Flat (2,3,5)-Distributions and Chazy's Equations

n the geometry of generic 2-plane fields on 5-manifolds, the local equivalence problem was solved by Cartan who also constructed the fundamental curvature invariant. For generic 2-plane fields or (2,3,5)-distributions determined by a single function of the form F(q), the vanishing condition for the...

Повний опис

Збережено в:
Бібліографічні деталі
Опубліковано в: :Symmetry, Integrability and Geometry: Methods and Applications
Дата:2016
Автор: Randall, M.
Формат: Стаття
Мова:Англійська
Опубліковано: Інститут математики НАН України 2016
Онлайн доступ:https://nasplib.isofts.kiev.ua/handle/123456789/147724
Теги: Додати тег
Немає тегів, Будьте першим, хто поставить тег для цього запису!
Назва журналу:Digital Library of Periodicals of National Academy of Sciences of Ukraine
Цитувати:Flat (2,3,5)-Distributions and Chazy's Equations / M. Randall // Symmetry, Integrability and Geometry: Methods and Applications. — 2016. — Т. 12. — Бібліогр.: 25 назв. — англ.

Репозитарії

Digital Library of Periodicals of National Academy of Sciences of Ukraine
_version_ 1859678191749169152
author Randall, M.
author_facet Randall, M.
citation_txt Flat (2,3,5)-Distributions and Chazy's Equations / M. Randall // Symmetry, Integrability and Geometry: Methods and Applications. — 2016. — Т. 12. — Бібліогр.: 25 назв. — англ.
collection DSpace DC
container_title Symmetry, Integrability and Geometry: Methods and Applications
description n the geometry of generic 2-plane fields on 5-manifolds, the local equivalence problem was solved by Cartan who also constructed the fundamental curvature invariant. For generic 2-plane fields or (2,3,5)-distributions determined by a single function of the form F(q), the vanishing condition for the curvature invariant is given by a 6th order nonlinear ODE. Furthermore, An and Nurowski showed that this ODE is the Legendre transform of the 7th order nonlinear ODE described in Dunajski and Sokolov. We show that the 6th order ODE can be reduced to a 3rd order nonlinear ODE that is a generalised Chazy equation. The 7th order ODE can similarly be reduced to another generalised Chazy equation, which has its Chazy parameter given by the reciprocal of the former. As a consequence of solving the related generalised Chazy equations, we obtain additional examples of flat (2,3,5)-distributions not of the form F(q)=qm. We also give 4-dimensional split signature metrics where their twistor distributions via the An-Nurowski construction have split G₂ as their group of symmetries.
first_indexed 2025-11-30T17:03:05Z
format Article
fulltext Symmetry, Integrability and Geometry: Methods and Applications SIGMA 12 (2016), 029, 28 pages Flat (2, 3, 5)-Distributions and Chazy’s Equations Matthew RANDALL Department of Mathematics and Statistics, Faculty of Science, Masaryk University, Kotlářská 2, 611 37 Brno, Czech Republic E-mail: randallm@math.muni.cz Received September 23, 2015, in final form March 14, 2016; Published online March 18, 2016 http://dx.doi.org/10.3842/SIGMA.2016.029 Abstract. In the geometry of generic 2-plane fields on 5-manifolds, the local equivalence problem was solved by Cartan who also constructed the fundamental curvature invariant. For generic 2-plane fields or (2, 3, 5)-distributions determined by a single function of the form F (q), the vanishing condition for the curvature invariant is given by a 6th order non- linear ODE. Furthermore, An and Nurowski showed that this ODE is the Legendre transform of the 7th order nonlinear ODE described in Dunajski and Sokolov. We show that the 6th order ODE can be reduced to a 3rd order nonlinear ODE that is a generalised Chazy equation. The 7th order ODE can similarly be reduced to another generalised Chazy equation, which has its Chazy parameter given by the reciprocal of the former. As a consequence of solving the related generalised Chazy equations, we obtain additional examples of flat (2, 3, 5)- distributions not of the form F (q) = qm. We also give 4-dimensional split signature metrics where their twistor distributions via the An–Nurowski construction have split G2 as their group of symmetries. Key words: generic rank two distribution in dimension five; conformal geometry; Chazy’s equations 2010 Mathematics Subject Classification: 58A30; 53A30; 34A05; 34A34 1 Introduction The following 6th order nonlinear ODE 10F (6)F ′′ 3 − 80F ′′ 2 F (3)F (5) − 51F ′′ 2 F (4)2 + 336F ′′F (3)2F (4) − 224F (3)4 = 0, (1.1) arises in [6, Corollary 2.1] in the study of generic 2-plane fields on 5-manifolds. The generi- city condition here means F ′′(q) 6= 0 in (1.1). This ODE arises as the integrability condition for generic 2-plane fields on 5 manifolds determined by a function of a single variable of the form F (q). A generic 2-plane field D on a 5-manifold M is a maximally non-integrable rank 2 distribution. For further details, see [6, 18, 24, 25]. This determines a filtration of the tangent bundle given by D ⊂ [D,D] ⊂ [[D,D],D] = TM. The distribution [D,D] has rank 3 while the full tangent space TM has rank 5, hence such a geo- metry is also known as a (2, 3, 5)-distribution. Let Mxyzpq denote the 5-dimensional manifold with local coordinates given by (x, y, z, p, q). The generic 2-plane field or rank 2 distribution determined by a function F (q) of a single variable with F ′′(q) 6= 0 is given by D = span{∂q, ∂x + p∂y + q∂p + F (q)∂z}. The fundamental Cartan curvature invariant of this distribution is computed in [6] and is found to be the term in the left hand side of (1.1). It is known that equation (1.1) vanishes when mailto:randallm@math.muni.cz http://dx.doi.org/10.3842/SIGMA.2016.029 2 M. Randall F (q) = qm and m ∈ { −1, 13 , 2 3 , 2 } . In these cases, the vanishing of the fundamental curvature invariant associated to the distribution D on Mxyzpq implies that the group of local symmetries is the maximal possible given by the split real form of G2. This is the result of [6, Corollary 2.1]. The authors of [6] call such generic 2-plane fields with vanishing Cartan curvature invariant symmetric and it is known that such symmetric or flat distributions are locally equivalent to the flat model F (q) = q2. Nonetheless we are interested in the general solution to (1.1) and it turns out that the ODE can be solved completely and is related to the generalised Chazy equation. To see this, let E(q) = F ′′(q) so that the ODE becomes 4th order: 10E(4)E3 − 80E2E′E′′′ − 51E2E′′ 2 + 336EE′ 2 E′′ − 224E′ 4 = 0. Working locally on an open set of Mxyzpq, we may assume that E(q) is positive on that open set. Making the substitution E(q) = eG(q) (if E(q) < 0, take E(q) = −eG(q) instead) gives e4G(q) ( 10G′′′′ − 40G′′′G′ − 21(G′′)2 + 54G′′(G′)2 − 9(G′)4 ) = 0 and taking G′(q) = j(q) gives a 3rd order ODE 10j′′′ − 40j′′j − 21(j′)2 + 54j′j2 − 9j4 = 0. Rescaling the ODE by taking j(q) = I(q) 2 , we can put it into the normal form for the generalised Chazy equation (see [14]) I ′′′ − 2I ′′I + 3(I ′)2 − 4 36− ( 2 3 )2 (6I ′ − I2)2 = 0 (1.2) with the Chazy parameter k2 = ( 2 3 )2 = 4 9 . The generalised Chazy equation can be solved com- pletely and the solutions give us new families of flat (2, 3, 5)-distributions that are not of the form F (q) = qm. In this article we first review the solutions to Chazy’s equations in Sections 2 and 3. In Section 4 we discuss the relationship between (1.1) and a 7th order ODE studied by Dunajski and Sokolov in [16] and also exhibit a Legendre transform that relates equation (1.2) to another generalised Chazy equation with the Chazy parameter given by k2 = ( 3 2 )2 = 9 4 . We compute the solutions to (1.2) in Section 5 and present examples of flat (2, 3, 5)-distributions in Section 6 using Nurowski’s metric. These examples are all explicit. In [5], the authors associated to split signature conformal structures on a 4-manifold a circle bundle with the natural structure of a (2, 3, 5)-distribution. This construction encapsulates the configuration space of 2 surfaces rolling along one another without slipping and twisting. The authors in [5] then found new exam- ples of flat (2, 3, 5)-distributions that arise from rolling bodies, prompting further search in [8]. The solutions to (1.2) give examples of 4-dimensional split signature metrics that have their An–Nurowski twistor distributions having split G2 as its group of symmetries and we exhibit them in Section 7. Let us recall some facts about Chazy’s equation and its generalised version. 