betaFeedback

WorksPhysics and Astronomy

Existence of a third self-similar solution of the Yang–Mills equation in high even dimensions

Abstract

We prove that the spherically symmetric SO(d)SO(d) Yang–Mills equation admits a third smooth self-similar profile, in addition to the two known explicit profiles, in every even dimension d≥12d\geq 12. This complements the analogous existence theorem in odd dimensions d≥11d\geq 11 (arXiv:2610.36784). In contrast to the odd-dimensional case, where the classification of smooth self-similar profiles reduces to counting roots of an explicit resonance polynomial [Phys. Rev. D 114, 064021 (2026)], the even-dimensional proof requires solving a nonlinear connection problem. In d=10d=10, where one explicit profile degenerates to the zero solution, we give a formal argument for the existence of a second nontrivial smooth profile in addition to the single nontrivial explicit one.

Introduction

Self-similar solutions of the Yang–Mills equation provide natural models of finite-time singularity formation. For such solutions a fixed spatial profile is compressed into a shrinking region around the blowup point. Constructing and classifying these profiles is a first step towards understanding the possible local blowup dynamics. Their role in the evolution depends on their stability.

Under the spherically symmetric magnetic ansatz with gauge group SO(d)SO(d), the Yang–Mills equation in (d+1)(d+1)-dimensional Minkowski spacetime reduces to the radial nonlinear wave equation

ϕtt=ϕrr+d−3rϕr+d−2r2ϕ(1−ϕ2).(1)\phi_{tt}=\phi_{rr}+\frac{d-3}{r}\phi_r+\frac{d-2}{r^2}\phi(1-\phi^2). \tag*{(1)}

The equation is invariant under ϕ(t,r)↦ϕ(t/λ,r/λ)\phi(t,r)\mapsto\phi(t/\lambda,r/\lambda) for any λ>0\lambda>0. Solutions unchanged by this scaling are called self-similar. By time-translation invariance, for any TT they take the form

ϕ(t,r)=u(y),y=rT−t,t<T.\phi(t,r)=u(y),\qquad y=\frac{r}{T-t},\qquad t<T.

Substituting this ansatz into (1) gives the profile equation

y2(1−y2)u′′+((d−3)y−2y3)u′+(d−2)u(1−u2)=0.(2)y^2(1-y^2)u''+\bigl((d-3)y-2y^3\bigr)u'+(d-2)u(1-u^2)=0. \tag*{(2)}

The profiles of interest are nonconstant and smooth on 0≤y≤10\le y\le1, so the associated solutions are smooth in the causal past of (T,0)(T,0), where they blow up as their spatial scale T−tT-t shrinks to zero. Throughout the paper, profiles related by the reflection u↦−uu\mapsto-u are regarded as equivalent.

A standard way to solve this two-endpoint boundary-value problem is shooting. Near the origin, there is a one-parameter family of regular local solutions

u(y;a)=1−ay2+O(y4),a∈R.(3)u(y;a)=1-ay^2+O(y^4),\qquad a\in\mathbb{R}. \tag*{(3)}

The shooting method consists in continuing these solutions towards y=1y=1 and selecting the values of the parameter aa for which the resulting shot is smooth at the light cone y=1y=1.

As shown in [1], this shooting problem simplifies considerably when d≥10d\geq10. More precisely, for each fixed dd in this range and every a>0a>0, the corresponding solution u(y;a)u(y;a) extends to the whole interval 0<y<10<y<1, where it is positive, strictly decreasing, and has a finite limit

c(a):=lim⁡y→1−u(y;a).c(a):=\lim_{y\to1^-}u(y;a).

Moreover, the shooting map a↦c(a)a\mapsto c(a) is a strictly decreasing bijection from [0,∞)[0,\infty) onto (0,1](0,1], with c(0)=1c(0)=1 and lim⁡a→∞c(a)=0\lim_{a\to\infty}c(a)=0. Thus, the oscillations and repeated endpoint values that complicate shooting in lower dimensions do not occur here: each endpoint value in (0,1](0,1] corresponds to exactly one shooting parameter. Below, we take these results as input. It remains to determine the values of the shooting parameter aa for which the solution u(y;a)u(y;a) extends smoothly to y=1y=1.

Two such solutions are known explicitly:

u±(y)=1−α±y2y2+β±,a±=α±β±,0<a+<a−,(4)u_\pm(y)=1-\frac{\alpha_\pm y^2}{y^2+\beta_\pm},\qquad a_\pm=\frac{\alpha_\pm}{\beta_\pm},\qquad0<a_+<a_-, \tag*{(4)}

where, using r=3(d−4)(d−2)r=\sqrt{3(d-4)(d-2)},

α±=2±2r3(d−2),β±=2(d−4)±r3.\alpha_\pm=2\pm\frac{2r}{3(d-2)},\qquad\beta_\pm=\frac{2(d-4)\pm r}{3}.

The solutions u+u_+ and u−u_- exist for d≥5d\geq5 and d≥11d\geq11, respectively.

The question is whether there are any other smooth profiles. The analysis depends crucially on the parity of dd because the local behavior of solutions near y=1y=1 is governed by the indicial exponents 00 and

ρ=d−32.\rho=\frac{d-3}{2}.

The problem is therefore resonant when dd is odd, in which case ρ\rho is an integer, and nonresonant when dd is even, in which case ρ\rho is a half-integer. We discuss these two cases separately.

Odd dimensions For odd d=2ℓ+5≥11d=2\ell+5\geq11 the general local solution with finite endpoint value cc has the expansion

u(y)=c+∑n=1∞αn(1−y)n+Aℓ(c)(1−y)ℓ+1log⁡(1−y)+⋯ ,u(y)=c+\sum_{n=1}^{\infty}\alpha_n(1-y)^n+A_\ell(c)(1-y)^{\ell+1}\log(1-y)+\cdots,

where we omitted higher-order power–log terms (1−y)jlog⁡k(1−y)(1-y)^j\log^k(1-y). Here cc and αℓ+1\alpha_{\ell+1} are free parameters, while α1,…,αℓ\alpha_1,\ldots,\alpha_\ell are determined recursively by cc. The coefficient

Aℓ(c)=c(1−c2)Pℓ(c2),A_\ell(c)=c(1-c^2)P_\ell(c^2),

measures the obstruction to continuing the Taylor expansion through the resonant order (1−y)ℓ+1(1-y)^{\ell+1}. Here PℓP_\ell is an explicitly computable polynomial of degree ℓ\ell, defined up to a constant factor. If Pℓ(c2)=0P_{\ell}(c^{2}) = 0, all logarithmic terms vanish and the expansion reduces to a convergent Taylor series. Since the origin-regular shots have 0<c<10 < c < 1, they are smooth at y=1y = 1 if and only if Pℓ(c2)=0P_{\ell}(c^{2}) = 0. Combined with monotonicity of the shooting map, this reduces the classification of smooth self-similar profiles to counting distinct roots of Pℓ(c2)P_{\ell}(c^{2}) in 0<c2<10 < c^{2} < 1: each root determines one smooth profile, modulo reflection. Two roots correspond to the two explicit profiles. The existence of at least one additional root was recently proved in [3], establishing the existence of a third nonconstant smooth profile in every odd dimension d≥11d \ge11 and confirming the conjecture formulated in [1].

Even dimensions. Write d=2md = 2m with an integer m≥5m \ge5. Since ρ=m−32\rho= m - \frac{3}{2} is not an integer, each endpoint value cc determines a unique local solution u0(y;c)u_0(y;c) analytic at y=1y = 1. The general local solution with the same endpoint value has an additional free parameter, the amplitude of its leading nonanalytic term. This contrasts with the logarithmic term in odd dimensions, whose coefficient is determined by cc.

