Analytic Keplerian-type parametrization for general spinning compact binaries with the leading order spin-orbit interaction

Gihyuk Cho; Hyung Mok Lee

REVIEW 2 major objections 7 minor 32 references

Analytic Keplerian-type parametrization for general spinning compact binaries with the leading order spin-orbit interaction

T0 review · 2 major / 7 minor · reviewed 2026-08-14 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read A fully analytic Keplerian-type parametrization now describes spinning compact binaries with arbitrary eccentricity, mass ratio, and spin orientation.

desk verdict A careful closed-form extension of the quasi-Keplerian program to general masses and spins at leading spin-orbit order; the main caveat is an unproven real-root condition that needs a numerical check. read the letter →

arxiv 1908.02927 v1 pith:RYXRPMLU submitted 2019-08-08 gr-qc

Gihyuk Cho , Hyung Mok Lee This is my paper · ORCID

classification gr-qc

keywords spinningcompactbinariesspin-orbitcouplingpost-NewtonianapproximationKeplerianparametrizationJacobiellipticfunctionsprecessionADMgaugegravitationalwaves

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper claims a fully analytic, Keplerian-type parametrization for the conservative motion of a spinning compact binary in ADM gauge, working at leading post-Newtonian order and to first order in spin. The parametrization covers arbitrary eccentricity, mass ratio, and initial spin configuration: the tilt angles of the spin vectors, the precession of the orbital plane, the radial separation, and the orbital phase are all expressed in closed form through Jacobi elliptic functions whose argument is the eccentric anomaly. If the claim holds, one no longer needs to integrate the post-Newtonian equations for these binaries; the trajectory is obtained directly, and gravitational-wave templates for eccentric, precessing systems could be evaluated much faster.

What carries the argument

The load-bearing object is the cubic polynomial in Eq. (3.2) obtained by rewriting L·(S1×S2)/(L S1 S2) as ±(δ2 S2)√(A(x−x1)(x−x2)(x−x3)), whose roots x1, x2, x3 are the turning points of x = cos κ1 between which the square root is real. With x confined to [x2,x3], the substitution sin² y = (x−x2)/(x3−x2) converts the equation for dx/dt into an elliptic integral of the first kind, whose inversion gives the Jacobi elliptic function sn(Υ,β) in Eq. (3.13). The same substitution, applied to ξ1 and ϕ, produces elliptic integrals of the third kind, so the entire parametrization is carried by standard special functions of a single Keplerian parameter.

What would settle it

Take the equations of motion (2.6) with unequal masses and a strongly non-aligned spin configuration, integrate them numerically, and test whether the argument of the square root in Eq. (3.2) stays non-negative for all time; a configuration where it turns negative would be a counterexample to the claimed coverage.

Watch

Extended reading notes

Core claim

The central discovery is that the angular dynamics closes on a single variable. The paper shows that the evolution equations for the angles γ, κ1, and κ2 admit two constants of motion, σ1 and σ2, so all three angles are determined once x = cos κ1 is known. The quantity L·(S1×S2), the common factor in all three angular equations, is written as the square root of a cubic polynomial in x; integrating the resulting one-dimensional equation yields Eq. (3.13), where cos κ1 is expressed as x2 + (x3−x2) sn²(Υ,β) with the amplitude Υ proportional to ν + e sin ν. The absolute precession angle ξ1 and the orbital phase ϕ follow as elliptic integrals of the third kind in Eqs. (3.32) and (3.43), and the radial separation keeps the quasi-Keplerian form r = ar(1 − er cos u) with spin-dependent corrections. These pieces assemble, through Eq. (3.44), into a complete inertial-frame trajectory.

Load-bearing premise

The whole solution depends on the assumption that the cubic equation governing the tilt angle between the orbit and the first spin always has three real solutions within the physical range, a condition the paper supports with a heuristic argument and extends to equal masses by continuity.