2 Chazy’s equation The study of Chazy’s equation is a very rich subject and has received alot of attention because of its connection to other diverse fields such as integrable systems and modular forms. See for instance [2, 10, 11, 14]. We will review here some facts about Chazy’s equation we need for the paper. Chazy [12, 13] studied the nonlinear 3rd order ODE y′′′(x)− 2y(x)y′′(x) + 3(y′(x))2 = 0 (2.1) Flat (2, 3, 5)-Distributions and Chazy’s Equations 3 in the context of investigating its Painlevé property. Solutions to equation (2.1) turn out to depend on hypergeometric functions. For further details, see [2] or [14]. Treat x as a dependent variable of s so that x(s) = z2(s) z1(s) , where z1(s), z2(s) are linearly independent solutions to the second order hypergeometric diffe- rential equation s(1− s)z′′ + (c− (a+ b+ 1)s)z′ − abz = 0. (2.2) Here a, b, c are constants to be determined. The general solution to this ODE (2.2) is given by hypergeometric functions z(s) = µ 2F1(a, b; c; s) + ν 2F1(a− c+ 1, b− c+ 1; 2− c; s)s1−c. Here µ, ν are constants. A computation gives dx = z1ż2 − z2ż1 (z1)2 ds, where dot denotes derivative with respect to s. We deduce that d dx = (z1) 2 z1ż2 − z2ż1 d ds . Applying the derivative to Chazy’s solution for y given by y = 6 d dx log z1 = 6z1ż1 z1ż2 − z2ż1 , (2.3) we find that (2.1) is satisfied precisely when (a, b, c) is one of( 1 12 , 1 12 , 1 2 ) , ( 1 12 , 1 12 , 2 3 ) , ( 1 6 , 1 6 , 2 3 ) , provided both a and b are non-zero. The equations (2.2) for the first two values of (a, b, c) are related by a linear transformation of the form s 7→ 1− s, while the solutions for the second and third values are related by a quadratic transformation (see [23, equation (2)]) given by 2F1 ( 1 6 , 1 6 ; 2 3 ; s ) = 2F1 ( 1 12 , 1 12 ; 2 3 ; 4s(1− s) ) . The general solution to (2.1) thus depend on hypergeometric functions. If either one of a or b is zero (say b = 0), solutions to (2.1) can be easily and explicitly described. The solutions to (2.2) with b = 0 are given by z(s) = µ+ ν 2F1(1− c, a+ 1− c; 2− c; s)s1−c = µ− νπ(c− 1) sin(πc) P (1−c, a−c) c−1 (1− 2s)s1−c, where P (a1,b1) n is the Jacobi polynomial. Taking z1(s) = ν 2F1(1− c, a+ 1− c; 2− c; s)s1−c and z2(s) = µ, a computation shows that x(s) = µ ν 2F1(1− c, a+ 1− c; 2− c; s)s1−c 4 M. Randall and y(x(s)) = −6 ν µ 2F1(1− c, a+ 1− c; 2− c; s)s1−c. Switching back to the original independent variable x, this gives y(x) = −6 x as one solution to (2.1). This solution is invariant under translations of the form x 7→ x + C. In [12, 13], Chazy also observed that y = − 6 x+ C − B (x+ C)2 is a solution to (2.1). It is well-known that Chazy’s equation and its generalised version can be rewritten as a first order system. This provides different parametrisations of y, in addition to the solution (2.3) originally given by Chazy. This will be discussed in Section 5. The method discussed here can also be applied to the generalised Chazy equation. 3 Generalised Chazy equations The generalised Chazy equation is given by y′′′(x)− 2y′′(x)y(x) + 3(y′(x))2 − 4 36− k2 (6y′(x)− y(x)2)2 = 0 (3.1) for k 6= ±6. We have the following: Proposition 3.1. Let x(s) = z2(s) z1(s) where z1(s), z2(s) are linearly independent solutions to the hypergeometric differential equation (2.2) where (a, b, c) is one of( k − 6 12k , k + 6 12k , 1 2 ) , ( k − 6 6k , k + 6 6k , 2 3 ) , ( k − 6 12k , k + 6 12k , 2 3 ) . Then y(x(s)) = 6 d dx log z1 = 6z1ż1 z1ż2 − z2ż1 satisfies equation (3.1). Proof. Analogous to solving Chazy’s equation (2.1), we find that the generalised equation (3.1) holds provided 6ab(z1) 8 ( 6((a− b)k − 6(a+ b))((a− b)k + 6(a+ b))s2 + ( (24ab− 12(a+ b)c+ 5(a+ b) + (2c− 1))k2 + 432(a+ b)c− 180(a+ b)− 72c+ 36 ) s + (k − 6)(k + 6)(2c− 1)(3c− 2) ) = 0. For a, b 6= 0, solving the system of equations 6((a− b)k − 6(a+ b))((a− b)k + 6(a+ b)) = 0, (24ab− 12(a+ b)c+ 5(a+ b) + (2c− 1))k2 + 432(a+ b)c− 180(a+ b)− 72c+ 36 = 0, (k − 6)(k + 6)(2c− 1)(3c− 2) = 0, gives the list of (a, b, c) as above. We exclude the case where (a, b, c) = ( 0, 0, 12 ) . Note that interchanging a and b gives the same solution so that the full list is symmetric in a and b. � Flat (2, 3, 5)-Distributions and Chazy’s Equations 5 When either a or b is zero, we again get y(x) = − 6 x as a solution. In [12, 13], Chazy noted that y = k − 6 2(x+ C) − k + 6 2(x+B) (3.2) is also a solution to (3.1). As a corollary to Proposition 3.1, we have Corollary 3.2. Let q(s) = z2(s) z1(s) where z1(s), z2(s) are linearly independent solutions to the hypergeometric differential equation (2.2) with (a, b, c) one of( −2 3 , 5 6 , 1 2 ) , ( −4 3 , 5 3 , 2 3 ) , ( −2 3 , 5 6 , 2 3 ) . Then I(q(s)) = 6 d dq log z1 = 6z1ż1 z1ż2 − z2ż1 satisfies equation (1.2). A Painlevé type analysis of equation (3.1) as done in [14] shows that the leading orders for analytic solutions to (3.1) occur at −6, −3 + k 2 or −3− k 2 . This corresponds to solutions of (3.1) given by y(x) = −6 x , y(x) = −3 + k 2 x , y(x) = −3− k 2 x . These solutions are invariant under translations of the form x 7→ x+ C. In the case of k = ±2 3 obtained in (1.2), we have I(q) = −6 q , I(q) = −10 3q , I(q) = − 8 3q . Along with the zero solution I(q) = 0, these solutions correspond respectively (modulo constants of integration) to the well-known explicit solutions to (1.1): F (q) = q−1, F (q) = q 1 3 , F (q) = q 2 3 , F (q) = q2. For these functions of a single variable q the associated (2, 3, 5)-distributions have vanishing Cartan invariant and therefore have G2 as their local symmetry. 4 Relationship to ODE studied by Dunajski and Sokolov For the function y = y(t), the 7th order nonlinear ODE studied in [16] is given by 10 ( y(3) )3 y(7) − 70 ( y(3) )2 y(4)y(6) − 49 ( y(3) )2( y(5) )2 + 280 ( y(3) )( y(4) )2 y(5) − 175 ( y(4) )4 = 0. (4.1) This is the unique 7th order ODE admitting the submaximal contact symmetry group of dimen- sion ten (see [16, 21]) and its relationship to equation (1.1) was originally explored in [6]. It is instructive to consider the 6th order ODE (for the Legendre transformation later on): 10 ( H(2) )3 H(6) − 70 ( H(2) )2 H(3)H(5) − 49 ( H(2) )2( H(4) )2 + 280 ( H(2) )( H(3) )2 H(4) − 175 ( H(3) )4 = 0 (4.2) 6 M. Randall with H(t) = y′(t). Let us show that this ODE can be reduced to a generalised Chazy equation. Again working locally in an open set where y′′′(t) is non-zero, and assuming y′′′(t) to be positive, we can make the substitution ep(t) = y(3) to get ep(t) ( 10p(4) − 30p′p′′′ − 19(p′′)2 + 32(p′)2p′′ − 4(p′)4 ) = 0. (4.3) We note that this 4th order ODE historically appears in [9, Section XII, formula (12)], where it first arises as the obstruction to integrability for (2, 3, 5)-distributions of the form DF (q). This will be made clear below once we show that (4.2) is the Legendre transform of (1.1) [6] and we will discuss this further in Section 6. Thus, for v(t) = p′(t), we obtain the third order ODE 10v′′′ − 30vv′′ − 19(v′)2 + 32v′v2 − 4v4 = 0. (4.4) Rescaling v(t) by u(t) = 3 2v(t), we put (4.4) into the normal form u′′′ − 2u′′u+ 3(u′)2 − 4 36− ( 3 2 )2 (6u′ − u2)2 = 0. (4.5) We therefore see that the ODE that Dunajski and Sokolov study in [16] reduces to a generalised Chazy equation (4.5) with parameter k′ = ±3 2 , related to the generalised Chazy equation (1.2) just by taking the reciprocals (k′)2 = 1 k2 of the corresponding parameters. Let t(s) = w2(s) w1(s) where w1(s), w2(s) are linearly independent solutions to the hypergeometric differential equation (2.2) with (a, b, c) one of( −1 4 , 5 12 , 1 2 ) , ( −1 4 , 5 12 , 2 3 ) , ( −1 2 , 5 6 , 2 3 ) . The solution to (4.5) is then given by