For an origin-regular shot, continuation from the origin to the light cone selects both the endpoint value c(a)c(a) and the nonanalytic amplitude Bm(a)B_m(a). The resulting endpoint expansion takes the form

u(y;a)=uan(y;a)+Bm(a)(1−y)ρg(y;a),(5)u(y;a) = u_{\mathrm{an}}(y;a) + B_m(a)(1-y)^{\rho}g(y;a), \tag*{(5)}

where uanu_{\mathrm{an}} and gg are analytic at y=1y = 1, with uan(1;a)=c(a)u_{\mathrm{an}}(1;a) = c(a) and g(1;a)=1g(1;a) = 1. Here uanu_{\mathrm{an}} denotes the analytic component of the full solution, not in general the analytic solution u0(y;c(a))u_0(y;c(a)). The decomposition and its dependence on the shooting parameter are justified in Appendix § 4. By (5), the shot is smooth at the light cone if and only if Bm(a)=0B_m(a) = 0. Unlike in odd dimensions, this criterion cannot be reduced to an algebraic condition on the endpoint value cc. Instead, finding the zeros of BmB_m requires controlling the nonanalytic amplitude along the family of origin-regular shots. This is a genuinely nonlinear connection problem.

For m≥6m \ge6, establishing a third smooth profile amounts to finding a zero of BmB_m distinct from the two explicit zeros,

Bm(a+)=Bm(a−)=0.B_m(a_+) = B_m(a_-) = 0.

For m=5m = 5, only the explicit zero a+a_+ remains, and we seek a second nontrivial smooth profile. By continuity of BmB_m, it is enough to find a sign change in an interval containing no explicit zero. The argument splits into three cases:

  • For m=6,7,8m = 6,7,8, BmB_m crosses both explicit zeros in the same direction, forcing an additional zero in (a+,a−)(a_+,a_-).

  • For every integer m≥9m \ge9, the sign of BmB_m immediately beyond a−a_- differs from its sign for large aa, forcing an additional zero in (a−,∞)(a_-,\infty).

  • For m=5m = 5, the sign of B5B_5 immediately beyond a+a_+ differs from its sign for large aa, forcing an additional zero in (a+,∞)(a_+,\infty).

The first two cases are proved rigorously, while the third is established only at a formal level. Thus, we prove the existence of a third nontrivial smooth profile in every even dimension d=2m≥12d = 2m \ge12. For m=5m = 5, where one explicit profile degenerates to the zero solution, we show by formal methods the existence of a second nontrivial smooth profile in addition to the remaining explicit one.

The rest of the paper is organized as follows. In § 2, we determine the large-aa behavior of Bm(a)B_m(a), including the exceptional case m=5m = 5. In § 3, we determine the crossing directions at the explicit zeros and complete the existence proof for m≥6m \ge6. The technical details are collected in the appendices.

Large-aa behavior

The case m≥6m \ge6

We use the shorthands

ν=m2−6m+6,k=m−2−ν,(6)\nu= \sqrt{m^2 - 6m + 6}, \qquad k = m - 2 - \nu, \tag*{(6)}

so that 1<k<21 < k < 2. As before, ρ=m−32\rho= m - \frac{3}{2} is the half-integer light-cone indicial exponent.

The origin expansion (1.3) identifies y∼a−1/2y \sim a^{-1/2} as the scale on which the deviation from u=1u = 1 is no longer small. To resolve this shrinking region, let

x=a y,Ua(x)=u(y;a).x = \sqrt{a}\,y, \qquad U_a(x) = u(y;a).

The rescaled profile satisfies

x2(1−x2a)Ua′′+((2m−3)x−2x3a)Ua′+(2m−2)Ua(1−Ua2)=0,Ua(x)=1−x2+O(x4).x^2\left(1-\frac{x^2}{a}\right)U_a''+\left((2m-3)x-\frac{2x^3}{a}\right)U_a'+(2m-2)U_a(1-U_a^2)=0,\qquad U_a(x)=1-x^2+O(x^4).

For fixed xx, letting a→∞a \to\infty gives

x2Q′′+(2m−3)xQ′+(2m−2)Q(1−Q2)=0,Q(x)=1−x2+O(x4).(7)x^2Q''+(2m-3)xQ'+(2m-2)Q(1-Q^2)=0,\qquad Q(x)=1-x^2+O(x^4). \tag*{(7)}

The normalized solution QQ is the limiting core profile: Ua→QU_a \to Q on compact xx-intervals as a→∞a \to\infty. With τ=log⁡x\tau=\log x and q(τ)=Q(eτ)q(\tau)=Q(e^\tau), equation (7) becomes

q¨+(2m−4)q˙+(2m−2)q(1−q2)=0.\ddot{q}+(2m-4)\dot{q}+(2m-2)q(1-q^2)=0.

The origin expansion shows that the orbit enters the forward-invariant strip

0<q<1,−kq<q˙<0,0<q<1,\qquad-kq<\dot{q}<0,

whose invariance follows by checking the vector field on its boundary. Thus QQ is global, positive, and strictly decreasing to zero. Linearization at q=0q=0 gives

Q(x)∼Cmx−kas x→∞,Cm>0.(8)Q(x)\sim C_m x^{-k}\quad\text{as }x\to\infty,\qquad C_m>0. \tag*{(8)}

Indeed, the inequality q˙+kq>0\dot{q}+kq>0 makes ekτq(τ)e^{k\tau}q(\tau) strictly increasing and hence rules out decay along the faster mode.

The outer field inherits its amplitude and small-yy behavior from this tail in the overlap region a−1/2≪y≪1a^{-1/2}\ll y\ll1:

u(y;a)=Ua(ay)≃Q(ay)∼Cma−k/2y−k.u(y;a)=U_a(\sqrt{a}y)\simeq Q(\sqrt{a}y)\sim C_m a^{-k/2}y^{-k}.

To extend this approximation beyond the overlap region, we substitute the ansatz

u(y;a)≃Cma−k/2Fm(y)(9)u(y;a)\simeq C_m a^{-k/2}F_m(y) \tag*{(9)}

into (1.2) and retain the leading terms as a→∞a\to\infty. This gives the linear equation

y2(1−y2)Fm′′+((2m−3)y−2y3)Fm′+(2m−2)Fm=0.(10)y^2(1-y^2)F_m''+\bigl((2m-3)y-2y^3\bigr)F_m'+(2m-2)F_m=0. \tag*{(10)}

Matching the behavior Fm(y)∼y−kF_m(y)\sim y^{-k} for y→0y\to0 selects the solution

Fm(y)=y−k 2F1(−k2,1−k2;1+ν;y2).(11)F_m(y)=y^{-k}\,{}_2F_1\left(-\frac{k}{2},\frac{1-k}{2};1+\nu;y^2\right). \tag*{(11)}

Thus the core fixes the amplitude, while the linear outer equation fixes the shape. Applying the hypergeometric connection formula [4 Eq. (15.8.4)] to (11) gives

Fm(y)=Fm,an(y)+Kmy−k(1−y2)ρ2F1(m+ν2,m−1+ν2;1+ρ;1−y2),F_m(y)=F_{m,\mathrm{an}}(y)+\mathcal{K}_m y^{-k}(1-y^2)^\rho{}_2F_1\left(\frac{m+\nu}{2},\frac{m-1+\nu}{2};1+\rho;1-y^2\right),

where Fm,anF_{m,\mathrm{an}} is analytic at y=1y=1 and

Km=Γ(1+ν)Γ(−ρ)Γ(−k/2)Γ((1−k)/2).(12)\mathcal{K}_m=\frac{\Gamma(1+\nu)\Gamma(-\rho)}{\Gamma(-k/2)\Gamma((1-k)/2)}. \tag*{(12)}

Expanding at y=1y=1, we obtain

Fm(y)=Fm,an(y)+2ρKm(1−y)ρ(1+O(1−y)).(13)F_m(y)=F_{m,\mathrm{an}}(y)+2^\rho\mathcal{K}_m(1-y)^\rho(1+O(1-y)). \tag*{(13)}

Comparing this expansion with (5) and recalling the outer amplitude Cma−k/2C_m a^{-k/2}, we obtain the asymptotic behaviour for a→∞a\to\infty

Bm(a)∼2ρCmKma−k/2.(14)B_m(a)\sim2^\rho C_m\mathcal{K}_m a^{-k/2}. \tag*{(14)}

The matching and the passage to the light-cone coefficient are justified rigorously in Appendix § 5. It remains to determine the sign of Km\mathcal{K}_m. Since 1<k<21<k<2, both denominator Gamma factors in (12) are negative. Hence

sgn⁡Km=sgn⁡Γ(−ρ)=εm,εm:=(−1)m−1.(15)\operatorname{sgn}\mathcal{K}_m=\operatorname{sgn}\Gamma(-\rho)=\varepsilon_m,\qquad\varepsilon_m:=(-1)^{m-1}. \tag*{(15)}

Since CmC_m is positive, this is also the sign of Bm(a)B_m(a) for sufficiently large aa: negative when m=6,8,10,…m=6,8,10,\ldots, and positive when m=7,9,11,…m=7,9,11,\ldots.