Editorial extensions

If this is right

A single eccentric-anomaly parameter u drives the tilt angles, spin precession, orbital-plane precession, and relative separation, so the complete three-dimensional trajectory is obtained in closed form.
In the equal-mass limit the parametrization reduces to the known analytic equal-mass solution, which the authors use as a consistency check.
The precession period of the orbital plane can be expressed through a hypergeometric function, giving the number of orbital cycles per precession cycle.
For nearly equal masses the elliptic expressions reduce to elementary functions, providing a simpler approximate parametrization at order O((δ2−δ1)²).
The closed-form expressions can be used to produce quick time-domain waveform templates modulated by spin precession and orbital-plane swings.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

One check the paper does not report is a direct numerical integration of the equations of motion (2.6) over a grid of mass ratios and eccentricities; because the parametrization is closed form, such a comparison would be straightforward and would independently probe the root-reality assumption.
The paper notes that β and related elliptic parameters are typically very small; if that trend holds across parameter space, the elliptic functions could be replaced by elementary functions in most configurations, making waveform evaluation even faster.
Because the derivation separates conservative orbital motion from radiation reaction, the formulas could serve as a reference against which to measure radiation-reaction-driven inspiral in fully numerical evolutions of eccentric, precessing binaries.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

This is a careful, old-school analytic mechanics paper. It does what it says: for the ADM Hamiltonian truncated at Newtonian order plus the leading spin-orbit term, it gives closed-form Keplerian-type parametrizations for the orbital angular momentum precession, the spin directions, and the relative orbit, for arbitrary masses, spins, and eccentricity. The genuinely new piece is the general-mass, general-spin case; previous closed forms covered equal mass, quasi-circular, slowly spinning, or specially aligned configurations. The equal-mass limit reproduces Königsdörffer-Gopakumar, which is a good sanity check and strongly suggests the algebra is right.

The paper is strong in several ways. The reduction to elliptic integrals is standard and clearly laid out. The conserved quantities sigma1 and sigma2 are used cleanly, and the authors explicitly say they are dropping 1PN and radiation reaction, so there is no overclaiming. There is no fitting and no circularity; the only cross-check is the equal-mass limit, which is legitimate.

The soft spot is exactly the one the stress test flags. The whole construction depends on the cubic in Eq. (3.2) having three real roots with the physical branch between x2 and x3. The proof in Sec. IIIA is a plausibility argument: continuity of the sign, then an averaging claim about the second derivative of L·(S1×S2), then a continuity extension to equal mass. That is not a rigorous proof. The equal-mass case is separately handled in Section IV, so the continuity extension is less dangerous than it looks, but the general-mass root condition still needs either a real proof or, much more easily, a numerical scan over mass ratios and spin orientations. I would also ask for a direct numerical comparison of Eqs. (3.13), (3.32), and (3.43) against numerical integration of the original equations of motion. That would settle most doubts and would make the paper much more convincing.

One smaller issue: Eq. (3.4) is dimensionally off as printed, with c squared times r cubed in the numerator rather than in the denominator. The neighboring Eq. (3.5) has the expected 1/(c^2 r^3) structure, so I suspect a typographical or sign error, but it should be fixed. The nearly-equal-mass expansion is useful, though the order bookkeeping there is heavy.

Who is this for? People building quick time-domain waveform models for eccentric, precessing binaries at leading spin-orbit order, and anyone doing formal PN dynamics. It is not detector-ready; it is a building block. It deserves a serious referee. I would send it to review, and ask for the numerical root check and equation correction before acceptance.

Referee Report

2 major / 7 minor

Summary. The paper presents a closed-form, Keplerian-type parametrization for the conservative dynamics of a spinning compact binary in ADM coordinates, at leading order in the spin-orbit interaction. Using two conserved quantities, the authors reduce the evolution of the three angles between L, S1, and S2 to a single differential equation for x = cos kappa1, whose right-hand side contains the square root of a cubic polynomial. They integrate this to obtain x as a Jacobi elliptic function of the eccentric anomaly u, and then express the precession of the orbital angular momentum and the in-plane orbital phase in terms of elliptic integrals of the third kind. An almost-equal-mass expansion is also provided, and the exactly equal-mass limit is shown to reproduce a previous result. The claimed result is that this solves the three-dimensional motion for arbitrary mass ratio, eccentricity, and initial spin configuration.

