the two bodies’ or- 358
bits are Keplerian at every moment in time. The conditions (15) 359
317 2.3.2. Quadrupole dynamics imply (cf. Eqs. (7) and (8)) 360
318 The Applegate mechanism foresees that there are two kinds of s
vθ (T 0∗ )
!
GMTOT 9 Q0
319 deformations due to the magnetic activity, which can be classi- vθ (T 0 ) =
∗
1+ 2
, ω(T 0∗ ) = . (16)
320 fied into: (1) distortions, which modify the hydrostatic equilib- a 2 mc a a
321 rium in the deformed configuration; (2) transitions, which cause
322 changes from one fluid hydrostatic configuration to another.
2.4. Physical observables 361
323 Applegate (1992a) explains that distortions are not astro-
324 physically relevant, because the weak magnetic fields (∼ 105 − This section provides some physical observables related to spi- 362
325 106 G) cannot supply enough energy for the star deformations der systems. To achieve this objective, we must first solve Eqs. 363
326 to reproduce the orbital period modulation timescales (see after (10) – (14), as we need rS (t) and Q(t), which are already funda- 364
327 Eq. (23), for a more detailed discussion). Therefore, we consider mental physical quantities. We focus on the following quantities: 365
328 modifications due to the transitions, where the dynamics of a ro- orbits pertaining to the two bodies (see Sec. 2.4.1), orbital period 366
329 tating star strongly depends on the matter distribution within it (see Sec. 2.4.2), magnetic field intensity (see Sec. 2.4.3), lumi- 367
330 and its angular momentum J, which influences the quadrupole nosity variability (see Sec. 2.4.4). 368
331 moment variations in agreement with the observed timescales.
332 A fundamental role is played by its external layers, which
333 contribute to spin-up the star, making it more oblate (i.e., en- 2.4.1. Two body orbits 369
334 hancing its quadrupole moment; Applegate 1992a). Therefore, An important feature of a binary system is to understand the evo- 370
335 the magnetic activity permits to arise and develop a torque, lution of the orbit. Since the two bodies are tidally locked and 371
336 which acts on the spin of the companion star to extract angu- synchronized, we need to determine the separation rS (t) and the 372
337 lar momentum (Applegate 1992a; Applegate & Shaham 1994). quadrupole moment Q(t) to calculate ω(t) (cf. Eq. (8)), which in 373
338 From Eq. (26) in Applegate (1992a), the companion star’s turn provides the evolution of the polar angle θS (t). Therefore, 374
339 quadrupole moment changes according to the following formula: passing in cartesian coordinates, the orbit is obtained by plotting 375
340 376
1 ΩR3c ˙ (rS (t) cos θS (t), rS (t) sin θS (t)),
!
t ∈ [0, 1] (17)
Q̇ = J, (11)
3 Gmc where t is given by Eq. (6). 377
7
It is important to note that being Q directed along the x-axis, thanks 8
It is important to note that ∆P(t) = P(t)−P0 with P0 being a constant.
to the transformation (1), we have that Q(t) points along rS (t). Therefore, d∆P/dt = Ṗ(t). We use both notations in the paper.

Article number, page 5 of 13

 A&A proofs: manuscript no. ms

378 2.4.2. Orbital period where Pma is the magnetic period, being specific for each spider 420
system. It is computed by searching for the maxima (or min- 421
379 The orbital period P(t) can be estimated in two ways: (1) once we ima) in the ∆P(t) profile, whose average distances allows us to 422
380 solve the system, we determine the orbit and we then calculate it extract a mean value. We stress that Eq. (22), as derived by Ap- 423
381 ; (2) it can be obtained a-priori by fitting the observational data. plegate (1992a), provides just an estimate of the magnetic field 424
382 In this work, we use the latter approach, as we aim to reconstruct intensity. In the original formula, ∆P is interpreted as a (con- 425
383 the dynamics of the physical observables. stant) positive quantity. Since in our case ∆P(t) assumes both 426
384 In the a-priori approach we first calculate ∆P(t) = P(t) − P0 , positive and negative values, as well as zero, it is more appro- 427
385 where P0 is estimated via Kepler’s third law, namely priate to insert the absolute value to derive a realistic result. In 428
s addition, inspired by the ∆P(t) profile, it is more opportune to 429
4π2 a3 calculate (through Eq. (22)) the variations ∆B(t), rather than B(t). 430
P0 = . (18)
GMTOT We stress ∆B(t) = B(t) − B0 , where B(t) is due to the quadrupole 431
variations, whereas B0 stays for the unperturbed case. In order to 432
386 The formula used to fit ∆P(t) (generally expressed in seconds) is guarantee regular behaviours and attain negative values, we have 433
387 added the function sign of the ∆P(t) outside the square root. We 434
1 1 make the function continuous, but it is not differentiable in the 435
∆P(t) = − , (19)
g(t) f0 points crossing the zero line. This is the best we can achieve, as 436
we are ignorant about the magnetic activity occurring inside the 437
388 where f0 = 1/P0 and companion star. 438
1 1 We consider a rough estimate of the magnetic field B0 pro- 439
g(t) ≃ f0 + f1 (t − T 0 ) + f2 (t − T 0 )2 + ... + fn (t − T 0 )n posed by Applegate, which is obtained by substituting in Eq. 440
2 n!
n (22), rS (t) with its initial value a and ∆P(t) with the trend AP0 441
X 1
= fi (t − T 0 )i . (20) (cf. Eq. (21)). Therefore, performing these calculations, we have 442
i=0
i! 443
The fitting procedure is performed in the interval [t0 , t1 ], ex-
s
389 !2
Gm2c a AP0
390 pressed in MJD. Furthermore, the coefficients { fi }ni=0 represent B0 ∼ sgn(A) 10 . (23)
391 the higher-order frequency derivatives and n is the order of terms R4c Rc Pma
392 involved in the fit. The parameters { fi }ni=0 and T 0 are obtained as
393 a timing solution of a BTX model provided by the TEMPO or 2.4.4. Luminosity variability 444
394 TEMPO2 software. The value of n changes from a pulsar to an-
395 other, since it is the number of the reliably-measured time deriva- Another non-uniform periodic mechanism is related to the 445
396 tives of the orbital period, and depends on various aspects, such physics of the luminous variability inside the active star. This 446
397 as: the time interval covered by the radio observations, the typi- cycle is divided into two phases (Applegate & Patterson 1987; 447
398 cal uncertainty on the measured pulses’ time of arrival, the r.m.s Applegate 1992a; Applegate & Shaham 1994): 448
399 of the timing residuals, and more (see e.g., Ridolfi et al. 2016;
400 Freire et al. 2017). We also note that Ṗ(T 0∗ ) = − f1 / f02 , useful for (i) the angular momentum transfer leads to an enhancement of 449
401 computing ṙ(T 0∗ ) in Eq. (15). the kinetic energy, because it is spent to power the differ- 450
402 For what follows, it is useful to define the constant quantity ential rotation between the inner part and the outer layer of 451
the active star. This activity leads the star to spin-up at the 452
∆P(t) expenses to lower the luminosity; 453
A=⟨ ⟩[t ,t ] , (21)
P(t) 0 1 (ii) when the angular momentum transfer decreases, the active 454
star rotates as a solid body. This causes the star to spin-down, 455
403 where ⟨·⟩[t0 ,t1 ] is the time-average in the interval [t0 , t1 ].
with a consequent increment in its luminous intensity. 456