The exceptional dimension d=10d=10 (m=5m=5)

For m=5m=5, the limiting core profile is explicit:

Q(x)=11+x2.(16)Q(x)=\frac{1}{1+x^2}. \tag*{(16)}

In the overlap region a−1/2≪y≪1a^{-1/2}\ll y\ll1,

Q(ay)=a−1y−2−a−2y−4+O(a−3y−6).Q(\sqrt{a}y)=a^{-1}y^{-2}-a^{-2}y^{-4}+O(a^{-3}y^{-6}).

The leading outer term is therefore of order a−1a^{-1}. Since the cubic term in (2) first contributes at order a−3a^{-3}, the first two outer terms satisfy the linear equation

y2(1−y2)F′′+(7y−2y3)F′+8F=0.y^2(1-y^2)F''+(7y-2y^3)F'+8F=0.

One solution is

F5(y)=y−2+14.F_5(y)=y^{-2}+\frac{1}{4}.

Reduction of order gives a second linearly independent solution,

G(y)=−2F5(y)∫1ys−7(1−s2)5/2F5(s)2 ds.(17)G(y)=-2F_5(y)\int_1^y\frac{s^{-7}(1-s^2)^{5/2}}{F_5(s)^2}\,ds. \tag*{(17)}

The factor −2-2 normalizes the leading term at the origin to y−4y^{-4}, while the lower integration limit fixes the freedom to add a multiple of F5F_5 by imposing G(1)=0G(1)=0. As y→0y \to0,

G(y)=y−4+6y−2log⁡y+O(y−2).G(y)=y^{-4}+6y^{-2}\log y+O(y^{-2}).

Matching the leading core term selects a−1F5a^{-1}F_5, while the term −a−2y−4-a^{-2}y^{-4} fixes the coefficient of GG to be −a−2-a^{-2}. The accompanying logarithmic term is matched by the first inner correction and introduces a logarithmic dependence on aa in the coefficient of F5F_5. Since F5F_5 is analytic at y=1y=1, this does not affect B5(a)B_5(a).

Using F5(y)=5/4+O(1−y)F_5(y)=5/4+O(1-y) and 1−s2=2(1−s)(1+O(1−s))1-s^2=2(1-s)(1+O(1-s)) in the integral representation of GG, we obtain

G(y)=3225∫y1(1−s)5/2 ds (1+O(1−y)).G(y)=\frac{32\sqrt{2}}{5}\int_y^1(1-s)^{5/2}\,ds\,(1+O(1-y)).

Hence

G(y)∼64235(1−y)7/2as y→1−.(18)G(y)\sim\frac{64\sqrt{2}}{35}(1-y)^{7/2}\qquad\text{as }y\to1^{-}. \tag*{(18)}

and therefore

B5(a)∼−64235a−2as a→∞.(19)B_5(a)\sim-\frac{64\sqrt{2}}{35}a^{-2}\qquad\text{as }a\to\infty. \tag*{(19)}

Remark 2.1. A rigorous proof of (2.14) requires controlling the first inner correction and propagating the matching error to the outer region. These standard but lengthy estimates are omitted to avoid obscuring the main argument. We therefore state (2.14) as a formal asymptotic prediction.

Additional profiles from crossing directions

The direction in which BmB_m crosses an explicit zero is determined by the variation

w(y):=∂u(y;a)∂a∣a=a±,w(y):=\left.\frac{\partial u(y;a)}{\partial a}\right|_{a=a_\pm},

so that

u(y;a±+δa)=u±(y)+δa w(y)+O(δa2).u(y;a_\pm+\delta a)=u_\pm(y)+\delta a\,w(y)+O(\delta a^2).

Differentiating (1.2) with respect to aa at a=a±a=a_\pm gives

y2(1−y2)w′′+((2m−3)y−2y3)w′+(2m−2)(1−3u±2)w=0.(20)y^2(1-y^2)w''+\bigl((2m-3)y-2y^3\bigr)w'+(2m-2)(1-3u_\pm^2)w=0. \tag*{(20)}

The origin expansion u(y;a)=1−ay2+O(y4)u(y;a)=1-ay^2+O(y^4) gives w(y)=−y2+O(y4)w(y)=-y^2+O(y^4). Similarly, differentiating the endpoint expansion (1.5) at a=a±a=a_\pm and using Bm(a±)=0B_m(a_\pm)=0 gives

w(y)=wan(y)+Bm′(a±)(1−y)ρg(y;a±),(21)w(y)=w_{\rm an}(y)+B_m'(a_\pm)(1-y)^\rho g(y;a_\pm), \tag*{(21)}

where wanw_{\rm an} is analytic at y=1y=1. Since g(1;a±)=1g(1;a_\pm)=1, the leading fractional coefficient of ww is Bm′(a±)B_m'(a_\pm). This coefficient determines the crossing direction whenever it is nonzero, because

Bm(a±+δa)=δa Bm′(a±)+O(δa2).B_m(a_\pm+\delta a)=\delta a\,B_m'(a_\pm)+O(\delta a^2).

We carry out this analysis for both explicit profiles when m≥6m\ge6 and for the single explicit profile u+u_+ when m=5m=5.

For either branch, suppress the sign label and set

A=1+β,ξ=Ay2y2+β,w=−βAξf(ξ).(22)A = 1+\beta,\qquad\xi= \frac{Ay^2}{y^2+\beta},\qquad w=-\frac{\beta}{A}\xi f(\xi). \tag*{(22)}

This maps 0≤y≤10 \le y \le1 to 0≤ξ≤10 \le\xi\le1, and the origin condition w=−y2+O(y4)w=-y^2+O(y^4) becomes f(0)=1f(0)=1. Inserting (22) into (20) and using the explicit profile parameters gives

f′′+(m+1ξ−m−52ξ−1+12(ξ−A))f′+m+3A−12−4ξξ(ξ−1)(ξ−A)f=0.(23)f^{\prime\prime}+\left(\frac{m+1}{\xi}-\frac{m-\frac{5}{2}}{\xi-1}+\frac{1}{2(\xi-A)}\right)f^\prime+\frac{m+\frac{3A-1}{2}-4\xi}{\xi(\xi-1)(\xi-A)}f=0. \tag*{(23)}

Equation (23) is a Heun equation with four regular singular points at ξ=0,1,A,∞\xi=0,1,A,\infty.

The origin-regular solution has the Taylor expansion

f(ξ)=∑n=0∞fnξn,f0=1.(24)f(\xi)=\sum_{n=0}^{\infty}f_n\xi^n,\qquad f_0=1. \tag*{(24)}

Since A>1A>1, the nearest possible singularity is at ξ=1\xi=1, so the series converges for 0≤ξ<10\le\xi<1.