u = 6 d dt logw1. A similar leading order analysis as before shows that the leading orders occur at −6, −9 8 , −15 8 . This corresponds to solutions of (4.5) given by u(t) = −6 t , u(t) = − 9 4t , u(t) = −15 4t . Along with the zero solution u(t) = 0, these correspond respectively (modulo constants of integration) to solutions of (4.2) given by H(t) = t−2, H(t) = t 1 2 , H(t) = t− 1 2 , H(t) = t2. In [6, Proposition 2.2], it is shown that a Legendre transformation takes (1.1) to (4.1). Hence we may hypothesise that amongst all 3rd order generalised Chazy equations, only those with the parameters k′ = ±3 2 , k = ±2 3 have in addition solutions that can be obtained from the dual equation via a Legendre transform. Proposition 4.1 ([6, Proposition 2.2]). Consider the Legendre transformation F (q) +H(t) = qt. Then F (q) satisfies the ODE (1.1) iff H(t) satisfies the ODE (4.2). Flat (2, 3, 5)-Distributions and Chazy’s Equations 7 Proof. Applying the exterior derivative to the relation gives (F ′ − t)dq + (H ′ − q)dt = 0, so that we take F ′ = t, H ′ = q and applying d dq = 1 H ′′ d dt we obtain F ′′ = 1 H′′ , F (3) = − H(3) (H′′)3 , etc. A computation shows that the 6th order ODE (1.1) holds for F iff (4.2) holds for H. � In light of the solutions obtained by solving the generalised Chazy equations, we can pass to F (q) = ∫∫ e ∫ I(q) 2 dqdqdq, (4.6) where q = z2(s) z1(s) and I(q) = 6 d dq log z1 are given in Corollary 3.2. This gives F (q) = ∫∫ (z1) 3dqdq. Similarly, for the dual equation (4.2) under the Legendre transform we pass to H(t) = ∫∫ e ∫ 2u(t) 3 dtdtdt, where t = w2(s) w1(s) and u(t) = 6 d dt logw1 are solutions to (4.5). This gives H(t) = ∫∫ (w1) 4dtdt. We have Lemma 4.2. There exists a Legendre transformation between Chazy’s solutions of (1.2) and (4.5) given by taking w1(s) = z − 3 4 1 , w2(s) = (z1) − 3 4 ∫ (z1)(ż2z1 − ż1z2)ds. This defines a mapping q = z2(s) z1(s) 7→ t = w2(s) w1(s) = ∫ z1(ż2z1 − ż1z2)ds. If I(q) = 6 d dq log z1 solves (1.2) where z1(s) and z2(s) are given in Corollary 3.2, then u(t) = 6 d dt logw1 solves the dual ODE (4.5). Consequently, if F (q) = ∫∫ (z1) 3dqdq solves (1.1), then H(t) = ∫∫ (w1) 4dtdt = ∫∫ (z1) −2(ż2z1 − ż1z2)ds z1(ż2z1 − ż1z2)ds solves the 6th order ODE (4.2). For the converse, the Legendre transform is given by z1(s) = w − 4 3 1 , z2(s) = (w1) − 4 3 ∫ (w1) 2(ẇ2w1 − ẇ1w2)ds. 8 M. Randall This sends t = w2(s) w1(s) 7→ q = z2(s) z1(s) = ∫ (w1) 2(ẇ2w1 − ẇ1w2)ds. In particular, if u(t) = 6 d dt logw1 solves the dual ODE (4.5), then I(q) = 6 d dq log z1 solves (1.2). Hence, if H(t) = ∫∫ (w1) 4dtdt solves the 6th order ODE (4.2), then F (q) = ∫∫ (z1) 3dqdq = ∫∫ (w1) −2(ẇ2w1 − ẇ1w2)ds (w1) 2(ẇ2w1 − ẇ1w2)ds solves (1.1). Proof. We observe that as a consequence of Chazy’s solutions, the Legendre transform in Proposition 4.1 gives w2 w1 = t = F ′ = ∫ (z1) 3dq = ∫ z1(ż2z1 − ż1z2)ds, z2 z1 = q = H ′ = ∫ (w1) 4dt = ∫ (w1) 2(ẇ2w1 − ẇ1w2)ds, and therefore ẇ2w1 − ẇ1w2 (w1)2 = z1(ż2z1 − ż1z2), ż2z1 − ż1z2 (z1)2 = (w1) 2(ẇ2w1 − ẇ1w2). Together this yields (z1) 3 = (w1) −4, from which we deduce w1 = z − 3 4 1 and w2 = w1 ∫ z1(ż2z1 − ż1z2)ds = z − 3 4 1 ∫ z1(ż2z1 − ż1z2)ds. For the converse, we find z1 = w − 4 3 1 and z2 = z1 ∫ (w1) 2(ẇ2w1 − ẇ1w2)ds = w − 4 3 1 ∫ (w1) 2(ẇ2w1 − ẇ1w2)ds. The rest follows from a routine computation. � In [16, formula (8)], a family of solutions to (4.1) is found to be given by the algebraic curve (y + f(t))2 = (t− a)(t− b)3, with a 6= b, and f(t) a quadratic. This gives y = ± √ (t− a)(t− b)3 − f(t). We obtain a solution to (4.1) with y(3) = ±3 8 (t− b)6(a− b)3 (y + f)5 = ± 3(a− b)3 8(t− a)2(y + f) . We find that for this solution, it yields u(t) = 3 4 5b+ 3a− 8t (t− a)(t− b) = − 15 4(t− a) − 9 4(t− b) as a solution to the generalized Chazy’s equation with parameter k2 = 9 4 . This corresponds to the solution given by Chazy in (3.2). It will be interesting to determine the solutions of (4.5) from the general solution given by [16, formula (13)]. Flat (2, 3, 5)-Distributions and Chazy’s Equations 9 5 First order system and different parametrisations of Chazy’s equations In this section, we first show that the generalised Chazy equation is equivalent to solving a third order differential equation involving the Schwarzian derivative and a potential term V (s). It is well-known that solutions to the generalised Chazy equation (3.1) can be rewritten as a first order system. For further details, see [2]. The first order system provides different parametrisations of the solutions, in addition to the one given by (2.3). We compute the solutions to the generalised Chazy equation (1.2) with k = ±2 3 for the different parametrisations below and present them in Tables 1, 2 and 3. We also show how these solutions are related to one another by algebraic transformation of hypergeometric functions. Let Ω1, Ω2, Ω3 be functions of q. Let dot denote differentiation with respect to q. Then consider Ω̇1 = Ω2Ω3 − Ω1(Ω2 + Ω3) + τ2, Ω̇2 = Ω3Ω1 − Ω2(Ω3 + Ω1) + τ2, Ω̇3 = Ω1Ω2 − Ω3(Ω1 + Ω2) + τ2, (5.1) where τ2 = α2(Ω1 − Ω2)(Ω3 − Ω1) + β2(Ω2 − Ω3)(Ω1 − Ω2) + γ2(Ω3 − Ω1)(Ω2 − Ω3) and α, β, γ are constants. Introducing the parameter s(q) = Ω1 − Ω3 Ω2 − Ω3 , we find that Ω1 = −1 2 d dq log ṡ s(s− 1) , Ω2 = −1 2 d dq log ṡ s− 1 , Ω3 = −1 2 d dq log ṡ s . The system of equations (5.1) are satisfied iff {s, q}+ ṡ2 2 V (s) = 0, (5.2) where {s, q} = d dq ( s̈ ṡ ) − 1 2 ( s̈ ṡ )2 is the Schwarzian derivative of s(q) and V (s) = 1− β2 s2 + 1− γ2 (s− 1)2 + β2 + γ2 − α2 − 1 s(s− 1) . Switching independent and dependent variables in (5.2), we have {s, q} = −(ṡ)2{q, s}, so that the dual of (5.2) is {q, s} − 1 2 V (s) = 0. 10 M. Randall The general solution is given by q(s) = u2(s) u1(s) where u1, u2 are linearly independent solutions of the second order ODE u′′ + 1 4 V (s)u = 0. (5.3) The general solution of (5.3) suggests taking u(s) = (s− 1) 1−γ 2 s 1−β 2 z(s) (cf. [3]) to transform (5.3) into the hypergeometric differential equation (2.2) with a = 1 2 (1− α− β − γ), b = 1 2 (1 + α− β − γ), c = 1− β. In [2], it was determined that taking y = −2(Ω1 + Ω2 + Ω3) = d dq log ṡ3 s2(s− 1)2 (5.4) gives solutions to the generalised Chazy equation (3.1) with variable q whenever α = β = γ = 2 k or α = 2 k , β = γ = 1 3 and its cyclic permutations. For α = β = γ = 2 k this gives (a, b, c) =( k−6 2k , k−2 2k , k−2 k ) . For α = 2 k , β = γ = 1 3 with cyclic permutations, this gives respectively (a, b, c) = ( k − 6 6k , k + 6 6k , 2 3 ) , ( k − 6 6k , k − 2 2k , k − 2 k ) , ( k − 6 6k , k − 2 2k , 2 3 ) , with only the first coinciding with the list in Proposition 3.1. This suggests that the solutions to (5.2) are more general than the solution of the form y = 6 d dq log z1 given by Chazy [13]. We can express Chazy’s solution in terms of s(q) as follows. A computation of the Wronskian of linearly independent solutions z1, z2 to (2.2) gives W (z1, z2) = z1ż2 − z2ż1 = w0(s− 1)c−a−b−1s−c for some non-zero constant w0. The latter equality holds by solving the first order differential equation W ′ = −c− (a+ b+ 1)s s(1− s) W. See for instance [4]. From q(s) = z2(s) z1(s) , we find that d dq = (z1)2 z1ż2−z2ż1 d ds . Applying this derivative to s(q), we obtain s′(q) = d dq s(q) = (z1) 2 z1ż2 − z2ż1 = (z1) 2 W (z1, z2) . Hence s′′(q) = d dq s′(q) = s′(q) d ds ( (z1) 2 W (z1, z2) ) = s′(q) ( 2z1ż1 W (z1, z2) − (z1) 2W ′(z1, z2) W (z1, z2)2 ) = s′(q) ( 2s′(q) ż1 z1 − s′(q)W ′(z1, z2) W (z1, z2) ) = s′(q) ( 2s′(q) ż1 z1 + s′(q) c− (a+ b+ 1)s(q) s(q)(1− s(q)) ) = s′(q) ( 2s′(q) ż1 z1 + s′(q) c(1− s(q))− (a+ b+ 1− c)s(q) s(q)(1− s(q)) ) , Flat (2, 3, 5)-Distributions and Chazy’s Equations 11 and we get s̈ ṡ = 2ṡ ż1 z1 + c ṡ s − (a+ b+ 1− c) ṡ 1− s . Therefore we have y = 6 d dq log z1 = 6 (z1) 2 W (z1, z2) ż1 z1 = 6ṡ ż1 z1 = 3 s̈ ṡ − 3c ṡ s + 3(a+ b+ 1− c) ṡ 1− s = 3 d dq log ṡ− 3c d dq log s− 3(a+ b+ 1− c) d dq log(1− s) = 3 d dq log ṡ sc(1− s)a+b+1−c = 1 2 d dq log ṡ6 s6c(1− s)6(a+b+1−c) . A comparison of Chazy’s formula for y = 6 d dq log z1 with the formula for y in (5.4) suggests taking c = 2 3 , a+ b = 1 3 . This is satisfied by (a, b, c) = ( k−6 6k , k+6 6k , 2 3 ) in Proposition 3.1. For k = 2 3 as in (1.2), we get the following solutions for u given in Table 1. For this Table 1. (α, β, γ) (a, b, c) General solution to u′′ + 1 4V (s)u = 0 (3, 3, 3) (−4,−1,−2) c1 2s−1 s(s−1) + c2 s2(s−2) s−1( 3, 13 , 1 3 ) ( −4 3 , 5 3 , 2 3 ) c1(3s− 2)s 2 3 (s− 1) 1 3 + c2(3s− 1)s 1 3 (s− 1) 2 3( 1 3 , 3, 1 3 ) ( −4 3 ,−1,−2 ) c1 2s−3 s (s− 1) 1 3 + c2 s−3 s (s− 1) 2 3( 1 3 , 1 3 , 3 ) ( −4 3 ,−1, 23 ) c1 2s+1 s−1 s 1 3 + c2 s+2 s−1s 2 3 Chazy parameter, the hypergometric series truncate and the solutions to (5.3) can be given by elementary functions. We have 2F1(−4,−1;−2; s) = 1− 2s, 2F1 ( −4 3 , 5 3 ; 2 3 ; s ) = 3s2 − 4s+ 1 (1− s) 2 3 , 2F1 ( −4 3 ,−1;−2; s ) = 1− 2 3 s, 2F1 ( −4 3 ,−1; 2 3 ; s ) = 1 + 2s. Moreover, the solutions in each row are related to one another by algebraic transformations of hypergeometric functions. The solutions in the second, third and fourth rows of Table 1 can be obtained from the first by a cubic transformation of hypergeometric functions (see [23, formula (23)]). Explicitly, to show how the solution in the third row is related to the first, let ω be a cube root of unity (solution to ω2 + ω + 1 = 0) and consider the map s 7→ t = 3(2ω + 1)s(s− 1) (s+ ω)3 and the relation z̃(t) = (1 + ωs)−4z(s). Then z̃(t) is a solution to the hypergeometric differential equation (2.2) with (a, b, c) = ( −4 3 ,−1, −2 ) iff z(s) solves (2.2) with (a, b, c) = (−4,−1,−2). The solutions in the last three rows are related to one another by fractional linear transformations. The symmetry s 7→ 1 − s interchanges β and γ in V (s) and hence transforms the solution in the 3rd row to the solution 12 M. Randall given in the 4th row while s 7→ t = s s−1 and the relation z̃(t) = (1 − s)− 4 3 z(s) transforms the solution in the 4th row to those in the 2nd. A different parametrisation of Chazy’s equations (cf. [10]) is also given by y = −Ω1 − 2Ω2 − 3Ω3 = 1 2 d dq log ṡ6 s4(s− 1)3 . (5.5) A comparison with Chazy’s formula (2.3) yields (a, b, c) = ( k−6 12k , k+6 12k , 2 3 ) from Proposition 3.1. The solution of the form (5.5) solves the generalised Chazy equation (3.1) whenever (α, β, γ) in (5.2) is given by( 1 k , 1 3 , 1 2 ) , ( 1 k , 2 k , 1 2 ) or ( 1 k , 1 3 , 3 k ) . The solution to (5.2) with (α, β, γ) = ( 1 k , 1 3 , 1 2 ) is given by the Schwarz function J (see [2, 3]). Considering the symmetry s = 1−K brings (5.5) to y = −Ω1 − 3Ω2 − 2Ω3 = 1 2 d dq log −K̇6 (1−K)4K3 (5.6) and comparing with Chazy’s formula gives (a, b, c) = ( k−6 12k , k+6 12k , 1 2 ) from Proposition 3.1. The solution of the form (5.6) solves (3.1) whenever (α, β, γ) in (5.2) is given by( 1 k , 1 2 , 1 3 ) , ( 1 k , 1 2 , 2 k ) or ( 1 k , 3 k , 1 3 ) . The symmetry s = 1 −K permutes β and γ in the formula for V (s) in (5.2). For k = 2 3 , the solutions to (5.3) with the parametrisation by (5.5) are presented in Table 2. Table 2. (α, β, γ) (a, b, c) General solution to u′′ + 1 4V (s)u = 0( 3 2 , 1 3 , 1 2 ) ( 5 6 ,− 2 3 , 2 3 ) c1(s− 1) 1 4 s 1 2P 1 3 1 ( √ 1− s) + c2(s− 1) 1 4 s 1 2Q 1 3 1 ( √ 1− s)( 3 2 , 3, 1 2 ) ( −1 2 ,−2,−2 ) c1 (s−1) 3 4 s + c2 (s−1) 1 4 s (s2 + 4s− 8)( 3 2 , 1 3 , 9 2 ) ( −7 6 ,− 8 3 , 2 3 ) c1s 2 3 (s− 1) 11 4 2F1 ( 13 6 , 11 3 ; 4 3 ; s ) + c2s 1 3 (s− 1) 11 4 2F1( 11 6 , 10 3 ; 2 3 ; s) We also note that we have 2F1 ( 5 6 ,−2 3 ; 2 3 ; s ) = Γ ( 2 3 )3√ 3 π P (− 1 3 ,− 1 2) 2 3 (1− 2s), 2F1 ( −1 2 ,−2;−2; s ) = 1 8 ( 8− 4s− s2 ) , 2F1 ( −7 6 ,−8 3 ; 2 3 ; s ) = 10 7 Γ ( 2 3 )3√ 3 π P (− 1 3 ,− 9 2) 8 3 (1− 2s). The algebraic transformations relating the solutions between the rows are given as follows. The map that takes the solution from the second row to the solution in the first row of Table 2 is a composition of fractional linear transformations and cubic transformation due to Goursat (see [23, formulas (20)–(22)] and [17, formula (123)]). To see this, we first consider the map s 7→ t = s(9− s)2 (s+ 3)3 Flat (2, 3, 5)-Distributions and Chazy’s Equations 13 and the relation z̃(t) = ( 1 + s 3 )−2 z(s). Then z̃(t) satisfies (2.2) with (a, b, c) = ( −2 3 ,− 1 3 , 1 2 ) iff z(s) satisfies (2.2) with (a, b, c) =( −2,−1 2 , 1 2 ) . Now 2F1 ( −2 3 ,−1 3 ; 1 2 ; t ) = (1− t) 2 3 2F1 ( −2 3 , 5 6 ; 1 2 ; t t− 1 ) while the map s 7→ 1 − s takes the differential equation (2.2) with (a, b, c) = ( −2 3 , 5 6 , 1 2 ) to (a, b, c) = ( −2 3 , 5 6 , 2 3 ) . Moreover, we have 2F1 ( −2,−1 2 ; 1 2 ; 1− s ) = 1 3 ( 8− 4s− s2 ) = 8 3 2F1 ( −2,−1 2 ;−2; s ) , so that all together the composition gives the map s 7→ t̃ = −(s− 4)3 27s2 and the relation z̃(t̃) = s− 4 3 z(s). Thus z̃(t̃) satisfies (2.2) with (a, b, c) = ( −2 3 , 5 6 , 2 3 ) iff z(s) satisfies (2.2) with (a, b, c) = ( −2, −1 2 ,−2 ) . To obtain the solution given in the third row from those in the first row requires a trans- formation of degree 4 (see [23, formulas (25)–(27)] and [17, equation (131)]). Consider the map s 7→ t = − s(s+ 8)3 64(1− s)3 and the relation z̃(t) = (1− s) 5 2 z(s). Then z̃(t) satisfies (2.2) with (a, b, c) = ( 5 6 ,− 2 3 , 2 3 ) iff z(s) satisfies (2.2) with (a, b, c) = ( 11 6 , 10 3 , 2 3 ) . Finally, we use the Euler transformation, 2F1 ( −7 6 ,−8 3 ; 2 3 ; s ) = (1− s) 9 2 2F1 ( 11 6 , 10 3 ; 2 3 ; s ) to obtain the solution in the 3rd row. There is furthermore a degree 6 transformation relating the solution in the first row of Table 1 to the solution in the first row of Table 2 (cf. [23, equation (28)], [17, equation (134)] and [1, Table 1]). Consider the map s 7→ t = 27s2(s− 1)2 4(s2 − s+ 1)3 and the relation z̃(t) = (1− s+ s2)−2z(s). 14 M. Randall Then z̃(t) satisfies (2.2) with (a, b, c) = ( −1 3 ,− 2 3 ,− 1 2 ) iff z(s) satisfies (2.2) with (a, b, c) = (−4,−1,−2). Furthermore s 7→ 1 − s takes the solution to (2.2) with (a, b, c) = ( −1 3 ,− 2 3 ,− 1 2 ) to the solution with (a, b, c) = ( −1 3 ,− 2 3 , 1 2 ) and an Euler transformation followed by s 7→ 1 − s again takes this to the solution with (a, b, c) = ( 5 6 ,− 2 3 , 2 3 ) as given in row 1 of Table 2. Finally, consider the parametrisation given by y = −4Ω1 − Ω2 − Ω3 = d dq log ṡ3 (s− 1) 5 2 s 5 2 . (5.7) We find that (α, β, γ) is one of( 4 k , 1 k , 1 k ) , ( 2 3 , 1 k , 1 k ) , ( 2 3 , 1 6 , 1 6 ) . For (α, β, γ) = ( 4 k , 1 k , 1 k ) , we find that (a, b, c) = ( k−6 2k , k+2 2k , k−1 k ) . For k = 2 3 , we obtain (a, b, c) = (−4, 2,−1 2). Let us relate the solution to the differential equation (2.2) with (a, b, c) = (−4, 2,−1 2) to the solution given in the second row of Table 2. For (α, β, γ) = ( 1 2 , 1 k , 2 k ) , we find that (a, b, c) = ( k−6 4k , 3k−6 4k , k−1k ) . We have 2F1 ( 1 2 − 3 k , 1 2 + 1 k ; 1− 1 k ; s ) = 2F1 ( 1 4 − 3 2k , 1 4 + 1 2k ; 1− 1 k ; 4s(1− s) ) = (1− 4s(1− s))− 1 4 + 3 2k 2F 1 ( 1 4 − 3 2k , 3 4 − 3 2k ; 1− 1 k ; 4s(1− s) 4s(1− s)− 1 ) and thus for k = 2 3 , we have 2F1 ( −4, 2;−1 2 ; s ) = (1− 4s(1− s))2 2F1 ( −2,−3 2 ;−1 2 ; 4s(1− s) 4s(1− s)− 1 ) . Again s 7→ 1 − s brings the solution to (2.2) with (a, b, c) = ( −2,−3 2 ,− 1 2 ) to the solution with (a, b, c) = ( −2,−3 2 ,−2 ) and an Euler transform gives the solution with (a, b, c) = ( −2,−1 2 ,−2 ) . For (α, β, γ) = ( 2 3 , 1 k , 1 k ) , we find that (a, b, c) = ( k−6 6k , 5k−6 6k , k−1k ) . For k = 2 3 , this gives (a, b, c) = ( −4 3 ,− 2 3 ,− 1 2 ) . We have a degree 2 transformation of the solution to (2.2) with this value to the solution in the first row of Table 2 given by the map s 7→ t = − 1 4s(s− 1) and the relation z̃(t) = 1 4 (s(1− s))− 2 3 z(s). We have z̃(t) satisfying (2.2) with (a, b, c) = ( −2 3 , 5 6 , 2 3 ) iff z(s) satisfies (2.2) with (a, b, c) =( −4 3 ,− 2 3 ,− 1 2 ) . Let us summarise the solutions to (5.3) with parametrisation given by (5.7) in Table 3. We also note that we have 2F1 ( −4, 2;−1 2 ; s ) = −128 5 P (− 3 2 ,− 3 2) 4 (1− 2s) = −128s4 + 256s3 − 144s2 + 16s+ 1, 2F1 ( −4 3 ,−2 3 ;−1 2 ; s ) = −16 27 π22 2 3 Γ ( 2 3 )3P (− 3 2 ,− 3 2) 4 3 (1− 2s), 2F1 ( 0, 2 3 ; 5 6 ; s ) = 1. Flat (2, 3, 5)-Distributions and Chazy’s Equations 15 Table 3. (α, β, γ) (a, b, c) General solution