404 2.4.3. Magnetic field intensity The alternation of the two cycles gives rise to the observed lumi- 457
nous modulations. In addition, in the entire process there is also 458
405 In the Applegate mechanism, the magnetic energy is the main a weak dissipation, because the orbital period tends to weakly 459
406 source of support to provide the necessary torque for the ex- decrease with time. However, this dissipative effect will be not 460
407 change of angular momentum between the shells of the compan- taken into account in this model due to its negligible contribu- 461
408 ion star (and consequently changes in the quadrupole moment). tions. 462
409 In this scenario, the magnetic field does not decay in rapid times, The energy emitted by the companion star can be estimated 463
410 because the so far observations have not detected orbital period through (see Eq. (28) in Applegate 1992a, for details) 464
411 variations over short timescales. This phenomenon represents an
412 indirect probe for the internal magnetic field dynamics within ac- J˙2
∆E = Ωdr J˙ + ∆t, (24)
413 tive stars. Therefore, spider pulsars are natural laboratories and Is
414 privileged systems to investigate the magnetic dynamics in low-
415 mass stars, since this topic is still matter of discussion. where Ωdr = Ω(1 − η) with efficiency η = 0.66 (see Fig. 1 in 465
416 The variation of the subsurface (toroidal) magnetic field in- Yoshida 2019) and I s represents the moment of inertia of the 466
417 tensity with respect to that of the unperturbed (i.e., Q(t) = 0) outer layer considered as a shell, given by (Applegate 1992a) 467
418 configuration can be estimated through the following formula
2
419 (see Eq. (33) in Applegate 1992a, for further details) Is = M s R2c , (25)
s 3
Gm2c rS (t) 2 |∆P(t)|
!
where M s ≈ 0.1mc is the outer layer mass (Applegate 1992a). 468
∆B(t) ∼ sgn(∆P(t)) 10 4 , (22)
Rc Rc Pma Therefore, the luminosity modulation can be easily calculated as 469
Article number, page 6 of 13
 De Falco et al. (2024): Reconstruction of spider system’s observables

470 Equation (14) is the only differential relation left, which must 515
∆E be solved numerically. However, we can avoid this last inte- 516
∆L (t) = π , (26) gration by sampling rS (t) and ∆P(t) functions at several points. 517
Pma Then, we fit them with high-accurate polynomials with n + 1 518
471 where ∆L (t) represents the difference between the luminosity coefficients, using the same n of the TEMPO coefficients { fi }ni=0 , 519
472 due to the quadrupole variations and the constant luminosity of because this allows a drastic reduction of the approximation er- 520
473 the star without altering its shape. This formula permits to track rors. The fitting procedure occurs in the interval [0, 1], since this 521
474 the companion star’s luminosity with respect to the unperturbed gives more accurate results. Using the transformation (6) we can 522
475 configuration during the time evolution, and to monitor how it then map the ensuing polynomial in the interval [t0 , t1 ]. 523
476 changes in terms of the orbital period modulations. In this case, Eq. (11) can be analytically written as 524
477 Finally, we provide an estimate of the unperturbed luminos- m
478 ity, expressed by the following formula (using Eq. (26), where mc Ω X ti+1
Q(t) = − ai + Q0 , (31)
479 we substitute rS (t) with a and ∆P(t) with AP0 ) 18π i=0 i + 1
!2
1 Gm2c a AP0 where the coefficients ai are real numbers, which can be explic- 525
L0 = Ω . (27)
9 Rc Rc Pmod itly calculated when we have the polynomials of rS (t) and ∆P(t). 526
The integer m depends on the final polynomial order obtained by 527
multiplying the two aforementioned polynomials. 528
480 3. Methodology
481 The spider system dynamics is described by two coupled non- 3.2. Approximation accuracy 529
482 linear differential equations (10) – (14), where the analytical so-
483 lution is too difficult to be determined and therefore, numeri- To check the reliability of the result we have found, we need to 530
484 cal routines must be exploited. We note that if we follow the compare our approximation with the numerical solution of Eqs. 531
485 a-posteriori approach (see Sec. 2.4.2), we must have a precise (10) and (14). To this purpose, we use Mathematica 13.1 and 532
486 temporal trend for d∆P/dt, which is related to the subsurface Python 3 to confidently validate our calculations. 533
487 magnetic field of the active star. This is a very demanding task In Mathematica 13.1, we employ the function NDSolve 534
488 for several reasons (see Dobler et al. 2006; Browning 2008, and and exploiting the methods StiffnessSwitching and 535
489 references therein): (1) theoretical uncertainties about the mi- ExplicitRungeKutta, selecting a precision and accuracy of 536
490 crophysics inside low-mass stars; (2) high computational cost 10, and a maximum step size of 10−6 in the interval [0, 1]. 537
491 toperform magnetohydrodynamic (MHD) simulations; (3) exis- Then, we plot rS with several points (∼ 800), while Q(t) can be 538
492 tence of several models, based on different simplifications and displayed by employing much fewer points (∼ 100). 539
493 hypotheses. In Python 3, we use the integration routine dop853, being 540
494 In order to avoid the aforementioned issues, we reckon on the the Dormand-Prince algorithm implemented within the class of 541
495 a-priori approach (see Sec. 2.4.2), founded on having the func- explicit Runge-Kutta methods of eight order Dormand & Prince 542
496 tion ∆P(t) by fitting the observational data. However, even after (1980); Press et al. (2002). We select absolute 10−20 and relative 543
497 substituting the orbital period modulations (19) in the equations 10−10 tolerances within [0, 1] with a step-size of 10−7 . The two 544
498 of motion, the ensuing dynamics is still too cumbersome to be approaches are in agreement, since they give the same results. 545
499 analytically integrated. Therefore, if we want to avoid numerical The approximate radius differs from the numerical one, be- 546
500 integrations, some approximation schemes must be employed. cause the latter contains mild oscillations (since the orbit is 547
501 We present a methodology to derive an approximate analyt- quasi-circular) and an overall modulation on the time span [0, 1]; 548
502 ical solution of Eqs. (10) and (14) in Sec. 3.1, about which we whereas the former features only the modulation on [0, 1] (since 549
503 comment on the approximation accuracy with respect to the orig- the orbit is considered circular). Instead, the quadrupole moment 550
504 inal (unaffected) equations in Sec. 3.2. We conclude by specify- between the two approaches coincides with mean relative errors 551
505 ing the inputs and outputs of our model in Sec. 3.3. (MREs) < 10−5 %. Therefore, our solution is consistent. 552