Near the light cone, the origin-regular solution defined by (24) has the representation

f(ξ)=fan(ξ)+C(1−ξ)ρh(ξ),h(1)=1,(25)f(\xi)=f_{\rm an}(\xi)+C(1-\xi)^\rho h(\xi),\qquad h(1)=1, \tag*{(25)}

where fanf_{\rm an} and hh are analytic at ξ=1\xi=1. Comparing this with (21) using the transformation (22), we obtain, after restoring the branch labels,

Bm′(a±)=−β±A±(2β±A±)ρCm,±.(26)B_m^\prime(a_\pm)=-\frac{\beta_\pm}{A_\pm}\left(\frac{2\beta_\pm}{A_\pm}\right)^\rho C_{m,\pm}. \tag*{(26)}

Our task is to determine the sign of the connection coefficient CC. To this end, we differentiate (25) m−1m-1 times with respect to ξ\xi. Recalling that ρ=m−32\rho=m-\frac{3}{2}, we obtain

f(m−1)(ξ)=εm(∏j=0m−2(ρ−j))C(1−ξ)−1/2+O(1),εm:=(−1)m−1,(27)f^{(m-1)}(\xi)=\varepsilon_m\left(\prod_{j=0}^{m-2}(\rho-j)\right)C(1-\xi)^{-1/2}+O(1),\qquad\varepsilon_m:=(-1)^{m-1}, \tag*{(27)}

where the product is positive. Thus, if C≠0C\ne0, the derivative diverges with sign εmsgn⁡C\varepsilon_m\operatorname{sgn}C.

On the other hand, differentiating the Taylor series (24) m−1m-1 times gives

f(m−1)(ξ)=∑n≥m−1n!(n−m+1)!fnξn−m+1.(28)f^{(m-1)}(\xi)=\sum_{n\ge m-1}\frac{n!}{(n-m+1)!}f_n\xi^{n-m+1}. \tag*{(28)}

The analysis of the recurrence relation in Appendix C shows that all sufficiently large Taylor coefficients fnf_n have the same sign σ∈{−1,1}\sigma\in\{-1,1\} and that f(m−1)f^{(m-1)} diverges as ξ→1−\xi\to1^-. By (28), this divergence has sign σ\sigma, since the finitely many terms preceding the tail remain bounded. Comparing with (27), we conclude that C≠0C\ne0 and

sgn⁡C=σεm.(29)\operatorname{sgn} C=\sigma\varepsilon_m. \tag*{(29)}

Since the prefactor in (26) is negative, we obtain

sgn⁡Bm′(a±)=−σεm.(30)\operatorname{sgn} B_m^\prime(a_\pm)=-\sigma\varepsilon_m. \tag*{(30)}

The appendix establishes σ=−1\sigma=-1 on the plus branch for every m≥5m \ge5 and on the minus branch for m=6,7,8m=6,7,8, whereas σ=1\sigma=1 on the minus branch for m≥9m \ge9. The resulting crossing signs are summarized below, together with the intervals containing an additional zero. For m≥9m \ge9, the conclusion also uses (14) and (15); for m=5m=5, it relies on the formal prediction (19).

mmsgn⁡Bm′(a+)\operatorname{sgn} B'_m(a_+)sgn⁡Bm′(a−)\operatorname{sgn} B'_m(a_-)interval containing a new zero
6,7,86,7,8εm\varepsilon_mεm\varepsilon_m(a+,a−)(a_+,a_-)
m≥9m \ge9εm\varepsilon_m−εm-\varepsilon_m(a−,∞)(a_-,\infty)
55+1+1—(a+,∞)(a_+,\infty)

Table 3.12.

For m≥6m \ge6, these sign changes and continuity of Bm(a)B_m(a) yield the following existence result.

Theorem 3.1. For every integer m≥6m \ge6 (spatial dimension d=2md=2m), there are at least three nonconstant smooth self-similar profiles on [0,1][0,1], modulo reflection. Besides u±u_{\pm}, there is a profile u⋆u_{\star} with shooting parameter

a⋆∈(a+,a−)(m=6,7,8),a⋆∈(a−,∞)(m≥9).a_{\star} \in(a_+,a_-) \quad(m=6,7,8), \qquad a_{\star} \in(a_-,\infty) \quad(m \ge9).

Proof. For m=6,7,8m=6,7,8, the derivatives of BmB_m at the two explicit zeros have the same nonzero sign, hence Bm(a++δ)B_m(a_+ + \delta) and Bm(a−−δ)B_m(a_- - \delta) have opposite signs for sufficiently small δ>0\delta>0. Continuity therefore gives an additional zero in (a+,a−)(a_+,a_-).

For m≥9m \ge9, we have sgn⁡Bm′(a−)=−εm\operatorname{sgn} B'_m(a_-)=-\varepsilon_m, so Bm(a−+δ)B_m(a_-+\delta) has sign −εm-\varepsilon_m for sufficiently small δ>0\delta>0. On the other hand, (14) and (15) give sign εm\varepsilon_m for Bm(a)B_m(a) at a sufficiently large parameter. Continuity therefore gives an additional zero in (a−,∞)(a_-,\infty).

In both cases, Theorem A.2 implies smoothness.

For m=5m=5, the crossing sign B5′(a+)>0B'_5(a_+)>0 gives B5(a++δ)>0B_5(a_+ + \delta)>0 for sufficiently small δ>0\delta>0, whereas (19) predicts that B5(a)<0B_5(a)<0 for large aa. Taking this prediction as input, continuity forces an additional zero in (a+,∞)(a_+,\infty). Theorem A.2 then gives a second nontrivial smooth profile. This conclusion remains formal because the error estimate needed to justify (19) has not been established.

Endpoint coordinates

We describe a shot near the light cone by two numbers: its limiting value c=u(1)c=u(1) and the amplitude BB of the leading fractional-power term. Smooth profiles are precisely those for which B=0B=0. We construct these endpoint coordinates and show that they depend analytically on the shooting parameter.

Local behaviour at the light cone

For a fixed integer m≥5m \ge5, put

s=1y−1,v(s)=u(y),ρ=m−32.s=\frac{1}{y}-1,\qquad v(s)=u(y),\qquad\rho=m-\frac{3}{2}.

The light cone corresponds to s=0s=0, and the profile equation becomes

s(2+s)v′′−2(ρ−1)(1+s)v′+(2m−2)v(1−v2)=0.(31)s(2+s)v''-2(\rho-1)(1+s)v'+(2m-2)v(1-v^2)=0. \tag*{(31)}

The indicial exponents are 00 and ρ\rho.

We use the following theorem of Breitenlohner–Forgács–Maison (BFM) [2].

Theorem A.1 (BFM). The system

sX′=sf(s,X,Y),sY′=−ΛY+sg(s,X,Y),sX' = sf(s, X, Y), \qquad sY' = -\Lambda Y + sg(s, X, Y),

where ff and gg are analytic and Λ\Lambda is a positive constant, has a local solution near s=0s = 0 satisfying

X(0)=c,Y(0)=0.X(0) = c, \qquad Y(0) = 0.

The solution is analytic in ss, cc, and any additional parameters entering the equations analytically, and is unique among solutions continuous at s=0s = 0 with these limiting values.

We apply this theorem twice: first to construct the analytic solution with prescribed endpoint value cc, and then to include the independent fractional-power correction.

The analytic solution. We first construct the solution analytic at s=0s = 0 with v(0)=cv(0) = c. To bring equation (A.1) into BFM form, we seek a polynomial

P(s,v)=∑j=0m−2sjpj(v)P(s,v) = \sum_{j=0}^{m-2} s^j p_j(v)

which approximates the relation between v′v' and vv along analytic solutions. Substituting

v′=P(s,v),v′′=Ps+PvPv' = P(s,v), \qquad v'' = P_s + P_vP

into (A.1) leaves the residual

R[P](s,v):=s(2+s)(Ps+PvP)−2(ρ−1)(1+s)P+(2m−2)v(1−v2).\mathcal{R}[P](s,v) := s(2+s)(P_s+P_vP) - 2(\rho-1)(1+s)P + (2m-2)v(1-v^2).