to u′′ + 1 4V (s)u = 0( 6, 32 , 3 2 ) ( −4, 2,−1 2 ) c1(2s− 1)(s(s− 1)) 5 4 + c2 128s4−256s3+144s2−16s−1 (s(s−1)) 1 4( 2 3 , 3 2 , 3 2 ) ( −4 3 ,− 2 3 ,− 1 2 ) c1(s(s− 1)) 5 4 2F1 ( 5 3 , 7 3 ; 5 2 ; s ) + c2 (s−1) 5 4 s 1 4 2F1 ( 1 6 , 5 6 ;−1 2 ; s ) ( 2 3 , 1 6 , 1 6 ) ( 0, 23 , 5 6 ) c1(s(s− 1)) 7 12 2F1 ( 1 3 , 1; 7 6 ; s ) + c2s 5 12 (1− s) 5 12 Comparing the parametrisation (5.7) with solutions of the form (2.3), we obtain (a, b, c) =( 0, 23 , 5 6 ) . For this solution given in the third row, we obtain solutions of the form y(q) = −6 q . Up to fractional linear transformations in the variable s, we have 7 classes of solutions to (1.2) determined by the solutions to (5.2). They are given by the solutions with (α, β, γ) either (3, 3, 3) or ( 3, 13 , 1 3 ) for I(q) given by (5.4), (α, β, γ) one of ( 3 2 , 1 3 , 1 2 ) , ( 3 2 , 1 3 , 9 2 ) or ( 3 2 , 3, 1 2 ) for I(q) given by (5.5) and (α, β, γ) either ( 6, 32 , 3 2 ) or ( 2 3 , 3 2 , 3 2 ) for I(q) given by (5.7). When α = β = γ = 0, we have τ = 0. The first order system (5.1) is the classical Darboux– Halphen system and y = −2(Ω1 + Ω2 + Ω3) satisfies Chazy’s equation (2.1). This system arises as the anti-self-dual Ricci-flat equations for Bianchi-IX metrics (see [7, 22]). 6 Examples of flat (2, 3, 5)-distributions In [19], Nurowski associated to (2, 3, 5)-distributions a conformal class of metrics of signatu- re (2, 3). The fundamental curvature invariant of (2, 3, 5)-distributions appears as the Weyl ten- sor of Nurowski’s metric. For (2, 3, 5)-distributions DF (q) determined by a single function F (q), the metric is described in [6, 19]. The distribution DF (q) on Mxyzpq is encoded by the annihilator of the three 1-forms ω1 = dy − pdx, ω2 = dp− qdx, ω3 = dz − F (q)dx, (6.1) and supplemented by the 1-forms ω4 = dq, ω5 = dx. The coframe on Mxyzpq is given by θ1 = ω1 − 1 F ′′ ( F ′ω2 − ω3 ) , θ2 = 1 F ′′ ( F ′ω2 − ω3 ) , θ3 = ( 1− F ′F (3) 4(F ′′)2 ) ω2 + F (3) 4(F ′′)2 ω3, θ4 = ( 7(F (3))2 − 4F ′′F (4) 40(F ′′)3 )( F ′ω2 − ω3 ) + ω4 − ω5, θ5 = −ω4, (6.2) (cf. [6]) and Nurowski’s metric gDF (q) = 2θ1θ5 − 2θ2θ4 + 4 3 θ3θ3 (6.3) has vanishing Weyl tensor (and hence conformally flat) iff DF (q) has the split real form of G2 as its group of local symmetries. There is a more elegant way to present Nurowski’s metric for 16 M. Randall the distribution DF (q). Equivalently, we can encode the distribution by ω̃1 = ω1 = dy − pdx, ω̃2 = 1 F ′′ (F ′ω2 − ω3) = 1 F ′′ ( F ′(dp− qdx)− (dz − F (q)dx) ) , ω̃3 = ω2 = dp− qdx, (6.4) with ω̃4 = ω4 and ω̃5 = ω5. The coframe is then given by θ1 = ω̃1 − ω̃2, θ2 = ω̃2, θ3 = ω̃3 − F (3) 4F ′′ ω̃2, θ4 = ( 7(F (3))2 − 4F ′′F (4) 40(F ′′)2 ) ω̃2 + ω̃4 − ω̃5, θ5 = −ω̃4. Let us take F (q) = ∫∫ e 1 2 ∫ I(q)dqdqdq, as in (4.6). This gives F ′′ = e 1 2 ∫ I(q)dq. Nurowski’s metric now has a very simple form given by gDF (q) = 2θ1θ5 − 2θ2θ4 + 4 3 θ3θ3 = 2ω̃2ω̃5 − 2ω̃1ω̃4 + 4 3 (ω̃3) 2 − I 3 ω̃2ω̃3 + 1 10 ( I ′ − I2 6 ) (ω̃2) 2. (6.5) The Ricci tensor of the above metric is Rabθ aθb = 9 120 ( 6I ′ − I2 ) (ω̃4) 2. We can consider conformal rescalings of the metric such that ĝDF (q) = Ω2gDF (q) is Ricci flat. It turns out that if we take Ω = ν(q)−1 > 0, then the Ricci tensor of the rescaled metric ĝDF (q) is given by Rabθ aθb = 3 40ν ( 40ν ′′ + ( 6I ′ − I2 ) ν ) (ω̃4) 2, (6.6) so the appropriate conformal scale ν(q) can be found by solving the differential equation in (6.6) (cf. [24, Proposition 35]). In the first part of this section we consider the conformally flat met- rics (6.5) obtained by solving (1.2) using the solution (2.3). Next, we then consider the solutions obtained from different parametrisations of the generalised Chazy equation given by (5.4), (5.5) and (5.7). We also consider conformally flat metrics obtained from solving the Legendre trans- form of (1.2). This involves computing the coframe for the metric under the Legendre transform. Finally, we consider the metrics obtain from Chazy’s solutions given by (3.2). The metrics asso- ciated to (2, 3, 5)-distributions DF (q) of the form F (q) = qm where m ∈ { −1, 13 , 2 3 , 2 } are given in [18]. 6.1 Chazy’s solution In order to express Nurowski’s metric associated to flat (2, 3, 5)-distributions obtained from solving (1.2), we have to switch independent variable s and dependent variable q. In other words we pass to coordinates (x, y, z, p, s) with q(s) = z2(s) z1(s) where z1(s), z2(s) are given in Corollary 3.2. Flat (2, 3, 5)-Distributions and Chazy’s Equations 17 Let us first consider solutions of the form I(q(s)) = 6 d dq log z1(s) as in Corollary 3.2. Observe that∫ I(q)dq = 6 log z1 and so F (q) = ∫∫ e 1 2 ∫ I(q)dqdqdq = ∫∫ (z1) 3dqdq. For this parametrisation we have dq = z1ż2 − z2ż1 (z1)2 ds, so that F (q(s)) = ∫ (∫ z1(z1ż2 − z2ż1)ds ) z1ż2 − z2ż1 (z1)2 ds. Let us denote K(s) = ∫ z1(z1ż2 − z2ż1)ds. Theorem 6.1. Let z1(s), z2(s) be two linearly independent solutions to (2.2) with (a, b, c) given by one of the list in Corollary 3.2. Let DC denote the (2, 3, 5)-distribution on Mxyzps associated to the annihilator of ω̃1 = dy − pdx, ω̃2 = K(s) (z1)3 ( dp− z2 z1 dx ) − 1 (z1)3 ( dz − (∫ K(s) z1ż2 − z2ż1 (z1)2 ds ) dx ) , ω̃3 = dp− z2 z1 dx. Supplement by the 1-forms ω̃4 = z1ż2 − z2ż1 (z1)2 ds, ω̃5 = dx. Then Nurowski’s metric (6.5) gDC = 2ω̃2ω̃5 − 2ω̃1ω̃4 + 4 3 (ω̃3) 2 − 2z1ż1 z1ż2 − z2ż1 ω̃2ω̃3 + 3 5 ( (z1) 3z̈1 (z1ż2 − z2ż1)2 − (z1) 3ż2(z1z̈2 − z2z̈1) (z1ż2 − z2ż1)3 ) (ω̃2) 2 has vanishing Weyl tensor (and hence conformally flat) and DC has the split real form of G2 as its group of local symmetries. 18 M. Randall Let us provide an explicit example given by Corollary 3.2. It turns out for the values of (a, b, c) obtained in Corollary 3.2, the solutions can be given by elementary functions. For (a, b, c) = ( −2 3 , 5 6 , 1 2 ) , the solutions to the hypergeometric differential equation (2.2) are given by z(s) = µ(s− 1) 1 6P 1 3 1 ( √ s) + ν(s− 1) 1 6Q 1 3 1 ( √ s), where Pm ` and Qm ` are the associated Legendre functions. This suggest passing further to the variable r = √ s, in which case (2.2) with (a, b, c) = ( −2 3 , 5 6 , 1 2 ) becomes 1 4 (1− r2)z′′(r)− 1 3 rz′(r) + 5 9 z(r) = 0. (6.7) The general solution is now given by the elementary functions z(r) = c1(r − 1) 1 3 (3r + 1) + c2(r + 1) 1 3 (3r − 1). There is thus a 4-dimensional space of solutions given by z1(r) = c1(r − 1) 1 3 (3r + 1) + c2(r + 1) 1 3 (3r − 1), z2(r) = c3(r − 1) 1 3 (3r + 1) + c4(r + 1) 1 3 (3r − 1), where c1c4− c2c3 6= 0. We also note that in the case (a, b, c) = ( −2 3 , 5 6 , 2 3 ) , the change of variable r = √ 1− s brings (2.2) to (6.7), so that the two hypergeometric ODEs with (a, b, c) = ( −2 3 , 5 6 , 1 2 ) and ( −2 3 , 5 6 , 2 3 ) can be brought to the same equation (6.7) by a coordinate transformation. Moreover, since 2F1 ( −2 3 , 5 6 ; 2 3 ; 4s(1− s) ) = 2F1 ( −4 3 , 5 3 ; 2 3 ; s ) , if we take r = √ 1− 4s(1− s) = 2s − 1, we can pass from (2.2) with (a, b, c) = ( −4 3 , 5 3 ; 2 3 ) to equation (6.7). Also compare with the second row of Table 1. We now pass to coordinates on Mxyzpr and take for a simple example z1(r) = c1(r − 1) 1 3 (3r + 1), z2(r) = c2(r + 1) 1 3 (3r − 1) where c1, c2 are non-zero constants. We also make use of ds = 2rdr. Here we have K(r) = −16(c1) 2c2(r − 1) 2 3 (r + 1) 1 3 . We have Proposition 6.2. Let c1, c2 be non-zero constants. Let D0 denote the (2, 3, 5)-distribution on Mxyzpr associated to the annihilator of ω̃1 = dy − pdx, ω̃2 = −16c2(r + 1) 1 3 c1(r − 1) 1 3 (3r + 1)3 dp− 1 (c1)3(r − 1)(3r + 1)3 dz + 16(c2) 2(r + 1) 2 3 (c1)2(r − 1) 2 3 (3r + 1)3 dx, ω̃3 = dp− c2(r + 1) 1 3 (3r − 1) c1(r − 1) 1 3 (3r + 1) dx, and supplemented by the 1-forms ω̃4 = − 16c2 3c1(3r + 1)2(r − 1) 4 3 (r + 1) 2 3 dr, ω̃5 = dx. Flat (2, 3, 5)-Distributions and Chazy’s Equations 19 Then Nurowski’s metric (6.5) gD0 = 2ω̃2ω̃5 − 2ω̃1ω̃4 + 4 3 (ω̃3) 2 + c1(3r − 2)(3r + 1)(r − 1) 1 3 (r + 1) 2 3 2c2 ω̃2ω̃3 + 3(c1) 2(r − 1) 5 3 (3r + 1)4(r + 1) 1 3 64(c2)2 (ω̃2) 2 has vanishing Weyl tensor (and hence conformally flat) and D0 has the split real form of G2 as its group of local symmetries. The