Funkcialaj Ekvacioj, 41 (1998) 483-503 Higher Order Painleve Equations of Type $A_{l}^{(1)}$ By Masatoshi NOUMI and Yasuhiko YAMADA (Kobe University, Japan) Absffact. of type A series of systems of nonlinear equations with affine Weyl group symmetry is studied. This series gives a generalization of Painleve equations to higher orders. $A_{l}^{(1)}$ and 1. $P_{¥mathrm{V}}$ $P_{¥mathrm{I}¥mathrm{V}}$ Introduction In this paper we propose a series of systems of nonlinear differential equations which have symmetry under the affine Weyl groups of type $(l ¥geq 2)$ . These systems are thought of as higher order analogues of the Painleve . We will study in particular their Hamiltonian structures equations and properties of their -functions. and some basic , we consider a differential system for $(f+1)$ unknown For each $/=2,3$ , functions , , containing complex parameters , _’ which correspond . In what follows, we to the simple roots of the affine root system of type $+a_{l}=k$ . Our differential system is defined as follows according to set $a_{0}+$ the parity of : For $=2n(n =1,2, ¥ldots)$ $A_{l}^{(1)}$ $P_{¥mathrm{V}}$ $P_{¥mathrm{I}¥mathrm{V}}$ $¥mathrm{r}$ $¥ldots$ $f_{0}$ $¥ldots,f_{l}$ $¥alpha_{0}$ $¥alpha_{l}$ $A_{I}^{(1)}$ ? $l$ $l$ (0. 1) where (0.1) $A_{2n}^{(1)}$ $0¥leq j¥leq 2n$ $A_{2n+1}^{(1)}$ : : $f_{j}^{¥prime}=f_{j}(¥sum_{1¥leq r¥leq n}f_{j+2r-1}-¥sum_{1¥leq r¥leq n}f_{j+2r})+¥alpha_{j}$ , and for $f=2n+1(n =1,2, ¥ldots)$ $f_{j}^{¥prime}=f_{j}(¥sum_{1¥leq r¥leq s¥leq n}f_{j+2r-1}f_{j+2s}-¥sum_{1¥leq r¥leq s¥leq n}f_{j+2r}f_{j+2s+1)}$ $+(¥frac{k}{2}-¥sum_{1¥leq r¥leq n}¥alpha_{j+2r})f_{j}+¥alpha_{j}(¥sum_{1¥leq r¥leq n}f_{j+2r})$ , where $0¥leq j¥leq 2n+1$ . In (0.1) and (0.2), ’ stands for the derivation $d/dt$ with for small 1 is respect to an independent variable . A table of formulas of are obtained given in Appendix for convenience. Formulas for the other from those simply by the rotation of indices. It can be shown that the differential systems for $1=2$ and $=3$ are equivalent to the fourth and the fifth Painleve equations, respectively (see [9]). $t$ $f_{0}^{¥prime}$ $f_{j}^{¥prime}$ $l$ 484 Masatoshi NOUMI and Yasuhiko YAMADA This paper is organized as follows. In Section 1, we discuss symmetry of our system under the affine Weyl group of type , by describing explicit Backlund transformations. After formulating a Poisson structure for our system, we will construct in Section 3 certain canonical coordinates with respect to the Poisson structure, and show that