This measures the failure of the relation v′=P(s,v)v' = P(s,v) to satisfy the profile equation. We choose the coefficients pj(v)p_j(v) to cancel successive powers of ss, so that R[P]=O(sm−1)\mathcal{R}[P] = O(s^{m-1}). Cancellation of the constant term gives

p0(v)=2m−22m−5v(1−v2),p_0(v) = \frac{2m-2}{2m-5}v(1-v^2),

while cancellation of the coefficient of ss gives

2(2−ρ)p1+2p0p0′−2(ρ−1)p0=0,2(2-\rho)p_1 + 2p_0p_0' - 2(\rho-1)p_0 = 0,

which determines p1p_1 from p0p_0. More generally, suppose that p0,…,pj−1p_0,\ldots,p_{j-1} have been chosen so that

Pj−1(s,v)=∑k=0j−1skpk(v),R[Pj−1]=sjrj(v)+O(sj+1).P_{j-1}(s,v) = \sum_{k=0}^{j-1}s^k p_k(v), \qquad\mathcal{R}[P_{j-1}] = s^j r_j(v) + O(s^{j+1}).

Then

R[Pj−1+sjpj]=sj[rj+2(j+1−ρ)pj]+O(sj+1),\mathcal{R}[P_{j-1}+s^j p_j] = s^j\left[r_j+2(j+1-\rho)p_j\right] + O(s^{j+1}),

so the next coefficient is determined by

pj(v)=−rj(v)2(j+1−ρ).p_j(v) = -\frac{r_j(v)}{2(j+1-\rho)}.

The denominator is nonzero because ρ\rho is a half-integer. Continuing through j=m−2j=m-2, we obtain

R[P]=sm−1R(s,v),\mathcal{R}[P]=s^{m-1}R(s,v),

with RR analytic. Having constructed PP, we introduce the remainder YY by

v′=P(s,v)+sm−2Y.v'=P(s,v)+s^{m-2}Y.

This brings (A.1) into BFM form:

sv′=s(P(s,v)+sm−2Y),sY′=−12Y+sG(s,v,Y),(32)\begin{aligned} sv'&=s\left(P(s,v)+s^{m-2}Y\right),\\ sY'&=-\frac{1}{2}Y+sG(s,v,Y), \tag*{(32)} \end{aligned}

where

G(s,v,Y)=(ρ−12+s−Pv(s,v))Y−R(s,v)2+sG(s,v,Y)=\left(\frac{\rho-1}{2+s}-P_v(s,v)\right)Y-\frac{R(s,v)}{2+s}

is analytic. Here the coefficient −12-\frac{1}{2} follows from ρ−1−(m−2)=−12\rho-1-(m-2)=-\frac{1}{2}. Theorem A.1 gives a solution with v(0)=cv(0)=c and Y(0)=0Y(0)=0, analytic in ss and cc. We denote the resulting analytic solution of Eq. (A.1) by v0(s;c)v_0(s;c). Since the Taylor recurrence for (A.1) has nonzero divisors 2n(n−ρ)2n(n-\rho) for every integer n≥1n\ge1, this is the unique analytic solution with endpoint value cc.

The fractional-power family. To include the second free parameter, we write

v(s;c,B)=v0(s;c)+sρZ(σ;c,B),σ=s.(33)v(s;c,B)=v_0(s;c)+s^\rho Z(\sigma;c,B),\qquad\sigma=\sqrt{s}. \tag*{(33)}

In terms of σ\sigma, the factor sρs^\rho becomes the integer power σ2m−3\sigma^{2m-3}. Substituting into (A.1) and writing W=dZ/dσW=dZ/d\sigma, we obtain

σdZdσ=σW,σdWdσ=−(2ρ+1)W+σH(σ,Z,W;c),(34)\begin{aligned} \sigma\frac{dZ}{d\sigma}&=\sigma W,\\ \sigma\frac{dW}{d\sigma}&=-(2\rho+1)W+\sigma H(\sigma,Z,W;c), \tag*{(34)} \end{aligned}

where,

(2+σ2)H=2(ρ−1)σW+4[ρ(ρ−1)−2(m−1)(1−3v02)]Z+24(m−1)v0σ2ρZ2+8(m−1)σ4ρZ3.(2+\sigma^2)H=2(\rho-1)\sigma W+4\left[\rho(\rho-1)-2(m-1)(1-3v_0^2)\right]Z+24(m-1)v_0\sigma^{2\rho}Z^2+8(m-1)\sigma^{4\rho}Z^3.

Since 2ρ=2m−32\rho=2m-3 is an integer, HH is analytic. The BFM theorem gives a unique solution with

Z(0;c,B)=B,W(0;c,B)=0,Z(0;c,B)=B,\qquad W(0;c,B)=0,

analytic in (σ,c,B)(\sigma,c,B). Thus cc fixes the endpoint value, while BB fixes the leading fractional-power correction. For B=0B=0, uniqueness gives Z=W=0Z=W=0, and the solution reduces to v0(s;c)v_0(s;c).

To recover the decomposition used in the main text, split ZZ into its even and odd parts:

Z(σ;c,B)=Ze(σ2;c,B)+σZo(σ2;c,B).Z(\sigma;c,B)=Z_e(\sigma^2;c,B)+\sigma Z_o(\sigma^2;c,B).

Since ρ+12=m−1\rho+\frac{1}{2}=m-1 is an integer, the odd part contributes only integer powers of ss. Moreover, ZeZ_e vanishes at B=0B=0, so it has an analytic factor BB. Consequently, the endpoint family can be written as

v(s;c,B)=van(s;c,B)+BsρG0(s;c,B),G0(0;c,B)=1.(35)v(s;c,B)=v_{\mathrm{an}}(s;c,B)+Bs^\rho G_0(s;c,B),\qquad G_0(0;c,B)=1. \tag*{(35)}

where both functions vanv_{\mathrm{an}} and G0G_0 are analytic in ss and the parameters. The analytic component need not be a solution, but equals v0v_0 when B=0B=0.

Returning to yy, we obtain

u(y)=uan(y;c,B)+B(1−y)ρg(y;c,B),u(y)=u_{\mathrm{an}}(y;c,B)+B(1-y)^{\rho}g(y;c,B),

where

uan(1;c,B)=c,g(1;c,B)=1.u_{\mathrm{an}}(1;c,B)=c,\qquad g(1;c,B)=1.

Identification of origin-regular shots

We show that every origin-regular shot belongs to the family constructed above. Fix one such shot, suppress its parameter aa, and let cc be its endpoint value, which exists by [1]. Subtracting the equation for v0(s;c)v_0(s;c) from that for vv shows that η=v−v0(s;c)\eta=v-v_0(s;c) satisfies the linear equation

s(2+s)η′′−2(ρ−1)(1+s)η′+Q(s)η=0,Q(s)=(2m−2)(1−v2−vv0−v02),s(2+s)\eta''-2(\rho-1)(1+s)\eta'+Q(s)\eta=0,\qquad Q(s)=(2m-2)(1-v^2-vv_0-v_0^2),

with η(s)→0\eta(s)\to0 as s→0+s\to0^+.

We seek a solution η1=sρz\eta_1=s^\rho z with z(0)=1z(0)=1. Substitution gives

s(2+s)z′′+2(ρ+1+s)z′+[Q(s)−ρ(ρ−1)]z=0.s(2+s)z''+2(\rho+1+s)z'+[Q(s)-\rho(\rho-1)]z=0.

An elementary local existence argument gives such a solution with bounded z′z', so η1(s)=sρ(1+O(s))\eta_1(s)=s^\rho(1+O(s)). Reduction of order yields a second solution, linearly independent of η1\eta_1,

η2(s)=η1(s)∫srρ−1(2+r)ρ−1η1(r)2 dr∼−2ρ−1ρ,s→0+.\eta_2(s)=\eta_1(s)\int^s\frac{r^{\rho-1}(2+r)^{\rho-1}}{\eta_1(r)^2}\,dr\sim-\frac{2^{\rho-1}}{\rho},\qquad s\to0^+.