Ricci tensor for this metric is Rabθ aθb = 6 r2 − 1 drdr. Rescaling this metric by Ω = 1 ν = 4 3 (3r + 1)(r − 1) 1 3 a1(r − 1) 1 3 − a2(r + 1) 1 3 , where a1 and a2 are constants, the conformally rescaled metric ĝD0 = Ω2gD0 given by ĝD0 = 16(3r + 1)2(r − 1) 2 3 9(a1(r − 1) 1 3 − a2(r + 1) 1 3 )2 ( 2ω̃2ω̃5 − 2ω̃1ω̃4 + 4 3 (ω̃3) 2 + c1(3r − 2)(3r + 1)(r − 1) 1 3 (r + 1) 2 3 2c2 ω̃2ω̃3 + 3(c1) 2(r − 1) 5 3 (3r + 1)4(r + 1) 1 3 64(c2)2 (ω̃2) 2 ) is both Ricci-flat and conformally flat. 6.2 Other parametrisations of the generalised Chazy equation Instead of choosing I(q) = 6 d dq log z1, we can consider the parametrisation I(q) = d dq log ṡ3 s2(s− 1)2 given in (5.4). In this case F (q) = ∫∫ e 1 2 ∫ I(q)dqdqdq = ∫∫ ṡ 3 2 s(s− 1) dqdq and the corresponding metric associated to the (2, 3, 5)-distribution DF (q) gives the following Theorem 6.3. Let s(q) be a solution to (5.2) with (α, β, γ) given by one of (3, 3, 3), ( 3, 1 3 , 1 3 ) , ( 1 3 , 3, 1 3 ) , ( 1 3 , 1 3 , 3 ) . Let Ds denote the (2, 3, 5)-distribution on Mxyzpq associated to the annihilator of ω1 = dy − pdx, ω2 = dp− qdx, ω3 = dz − (∫∫ ṡ 3 2 s(s− 1) dqdq ) dx. Supplement by the 1-forms ω4 = dq, ω5 = dx, 20 M. Randall and take the coframe on Mxyzpq to be given by (θ1, θ2, θ3, θ4, θ5) as in (6.2) where F (q) = ∫∫ ṡ 3 2 s(s− 1) dqdq. Then Nurowski’s metric gDs = 2θ1θ5 − 2θ2θ4 + 4 3 θ3θ3 has vanishing Weyl tensor (and hence conformally flat) and Ds has the split real form of G2 as its group of local symmetries. To obtain explicit examples, it is useful to switch the independent variable q and the depen- dent variable s. We pass to the variables (x, y, z, p, s) with q = u2(s) u1(s) where u1 and u2 are linearly independent solutions of (5.3) given in Table 1. Note that up to fractional linear transforma- tions in the variable s, we only need to consider the solutions to (5.2) with the values of (α, β, γ) given by either (3, 3, 3) or ( 3, 13 , 1 3 ) for the parametrisation given by (5.4). Note the symmetry permuting β and γ. A computation shows that W (u1, u2) = u1u̇2 − u̇1u2 is constant, which we can normalise to set W (u1, u2) = 1. We have dq = u1u̇2 − u2u̇1 (u1)2 ds = 1 (u1)2 ds and ṡ = 1 q′(s) = (u1) 2, so that F (q(s)) = ∫ (∫ (u1) 3 s(s− 1) 1 (u1)2 ds ) 1 (u1)2 ds = ∫ (∫ u1 s(s− 1) ds ) 1 (u1)2 ds. Let us denote K(s) = ∫ u1 s(s− 1) ds. Theorem 6.4. Let u1(s), u2(s) be two linearly independent solutions to (5.3) subject to the constraint W (u1, u2) = 1 with (α, β, γ) given by Table 1. Let Ds denote the (2, 3, 5)-distribution on Mxyzps associated to the annihilator of ω1 = dy − pdx, ω2 = dp− u2(s) u1(s) dx, ω3 = dz − (∫ K(s) (u1)2 ds ) dx. We pass to the annihilator 1-forms given by (6.4) to obtain ω̃1 = ω1 = dy − pdx, ω̃2 = s(s− 1) (u1)3 (Kω2 − ω3), ω̃3 = ω2 = dp− u2(s) u1(s) dx, with ω̃4 = ω4 = 1 (u1)2 ds, ω̃5 = ω5 = dx. Flat (2, 3, 5)-Distributions and Chazy’s Equations 21 Take the coframe on Mxyzps to be given by θ1 = ω̃1 − ω̃2, θ2 = ω̃2, θ3 = ω̃3 − (u1) 2 4s(s− 1) ( 3 u̇1 u1 s(s− 1)− (2s− 1) ) ω̃2, θ4 = (u1) 4 40 ( 4s2 − 4s− 1 s2(s− 1)2 + 3V (s)− 10(2s− 1) s(s− 1) u̇1 u1 + 15 ( u̇1 u1 )2 ) ω̃2 + ω̃4 − ω̃5, θ5 = −ω̃4. Then Nurowski’s metric gDs = 2θ1θ5 − 2θ2θ4 + 4 3 θ3θ3 = 2ω̃2ω̃5 − 2ω̃1ω̃4 + 4 3 (ω̃3) 2 + 2u1(3s(1− s)u′1 + (2s− 1)u1) 3s(s− 1) ω̃2ω̃3 − (u1) 4(9V (s)s2(s− 1)2 − 8(s2 − s+ 1)) 60(s− 1)2s2 (ω̃2) 2 has vanishing Weyl tensor (and hence conformally flat) and Ds has the split real form of G2 as its group of local symmetries. The analogous results hold for I(q) given by the formulas in (5.5), (5.6) and (5.7). If we take I(q) = d dq log ṡ3 s2(s− 1) 3 2 , as in (5.5), we have the corresponding (2, 3, 5)-distribution associated to F (q) = ∫∫ ṡ 3 2 s(s− 1) 3 4 dqdq. Theorem 6.5. Let s(q) be a solution to (5.2) with (α, β, γ) given by ( 3 2 , 1 3 , 1 2 ) , ( 3 2 , 3, 1 2 ) or( 3 2 , 1 3 , 9 2 ) . Let Ds1 denote the (2, 3, 5)-distribution on Mxyzpq associated to the annihilator of ω1 = dy − pdx, ω2 = dp− qdx, ω3 = dz − (∫∫ ṡ 3 2 s(s− 1) 3 4 dqdq ) dx. Supplement by the 1-forms ω4 = dq, ω5 = dx, and take the coframe on Mxyzpq to be given by (θ1, θ2, θ3, θ4, θ5) as in (6.2). Then Nurowski’s metric (6.3) has vanishing Weyl tensor (and hence conformally flat) and Ds1 has the split real form of G2 as its group of local symmetries. Similarly, for I(q) given by (5.7), that is to say I(q) = d dq log ṡ3 s 5 2 (s− 1) 5 2 , we have the corresponding (2, 3, 5)-distribution associated to F (q) = ∫∫ ṡ 3 2 s 5 4 (s− 1) 5 4 dqdq. 22 M. Randall Theorem 6.6. Let s(q) be a solution to (5.2) with (α, β, γ) given by ( 6, 32 , 3 2 ) , ( 2 3 , 3 2 , 3 2 ) or( 2 3 , 1 6 , 1 6 ) . Let Ds2 denote the (2, 3, 5)-distribution on Mxyzpq associated to the annihilator of ω1 = dy − pdx, ω2 = dp− qdx, ω3 = dz − (∫∫ ṡ 3 2 s 5 4 (s− 1) 5 4 dqdq ) dx. Supplement by the 1-forms ω4 = dq, ω5 = dx, and take the coframe on Mxyzpq to be given by (θ1, θ2, θ3, θ4, θ5) as in (6.2). Then Nurowski’s metric (6.3) has vanishing Weyl tensor (and hence conformally flat) and Ds2 has the split real form of G2 as its group of local symmetries. 6.3 Legendre transformed coframe The Legendre transform of Proposition 4.1 takes the 1-forms (6.1) to ω1 = dy − pdx, ω2 = dp−H ′dx, ω3 = dz − (tH ′ −H)dx, ω4 = H ′′dt, ω5 = dx, where H = H(t) with H ′′ 6= 0 on Mxyzpt and the coframe (6.2) to θ1 = ω1 −H ′′(tω2 − ω3), θ2 = H ′′(tω2 − ω3), θ3 = ( 1 + t H ′′′ 4H ′′ ) ω2 − H ′′′ 4H ′′ ω3, θ4 = 4H ′′H ′′′′ − 5(H ′′′)2 40(H ′′)3 (tω2 − ω3) + ω4 − ω5, θ5 = −ω4. Note that our H(t) is related to Θ(x5) of [6] via Θ55 = −H, t = x5. The Nurowski metric g = 2θ1θ5 − 2θ2θ4 + 4 3 θ3θ3 has the only non-vanishing component of the Weyl tensor given by the left hand side term of the dual ODE (4.2). This accounts for the appearance of equation (4.3) in [9]. The solutions of the dual generalised Chazy ODE (4.5) with parameter ±3 2 give us further examples of flat (2, 3, 5)-distributions. We pass to (x, y, z, p, s) as before, with t(s) = w2(s) w1(s) where w1(s), w2(s) are linearly independent solutions to (2.2) with (a, b, c) one of( −1 4 , 5 12 , 1 2 ) , ( −1 4 , 5 12 , 2 3 ) , ( −1 2 , 5 6 , 2 3 ) . Here we have taken k = 3 2 . Note that the equations (2.2) for (a, b, c) = ( −1 4 , 5 12 , 1 2 ) and( −1 4 , 5 12 , 2 3 ) are equivalent by a linear transformation and thus the solutions to each equation can be expressed as linear combinations of the other, while the solutions for (a, b, c) = ( −1 4 , 5 12 , 2 3 ) and ( −1 2 , 5 6 , 2 3 ) are equivalent by a quadratic transformation as before. However, the author does not know if the solutions in these cases can be expressed by elementary functions. We consider once again solutions to (4.5) of the form u(t) = 6 d dt logw1. This gives H(t) = ∫∫ (w1) 4dtdt, H ′(t) = ∫ (w1) 4dt, H ′′(t) = (w1) 4. For this parametrisation we have dt = w1ẇ2 − w2ẇ1 (w1)2 ds Flat (2, 3, 5)-Distributions and Chazy’s Equations 23 and so H = ∫∫ (w1) 2(w1ẇ2 − w2ẇ1)ds w1ẇ2 − w2ẇ1 (w1)2 ds, H ′ = ∫ (w1) 2(w1ẇ2 − w2ẇ1)ds. We therefore obtain ω1 = dy − pdx, ω2 = dp− ∫ (w1) 2(w1ẇ2 − w2ẇ1)dsdx, ω3 = dz − ( w2 w1 ∫ (w1) 2(w1ẇ2 − w2ẇ1)ds − ∫∫ (w1) 2(w1ẇ2 − w2ẇ1)ds w1ẇ2 − w2ẇ1 (w1)2 ds ) dx, ω4 = (w1) 2(w1ẇ2 − w2ẇ1)ds, ω5 = dx and the adapted coframe for Nurowski’s metric is θ1 = ω1 − (w1) 4 ( w2 w1 ω2 − ω3 ) , θ2 = (w1) 4 ( w2 w1 ω2 − ω3 ) , θ3 = ( 1 + w2ẇ1 w1ẇ2 − w2ẇ1 ) ω2 − w1ẇ1 w1ẇ2 − w2ẇ1 ω3, θ4 = 2(ẅ1ẇ2 − ẅ2ẇ1) 5(w1ẇ2 − w2ẇ1)3 ( w2 w1 ω2 − ω3 ) + ω4 − ω5, θ5 = −ω4. (6.8) Equivalently, we can take ω̃1 = ω1, ω̃2 = (w1) 4 ( w2 w1 ω2 − ω3 ) , ω̃3 = ω2 with ω̃4 = ω4 and ω̃5 = ω5. The coframe is then given by θ1 = ω̃1 − ω̃2, θ2 = ω̃2, θ3 = ω̃3 + ẇ1 (w1)3(w1ẇ2 − w2ẇ1) ω̃2, θ4 = 2(ẅ1ẇ2 − ẅ2ẇ1) 5(w1)4(w1ẇ2 − w2ẇ1)3 ω̃2 + ω̃4 − ω̃5, θ5 = −ω̃4. (6.9) Proposition 6.7. The Nurowski metric g = 2θ1θ5 − 2θ2θ4 + 4 3 θ3θ3 given by the above coframe (6.8) for w1(s), w2(s) linearly independent solutions to the hyperge- ometric differential equation (2.2) with (a, b, c) one of( −1 4 , 5 12 , 1 2 ) , ( −1 4 , 5 12 , 2 3 ) , ( −1 2 , 5 6 , 2 3 ) are all conformally flat. For each (a, b, c) there is a 4-dimensional family of solutions. Up