506 3.1. Approximation strategy 3.3. Inputs and outputs of the model 553
507 Analysing better the problem, we should bare in mind that the Our approach relies on the analytical (high-accurate approxi- 554
508 motion is quasi-circular. Therefore, we can assume the validity mate) formulae of rS (t) and Q(t). This permits to fast compute 555
509 of Kepler’s third law in each point. Differentiating it, we obtain, the evolution of the associated physical observables reported in 556
510 at the first order in ∆P and ∆rS : Sec. 2.4. Our methodology is also flexible, because it can be used 557
in an opposite manner. Indeed, knowing the trend of some phys- 558
∆P 3 ∆rS
= . (28) ical variables (e.g., the luminosity), we could extract the orbital 559
P 2 rS period modulations and then determine all the other quantities. 560
511 Substituting the following relations: The input values of our model are: 561

P(t) = P0 + ∆P(t), rS (t) = a + ∆rS (t)

n o
(29) m p , mc , Rc , Pma , f0 , . . . , fn , T 0 , (32)
512 into Eq. (28) and neglecting second order terms, we have where m p can be set equal to the value of a standard NS (i.e., 562
2 a∆P(t) m p = 1.4M⊙ ) and if we know the spider class, we can assign an 563
∆rS (t) = , (30) average value of mc and Rc . The initial orbital separation a can 564
3 2P0
be calculated via the third Kepler’s law (cf. Eq. (18)), knowing 565
513 which completely determines the function rS (t) in terms of the P0 = 1/ f0 , m p , mc . Depending on the goal, we generally have 566
514 orbital period modulations ∆P(t), taken from the observations. n + 6 input parameters, which can eventually be lowered to n + 3. 567
Article number, page 7 of 13
 A&A proofs: manuscript no. ms

568 The output parameters of our model are: layers, altering its convection properties and potentially leading 621
n o to enhanced material loss. While mass-loss rates in black widows 622
rS (t), θS (t), Q(t), ∆B(t), ∆L (t) . (33) and redbacks depend on multiple factors, including the intensity 623
of irradiation and orbital evolution, models indicate that redback 624
companions, being more massive, may sustain a stronger mag- 625
569 4. Results netic field, which provides greater resistance to ablation-driven 626
mass loss (Conrad-Burton et al. 2023). 627
570 As an application of our model, we consider the spider systems:
The magnetic fields of redbacks exhibit larger fluctuations, 628
571 47 Tuc W (redback) and 47 Tuc O (black widow). The input
with shifts of up to 150 kG, compared to the more stable 20 629
572 parameters and the TEMPO coefficients of these two physical sys-
kG variations in black widows. However, the magnetic activity 630
573 tems are reported in Table A.1. In Fig. 3 we display the orbital
in redbacks is observed to persist for shorter timescales than in 631
574 period modulations of these two sources (see Ridolfi et al. 2016;
black widows (see Table A.1). This behaviour fits within the 632
575 Freire et al. 2017, for more details). The polynomial approxi-
irradiation-driven evolution scenario, where the pulsar’s rela- 633
576 mations described in Sec. 3.1 are reported in Appendix A. In
tivistic wind and high-energy radiation interact with the compan- 634
577 this section, we first compare the results from the two sources in
ion’s magnetosphere (Conrad-Burton et al. 2023). In redbacks, 635
578 Sec. 4.1 and then in Sec. 4.2 we analyse how we determine the
the stronger magnetic field can counteract the pulsar wind for 636
579 associated parameters, once we detect a spider source.
longer durations, delaying the complete stripping of the com- 637
panion. In contrast, the weaker fields of black widow compan- 638
580 4.1. Comparison between 47 Tuc W and 47 Tuc O ions result in a more effective removal of material, leading to 639
more rapid and extreme ablation (Podsiadlowski 1991; Chen 640
581 We analyse 47 Tuc W and 47 Tuc O’s dynamics via our model et al. 2013; Conrad-Burton et al. 2023). Interestingly, while abla- 641
582 through the related physical variables’ profiles (33), highlighting tion is commonly associated with more severe mass loss in black 642
583 common features (see Sec. 4.1.1) and diversities (see Sec. 4.1.2). widows, it is expected that redbacks’ companions can intercept 643
a larger fraction of the pulsar’s spin-down luminosity, possibly 644
584 4.1.1. Analogies as a result of a different geometric configuration (Chen et al. 645
2013; Conrad-Burton et al. 2023). This effect likely stems from 646
585 The radius follows the same trend of the orbital period modula- the more substantial convective envelope and deeper energy de- 647
586 tions (cf. Eq. (29) and (30)), but with very mild oscillations, see position in redback companions, which alters their thermal and 648
587 Fig. 4. This feature transmits also to the orbits, which, besides to magnetic structures. 649
588 be quasi-circular, admit a very narrow advancement with respect From a modeling perspective, a convincing explanation of 650
589 to the long time baseline. Instead, the quadrupole moment be- the connection between the directly observed surface and bulk 651
590 haves in the opposite way (see Fig. 5), because the orbit shrinks magnetic field properties of these systems is still missing. The 652
591 (enlarges) as the quadrupole increases (decreases). state of the art on the most relevant studies on this topic follows 653
592 The magnetic field variability, compared to other observ- two distinct approaches, mainly based on three-dimensional 654
593 ables, features a very oscillating trend scanned by the magnetic MHD simulations (Browning 2008; Yadav et al. 2015; MacDon- 655
594 activity period Pma , see Fig. 6. The variable magnetic field B(t) ald & Mullan 2017) and one-dimensional stellar evolution analy- 656
595 crosses the zero line at the points where ∆P(t) nullifies, corre- ses (Feiden & Chaboyer 2014). We can see that in the aforemen- 657
596 sponding to moments when the magnetic field matches the initial tioned works there is an ongoing disagreement about the order 658
597 star’s intensity. Since the provided formula is a rough estimate of of magnitude of the field intensity in fully convective stars: some 659
598 the subsurface magnetic field strength (with an offset value B0 ), authors require ∼MG fields, while others insist on possible upper 660
599 we anticipate that MHD simulations could offer a similar but limits of ∼10 kG. Given this lack of consensus, our 20-100 kG 661
600 more detailed and accurate representation. subsurface field variation (see Fig. 6) at a depth ∼ 0.1Rc (Apple- 662
601 Finally, the luminosity is ruled by the quadrupole moment gate 1992a) seems to be reasonable. Currently, there are no simu- 663
602 (cf. Eq. (26)), where the zero points coincide with the luminosity lations specifically focused on this aspect and up to now there are 664
603 of the star initially observed, see Fig. 7. For the calculation of only some indications that the subsurface magnetic field could be 665
604 the magnetic field and luminosity we use their original formulae larger than the surface one by a factor of ∼ 2 (see Fig. 4 in Yadav 666
605 (cf. Eqs. (22) and (26)) without any approximation for the radius et al. 2015, for details). 667
606 rS (t). We express the luminosity variability in solar luminosity One of the novel aspects of our model with respect to the 668
607 units, corresponding to L⊙ = 3.83 × 1033 erg/s. We note that literature relies on the temporal dynamics of the ∆B(t) plot. The 669
608 the results we have found are in agreement with the estimates long-term evolution of the surface fields is sometimes observed, 670
609 reported in Applegate (1992a), namely ⟨∆L (t)⟩[t0 ,t1 ] /L0 ∼ 0.3. while the suggested drastic variation of the field intensity by a 671
factor of ∼10 with a characteristic oscillatory shape seems to 672
610 4.1.2. Differences be not seen in any other observations or theoretical modeling of 673
fully-convective stars outside spider systems. It is true that our 674
611 The described similarities are mainly due to the underlying equal profile is based on a naive formula of the magnetic field, but 675
612 mathematical structure, while the discrepancies arise from the the conventional dynamo models interplay with the quadrupole 676
613 different physical nature of the two spider classes. variations in these stars, possibly generating new behaviours. 677
614 Observationally, the orbits associated with 47 Tuc W appear Regarding luminosity, redback companions generally exhibit 678
615 to be tighter than those of 47 Tuc O (see Fig. 4). This trend higher optical luminosities than black widows, as shown in Fig. 679
616 is consistent with theoretical models of spider pulsars, which 7. This trend arises from a combination of factors (Roberts 680
617 suggest that the pulsar wind significantly impacts the compan- 2012; Gentile et al. 2014; Roberts et al. 2018; Strader et al. 681
618 ion’s structure and mass-loss process (Chen et al. 2013; Wang 2019; Sullivan & Romani 2024): (i) the intrinsic luminosity of 682
619 et al. 2021; Conrad-Burton et al. 2023). The high-energy photons the star, which is modulated by pulsar heating; (ii) the pulsar 683
620 emitted by the pulsar deposit energy in the companion’s outer wind interaction with the companion and the surrounding ma- 684
Article number, page 8 of 13
 De Falco et al. (2024): Reconstruction of spider system’s observables