Since η→0\eta\to0, its η2\eta_2 component vanishes, so η=Bη1\eta=B\eta_1 for a unique BB.

Setting σ=s\sigma=\sqrt{s}, we therefore obtain

Z(σ):=σ−2ρη(σ2)=Bz(σ2)=B+O(σ2),W(σ):=dZdσ=2Bσz′(σ2)=O(σ).Z(\sigma):=\sigma^{-2\rho}\eta(\sigma^2)=Bz(\sigma^2)=B+O(\sigma^2),\qquad W(\sigma):=\frac{dZ}{d\sigma}=2B\sigma z'(\sigma^2)=O(\sigma).

These functions satisfy (A.4). The uniqueness assertion in Theorem A.1 therefore identifies the shot with the member of the family constructed above corresponding to (c,B)(c,B).

Dependence on the shooting parameter

The endpoint value and fractional amplitude are genuine coordinates for the local solution, not merely formal expansion coefficients. To see this, fix a small s0>0s_0>0 and consider

(c,B)⟼(v(s0;c,B),∂sv(s0;c,B)).(c,B)\longmapsto\bigl(v(s_0;c,B),\partial_s v(s_0;c,B)\bigr).

The endpoint expansion gives

∂cv=1+O(s),∂s∂cv=O(1),\partial_c v=1+O(s),\qquad\partial_s\partial_c v=O(1),

and

∂Bv=sρ(1+O(s)),∂s∂Bv=ρsρ−1(1+O(s)).\partial_B v=s^\rho(1+O(s)),\qquad\partial_s\partial_B v=\rho s^{\rho-1}(1+O(s)).

Consequently,

det⁡∂(v,vs)∂(c,B)=ρs0ρ−1(1+O(s0))≠0.\det\frac{\partial(v,v_s)}{\partial(c,B)}=\rho s_0^{\rho-1}(1+O(s_0))\ne0.

The analytic inverse function theorem shows that (c,B)(c,B) depend locally real analytically on the interior value and derivative. Since the origin-regular shot depends analytically on aa at every interior point, we obtain the following result.

Theorem A.2. The endpoint parameters (c,B)(c,B) depend locally real analytically on the value and derivative of the solution at a fixed interior point sufficiently close to y=1y=1. In particular, for the origin-regular shooting family, c(a)c(a) and Bm(a)B_m(a) are locally real analytic functions of a>0a>0. An origin-regular shot is smooth at y=1y=1 if and only if Bm(a)=0B_m(a)=0. In that case, it is analytic there.

The smoothness criterion follows directly from the endpoint expansion. When B=0B=0, the solution coincides with the analytic solution v0v_0. When B≠0B\ne0, the term B(1−y)ρg(y)B(1-y)^\rho g(y), with g(1)=1g(1)=1, prevents smoothness because ρ\rho is a half-integer.

Near the zero solution, linearizing the map from interior data to BB gives the fractional-power coefficient of the corresponding linearized solution. This observation is used in Appendix B to transfer the outer approximation to the endpoint amplitude.

Matching the core to the outer solution

For m≥6m\ge6, we justify the large-aa approximation

u(y;a)∼Cma−k/2Fm(y)u(y;a)\sim C_m a^{-k/2}F_m(y)

and its consequence for the fractional-power coefficient,

Bm(a)∼2ρCmKma−k/2.B_m(a)\sim2^\rho C_m K_m a^{-k/2}.

The main point is that the nonlinear core fixes the amplitude of the outer solution, while nonlinear effects outside the core produce only a smaller correction.

Matching in the overlap region. Choose a large number RR and match at

yR=Ra.y_R=\frac{R}{\sqrt{a}}.

For large aa, this point lies well inside the light cone but outside the core. Convergence to the core profile gives

u(yR;a)≃Q(R),yRuy(yR;a)≃RQ′(R).u(y_R;a)\simeq Q(R),\qquad y_Ru_y(y_R;a)\simeq RQ'(R).

Let

La(y)=αa,RFm(y)+βa,RGm(y)L_a(y)=\alpha_{a,R}F_m(y)+\beta_{a,R}G_m(y)

be the solution of the linear outer equation with these same value and derivative at yRy_R. Here

Fm(y)∼y−k,Gm(y)∼y−ℓ,ℓ=m−2+ν>k,F_m(y)\sim y^{-k},\qquad G_m(y)\sim y^{-\ell},\qquad\ell=m-2+\nu>k,

as y→0y\to0. Matching the two powers gives, for fixed RR as a→∞a\to\infty,

ak/2αa,R⟶Cm,R:=RkℓQ(R)+RQ′(R)ℓ−k,ak/2βa,R⟶0.a^{k/2}\alpha_{a,R}\longrightarrow C_{m,R}:=R^k\frac{\ell Q(R)+RQ'(R)}{\ell-k},\qquad a^{k/2}\beta_{a,R}\longrightarrow0.

The core tail Q(R)∼CmR−kQ(R)\sim C_mR^{-k} implies Cm,R→CmC_{m,R}\to C_m as R→∞R\to\infty. Thus the core selects the outer mode FmF_m with amplitude Cma−k/2C_ma^{-k/2}; the contribution of the faster mode GmG_m is negligible at fixed y>0y>0.

Size of the nonlinear correction. At the matching point, the field is already small: u(yR;a)=O(R−k)u(y_R;a)=O(R^{-k}). The cubic term is therefore smaller than the linear term by a factor O(R−2k)O(R^{-2k}). To check that this small correction does not accumulate as we move outwards, write the equation in integral form:

u(y;a)−La(y)=∫yRyGm(y,s)u(s;a)3 ds.u(y;a)-L_a(y)=\int_{y_R}^{y}G_m(y,s)u(s;a)^3\,\mathrm{d}s.

For yR≤s≤y≤y0y_R\le s\le y\le y_0, where y0>0y_0>0 is fixed and small, the linear Green function satisfies

∣Gm(y,s)∣≤Cy−ksk−1.\lvert G_m(y,s)\rvert\le Cy^{-k}s^{k-1}.

Using the matched decay u(s;a)=O(a−k/2s−k)u(s;a)=O(a^{-k/2}s^{-k}) in this integral gives

∣u(y;a)−La(y)∣≤CR−2ka−k/2y−k.\lvert u(y;a)-L_a(y)\rvert\le CR^{-2k}a^{-k/2}y^{-k}.

For sufficiently large RR, this estimate also makes the assumed decay self-consistent. The derivative is controlled in the same way at y=y0y=y_0.

Hence the nonlinear correction can be made arbitrarily small relative to the leading outer field by choosing RR large. Taking a→∞a\to\infty first and then R→∞R\to\infty, we obtain

ak/2(u(y0;a),uy(y0;a))⟶Cm(Fm(y0),Fm′(y0)).a^{k/2}\bigl(u(y_0;a),u_y(y_0;a)\bigr)\longrightarrow C_m\bigl(F_m(y_0),F'_m(y_0)\bigr).

Away from the origin and the light cone, ordinary continuous dependence propagates this convergence. Thus

ak/2u(⋅;a)⟶CmFma^{k/2}u(\mathord{\cdot};a)\longrightarrow C_mF_m

together with the first derivative on every closed subinterval of (0,1)(0,1).

Transfer to the light cone. Convergence at interior points does not by itself determine the fractional-power coefficient at the singular endpoint. For this last step, we use the endpoint coordinates constructed in Appendix A.

Fix y1<1y_1<1 sufficiently close to the light cone. Near the zero solution, the fractional-power coefficient is an analytic function B\mathcal{B} of the interior data (u(y1),uy(y1))(u(y_1),u_y(y_1)). Its linear term gives the fractional-power coefficient of the corresponding solution of the linearized equation. Since

(u(y1;a),uy(y1;a))=Cma−k/2(Fm(y1),Fm′(y1))+o(a−k/2),\bigl(u(y_1;a),u_y(y_1;a)\bigr)=C_ma^{-k/2}\bigl(F_m(y_1),F'_m(y_1)\bigr)+o(a^{-k/2}),

we need only the fractional-power coefficient of FmF_m. By (2.8), this is 2ρKm2^\rho K_m. Consequently,

Bm(a)=2ρCmKma−k/2+o(a−k/2),B_m(a)=2^\rho C_mK_ma^{-k/2}+o(a^{-k/2}),

which proves (2.9).