to fractional linear transformation in the variable s there are 2 distinct classes given by the last two entries. In addition, the Legendre transformation of Lemma 4.2 given by w1 = (z1) − 3 4 , w2 = (z1) − 3 4 ∫ z1(z1ż2 − z2ż1)ds 24 M. Randall takes the coframe (6.9) to the coframe in Theorem 6.1 and conversely so. The Legendre transform also applies to the coframes given in Theorems 6.3, 6.5 and 6.6. We also have the analogous results of Section 6.2. Up to fractional linear transformations in s we have 7 classes of solutions to the generalised Chazy equation (4.5) determined by s(t) satisfying (5.2). These are given by the parametrisations in Section 5 and the corresponding values for (α, β, γ) can be computed for the parameter k = 3 2 . For the parametrisation H(t) = ∫∫ ṡ2 s 4 3 (s− 1) 4 3 dtdt, the values for (α, β, γ) are given by either ( 4 3 , 4 3 , 4 3 ) or ( 4 3 , 1 3 , 1 3 ) . For H(t) = ∫∫ ṡ2 s 4 3 (s− 1) dtdt, (α, β, γ) takes the values of ( 2 3 , 1 2 , 1 3 ) , ( 2 3 , 1 2 , 4 3 ) or ( 2 3 , 2, 1 3 ) . For H(t) = ∫∫ ṡ2 s 5 3 (s− 1) 5 3 dtdt, we obtain ( 8 3 , 2 3 , 2 3 ) or ( 2 3 , 2 3 , 2 3 ) . The Legendre transform therefore provides seven further classes of flat Nurowski metrics up to fractional linear transformations in s. 6.4 Additional examples The solution (3.2) for k = ±2 3 gives I(q) = − 8 3(q+C) − 10 3(q+B) . Hence, the metric (6.5) on Mxyzpq given by gDF (q) = 2ω̃2ω̃5 − 2ω̃1ω̃4 + 4 3 (ω̃3) 2 + ( 8 9(q + C) + 10 9(q +B) ) ω̃2ω̃3 + 4(B − C)2 27(q +B)2(q + C)2 (ω̃2) 2 is conformally flat. In the dual coframe the flat metric (6.5) on Mxyzpt is given by g = 2ω̃2ω̃5 − 2ω̃1ω̃4 + 4 3 (ω̃3) 2 + 4 9 u(t)e ∫ − 2 3 u(t)dtω̃2ω̃3 − 2 135 (9u̇(t)− 4u(t)2)e ∫ − 4 3 u(t)dt(ω̃2) 2, (6.10) where u(t) satisfies the generalised Chazy equation (4.5) with parameter k = ±3 2 . The solu- tion (3.2) for k = ±3 2 gives u(t) = − 15 4(t−a) − 9 4(t−b) and substituting this into (6.10) gives the conformally flat metric g = 2ω̃2ω̃5 − 2ω̃1ω̃4 + 4 3 (ω̃3) 2 − 256 3 (8t− 5b− 3a)(t− a) 3 2 (t− b) 1 2 ω̃2ω̃3 + 65536 3 (t− b)2(t− a)3(4t− 3a− b)(ω̃2) 2. To summarise the results of this section, we first presented different examples of Nurowski metrics that are conformally flat up to fractional linear transformation in the variable s. Two examples are given in Theorem 6.3, three examples are given in Theorem 6.5 and two more in Theorem 6.6. Seven additional examples are obtained from the Legendre transform as in Proposition 4.1. We also have 2 additional examples from the solutions of the form (3.2). Finally, there are examples associated to distributions of the form F (q) = qm, where m ∈ { −1, 13 , 2 3 , 2 } and passing to the dual coframe, distributions of the form H(t) = tm, where m ∈ { −2,−1 2 , 1 2 , 2 } . Flat (2, 3, 5)-Distributions and Chazy’s Equations 25 7 An–Nurowski circle twistor bundle In [5], An and Nurowski showed how to associate to a split signature conformal structure [g] on a 4-manifold M4 a natural (2, 3, 5)-distribution. 4-dimensional split signature conformal structures admit real self-dual totally null 2-planes. The bundle of such 2-planes is a circle bundle over M4 with fibres S1 [5]. This is called the circle twistor bundle T(M4) and it has a rank 2 distribution given by lifting horizontally the null 2-planes on M4. This distribution is non-integrable, i.e., defines a (2, 3, 5)-distribution whenever the self-dual part of the Weyl tensor of g on M4 is non-vanishing. Moreover in [6], the authors presented split signature conformal structures on M4 that give rise to (2, 3, 5)-distributions of the form DF (q) on T(M4). Such split signature metrics are called Plebański’s second heavenly metrics in [6]. Following [6, Section 3], we can find these metrics that have a flat circle twistor bundle. Such circle twistor bundles have split G2 as their group of symmetries. Let (w, x, y, z) be local coordinates on M4. Let Θ = Θ(w, x, y, z) be an arbitrary function of 4 variables (second heavenly function of Plebański). Let (ei) be an orthonormal frame on M4 and (θj) the dual coframe satisfying θj(ei) = δj i. The split signature Plebański metric is given by g = gijθ i ⊗ θj = 2θ1θ2 + 2θ3θ4, where θiθj = 1 2θ i ⊗ θj + 1 2θ j ⊗ θi and θ1 = dx−Θyydw + Θxydz, θ2 = dw, θ3 = dy −Θxxdz + Θxydw, θ4 = dz. Hence g12 = g34 = 1 and all other components are zero. Such split signature metrics admit a real parallel spinor [15]. A computation shows that the connection 1-forms we need are given by Γ1 1 = −Θyyxθ 2 + Θyxxθ 4, Γ1 3 = −Θyyyθ 2 + Θyyxθ 4, Γ3 1 = Θyxxθ 2 −Θxxxθ 4, Γ3 3 = Θyyxθ 2 −Θyxxθ 4. Using [6] and Nurowski’s notes [20], we find that the (2, 3, 5)-distribution on T(M4) is annihi- lated by the following three 1-forms: ω3 = dξ + Γ3 1 + ( Γ3 3 − Γ1 1 ) ξ − Γ1 3ξ 2 = dξ + ( Θyxx + 2Θyyxξ + Θyyyξ 2 ) θ2 − ( Θxxx + 2Θyxxξ +Hyyxξ 2 ) θ4 = dξ + ( Θyxx + 2Θyyxξ + Θyyyξ 2 ) dw − ( Θxxx + 2Θyxxξ + Θyyxξ 2 ) dz and ω4 = ξθ4 + θ2 = ξdz + dw, ω5 = θ3 − ξθ1 = dy −Θxxdz + Θxydw − ξ(dx−Θyydw + Θxydz) = dy − ξdx− (Θxx + ξΘxy)dz + (Θxy + ξΘyy)dw. The distribution is therefore annihilated by the 1-forms ω̃3 = dξ − ( Θxxx + 3Θyxxξ + 3Θyyxξ 2 + Θyyyξ 3 ) dz, ω̃4 = ξdz + dw, ω̃5 = dy − ξdx− ( Θxx + 2ξΘxy + ξ2Θyy ) dz. Following [6], we now pass to the new coordinates (x̃, ỹ, z̃, p̃, t̃) on T(M4) x 7→ t̃, w 7→ ỹ z 7→ x̃, −ξ 7→ p̃, y 7→ z̃ − p̃t̃. 26 M. Randall We obtain the distribution annihilated by the following 1-forms: ω̃3 = −dp̃− Ãdx̃, ω̃4 = −p̃dx̃+ dỹ, ω̃5 = dz̃ − p̃dt̃− t̃dp̃+ p̃dt̃− B̃dx̃ = dz̃ − t̃dp̃− B̃dx̃ = dz̃ + (t̃Ã− B̃)dx̃, where à and B̃ are coordinate transforms of the functions A(w, x, y, z, ξ) = Θxxx + 3Θyxxξ + 3Θyyxξ 2 + Θyyyξ 3, B(w, x, y, z, ξ) = Θxx + 2ξΘxy + ξ2Θyy respectively. This suggests taking à = −H ′(t), B̃ = −H(t) to obtain the Legendre transformed 1-forms in Section 6.3. Passing back to coordinates (w, x, y, z, ξ) on T(M4), this gives −H ′(x) = Θxxx + 3Θyxxξ + 3Θyyxξ 2 + Θyyyξ 3, −H(x) = Θxx + 2ξΘxy + ξ2Θyy, so that Θ(x) = − ∫∫ H(x)dxdx will satisfy the condition. We have Θxx = −H(x). We have the following theorem. Theorem 7.1. The An–Nurowski twistor distribution D on the circle twistor bundle T(M4)→ M4 of (M4, g) with the Plebański metric g = dwdx+ dzdy +H(x)dz2 and the function H(x) has split G2 as its group of local symmetries provided that H(x) is one of the following up to fractional linear transformations in s: 1. The function H(x) is given by H(x) = ∫∫ ṡ2 s 4 3 (s− 1) 4 3 dxdx, where s(x) is a solution to the 3rd order ODE (5.2) {s, x}+ ṡ2 2 V (s) = 0 with (α, β, γ) given by either ( 4 3 , 4 3 , 4 3 ) or ( 4 3 , 1 3 , 1 3 ) . 2. The function H(x) is given by H(x) = ∫∫ ṡ2 s 4 3 (s− 1) dxdx, where s(x) is a solution to (5.2) with (α, β, γ) one of ( 2 3 , 1 2 , 1 3 ) , ( 2 3 , 1 2 , 4 3 ) or ( 2 3 , 2, 1 3 ) . Flat (2, 3, 5)-Distributions and Chazy’s Equations 27 3. The function H(x) is given by H(x) = ∫∫ ṡ2 s 5 3 (s− 1) 5 3 dxdx, where s(x) is a solution to (5.2) with (α, β, γ) either ( 8 3 , 2 3 , 2 3 ) or ( 2 3 , 2 3 , 2 3 ) . 4. The function H(x) is given by H(x) = xm where m ∈ { −2,−1 2 , 1 2 , 2 } . 5. The function H(x) is given by H(x) = − 1 192 √ x+ C(4x+ 3B + C)√ x+B(B − C)3 . This corresponds to the solution obtained from (3.2). 6. The function H(x) is the Legendre transform of the function F (q) with q = H ′(x) and H(x) = qx− F (q) = xH ′(x)− F (H ′(x)). In this case F (q) can be given by one of the following: (a) F (q) = ∫∫ ṡ 3 2 s(s− 1) dqdq where s(q) is again a solution to the 3rd order ODE (5.2) with (α, β, γ) one of (3, 3, 3) or( 3, 13 , 1 3 ) . (b) F (q) = ∫∫ ṡ 3 2 s(s− 1) 3 4 dqdq, where s(q) is a solution to (5.2) with (α, β, γ) one of ( 3 2 , 1 3 , 1 2 ) , ( 3 2 , 3, 1 2 ) or ( 3 2 , 1 3 , 9 2 ) . (c) F (q) = ∫∫ ṡ 3 2 s 5 4 (s− 1) 5 4 dqdq, where s(q) is a solution to (5.2) with (α, β, γ) one of ( 6, 32 , 3 2 ) or ( 2 3 , 3 2 , 3 2 ) . (d) F (q) = qm, where m ∈ { −1, 13 , 2 3 , 2 } . (e) F (q) = −1 6 (q +B) 1 3 (q + C) 2 3 (B − C)2 , again corresponding to the solution (3.2). Acknowledgements This work is inspired by the paper of [6]. The author would like to thank Daniel An, Pawe l Nurowski, Travis Willse and the anonymous referees for comments. Part of this work is sup- ported by the Grant agency of the Czech Republic P201/12/G028. 