5 10 47 TUC O
47 TUC W
4 8
3 6

ΔP (ms)
ΔP (ms)

2
4
1
0 2
-1 0
-2
-2
51 500 52 000 52 500 53 000 53 500 54 000 54 500 55 000 48 000 50 000 52 000 54 000 56 000
t (MJD) t (MJD)

Fig. 3. Orbital period modulations of 47 Tuc W and 47 Tuc O, obtained from the fitting of the observational data (cf. Table A.1). The vertical
dashed line marks the position of T 0 .

5. × 10-6
2. × 10-6 4. × 10-6

rS (t)-a (1010 cm)

rS (t)-a (1010 cm)

3. × 10-6
1. × 10-6
2. × 10-6
0 1. × 10-6
0
-1. × 10-6
-1. × 10-6
51 500 52 000 52 500 53 000 53 500 54 000 54 500 55 000 48 000 50 000 52 000 54 000 56 000
t (MJD) t (MJD)

Fig. 4. Discrepancy of the actual separation among the bodies with respect to the initial datum. The vertical dashed line marks T 0 . The order of
the sources is placed as in Fig. 3.

1.0004
1.000
1.0002
1.0000
Q/Q0
Q/Q0

0.9998 0.995
0.9996
0.9994 0.990
0.9992
51 500 52 000 52 500 53 000 53 500 54 000 54 500 55 000 48 000 50 000 52 000 54 000 56 000
t (MJD) t (MJD)

Fig. 5. Evolution of the quadrupole moment Q(t)/Q0 , where the horizontal dashed line is set at 1 and the vertical dashed line marks T 0 . The order
of the sources is placed as in Fig. 3.

40 100
ΔB(t) (kG)
ΔB(t) (kG)

20 50

0
0
-20
-50
-40
51 500 52 000 52 500 53 000 53 500 54 000 54 500 55 000 48 000 50 000 52 000 54 000 56 000
t (MJD) t (MJD)

Fig. 6. Variation of magnetic field intensity ∆B(t). The horizontal dashed line is B0 (cf. Eq. (23)), whereas the vertical dashed line marks T 0 . The
order of the sources is placed as in Fig. 3.

10 2

5 0

0 -2
Δ𝓛 (L⦿)
Δ𝓛 (L⦿)

-5 -4

-10 -6
-8
-15
-10
-20
51 500 52 000 52 500 53 000 53 500 54 000 54 500 55 000 48 000 50 000 52 000 54 000 56 000
t (MJD) t (MJD)

Fig. 7. Variation of luminosity ∆L (t). The horizontal dashed line is L0 (cf. Eq. (27)), whereas the vertical dashed line marks T 0 . The order of the
sources is placed as in Fig. 3.