Taylor coefficient estimates

We establish the eventual coefficient signs and the divergence of f(m−1)f^{(m-1)} used in Section 3. Throughout, ρ=m−32\rho=m-\frac{3}{2}, and AA denotes the parameter of the explicit branch under consideration.

The recurrence and a comparison identity

Substituting the Taylor series

f(ξ)=∑n≥0fnξnf(\xi)=\sum_{n\ge0}f_n\xi^n

into the Heun equation (23) gives the recurrence

Dnfn+1=Tnfn−Cnfn−1,n≥1,(36)D_nf_{n+1}=T_nf_n-C_nf_{n-1},\qquad n\ge1, \tag*{(36)}

where

Dn=A(n+1)(n+m+1),D_n=A(n+1)(n+m+1),
Tn=(A+1)n2+(m+5A+12)n−m−3A−12,T_n=(A+1)n^2+\left(m+\frac{5A+1}{2}\right)n-m-\frac{3A-1}{2},
Cn=(n+3)(n−2).(37)C_n=(n+3)(n-2). \tag*{(37)}

The starting coefficients are

f0=1,f1=−3A+2m−12A(m+1),f_0=1,\qquad f_1=-\frac{3A+2m-1}{2A(m+1)},

and

f2=(A−1)(2m−1−3A)2A2(m+1)(m+2),f3=15A+2m+116A(m+3)f2.f_2=\frac{(A-1)(2m-1-3A)}{2A^2(m+1)(m+2)},\qquad f_3=\frac{15A+2m+11}{6A(m+3)}f_2.

Let gng_n denote either fnf_n or −fn-f_n. It satisfies the same recurrence. To obtain lower bounds on gn/gn−1g_n/g_{n-1}, choose a positive sequence rnr_n and set

en=gn−rngn−1,En=Tn−Cnrn−Dnrn+1.(38)e_n=g_n-r_ng_{n-1},\qquad E_n=T_n-\frac{C_n}{r_n}-D_nr_{n+1}. \tag*{(38)}

Then (C.1) gives the exact identity

Dnen+1=Cnrnen+Engn.(39)D_ne_{n+1}=\frac{C_n}{r_n}e_n+E_ng_n. \tag*{(39)}

Suppose that for some N≥3N\ge3,

gN−1>0,eN>0,En>0(n≥N).g_{N-1}>0,\qquad e_N>0,\qquad E_n>0\quad(n\ge N).

Then gN=rNgN−1+eN>0g_N=r_Ng_{N-1}+e_N>0, and (C.4) gives eN+1>0e_{N+1}>0, since Cn,Dn>0C_n,D_n>0 for n≥3n\ge3. Induction therefore yields

gn>rngn−1>0(n≥N).(40)g_n>r_ng_{n-1}>0\qquad(n\ge N). \tag*{(40)}

Iterating this inequality gives

gn>gN−1∏j=Nnrj.g_n>g_{N-1}\prod_{j=N}^{n}r_j.

If

rn=1−αn+O(n−2),r_n=1-\frac{\alpha}{n}+O(n^{-2}),

then taking logarithms and summing gives

log⁡∏j=Nnrj=−αlog⁡n+O(1).\log\prod_{j=N}^{n}r_j=-\alpha\log n+O(1).

Consequently,

gn>cmn−α(n≥N)g_n > c_m n^{-\alpha} \qquad(n \ge N)

for some cm>0c_m > 0. Thus it suffices to verify the starting inequalities gN−1>0g_{N-1} > 0 and eN>0e_N > 0, together with En>0E_n > 0 for n≥Nn \ge N. Here cm>0c_m > 0 is independent of nn.

We treat the branches separately below. In each case, the comparison identity determines the eventual coefficient sign. For the negative tails, the resulting polynomial lower bound also proves endpoint divergence. For the positive tail, we first exclude analyticity at ξ=1\xi= 1 by comparing coefficient ratios, and then deduce divergence from the endpoint expansion.

Negative coefficient tails

The plus branch for m≥5m \ge5. Set gn=−fng_n = -f_n and choose

rn=n−2n+m−52,n≥3.r_n = \frac{n-2}{n+m-\frac{5}{2}}, \qquad n \ge3.

On this branch A>2m−3A > 2m - 3, so g2,g3>0g_2,g_3 > 0. Moreover,

g3g2−r3=18Am−21A+4m2+24m+116A(m+3)(2m+1)>0.\frac{g_3}{g_2} - r_3 = \frac{18Am - 21A + 4m^2 + 24m + 11}{6A(m+3)(2m+1)} > 0.

Thus the comparison starts at n=3n = 3.

A direct calculation gives

En=Smn+Qm2(2n+2m−3),E_n = \frac{S_m n + Q_m}{2(2n + 2m - 3)},

where

Sm=A(10m−17)−16m+32,Qm=A(13−2m)−16m2+56m−48.S_m = A(10m - 17) - 16m + 32, \qquad Q_m = A(13 - 2m) - 16m^2 + 56m - 48.

Using A>2m−3A > 2m - 3, we find

Sm>20m2−80m+83>0,3Sm+Qm>40m2−152m+162>0.S_m > 20m^2 - 80m + 83 > 0, \qquad3S_m + Q_m > 40m^2 - 152m + 162 > 0.

Hence

Smn+Qm=(n−3)Sm+(3Sm+Qm)>0(n≥3),S_m n + Q_m = (n - 3)S_m + (3S_m + Q_m) > 0 \qquad(n \ge3),

so En>0E_n > 0 for all n≥3n \ge3. The comparison identity therefore gives

gn>rngn−1>0,n≥3.g_n > r_n g_{n-1} > 0, \qquad n \ge3.

Since

rn=1−ρ+1n+O(n−2),r_n = 1 - \frac{\rho+ 1}{n} + O(n^{-2}),

it follows that

−fn≥cmn−ρ−1(41)-f_n \ge c_m n^{-\rho- 1} \tag*{(41)}

for some cm>0c_m > 0 and all $n \ge 3.

The minus branch for m=6,7,8m=6,7,8. Set gn=−fng_n=-f_n and choose

rn=n−16n−16+m−12,n≥66.r_n=\frac{n-16}{n-16+m-\frac{1}{2}},\qquad n\ge66.

The comparison starts later because f2,f3>0f_2,f_3>0. For each m=6,7,8m=6,7,8, we use exact arithmetic in Maple to verify the AI-suggested starting inequalities

g65>0,g66g65−r66>11000,(42)g_{65}>0,\qquad\frac{g_{66}}{g_{65}}-r_{66}>\frac{1}{1000}, \tag*{(42)}

so e66>0e_{66}>0. The same calculation shows that

2(n−16)(2n+2m−31)En2(n-16)(2n+2m-31)E_n

is a polynomial in n−66n-66 with strictly positive coefficients. Hence En>0E_n>0 for all n≥66n\ge66, and the comparison identity gives

gn>rngn−1>0,n≥66.g_n>r_ng_{n-1}>0,\qquad n\ge66.

Since

rn=1−ρ+1n+O(n−2),r_n=1-\frac{\rho+1}{n}+O(n^{-2}),

we again obtain (C.6).

Nonvanishing of CC and endpoint divergence. In all the negative-tail cases, the eventual coefficient sign is σ=−1\sigma=-1. Moreover,

n!(n−m+1)!∼nm−1,m−2−ρ=−12.\frac{n!}{(n-m+1)!}\sim n^{m-1},\qquad m-2-\rho=-\frac{1}{2}.