our system can be equivalently written as a Hamiltonian system with a polynomial Hamiltonian. In the final we introduce a family of -functions for our system and discuss their Backlund transformations. In particular we will show that the variables are expressed $A_{l}^{(1)}$ $¥mathrm{s}¥mathrm{e}¥mathrm{c}¥mathrm{t}¥mathrm{i}¥mathrm{o}¥mathrm{n}_{>}$ $¥mathrm{r}$ $f_{j}$ multiplicatively in terms of the -functions and their Backlund transformations. This shows that our -functions are consistent with those introduced in [8] from the viewpoint of discrete dynamical systems. $¥mathrm{r}$ $¥mathrm{r}$ 1. Affine Weyl group symmetry be the field of rational functions in and $f=$ . Then the differential system (0. 1) (resp. (0.2)) defines the structure of a differential field on $C(¥alpha; f)$ . We say that an automorphism of $C(¥alpha; f)$ is a Backfund transformation of the differential system if it commutes with the derivation ’. For each $i=0,1$ , of $C(¥alpha; f)$ as , we define an automorphism follows: Let $C(¥alpha; f)$ $¥alpha=$ $(¥alpha_{0}, ¥ldots, ¥alpha_{l})$ $(f_{0}, ¥ldots,f_{l})$ $f$ $s_{i}$ _’ $s_{i}(¥alpha_{i})=-¥alpha_{i}$ (1.1) $s_{i}(f_{i})=f_{i}$ , , $s_{i}(¥alpha_{j})=¥alpha_{j}+¥alpha_{i}(j=i¥pm 1)$ $s_{i}(f_{j})=f_{j}¥pm¥frac{¥alpha_{i}}{f_{i}}(j=i¥pm 1)$ where the indices 0, 1, . . . define an automorphism (1.2) $f$ ’ $¥pi$ , , $s_{i}(¥alpha_{j})=¥alpha_{j}(j¥neq i, i¥pm 1)$ $s_{i}(f_{j})=f_{j}(j¥neq i, i¥pm 1)$ , , are understood as elements of $Z/(l+1)Z$ . We also of $C(¥alpha; f)$ $¥pi(¥alpha_{j})=¥alpha_{j+1}$ , by $¥pi(f_{j})=f_{j+1}$ . Theorem 1.1. ([8]). The automorphisms and , define a representation of the extended affine Weyl group . Namely, they satisfy the commutation relations type $s_{0},s_{1}$ _’ $s_{l}$ $¥pi$ described above $¥tilde{W}=¥langle s_{0}, ¥_’ s_{l}, ¥pi¥rangle$ of $A_{l}^{(1)}$ (1.3) for $i,j$ (1.4) $s_{i}^{2}=1$ $=0,1$ , , $s_{i}s_{j}=s_{j}s_{i}(j¥neq i, i¥pm 1)$ $f$ _’ , $s_{i}s_{j}s_{i}=s_{j}s_{i}s_{j}(j=i¥pm 1)$ and $¥pi^{l+1}=1$ , $¥pi s_{i}=s_{i+1}¥pi(i=0,1, -’ l)$ . Note that the extended affine Weyl group $¥tilde{W}=W¥mathrm{x}¥{1, ¥pi, ¥_’ ¥pi^{l}¥}$ is the extension of the ordinary affine Weyl group by a of type cyclic group of order 1+1 generated by the diagram rotation . $ W=¥langle s_{0}, ¥ldots,s_{l}¥rangle$ $A_{l}^{(1)}$ $¥pi$ Higher Order Painleve Equations of Type 485 $A_{l}^{(1)}$ Theorem 1.2. The action of defined as above commutes with the $C( ¥ alpha; f)$ . derivation of the differential fiefd $¥tilde{W}$ In this