Significance. If established, the result is a useful addition to the post-Newtonian toolkit: it gives an explicit, fast-to-evaluate analytic description of spin-precessing binaries at leading spin-orbit order, and it generalizes earlier equal-mass or quasi-circular results. The derivation is mostly algebraic, and the recovery of the independent equal-mass limit in Sec. IV is a valuable consistency check. However, the paper's central claim of full generality rests on two points that are not currently established: a rigorous proof that the relevant cubic always has three real roots in the accessible physical range, and a clean set of dimensionally consistent master equations. These issues are load-bearing because they determine whether Eqs. (3.13), (3.32), and (3.43) actually cover all initial data claimed in the abstract.

major comments (2)

[Sec. III A, paragraph after Eq. (3.2)] The proof that the cubic in Eq. (3.2) has three real roots, with the physical branch x2 <= cos kappa1 <= x3, is heuristic rather than rigorous. The argument assumes L dot (S1 x S2) never changes sign, then invokes monotonicity, time reversal, and an unproved averaging statement about the second derivative to conclude that the invariant must cross zero; this does not exclude the possibility that the quantity approaches a nonzero constant asymptotically. The equal-mass case is added by continuity even though A -> 0 and x1 -> infinity. Because Eq. (3.13) is only valid on a branch where the cubic has three real roots, the claimed coverage of arbitrary mass ratio and initial spin orientation is not established. I request a rigorous proof of the root condition or, failing that, a systematic numerical survey over mass ratios, spin magnitudes and orientations, and eccentricities; such a survey should also compare the analytic expressions (3.13), (3.32), and (3.43) with direct numerical integration of Eqs. (2.6).
[Sec. III A, Eqs. (3.4)-(3.5)] The displayed master equations are not mutually consistent. Equation (3.4) as printed contains c^2 r^3 in the numerator, which would rearrange to a right-hand side with c^2 r^3 dt in Eq. (3.5), whereas Eq. (3.5) has dt/(c^2 r^3). In addition, substitution of Eq. (3.2) into Eq. (2.12b) gives a factor delta2^2 S2^2 multiplying the square root, and this factor is absent from both displayed equations; it would propagate into the elliptic argument Upsilon in Eq. (3.13). Please correct these equations and demonstrate explicitly that Eq. (3.13) follows from the elliptic integral evaluation in Eq. (3.12).

minor comments (7)

[Throughout] There are numerous typographical errors, including 'bianary' in the abstract, 'Kepelerian' in the heading of Sec. III A, 'configuration' rendered as 'conﬁguration', 'witout' on page 1, 'secion' on page 1, and 'staightforwardly' in Sec. IV; the text should be proofread.
[Sec. II B] Figures 1 and 2 are referenced but not included in the manuscript; the final version should include them with clear labels for xi1, xi2, kappa1, kappa2, and Delta psi.
[Sec. III A, Eq. (3.1)] The notation in Eq. (3.1) uses cos2kappa1 and similar expressions, which should be written as cos^2 kappa1 to avoid confusion with cos(2 kappa1).
[Sec. V] The statement that 'rough numerical estimations suggest that a couple of the constants ... tend to be very small' is not supported by any presented data; either remove the claim or give quantitative details.
[Sec. III C, Eqs. (3.40)-(3.43)] The radial expression (3.42) is obtained from the Hamiltonian (2.1), which deliberately omits the 1PN orbital corrections; because the spin-orbit terms retained in Eq. (3.40) are of the same nominal order as the omitted 1PN terms, the use of Eq. (3.42) in deriving Eq. (3.43) should be explicitly labeled as a deliberate truncation so that the result is not mistaken for a complete 1.5PN expression.
[Sec. IV, Eqs. (4.7)-(4.8)] The expression for Xi2 in Eq. (4.8) appears to contain a factor 1/sigma2 (or a misplaced factor) that would diverge at sigma2 = 0; please verify the sigma2 -> 0 limit of the almost-equal-mass expansion.
[References] The reference list contains incomplete or inconsistent entries (e.g., reference [4] has an obviously wrong volume/page range), and some entries lack full author lists; please update the bibliography.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the analytic parametrization follows by direct integration from the stated Hamiltonian; the equal-mass comparison with [16] is a cross-check, not an input.