685 terial, particularly in X-ray bands; (iii) the gamma-ray emission magnetic field; and (v) non-thermal radiation from the intra- 689
686 from pulsar magnetospheric processes such as curvature radia- binary shock formed between the pulsar wind and the ablated 690
687 tion and inverse Compton scattering; (iv) occasional mass trans- material. 691
688 fer episodes, where infalling matter is energized by the pulsar’s

Article number, page 9 of 13

 A&A proofs: manuscript no. ms

692 In our case, redbacks reach total luminosities of L0 ∼ 1035 This paper constitutes the first attempt to make the Applegate 752
693 erg/s, while black widows are typically a few orders of mag- mechanism dynamical. To the best of our knowledge, we provide 753
694 nitude dimmer (L0 ∼ 1031 erg/s). The larger companion star for the first time the spider observables’ evolution. Combining 754
695 and stronger irradiation in redbacks contribute to a higher op- information derived by pulsar timing observations and assuming 755
696 tical luminosity. However, in the X-ray and gamma-ray bands, the Applegate mechanism as true, we are able to track the evo- 756
697 black widows can still exhibit comparable or even greater lu- lution in time (within a determined timeframe) of some physical 757
698 minosities due to the more efficient formation of intra-binary quantities (cf. Eq. (33)). 758
699 shocks (Conrad-Burton et al. 2023). These findings suggest that Voisin et al. (2020a,b) considerably improved the treatment 759
700 spider systems provide a unique laboratory for studying the ef- of the quadrupole deformations while also adding relativis- 760
701 fects of extreme irradiation on stellar magnetism, convection, tic effects, but they only provided the dynamical evolution of 761
702 and mass-loss processes. Future observational constraints and the quadrupole moment. Another important difference between 762
703 detailed simulations will be crucial in further refining our un- Voisin et al. (2020a,b) and our approach relies on the final goals. 763
704 derstanding of the evolution of these exotic binary systems. Voisin et al. (2020a,b) proposed a detailed model for describing 764
the motion of spider binary systems to accurately estimate the 765
observed ∆P variations with the ultimate objective of improving 766
705 4.2. Parameters’ determination
the timing solution pertaining to these gravitational sources. 767
706 In the current study, it is of essential importance to understand On the other side, our work employs a reverse approach: 768
707 how to determine the parameters associated with a spider system rather than finding a physical justification for the ∆P variations, 769
708 using our model, which is fully based on the Applegate mech- we take them from the long-term observations and, by relying 770
709 anism. This process is critical for extracting detailed informa- on the validity of the Applegate mechanism, we reconstruct the 771
710 tion about the gravitational source under investigation. Broadly dynamics of the related physical variables. Therefore, we follow 772
711 speaking, two main approaches can be identified. an observational and deductive analysis instead of a theoretical 773
712 The first route relies on the fact that when a new spider and inductive one. Even though our model is very simple, we 774
713 source is discovered, we can generally fit the orbital period mod- emphasize that it could also be applied to more refined frame- 775
714 ulations, thus obtaining the parameters T 0 , the TEMPO coeffi- works. 776
715 cients { fi }nI=0 , and Pma . Now, assuming that the NS has a canon- To achieve our goal, we have first derived the equations of 777
716 ical mass of m p = 1.4M⊙ and that the system is observed nearly motion (10) – (14) pertaining to the dynamics of spider binaries, 778
717 edge-on, the mass of the companion star, mc , can then be esti- based on the Applegate works. Then, we have considered the 779
718 mated (see e.g., Ridolfi et al. 2016; Freire et al. 2017, for more function ∆P(t), reconstructed by fitting the observational data on 780
719 details). This step is crucial to determine the spider system na- the orbital period modulations. However, the resolution of this 781
720 ture, namely whether it is a redback or a black widow. Sub- problem can be accomplished most likely only numerically and 782
721 sequently, we can estimate the companion star’s radius, Rc , by this can be excessively time consuming. To this end, we have de- 783
722 adopting the same argument detailed in Appendix A. Using Ke- veloped a mathematical procedure (see Sec. 3), based on approx- 784
723 pler’s third law, the initial orbital separation, a, can then be cal- imating the quasi-circular orbit with a circular one, even though 785
724 culated. By applying these criteria, we obtain a comprehensive it presents mild oscillations on small timescales. This strategy 786
725 measurement of all the independent parameters. has allowed us to obtain rS (t) without solving Eq. (10). Substitut- 787
726 An alternative strategy shares with the above procedure ing this function and ∆P(t) into Eq. (14) we have that the ensuing 788
727 the determination of the parameters T 0 , { fi }ni=0 , and Pma from differential equation is still difficult to solve analytically. 789
728 the observations, while leaving the remaining four parameters Therefore, to avoid any kind of numerical integration, we 790
729 m p , mc , Rc , a to be inferred. They could be potentially deter- have approximated rS (t) and ∆P(t) with high-accurate polyno- 791
730 mined if we have experimental data pertaining to a physical ob- mials. This has allowed us to have an analytical expression of 792
731 servable, such as for example the luminosity profile (26), which Q(t), whose functional formula has been reported in Eq. (31). 793
732 encapsulates a combination of all these unknowns. By fitting This result is in good agreement with the corresponding func- 794
733 such data and extracting the best-fit parameters, with the annex tion computed by numerically integrating together Eqs. (10) – 795
734 constraints on the variation ranges and the Kepler’s third law va- (14) without making any simplifying hypothesis. Through these 796
735 lidity, a consistent set of values can be extracted. Depending on formulae, it has been possible to easily obtain the evolution of 797
736 the available data, one or the other strategy could be used. Obvi- the related physical observables, which are (see Sec. 2.4): orbits, 798
737 ously, the available data varies from system to system. quadrupole moment, magnetic field, and luminosity. 799
In Fig. 3, we have displayed the orbital period modulations 800
738 5. Conclusions of the redback 47 Tuc W and black widow 47 Tuc O, whereas in 801
Figs. 4, 5, 6, and 7 we have shown the evolution of the related 802
739 This article deals with spider binary systems, formed by a pul- physical observables via our model. We have discussed analo- 803
740 sar and a low-mass companion star, classified in redbacks (mc ∼ gies and differences among the two sources (see Sec. 4.1), being 804
741 0.1 − 0.4M⊙ ) and black widows (mc ≪ 0.1M⊙ ). One of their dis- representatives of the redbacks and black widows. We have con- 805
742 tinctive features on which we concentrate is the long-term un- textualized these results in a more general physical picture. 806
743 predictable variations in the orbital period and its first deriva- The advantages of our approach are: (1) making use of sim- 807
744 tive (Roberts et al. 2014). Among the different contributions, ple formulae; (2) having the evolution of the above described 808
745 which can account for orbital period modulations, it has been physical variables, which allow to be better modeled through fu- 809
746 clearly shown that the most reasonable explanation is due to the ture upgraded descriptions; (3) having insight into the subsurface 810
747 Applegate mechanism (Applegate & Patterson 1987; Applegate magnetic activity inside the companion star, which is still not 811
748 1992a,b). This description accounts for orbital timing variations a clear topic; (4) employing our analytical formulae to tightly 812
749 via the companion star’s quadrupole moment changes induced constraint the model parameters by using not only the orbital pe- 813
750 by the magnetic dynamo action, which in turn is communicated riod modulations, but also the profile of other observables (e.g., 814
751 to the orbital motion through quadrupole-gravity coupling. Zhao & Heinke 2023, use the luminosity profile as further in- 815
Article number, page 10 of 13
 De Falco et al. (2024): Reconstruction of spider system’s observables