sense, our differential system (0.1) (resp. (0.2)) admits the action of type , as a group of Backlund of the extended affine Weyl group transformations. Theorem 1.2 can be checked essentially by direct computations. We will explain below how such computations can be carried out. $¥tilde{W}$ Remark 1.3. A $l(1)$ Let $A$ $=(a_{ij})_{0¥leq i,j¥leq l}$ $A_{l}^{(1)}$ be the generalized Cartan matrix of type : (1.5) $a_{jj}=2$ Then the action of (1.6) , $a_{ij}=-1(j =i¥pm 1)$ , $s_{0}$ , _’ $s_{l}$ on the simple roots We introduce an $(l+1)¥times(l+1)$ (1.7) $u_{ij}=¥pm 1(j =i¥pm 1)$ , matrix Then the Backlund transformations $s_{0}$ $s_{i}(f_{j})=f_{j}+¥frac{¥alpha_{i}}{f_{i}}u_{ij}$ $¥alpha_{0}$ , _’ $(¥mathrm{i}¥mathrm{j} =0,1, ¥_’ ¥mathit{1})$ $s_{i}(¥alpha_{j})=¥alpha_{j}-¥alpha_{i}a_{ij}$ (1.8) $a_{ij}=0(j ¥neq i, i¥pm 1)$ . , . $U=(u_{ij})_{0¥leq i,j¥leq l}$ by setting $u_{ij}=0(j ¥neq i¥pm 1)$ _’ $s_{l}$ is described as $¥alpha_{l}$ . are determined by the formulas $(¥mathrm{i}¥mathrm{j} =0,1, ¥_’ I)$ . For a general treatment of affine Weyl group symmetry in terms of these data $A$ and $U$, as well as discrete dynamical systems arising from Backlund transformations, we refer the reader to [8]. For practical computations of Backlund transformations, it is convenient to $(i =0, ¥_’ f)$ defined by use the Demazure operators $¥Delta_{i}$ (1.9) The action of $¥Delta_{i}(¥varphi)=¥frac{1}{¥alpha_{i}}(s_{i}(¥varphi)-¥varphi)$ $¥Delta_{i}$ $(¥varphi¥in C(¥alpha;f))$ . on a product can be determined by the twisted Leibniz rule (1.10) $¥Delta_{i}(¥varphi¥psi)=¥Delta_{i}(¥varphi)¥psi+s_{i}(¥varphi)¥Delta_{i}(¥psi)$ We remark that $¥Delta_{i}(¥varphi)=0$ $¥Delta_{i}(¥varphi¥psi)=¥varphi¥Delta_{i}(¥psi)$ when $¥varphi$ . is -invariant and that one has if and only if is -invariant. Note that (1.6) and (1.8) can be $¥varphi$ $¥mathrm{s}_{¥mathrm{i}}$ $¥mathrm{s}_{¥mathrm{i}}$ rewritten as (1.11) $¥Delta_{i}(¥alpha_{j})=-a_{ij}$ , $¥Delta_{i}(f_{j})=¥frac{u_{ij}}{f_{i}}(i,j=0, -’ ¥mathit{1})$ , Masatoshi NOUMI and Yasuhiko YAMADA 486 Namely one has in terms of the Demazure operators. $¥Delta_{i}(¥alpha_{i})=-2$ (1.12) $¥Delta_{i}(f_{i})=0$ , , $¥Delta_{i}(¥alpha_{j})=1(j=i¥pm 1)$ , $¥Delta_{i}(f_{j})=¥frac{¥pm 1}{f_{i}}(j=i¥pm 1)$ $¥Delta_{i}(¥alpha_{j})=0(j¥neq i, i¥pm 1)$ , , $¥Delta_{i}(f_{j})=0(j¥neq i, i¥pm 1)$ . It is also well known that the relations (1.3) imply (1.13) for $¥Delta_{i}^{2}=0$ $i,j$ $=0$ , $¥ldots$ , $l$ , , $¥Delta_{i}¥Delta_{j}=¥Delta_{j}¥Delta_{i}(j¥neq i¥pm 1)$ $¥Delta_{i}¥Delta_{j}¥Delta_{i}=¥Delta_{j}¥Delta_{i}¥Delta_{j}(j=i¥pm 1)$ , . , , we denote by $F_{j}=F_{j}(¥alpha; f)$ Theorem 1.2. For each $j=0$ , the polynomial appearing on the right-hand side of (0.1) or (0.2). Then it is if and only if easy to see that Proof of $f$ $¥ldots$ $s_{i}(f_{j})^{¥prime}=s_{i}(f_{j}^{¥prime})$ (1.14) $(i,j =0, ¥ldots, f)$ $F_{j}-¥frac{¥alpha_{i}}{f_{i}^{2}}F_{i}u_{ij}=s_{i}(F_{j})$ or equivalently, (1.15) $¥Delta_{i}(F_{j})=-¥frac{u_{ij}}{f_{i}^{2}}F_{i}$ in terms of the Demazure operators. check (1.15) for $j=0$ : (1.16) $¥Delta_{1}(F_{0})=¥frac{F_{1}}{f_{1}^{2}}$ , $(i,j =0, ¥ldots, l)$ By the rotation symmetry, it is enough to $¥Delta_{l}(F_{0})=-¥frac{F_{1}}{f_{l}^{2}}$ , $¥Delta_{i}(F_{0})=0(i¥neq 0, f)$ We will show for example the equality $2n+1$ , separately. When $f=2n$ , we have $¥Delta_{1}(F_{0})=F_{1}/f_{1}^{2}$ (1.17) Since (1.18) . for $f=2n$ and $F_{0}=f_{0}(f_{1}-f_{2}+¥cdots-f_{2n})+¥alpha_{0}$ . $¥Delta_{1}(f_{j})=0$ for $j¥neq 0,2$ , we have $¥Delta_{1}(F_{0})=¥Delta_{1}(f_{0})(¥sum_{i=1}^{2n}(-1)^{i-1}f_{i})-s_{1}(f_{0})¥Delta_{1}(f_{2})+¥Delta_{1}(¥alpha_{0})$ $=-¥frac{1}{f_{1}}(¥sum_{i=1}^{2n}(-1)^{i-1}f_{i})-(f_{0}-¥frac{¥alpha_{1}}{f_{1}})¥frac{1}{f_{1}}+1$ $=¥frac{1}{f_{1}}(¥sum_{i=2}^{2n}(-1)^{i}f_{i}-f_{0})+¥frac{¥alpha_{1}}{f_{1}^{2}}=¥frac{1}{f_{1}^{2}}F_{1}$ . $l$ $=$ Higher Order Painleve Equations When (1.19) $l$ of Type $=2n+1$ , we have $F_{0}=f_{0}(f_{1}f_{2}+f_{1}f_{4}+¥cdots+f_{2n-1}f_{2n}-f_{2}f_{3}¥_¥cdots-f_{2n}f_{2n+1})$ $+(¥frac{k}{2}-¥alpha_{2}-¥alpha_{4}¥_¥cdots-¥alpha_{2n})f_{0}+¥alpha_{0}(f_{2}+f_{4}+¥cdots+f_{2n})$ Hence, 487 $A_{l}^{(1)}$ $¥Delta_{1}(F_{0})$ . is computed as follows: $¥Delta_{1}(f_{0})(¥sum_{1¥leq r¥leq s¥leq n}f_{2r-1}f_{2s}-¥sum_{1¥leq r¥leq s¥leq n}f_{2r}f_{2s+1})+s_{1}(f_{0})¥Delta_{1}(f_{2})(f_{1}-¥sum_{r=1}^{n}f_{2r+1})$ $-¥Delta_{1}(¥alpha_{2})f_{0}+s_{1}(¥frac{k}{2}-¥sum_{r=1}^{n}¥alpha_{2r})¥Delta_{1}(f_{0})+¥Delta_{1}(¥alpha_{0})(¥sum_{r=1}^{n}f_{2r})+s_{1}(¥alpha_{0})¥Delta_{1}(f_{2})$ $=-¥frac{1}{f_{1}}(¥sum_{1¥leq r¥leq s¥leq n}f_{2r-1}f_{2s}-¥sum_{1¥leq r¥leq s¥leq n}f_{2r}f_{2s+1})$ $+(f_{0}-¥frac{¥alpha_{1}}{f_{1}})¥frac{1}{f_{1}}(f_{1}-¥sum_{r=1}^{n}f_{2r+1})-f_{0}+(¥frac{k}{2}-¥sum_{r=1}^{n}¥alpha_{2r+1}-¥alpha_{0})¥frac{1}{f_{1}}$ $+(¥sum_{r=1}^{n}f_{2r})+(¥alpha_{0}+¥alpha_{1})¥frac{1}{f_{1}}$ $=¥frac{1}{f_{1}}(¥sum_{1¥leq r¥leq s¥leq n}f_{2r}f_{2s+1}-¥sum_{2¥leq r¥leq s¥leq n}f_{2r-1}f_{2s}-¥sum_{r=1}^{n}f_{2r+1}f_{0})$ $+(¥frac{k}{2}-¥sum_{r=1}^{n}¥alpha_{2r+1})¥frac{1}{f_{1}}+¥frac{¥alpha_{1}}{f_{1}^{2}}(¥sum_{r=1}^{n}f_{2r+1})=¥frac{1}{f_{1}^{2}}F_{1}$ . The other formulas in (1.16) can be checked similarly. 