full rationale

The derivation is self-contained. Starting from the ADM Hamiltonian (2.1) and the Poisson-bracket equations (2.5)-(2.6), the authors derive conserved quantities sigma_1 and sigma_2 in Eq. (2.13) and reduce the angle dynamics to a single cubic polynomial in Eq. (3.2). The subsequent parametrizations in Eqs. (3.13), (3.32), and (3.43a) are algebraic and elliptic-function integrations of this system; initial conditions enter only through alpha, xi_1(0), and phi_0. No free parameter is fitted to data, and no fitted quantity is later renamed as a prediction. The equal-mass comparison in Sec. IV uses the prior work [16] only as a sanity check, not as a load-bearing input to the general solution. The heuristic real-root argument and the continuity extension to equal mass are potential mathematical gaps, but they are not circular: they do not assume the conclusion and do not import an unverified self-citation. There is no load-bearing self-citation, no ansatz smuggled in via citation, and no renaming of a known result as a new derivation.

Assumptions & free parameters 0 free parameters · 5 assumptions · 0 invented entities

No data-fitting parameters are introduced; the conserved quantities sigma1, sigma2 and initial anomaly alpha are determined by initial conditions, not fitted to observations. No new physical entities are postulated; the non-inertial frames and Euler angles are coordinate choices. The axioms are standard background assumptions or the specific conservative dynamics model used.

assumptions (5)

domain assumption The ADM-gauge Hamiltonian (2.1) with Newton-Wigner-Pryce spin condition is the correct conservative dynamics at leading spin-orbit order.
The paper takes Eq. (2.1) from [24] rather than deriving it; the entire solution rests on this Hamiltonian.
domain assumption The binary motion is conservative and time-reversal symmetric, with no radiation reaction.
Invoked in Sec. III A to exclude dissipative possibilities in the proof of real cubic roots.
domain assumption Total angular momentum J is conserved and can be aligned with an inertial z-axis.
Used in Sec. II B to set the frame in Fig. 1 and in Eq. (3.20).
domain assumption Newtonian Keplerian expressions for r and t, r = a(1 - e cos u) and n(t - t0) = u - e sin u, are sufficient for integrating spin precession at leading spin-orbit order.
Used in Eq. (3.6) for the right-hand side of Eq. (3.5).
standard math Standard properties of Jacobi elliptic functions and elliptic integrals of the first and third kind.
Used throughout Sec. III for the analytic integration, e.g., Eq. (3.11) and Eq. (3.31).

how reviews work

0 comments

Cite this review

Pith. "Pith review of Analytic Keplerian-type parametrization for general spinning compact binaries with the leading order spin-orbit interaction." pith.science (2026). https://pith.science/paper/RYXRPMLU

@misc{pith2026190802927,
  author       = {Pith},
  title        = {Pith review of: Analytic Keplerian-type parametrization for general spinning compact binaries with the leading order spin-orbit interaction},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/RYXRPMLU}},
  note         = {Machine review of arXiv:1908.02927}
}

read the original abstract

We derive a fully analytic Keplerian-type parametrization solution to conservative motion of spinning binary in ADM gauge. This solution is able to describe three dimensional motion of binaries of arbitrary eccentricity, mass ratio and initial configuration of spin angular momentum up to the leading order of post-Newtonian(PN) approximation and a linear order in spin. Based on our results waveforms can be quickly computed with high accuracy.

Figures

Figures reproduced from arXiv: 1908.02927 by the authors.

**Figure 2.** The geometry of spin and angular momentum vec [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗

**Figure 3.** The shape of the function inside the square-root of [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

32 extracted references · 24 canonical work pages

[1]

Note that our assumption does not include the equal mass case, i.e., δ1 = δ2 in the above proof

Additionally, from the fact that cubic polynomicals of which all coeﬃcients are real, cannot have two real roots and a single complex root, we can conclude that x1 is also real. Note that our assumption does not include the equal mass case, i.e., δ1 = δ2 in the above proof. Since there is no reason for any discontinuity toward the equal mass case, we can ...

work page
[3]