816 formaion); (5) the vast application of our developments to more

817 updated or different descriptions of spider binary systems.
818 However, our treatment possesses also some evident limits:
819 (1) the model is very simple and must be improved under dif-
820 ferent naive aspects; (2) the magnetic field necessitates to be
821 modeled through more realistic formulae, based on MHD simu-
822 lations (Dobler et al. 2006; Browning 2008), to clarify the link
823 between surface and subsurface magnetic fields, still not fully
824 treated (Morin 2012); (3) it is not adequate to reproduce dissipa-
825 tive phenomena (Lanza 2006), as well as new effects as the gen-
826 eral relativistic corrections and the three-dimensional motion of
827 the two-bodies (Voisin et al. 2020a,b); (4) the model prediction
828 power is limited only within the observational period [t0 , t1 ].
829 The future perspectives can branch out into several routes.
830 First, the methodology and results of this article could be ap-
831 plied to catalog all the available spider systems and to perform
832 more accurate analyses, in order to extract relevant information
833 on stellar evolution, and to classifyi the pulsar population in a
834 more methodical fashion. Another possible development is to
835 derive a more handy equation from actual MHD simulations to
836 better model the magnetic activity inside the companion star.

837 Acknowledgments
838 The authors greatly thank Oleg Kochukhov for valuable com-
839 ments on our results. V.D.F. is grateful to Gruppo Nazionale di
840 Fisica Matematica of Istituto Nazionale di Alta Matematica (IN-
841 DAM) for support. V.D.F. acknowledges the support of INFN
842 sez. di Napoli, iniziativa specifica TEONGRAV. V.D.F. is grate-
843 ful to both the SRT – Sardinia Radio Telescope and the Max
844 Planck Institute für Radioastronomie in Bonn for the hospitality.
845 [Link]. is grateful to Scuola Superiore Meridionale for hospital-
846 ity. A.R. is supported by the Italian National Institute for Astro-
847 physics (INAF) through an ‘IAF - Astrophysics Fellowship in
848 Italy’ fellowship (Codice Unico di Progetto: C59J21034720001;
849 Project ‘MINERS’). AR also acknowledges continuing valuable
850 support from the Max-Planck Society.

Article number, page 11 of 13

 A&A proofs: manuscript no. ms

851 Appendix A: Polynomial approximations A.2. Black widow: 47 Tuc O 891

852 We provide the polynomial approximations of ∆P(t), rS (t), Q(t) We approximate ∆P(t) with the polynomial (in ms unit) 892
853 with the related MRE for the spider systems 47 Tuc W (see Eq.
854 (A.1)) and 47 Tuc O (see Eq. (A.2)). The input parameters of P2 (t) = 2.98 × 106 t12 − 1.78 × 107 t11 + 4.66 × 107 t10
855 these two physical systems are reported in Table A.1. − 7.07 × 107 t9 + 6.88 × 107 t8 − 4.48 × 107 t7
856 We assume that the pulsar mass is the standard value m p = + 1.98 × 107 t6 − 5.89 × 106 t5 + 1.15 × 106 t4
857 1.4M⊙ . The companion star’s mass, mc is taken from works cited
858 in the caption of Table A.1, assuming that the binary system is − 1.40 × 105 t3 + 9.72 × 103 t2 − 3.42 × 102 t + 4.59, (A.4)
859 seen by the observer almost edge on, namely sin i = 1 with i
860 inclination of the observer with respect to the z-axis. Instead, the where t ∈ [0, 1], whose related MRE is ∼ ×10−3 %, being very 893
861 initial binary separation a is calculated via the Kepler’s third law accurate, as P2 (t) is of twelfth order (see Sec. 3.1, for details). 894
862 (cf. Eq. (18)), employing f0 = 1/P0 , m p , mc . The approximation (A.4) is exploited for computing the ra- 895
dius rS (t) with a MRE of ∼ 10−14 %. The estimation of the 896
863 Regarding the companion star’s radius (at rest), Rc , there is quadrupole moment (11) is performed by approximating the ra- 897
864 no guidance on its calculation in the referenced papers on the dius rS (t) via the following polynomial (in cm unit and t ∈ [0, 1]) 898
865 two sources. To estimate it, we follow this strategy. In black wid-
866 ows, the companion can reach extremely low masses, as seen R2 (t) = 1.48 × 1010 t12 − 8.79 × 1010 t11 + 2.30 × 1011 t10
867 in 47 Tuc O, suggesting it may be a brown dwarf, whose ra-
868 dius typically falls within the range 0.064 − 0.113R⊙ (Sorahana − 3.50 × 1011 t9 + 3.41 × 1011 t8 − 2.22 × 1011 t7
869 et al. 2013). For our purposes, we adopt an average radius of + 9.81 × 1010 t6 − 2.92 × 1010 t5 + 5.69 × 109 t4 − 6.91 × 108 t3
870 Rc = 0.08R⊙ . In contrast, for redbacks, the companion is a main-
871 sequence star, allowing us to estimate its radius using the follow- + 4.81 × 107 t2 − 1.69 × 106 t + 7.74 × 1010 . (A.5)
872 ing formula (see Table A.1 and Demory et al. 2009): The associated MRE is of ∼ 10−12 %. Therefore, the MRE on 899
Q(t) is ∼ 10−4 %, being in perfect agreement with the original 900
mc
!0.8 formula. Finally, the quadrupole moment is approximated with 901
Rc = R⊙ = 0.20R⊙ . (A.1) respect to the original solution with a MRE of 0.03%. 902
M⊙