2. $¥square $ Poisson structure $¥{, ¥}$ By using the matrix $U$ defined in (1.7), we introduce the Poisson bracket on $C(¥alpha; f)$ as follows: (2. 1) Note that (2.2) $¥{¥varphi, ¥psi¥}=¥sum_{0¥leq i,j¥leq l}¥frac{¥partial¥varphi}{¥partial f_{i}}u_{ij}¥frac{¥partial¥psi}{¥partial f_{j}}$ $¥{, ¥}$ is a $¥mathrm{C}(¥mathrm{a})$ . -bilinear skewsymmetric form such that $¥{f_{i},f_{j}¥}=u_{ij}$ $(¥mathrm{i},¥mathrm{j} =0,1, ¥_’ l)$ , 488 Masatoshi NOUMI and Yasuhiko YAMADA namely, (2.3) for $¥{f_{i},f_{j}¥}=¥pm 1(j=i¥pm 1)$ $i,j$ $=0,1$ , $¥_’$ /. , $¥{f_{i},f_{j}¥}=0(j¥neq i¥pm 1)$ By direct calculations, one can show invarian Proposition 2.1. The skewsymmetric form ¥ ¥ above defines a $C( ¥ alpha; f)$ . Namely, one has Poisson structure on the differential fiefd , , (1) , (2) . (3) $w(¥{¥varphi, ¥psi¥})=¥{w(¥varphi), w(¥psi)¥}$ for any , $¥psi¥in C(¥alpha; f)$ and $ {, $¥{¥varphi, ¥psi_{1}¥psi_{2}¥}=¥psi_{1}¥{¥varphi, ¥psi_{2}¥}+¥{¥varphi, ¥psi_{1}¥}¥psi_{2}$ $¥tilde{W}-$ }$ $¥tau$ $¥{¥varphi_{1}¥varphi_{2}, ¥psi¥}=¥varphi_{1}¥{¥varphi_{2}, ¥psi¥}+¥{¥varphi_{1}, ¥psi¥}¥varphi_{2}$ $¥{¥varphi_{1}, ¥{¥varphi_{2}, ¥varphi_{3}¥}¥}+¥{¥varphi_{2}, ¥{¥varphi_{3}, ¥varphi_{1}¥}¥}+¥{¥varphi_{3}, ¥{¥varphi_{1}, ¥varphi_{2}¥}¥}=0$ $w¥in¥tilde{W}$ $¥varphi$ We remark that our Poisson bracket has a nontrivial radical. Regard -dimensional vector space $E=$ as a skewsymmetric form on the . Then its radical is precisely the subspace of all linear combisuch that ¥ for with coefficients in nations $i=0$ , _’ . It is a one-dimensional subspace generated by $¥{, ¥}$ $(¥mathrm{f}+1)$ $¥oplus_{j=0}^{l}C(¥alpha)f_{j}$ $c_{¥mathit{0}}f_{¥mathit{0}}+$ ? $ sum_{j=0}^{l}u_{ij}c_{j}=0$ $C(¥alpha)$ $+c_{l}f_{l}$ $¥mathrm{a}1¥mathrm{I}$ $l$ (2.4) if $g=f_{0}+f_{1}+-+f_{2n}$ $=2n$ , $l$ and is a two-dimensional subspace generated by (2.5) if $g_{0}$ $g_{0}=f_{0}+f_{2}+-+f_{2n}$ $¥mathit{1}=2n+1$ , $g_{1}$ are , $g_{1}=f_{1}+f_{3}+-+f_{2n+1}$ , respectively. Note that is -invariant, and that $g_{0}+g_{1}$ is $g$ $¥tilde{W}-$ $¥mathrm{W}$ , invariant when $f=2n$ , and that invariant when $=2n+1$ . $¥tilde{W}-$ $l$ We will describe below our differential systems (0.1) and (0.2) by means of the Poisson structure introduced above. In order to define a“Hamiltonian” we need to fix some notation. the -th fundamental weight of the For each $i=1$ , , , we denote by