It follows from (C.6) that

n!(n−m+1)!fn≤−c~mn−1/2\frac{n!}{(n-m+1)!}f_n\le-\widetilde c_m n^{-1/2}

for all sufficiently large nn, with c~m>0\widetilde c_m>0. Consequently, the tail of the differentiated Taylor series

f(m−1)(ξ)=∑n≥m−1n!(n−m+1)!fnξn−m+1f^{(m-1)}(\xi)=\sum_{n\ge m-1}\frac{n!}{(n-m+1)!}f_n\xi^{n-m+1}

tends to mathord−∞mathord{-}\infty as ξ→1−\xi\to1^-. The finitely many preceding terms remain bounded, so

f(m−1)(ξ)⟶−∞as ξ→1−.f^{(m-1)}(\xi)\longrightarrow-\infty\qquad\text{as }\xi\to1^-.

The endpoint expansion (3.8) therefore implies C≠0C\ne0.

The minus branch for m≥9m\ge9

Set gn=fng_n=f_n, write

Rm=23(m−2)(m−1),A=4m−5−Rm3,R_m=2\sqrt{3(m-2)(m-1)},\qquad A=\frac{4m-5-R_m}{3},

and choose

rn=n+1A(n+52)(1+13n2),n≥3.(43)r_n=\frac{n+1}{A\left(n+\frac{5}{2}\right)}\left(1+\frac{1}{3n^2}\right),\qquad n\ge3. \tag*{(43)}

We first prove positivity of the coefficients and then establish nonvanishing of CC and endpoint divergence.

Starting the comparison. Here g2>0g_2 > 0, and

g3g2−r3=5(297A−50m−51)594A(m+3).\frac{g_3}{g_2} - r_3 = \frac{5(297A - 50m - 51)}{594A(m+3)}.

Substituting the formula for AA gives 297A−50m−51>0297A - 50m - 51 > 0 for m≥9m \ge9. Thus g3/g2>r3g_3/g_2 > r_3, initializing the comparison at n=3n = 3.

Positivity of the gap. We first establish the lower bound

A>L(m):=2213+423(m−9),m≥9.A > L(m) := \frac{22}{13} + \frac{4}{23}(m - 9), \qquad m \ge9.

Indeed, direct calculation gives A(9)=(31−442)/3>22/13A(9) = (31 - 4\sqrt{42})/3 > 22/13 and

A′(m)=23(2−6m−9Rm)>423.A'(m) = \frac{2}{3}\left(2 - \frac{6m - 9}{R_m}\right) > \frac{4}{23}.

Integration yields the claim.

For (C.8), direct calculation gives

En=Pm(n;A)6(n+1)(2n+7)(3n2+1).E_n = \frac{P_m(n; A)}{6(n+1)(2n+7)(3n^2+1)}.

where

Pm(n;A)=(174A−18m−111)n4+(1137A−228m−144)n3+(1860A−228m−70)n2+(24A−76m−48)n−63A−74m−11.P_m(n; A) = (174A - 18m - 111)n^4 + (1137A - 228m - 144)n^3 + (1860A - 228m - 70)n^2 + (24A - 76m - 48)n - 63A - 74m - 11.

Since

∂APm(n;A)=3(n+3)(2n+7)(29n2+n−1)>0(n≥3),\partial_A P_m(n; A) = 3(n+3)(2n+7)(29n^2+n-1) > 0 \qquad(n \ge3),

it suffices to substitute A=L(m)A = L(m). At this value,

Pm(n;L(m))=J(n)13+2(m−9)23K(n−3),P_m(n; L(m)) = \frac{J(n)}{13} + \frac{2(m-9)}{23}K(n-3),

where

J(n)=279n4−3534n3+13334n2−8988n−10187J(n) = 279n^4 - 3534n^3 + 13334n^2 - 8988n - 10187

and

K(j)=141j4+1344j3+5580j2+11594j+8452.K(j) = 141j^4 + 1344j^3 + 5580j^2 + 11594j + 8452.

All coefficients of KK are positive, as are J(3)=10036J(3) = 10036, J(4)=12453J(4) = 12453, J(5)=10848J(5) = 10848, and

J(6+j)=279j4+3162j3+9986j2+10404j+14149(j≥0).J(6+j) = 279j^4 + 3162j^3 + 9986j^2 + 10404j + 14149 \qquad(j \ge0).

Hence En>0E_n > 0 for every integer n≥3n \ge3. The comparison identity therefore gives

gn>rngn−1>0,n≥3.g_n > r_n g_{n-1} > 0, \qquad n \ge3.

Since gn=fng_n = f_n, the eventual coefficient sign is σ=1\sigma= 1. Moreover, the comparison product is asymptotic to a positive constant times A−nn−3/2A^{-n}n^{-3/2}, so

gn>cmA−nn−3/2n≥3,g_n > c_m A^{-n}n^{-3/2} \qquad n \ge3,

where cm>0c_m > 0 is independent of nn.

Nonvanishing of CC and endpoint divergence. Returning to fn=gnf_n = g_n, we have

fn>cmA−nn−3/2.f_n > c_m A^{-n} n^{-3/2}.

This bound does not by itself force divergence at ξ=1\xi= 1.

We prove that C≠0C \ne0 by contradiction. Suppose that C=0C = 0. Then ff is analytic at ξ=1\xi= 1, so its Taylor series at 0 has radius at least AA. The lower bound above shows that the radius is exactly AA. Thus ff is singular at ξ=A\xi= A, where the local exponents are 0 and 12\frac{1}{2}, and hence has a nonzero square-root component there. Standard coefficient asymptotics give a ratio expansion whose n−2n^{-2} coefficient is determined by the recurrence:

fnfn−1=1A(1−32n+2m−4−3A4(A−1)n2+O(n−3)).\frac{f_n}{f_{n-1}} = \frac{1}{A}\left(1-\frac{3}{2n}+\frac{2m-4-3A}{4(A-1)n^2}+O(n^{-3})\right).

On the other hand,

rn=1A(1−32n+4912n2+O(n−3)).r_n = \frac{1}{A}\left(1-\frac{3}{2n}+\frac{49}{12n^2}+O(n^{-3})\right).

The n−2n^{-2} coefficient in the ratio expansion is strictly smaller than that in rnr_n, since

4912−2m−4−3A4(A−1)=58A−6m−3712(A−1)>0,\frac{49}{12}-\frac{2m-4-3A}{4(A-1)}=\frac{58A-6m-37}{12(A-1)}>0,

where the last inequality follows from A>L(m)A > L(m). Thus fn/fn−1<rnf_n/f_{n-1} < r_n for all sufficiently large nn, contrary to the comparison inequality. Hence C≠0C \ne0.

By (27) and eventual positivity of the Taylor coefficients,

f(m−1)(ξ)⟶+∞asξ→1−.f^{(m-1)}(\xi) \longrightarrow+\infty\qquad\text{as} \qquad\xi\to1^{-}.

This completes the coefficient and divergence estimates used in § 3.

Acknowledgements

AI assistance was used in Appendix C to explore recurrence relations, choose comparison ratios, initialize the induction arguments, and perform exact-arithmetic calculations. The author takes full responsibility for the mathematical content and final text.

References

  1. [1]P. Bizoń, I. Głogić, and A. Wasserman, Three self-similar solutions of Yang–Mills equations in high odd dimensions, Phys. Rev. D 114, 064021 (2026).
  2. [2]P. Breitenlohner, P. Forgács, and D. Maison, Static Spherically Symmetric Solutions of the Einstein–Yang–Mills Equations, Commun. Math. Phys. 163 (1994), 141–172.DOI
  3. [3]P. Bizoń and R. Kluszczyński, Existence of a third self-similar solution of the Yang–Mills equation in high odd dimensions, arXiv:2610.36784 (2026).arxiv.org/abs/2610.36784
  4. [4]NIST Digital Library of Mathematical Functions, https://dlmf.nist.gov/, Eq. (15.8.4).

Paper details

vlnc:W_gwcwwzx2
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 4

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

AI assistance

View statement
Looking for related papers…
Contents