(J +L +σ2S2) +O(δ2), (4.2) x2−x3 α2 +x2 =−δ 2S1S2 √ (1−σ2

work page
[4]

(S2 1 + 2σ1S1S2 + (1−σ2 2)S2 2) δ1 (S2 1 + 2σ1S1S2 +S2

work page
[5]

(−J +L +σ2S2) +O(δ2), (4.3) β =δ 4LS 1S2 √ (1−σ2

work page
[6]

(S2 1 + 2σ1S1S2 + (1−σ2 2)S2 2) δ1 (S2 1 + 2σ1S1S2 +S2 2)2 +O(δ2). (4.4) In order to get the Keplerian parametrizations for ξ1 and ϕ expanded up to the ﬁrst order in δ, it is essential to get the ﬁrst order approximant of the type Π(A, am(Υ,β ), β). With A∼ O(δ1), β∼O(δ1) and Υ∼O(δ0), we have Π(A, am(Υ,β ), β) = ( 1 +A 2 ) Υ−A 4 sin 2Υ +O(δ2). (4.5) Then ...

work page
[7]

Mathisson, Gen

M. Mathisson, Gen. Relativ. Gravit. 42, 989 (2010), ISSN 0001-7701

work page 2010
[8]

Corinaldesi and A

E. Corinaldesi and A. Papapetrou, Tech. Rep. 1097 (1951)

work page 1951
[9]

Description, Proc

C.-o.-m. Description, Proc. Boy. Soc. Lond. A 8, 197 (1977), ISSN 1537-5323

work page 1977

Show all 32 references

[10]

Bailey and W

I. Bailey and W. Israel, Commun. Math. Phys. 588, 577 (1983), ISSN 0010-3616

1983
[11]

Blanchet, Living Rev

L. Blanchet, Living Rev. Relativ. 17 (2014)

2014
[12]

Bernard, L

L. Bernard, L. Blanchet, G. Faye, and T. Marchand, Phys. Rev. D 97, 044037 (2018), ISSN 24700029, arXiv:1711.00283v2

2018 arXiv
[13]

L. ´A. Gergely, Z. I. Perj´ es, and M. Vas´ uth, Phys. Rev. D 58, 124001 (1998), ISSN 15502368, 9808063

1998
[14]

L. ´A. Gergely, Phys. Rev. D 61, 024035 (1999), ISSN 15502368

1999
[15]

Blanchet, A

L. Blanchet, A. Buonanno, and G. Faye, Phys. Rev. D 74, 104034 (2006), ISSN 15507998, 0605140

2006
[16]

G. Faye, L. Blanchet, and A. Buonanno, Phys. Rev. D 74, 104033 (2006), ISSN 1550-7998, 0605139

2006
[17]

Damour, P

T. Damour, P. Jaranowski, and G. Sch¨ afer, Phys. Rev. D 89, 064058 (2014), ISSN 1550-7998, arXiv:1401.4548v1

2014 arXiv
[18]

Veitch, V

J. Veitch, V. Raymond, B. Farr, W. Farr, P. Graﬀ, S. Vitale, B. Aylott, K. Blackburn, N. Christensen, M. Coughlin, et al., Phys. Rev. D 91, 042003 (2015), ISSN 15502368, 1409.7215

2015 arXiv
[19]

B. Farr, C. P. L. Berry, W. M. Farr, C.-J. Haster, H. Mid- dleton, K. Cannon, P. B. Graﬀ, C. Hanna, I. Mandel, C. Pankow, et al., Astrophys. J. 825, 116 (2016), ISSN 1538-4357, 1508.05336

2016 arXiv
[20]

Klein, N

A. Klein, N. Cornish, and N. Yunes, Phys. Rev. D 88, 124015 (2013), ISSN 1550-7998

2013
[21]

Chatziioannou, A

K. Chatziioannou, A. Klein, N. Yunes, and N. Cor- nish, Phys. Rev. D 88, 063011 (2013), ISSN 1550-7998, 1307.4418

2013 arXiv
[22]

K¨ onigsd¨ orﬀer and A

C. K¨ onigsd¨ orﬀer and A. Gopakumar, Phys. Rev. D71, 024039 (2005), ISSN 1550-7998, 0501011

2005
[23]