References 903
873 The radii we have selected are in agreement with the Roche lobe
874 size RL of the two sources (Frank et al. 2002), as for 47 TUC W Alpar, M. A., Cheng, A. F., Ruderman, M. A., & Shaham, J. 1982, Nature, 300, 904
875 we have Rc /RL = 0.78, while for 47 TUC O it is Rc /RL = 0.63. 728
Applegate, J. H. 1992a, ApJ, 385, 621
905
906
Applegate, J. H. 1992b, in Astronomical Society of the Pacific Conference Se- 907
ries, Vol. 26, Cool Stars, Stellar Systems, and the Sun, ed. M. S. Giampapa & 908
876 A.1. Redback: 47 Tuc W J. A. Bookbinder, 343 909
Applegate, J. H. & Patterson, J. 1987, ApJ, 322, L99 910
877 We approximate ∆P(t) with the polynomial Applegate, J. H. & Shaham, J. 1994, ApJ, 436, 312 911
Balbus, S. A. & Brecher, K. 1976, ApJ, 203, 202 912
Baliunas, S. L. & Vaughan, A. H. 1985, Ann. Rev. Astron. Astrophys., 23, 379 913
Belczynski, K., Lorimer, D. R., Ridley, J. P., & Curran, S. J. 2010, MNRAS, 914
P1 (t) = −1.24 × 105 t9 + 5.67 × 105 t8 − 1.10 × 106 t7 407, 1245 915
+ 1.16 × 106 t6 − 7.31 × 105 t5 + 2.80 × 105 t4 Bhattacharya, D. & van den Heuvel, E. P. J. 1991, Phys. Rep., 203, 1
Bisnovatyi-Kogan, G. S. & Komberg, B. V. 1974, Soviet Ast., 18, 217
916
917
− 6.30 × 104 t3 + 7.62 × 103 t2 − 3.97 × 102 t + 4.74, (A.2) Blandford, R. & Teukolsky, S. A. 1976, ApJ, 205, 580 918
Bochenek, C., Ransom, S., & Demorest, P. 2015, ApJ, 813, L4 919
Bogdanov, S., Grindlay, J. E., & van den Berg, M. 2005, ApJ, 630, 1029 920
Bogovalov, S. V., Khangulyan, D., Koldoba, A. V., Ustyugova, G. V., & Aharo- 921
878 where t ∈ [0, 1], which can be cast in [t0 , t1 ] via Eq. (6), and nian, F. A. 2012, MNRAS, 419, 3426 922
879 P1 (t) has the dimension of ms. The related MRE is ∼ 10−5 %, Bogovalov, S. V., Khangulyan, D. V., Koldoba, A. V., Ustyugova, G. V., & Aha- 923
880 which is extremely accurate, since we have used a ninth-order ronian, F. A. 2008, MNRAS, 387, 63 924
881 polynomial (see discussion of Sec. 3.1, for details). Bours, M. C. P., Marsh, T. R., Parsons, S. G., et al. 2016, MNRAS, 460, 3873 925
Browning, M. K. 2008, ApJ, 676, 1262 926
882 We use the approximation (A.2) for calculating the radius Browning, M. K., Miesch, M. S., Brun, A. S., & Toomre, J. 2006, ApJ, 648, 927
883 rS (t) (cf. Eqs. (29) and (30)), committing a MRE of ∼ 10−13 %, L157 928
884 which is still very accurate. For the quadrupole moment, we em- Chandrasekhar, S. 1961, Hydrodynamic and hydromagnetic stability 929
Charbonneau, P. & Dikpati, M. 2000, ApJ, 543, 1027 930
885 ploy Eq. (11) and this polynomial approximation for rS (t) Chen, H.-L., Chen, X., Tauris, T. M., & Han, Z. 2013, ApJ, 775, 27 931
Choudhuri, A. R., Schussler, M., & Dikpati, M. 1995, A&A, 303, L29 932
Conrad-Burton, J., Shabi, A., & Ginzburg, S. 2023, Monthly Notices of the Royal 933
R1 (t) = −6.31 × 108 t9 + 2.89 × 109 t8 − 5.58 × 109 t7 Astronomical Society, 525, 2708 934
D’Antona, F. & Ergma, E. 1993, A&A, 269, 219 935
+ 5.91 × 109 t6 − 3.73 × 109 t5 + 1.43 × 109 t4 − 3.21 × 108 t3 Demory, B. O., Ségransan, D., Forveille, T., et al. 2009, A&A, 505, 205 936
Dobler, W., Stix, M., & Brandenburg, A. 2006, ApJ, 638, 336 937
+ 3.89 × 107 t2 − 2.02 × 106 t + 8.79 × 1010 , (A.3) Donati, J.-F., Collier Cameron, A., Semel, M., et al. 2003, MNRAS, 345, 1145 938
Dormand, J. & Prince, P. 1980, Journal of Computational and Applied Mathe- 939
matics, 6, 19 940
886 where t ∈ [0, 1] and R1 (t) is expressed in cm. The MRE for Falanga, M., Kuiper, L., Poutanen, J., et al. 2005, A&A, 444, 15 941
887 rS (t) is of ∼ 10−13 %. Using this approach, the MRE on Q(t) Feiden, G. A. & Chaboyer, B. 2014, ApJ, 789, 53 942
Frank, J., King, A., & Raine, D. J. 2002, Accretion Power in Astrophysics: Third 943
888 is ∼ ×10−3 %, which is still in good agreement. We conclude Edition, 398 944
889 that the quadrupole moment is approximated with respect to the Freire, P. C. C., Ridolfi, A., Kramer, M., et al. 2017, MNRAS, 471, 857 945
890 original solution with a MRE of 0.004%. Fruchter, A. S., Stinebring, D. R., & Taylor, J. H. 1988, Nature, 333, 237 946

Article number, page 12 of 13

 De Falco et al. (2024): Reconstruction of spider system’s observables

Table A.1. We show the input parameters (32) pertaining to 47 Tuc W and 47 Tuc O, as well as the quantities Q0 (cf. Eq. (15)), B0 (cf. Eq. (23)),
and L0 (cf. Eq. (27)). We express all masses and distances in terms of the solar mass M⊙ = 2 × 1033 g and the solar radius R⊙ = 6.96 × 1010 cm,
respectively. The input parameters and TEMPO coefficients t0 , t1 , T 0 , { fi }ni=0 are both taken from Ridolfi et al. (2016); Freire et al. (2017). For more
details on how a and Rc have been calculated/chosen, we refer to the beginning of Appendix A.

Parameter Unit 47 Tuc W 47 Tuc O

mp M⊙ 1.40 1.40
mc M⊙ 0.13 0.022
Rc R⊙ 0.20 0.08
a R⊙ 1.26 1.11
Pma yr 2.70 5.00
Q0 M⊙ R2⊙ 3.61 × 10−5 6.25 × 10−6
B0 kG 48.79 0.92
L0 L⊙ 34.78 4.81 × 10−3

TEMPO coefficients

t0 MJD 51214.216 47717.894

t1 MJD 54934.047 56388.106
T0 MJD 51585.3327 51600.1084
f0 s−1 8.71 × 10−5 9.59 × 10−5
f1 s−2 −1.27 × 10−18 −2.09 × 10−21
f2 s−3 4.06 × 10−26 −1.89 × 10−28
f3 s−4 6.30 × 10−33 3.94 × 10−35
f4 s−5 −9.18 × 10−40 −1.49 × 10−43
f5 s−6 6.27 × 10−47 −5.49 × 10−50
f6 s−7 −2.68 × 10−54 5.76 × 10−58
f7 s−8 7.41 × 10−62 5.478 × 10−65
f8 s−9 −1.22 × 10−69 −8.91 × 10−73
f9 s−10 9.30 × 10−78 −3.90 × 10−80
f10 s−11 – 8.57 × 10−88
f11 s−12 – 1.52 × 10−95
f12 s−13 – −4.21 × 10−103