28 M. Randall References [1] Ablowitz M.J., Chakravarty S., Halburd R., The generalized Chazy equation and Schwarzian triangle func- tions, Asian J. Math. 2 (1998), 619–624. [2] Ablowitz M.J., Chakravarty S., Halburd R., The generalized Chazy equation from the self-duality equations, Stud. Appl. Math. 103 (1999), 75–88. [3] Ablowitz M.J., Chakravarty S., Halburd R.G., Integrable systems and reductions of the self-dual Yang–Mills equations, J. Math. Phys. 44 (2003), 3147–3173. [4] Ablowitz M.J., Fokas A.S., Complex variables: introduction and applications, Cambridge Texts in Applied Mathematics, Cambridge University Press, Cambridge, 1997. [5] An D., Nurowski P., Twistor space for rolling bodies, Comm. Math. Phys. 326 (2014), 393–414, arXiv:1210.3536. [6] An D., Nurowski P., Symmetric (2, 3, 5) distributions, an interesting ODE of 7th order and Plebański metric, arXiv:1302.1910. [7] Atiyah M., Hitchin N., The geometry and dynamics of magnetic monopoles, M. B. Porter Lectures, Princeton University Press, Princeton, NJ, 1988. [8] Bor G., Hernández Lamoneda L., Nurowski P., The dancing metric, G2-symmetry and projective rolling, arXiv:1506.00104. [9] Cartan E., Les systèmes de Pfaff, à cinq variables et les équations aux dérivées partielles du second ordre, Ann. Sci. École Norm. Sup. (3) 27 (1910), 109–192. [10] Chakravarty S., Ablowitz M.J., Parameterizations of the Chazy equation, Stud. Appl. Math. 124 (2010), 105–135, arXiv:0902.3468. [11] Chakravarty S., Ablowitz M.J., Clarkson P.A., Reductions of self-dual Yang–Mills fields and classical sys- tems, Phys. Rev. Lett. 65 (1990), 1085–1087. [12] Chazy J., Sur les équations différentielles dont l’intégrale générale est uniforme et admet des singularités essentielles mobiles, C. R. Acad. Sci. Paris 149 (1910), 563–565. [13] Chazy J., Sur les équations différentielles du troisième ordre et d’ordre supérieur dont l’intégrale générale a ses points critiques fixes, Acta Math. 34 (1911), 317–385. [14] Clarkson P.A., Olver P.J., Symmetry and the Chazy equation, J. Differential Equations 124 (1996), 225–246. [15] Dunajski M., Anti-self-dual four-manifolds with a parallel real spinor, R. Soc. Lond. Proc. Ser. A Math. Phys. Eng. Sci. 458 (2002), 1205–1222, math.DG/0102225. [16] Dunajski M., Sokolov V., On the 7th order ODE with submaximal symmetry, J. Geom. Phys. 61 (2011), 1258–1262, arXiv:1002.1620. [17] Goursat É., Sur l’équation différentielle linéaire, qui admet pour intégrale la série hypergéométrique, Ann. Sci. École Norm. Sup. (2) 10 (1881), 3–142. [18] Leistner T., Nurowski P., Sagerschnig K., New relations between G2-geometries in dimensions 5 and 7, arXiv:1601.03979. [19] Nurowski P., Differential equations and conformal structures, J. Geom. Phys. 55 (2005), 19–49, math.DG/0406400. [20] Nurowski P., Notes, 2012, available at http://www.mat.univie.ac.at/~cap/esiprog/Nurowski.pdf. [21] Olver P.J., Equivalence, invariants, and symmetry, Cambridge University Press, Cambridge, 1995. [22] Tod K.P., Metrics with SD Weyl tensor from Painlevé-VI, Twistor Newsletter 35 (1992), 5–7, available at http://people.maths.ox.ac.uk/lmason/Tn/35/TN35-04.pdf. [23] Vidūnas R., Algebraic transformations of Gauss hypergeometric functions, Funkcial. Ekvac. 52 (2009), 139–180, math.CA/0408269. [24] Willse T., Highly symmetric 2-plane fields on 5-manifolds and 5-dimensional Heisenberg group holonomy, Differential Geom. Appl. 33 (2014), suppl., 81–111, arXiv:1302.7163. [25] Willse T., An explicit ambient metric of holonomy G∗2, arXiv:1411.7172. http://dx.doi.org/10.4310/AJM.1998.v2.n4.a1 http://dx.doi.org/10.1111/1467-9590.00121 http://dx.doi.org/10.1063/1.1586967 http://dx.doi.org/10.1007/s00220-013-1839-2 http://arxiv.org/abs/1210.3536 http://arxiv.org/abs/1302.1910 http://dx.doi.org/10.1515/9781400859306 http://dx.doi.org/10.1515/9781400859306 http://arxiv.org/abs/1506.00104 http://dx.doi.org/10.1111/j.1467-9590.2009.00463.x http://arxiv.org/abs/0902.3468 http://dx.doi.org/10.1103/PhysRevLett.65.1085 http://dx.doi.org/10.1007/BF02393131 http://dx.doi.org/10.1006/jdeq.1996.0008 http://dx.doi.org/10.1098/rspa.2001.0918 http://dx.doi.org/10.1098/rspa.2001.0918 http://arxiv.org/abs/math.DG/0102225 http://dx.doi.org/10.1016/j.geomphys.2011.02.021 http://arxiv.org/abs/1002.1620 http://arxiv.org/abs/1601.03979 http://dx.doi.org/10.1016/j.geomphys.2004.11.006 http://arxiv.org/abs/math.DG/0406400 http://www.mat.univie.ac.at/~cap/esiprog/Nurowski.pdf http://dx.doi.org/10.1017/CBO9780511609565 http://people.maths.ox.ac.uk/lmason/Tn/35/TN35-04.pdf http://dx.doi.org/10.1619/fesi.52.139 http://arxiv.org/abs/math.CA/0408269 http://dx.doi.org/10.1016/j.difgeo.2013.10.010 http://arxiv.org/abs/1302.7163 http://arxiv.org/abs/1411.7172 1 Introduction 2 Chazy's equation 3 Generalised Chazy equations 4 Relationship to ODE studied by Dunajski and Sokolov 5 First order system and different parametrisations of Chazy's equations 6 Examples of flat (2,3,5)-distributions 6.1 Chazy's solution 6.2 Other parametrisations of the generalised Chazy equation 6.3 Legendre transformed coframe 6.4 Additional examples 7 An–Nurowski circle twistor bundle References
id nasplib_isofts_kiev_ua-123456789-147724
institution Digital Library of Periodicals of National Academy of Sciences of Ukraine
issn 1815-0659
language English
last_indexed 2025-11-30T17:03:05Z
publishDate 2016
publisher Інститут математики НАН України
record_format dspace
spelling Randall, M.
2019-02-15T18:43:43Z
2019-02-15T18:43:43Z
2016
Flat (2,3,5)-Distributions and Chazy's Equations / M. Randall // Symmetry, Integrability and Geometry: Methods and Applications. — 2016. — Т. 12. — Бібліогр.: 25 назв. — англ.
1815-0659
2010 Mathematics Subject Classification: 58A30; 53A30; 34A05; 34A34
DOI:10.3842/SIGMA.2016.029
https://nasplib.isofts.kiev.ua/handle/123456789/147724
n the geometry of generic 2-plane fields on 5-manifolds, the local equivalence problem was solved by Cartan who also constructed the fundamental curvature invariant. For generic 2-plane fields or (2,3,5)-distributions determined by a single function of the form F(q), the vanishing condition for the curvature invariant is given by a 6th order nonlinear ODE. Furthermore, An and Nurowski showed that this ODE is the Legendre transform of the 7th order nonlinear ODE described in Dunajski and Sokolov. We show that the 6th order ODE can be reduced to a 3rd order nonlinear ODE that is a generalised Chazy equation. The 7th order ODE can similarly be reduced to another generalised Chazy equation, which has its Chazy parameter given by the reciprocal of the former. As a consequence of solving the related generalised Chazy equations, we obtain additional examples of flat (2,3,5)-distributions not of the form F(q)=qm. We also give 4-dimensional split signature metrics where their twistor distributions via the An-Nurowski construction have split G₂ as their group of symmetries.
This work is inspired by the paper of [6]. The author would like to thank Daniel An, Pawe l Nurowski, Travis Willse and the anonymous referees for comments. Part of this work is supported by the Grant agency of the Czech Republic P201/12/G028.
en
Інститут математики НАН України
Symmetry, Integrability and Geometry: Methods and Applications
Flat (2,3,5)-Distributions and Chazy's Equations
Article
published earlier
spellingShingle Flat (2,3,5)-Distributions and Chazy's Equations
Randall, M.
title Flat (2,3,5)-Distributions and Chazy's Equations
title_full Flat (2,3,5)-Distributions and Chazy's Equations
title_fullStr Flat (2,3,5)-Distributions and Chazy's Equations
title_full_unstemmed Flat (2,3,5)-Distributions and Chazy's Equations
title_short Flat (2,3,5)-Distributions and Chazy's Equations
title_sort flat (2,3,5)-distributions and chazy's equations
url https://nasplib.isofts.kiev.ua/handle/123456789/147724
work_keys_str_mv AT randallm flat235distributionsandchazysequations