Tessmer, Phys

M. Tessmer, Phys. Rev. D 80, 124034 (2009), ISSN 15507998, 0910.5931

2009 arXiv
[24]

Racine, Phys

´E. Racine, Phys. Rev. D 78, 044021 (2008), ISSN 1550- 7998, 0803.1820

2008 arXiv
[25]

Marsat, A

S. Marsat, A. Boh´ e, L. Blanchet, and A. Buonanno, Class. Quantum Gravity 31, 025023 (2014), ISSN 0264- 9381, 1307.6793

2014 arXiv
[26]

Chatziioannou, A

K. Chatziioannou, A. Klein, N. Cornish, and N. Yunes, Phys. Rev. Lett. 118, 051101 (2017), ISSN 0031-9007, 1606.03117

2017 arXiv
[27]

Arnowitt, S

R. Arnowitt, S. Deser, and C. W. Misner (2004), 0405109

2004
[28]

T. D. Newton and E. P. Wigner, Rev. Mod. Phys. 21, 400 (1949), ISSN 0034-6861

1949
[29]

M. H. L. Pryce, Proc. R. Soc. London. Ser. A. Math. Phys. Sci. 195, 62 (1948), ISSN 2053-9169

1948
[30]

Wex, Class

N. Wex, Class. Quantum Gravity 12, 983 (1995), ISSN 02649381

1995
[31]

Damour and N

T. Damour and N. Deruelle, Ann. l’Institut Henri Poincare 43, 107 (1985), ISSN 0246-0211

1985
[32]

Kesden, D

M. Kesden, D. Gerosa, R. O’Shaughnessy, E. Berti, and U. Sperhake, Phys. Rev. Lett. 114, 081103 (2015), ISSN 0031-9007, 1411.0674

2015 arXiv
[33]

Whittaker and G

E. Whittaker and G. Watson, A course of modern anal- ysis: an introduction to the general theory of inﬁnite processes and of analytic functions; with an account of the principal transcendental functions (3rd edition, 1920) (2009)

2009

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.

[1] [1]

Note that our assumption does not include the equal mass case, i.e., δ1 = δ2 in the above proof

Additionally, from the fact that cubic polynomicals of which all coeﬃcients are real, cannot have two real roots and a single complex root, we can conclude that x1 is also real. Note that our assumption does not include the equal mass case, i.e., δ1 = δ2 in the above proof. Since there is no reason for any discontinuity toward the equal mass case, we can ...

work page

[2] [3]

(J +L +σ2S2) +O(δ2), (4.2) x2−x3 α2 +x2 =−δ 2S1S2 √ (1−σ2

work page

[3] [4]

(S2 1 + 2σ1S1S2 + (1−σ2 2)S2 2) δ1 (S2 1 + 2σ1S1S2 +S2

work page

[4] [5]

(−J +L +σ2S2) +O(δ2), (4.3) β =δ 4LS 1S2 √ (1−σ2

work page

[5] [6]

(S2 1 + 2σ1S1S2 + (1−σ2 2)S2 2) δ1 (S2 1 + 2σ1S1S2 +S2 2)2 +O(δ2). (4.4) In order to get the Keplerian parametrizations for ξ1 and ϕ expanded up to the ﬁrst order in δ, it is essential to get the ﬁrst order approximant of the type Π(A, am(Υ,β ), β). With A∼ O(δ1), β∼O(δ1) and Υ∼O(δ0), we have Π(A, am(Υ,β ), β) = ( 1 +A 2 ) Υ−A 4 sin 2Υ +O(δ2). (4.5) Then ...

work page

[6] [7]

Mathisson, Gen

M. Mathisson, Gen. Relativ. Gravit. 42, 989 (2010), ISSN 0001-7701

work page 2010

[7] [8]

Corinaldesi and A

E. Corinaldesi and A. Papapetrou, Tech. Rep. 1097 (1951)

work page 1951

[8] [9]

Description, Proc

C.-o.-m. Description, Proc. Boy. Soc. Lond. A 8, 197 (1977), ISSN 1537-5323

work page 1977