947 Gentile, P. A., Roberts, M. S. E., McLaughlin, M. A., et al. 2014, ApJ, 783, 69 Roberts, M. S. E. 2013, in IAU Symposium, Vol. 291, Neutron Stars and Pulsars: 982
948 Guerra, C., Meliani, Z., & Voisin, G. 2024, A&A, 690, A75 Challenges and Opportunities after 80 years, ed. J. van Leeuwen, 127–132 983
949 Hall, D. S. 1990, in NATO Advanced Science Institutes (ASI) Series C, Vol. 319, Roberts, M. S. E., Al Noori, H., Torres, R. A., et al. 2018, in IAU Symposium, 984
950 95 Vol. 337, Pulsar Astrophysics the Next Fifty Years, ed. P. Weltevrede, B. B. P. 985
951 Han, E., Lopez-Valdivia, R., Mace, G., & Jaffe, D. 2023, in American Astro- Perera, L. L. Preston, & S. Sanidas, 43–46 986
952 nomical Society Meeting Abstracts, Vol. 241, American Astronomical Soci- Roberts, M. S. E., Mclaughlin, M. A., Gentile, P., et al. 2014, Astronomische 987
953 ety Meeting Abstracts, 429.07 Nachrichten, 335, 313 988
954 Harvey, D., Skillman, D. R., Patterson, J., & Ringwald, F. A. 1995, PASP, 107, Romani, R. W., Filippenko, A. V., Silverman, J. M., et al. 2012, ApJ, 760, L36 989
955 551 Ruderman, M., Shaham, J., Tavani, M., & Eichler, D. 1989, ApJ, 343, 292 990
956 Kaplan, D. L., Bhalerao, V. B., van Kerkwijk, M. H., et al. 2013, ApJ, 765, 158 Shaifullah, G., Verbiest, J. P. W., Freire, P. C. C., et al. 2016, MNRAS, 462, 1029 991
957 Kluzniak, W., Ruderman, M., Shaham, J., & Tavani, M. 1988, Nature, 334, 225 Smedley, S. L., Tout, C. A., Ferrario, L., & Wickramasinghe, D. T. 2015, MN- 992
958 Kochanek, C. S. 1992, ApJ, 385, 604 RAS, 446, 2540 993
959 Lanza, A. F. 2006, MNRAS, 369, 1773 Sorahana, S., Yamamura, I., & Murakami, H. 2013, ApJ, 767, 77 994
960 Lanza, A. F. & Rodonò, M. 1999, A&A, 349, 887 Stappers, B. W., Archibald, A. M., Hessels, J. W. T., et al. 2014, ApJ, 790, 39 995
961 Lanza, A. F. & Rodonò, M. 2002, Astronomische Nachrichten, 323, 424 Stappers, B. W., Bailes, M., Lyne, A. G., et al. 2001, MNRAS, 321, 576 996
962 Lanza, A. F., Rodono, M., & Rosner, R. 1998, MNRAS, 296, 893 Stappers, B. W., Bailes, M., Lyne, A. G., et al. 1996a, ApJ, 465, L119 997
963 Lazaridis, K., Verbiest, J. P. W., Tauris, T. M., et al. 2011, MNRAS, 414, 3134 Stappers, B. W., Bessell, M. S., & Bailes, M. 1996b, ApJ, 473, L119 998
964 MacDonald, J. & Mullan, D. J. 2009, ApJ, 700, 387 Sterne, T. E. 1939, MNRAS, 99, 451 999
965 MacDonald, J. & Mullan, D. J. 2017, ApJ, 850, 58 Strader, J., Swihart, S., Chomiuk, L., et al. 2019, ApJ, 872, 42 1000
966 Morin, J. 2012, EAS Publications Series, 57, 165 Sullivan, A. G. & Romani, R. W. 2024 [arXiv:2405.13889] 1001
967 Ng, C. W., Takata, J., Strader, J., Li, K. L., & Cheng, K. S. 2018, ApJ, 867, 90 van Buren, D. 1986, AJ, 92, 136 1002
968 Pallanca, C., Mignani, R. P., Dalessandro, E., et al. 2012, ApJ, 755, 180 van den Heuvel, E. P. J. & van Paradijs, J. 1988, Nature, 334, 227 1003
969 Pan, Z., Lu, J. G., Jiang, P., et al. 2023, Nature, 620, 961 Voisin, G., Breton, R. P., & Summers, C. 2020a, MNRAS, 492, 1550 1004
970 Papitto, A., Ferrigno, C., Bozzo, E., et al. 2013, Nature, 501, 517 Voisin, G., Clark, C. J., Breton, R. P., et al. 2020b, MNRAS, 494, 4448 1005
971 Parker, E. N. 1955, ApJ, 122, 293 Wang, S. Q., Wang, J. B., Wang, N., et al. 2021, ApjL, 922, L13 1006
972 Parker, E. N. 1979, ApJ, 230, 905 Warner, B. 1988, Nature, 336, 129 1007
973 Podsiadlowski, P. 1991, Nature, 350, 136 Wijnands, R. & van der Klis, M. 1998, Nature, 394, 344 1008
974 Press, W. H., Teukolsky, S. A., Vetterling, W. T., & Flannery, B. P. 2002, Numer- Yadav, R. K., Christensen, U. R., Morin, J., et al. 2015, ApJ, 813, L31 1009
975 ical recipes in C++ : the art of scientific computing Yoshida, S. 2019, MNRAS, 486, 2982 1010
976 Radhakrishnan, V. & Srinivasan, G. 1982, Current Science, 51, 1096 Zahn, J. P. 2008, in EAS Publications Series, Vol. 29, EAS Publications Series, 1011
977 Rappaport, S., Verbunt, F., & Joss, P. C. 1983, ApJ, 275, 713 ed. M. J. Goupil & J. P. Zahn, 67–90 1012
978 Rasio, F. A., Shapiro, S. L., & Teukolsky, S. A. 1989, ApJ, 342, 934 Zhao, J. & Heinke, C. O. 2023, MNRAS, 526, 2736 1013
979 Ridolfi, A., Freire, P. C. C., Torne, P., et al. 2016, MNRAS, 462, 2918
980 Roberts, M. S. E. 2012, Proceedings of the International Astronomical Union, 8,
981 127–132

Article number, page 13 of 13