finite root system of type $l$ $¥mathrm{i}$ $¥varpi_{i}$ $¥ldots$ $A_{l}$ (2.6) $¥varpi_{i}=¥frac{1}{l+1}¥{(l+1-i)¥sum_{r=1}^{i}r¥alpha_{r}+i¥sum_{r=i+1}^{l}(l+1-r)¥alpha_{r}¥}$ $=¥sum_{r=1}^{l}$ $¥min¥{i, r¥}$ $-¥frac{ir}{l+1}$ ) $¥alpha_{r}$ form the dual basis of the simple roots Note that , _’ such that , , with respect to the symmetric bilinear form $=1$ , _’ ). are then expressed as The simple affine roots , and set $¥alpha_{1}$ ( $¥ldots$ $¥mathrm{i},¥mathrm{j}$ (2.7) $¥varpi_{0}=0$ . $¥varpi_{1}$ $¥varpi_{l}$ $ $ $¥langle¥alpha_{i}, ¥alpha_{j}¥rangle=a_{ij}$ $¥langle, ¥rangle$ $¥alpha_{l}$ $ f$ $¥alpha_{0}$ _’ $¥alpha_{j}=-¥varpi_{j-1}+2¥varpi_{j}-¥varpi_{j+1}+¥delta_{j,0}k$ $¥alpha_{f}$ $(j =1, ¥ldots, f)$ , Higher Order Painleve Equations , in terms of the fundamental weights as weights fundamental act on the so, $¥varpi_{1}$ _’ $¥varpi_{l}$ of Type 489 $A_{l}^{(1)}$ We remark that the reflections . $¥ldots,s_{l}$ (2.8) $s_{0}(¥varpi_{j})=¥varpi_{j}+¥alpha_{0}$ , $s_{i}(¥varpi_{j})=¥varpi_{j}-¥delta_{i,j}¥alpha_{i}(i=1, ¥ldots, l)$ , , acts on , . Notice that the diagram rotation for each $j=1$ , nontrivially. ; it is a circle with $(l+1)$ nodes be the Dynkin diagram of type Let labeled by the elements of $Z/(f+1)Z$ . For each chain $C$ of , consisting of consecutive nodes $j,j+1$ , _’ $j+m-1$ $(m ¥leq f)$ , we denote by $¥chi(C)$ the alternating sum of corresponding fundamental weights: $f$ $¥pi$ _’ $¥varpi_{1}$ $¥ldots$ $¥varpi_{l}$ $A_{l}^{(1)}$ $¥Gamma$ $¥Gamma$ (2.9) $¥chi(C)=¥varpi_{j}-¥varpi_{j+1}+¥cdots+(-1)^{m-1}¥varpi_{j+m-1}$ . with , we denote by $¥chi(C)=¥sum_{i}¥chi(C_{i})$ the For each subdiagram $C$ of over all connected components . sum of $f+1$ , we denote by the set of all subsets For each $d=1$ , $K¥subset¥{0,1, ¥_’ ¥mathit{1}¥}$ with cardinality such that the connected components of the diagram $ K^{c}=¥Gamma$ , obtained by removing the nodes of $K$, are all chains of even ¥ ¥ is . We remark that the set nodes. For each $K¥in¥Psi_{d}$ , we set $K$ with a of nonempty if and only if $f+1-d$ is even, and that subset if and only if it has an expression $|K|=d$ then belongs to $K=¥{k_{1}, k_{2}, ¥_’ k_{d}¥}$ with a sequence $0¥leq k_{1}<k_{2}<¥cdots<k_{d}¥leq f$ such that $ C¥neq¥Gamma$ $¥Gamma$ $¥chi(C_{i})$ $C_{i}$ $¥Psi_{d}$ _’ $d$ $f_{K}= prod_{i in K}f_{i}$ $¥Psi_{d}$ $¥{0, ¥_’ ¥mathit{1}¥}$ $¥mathit{9}_{d}^{2}$ (2. 10) $(k_{1}, k_{2}, -’ k_{d})¥equiv(0,1,0, -)$ With this expression, (2. 11) $¥chi(K^{c})$ or $(1, 0, 1, ¥_)$ $¥mathrm{m}¥mathrm{o}¥mathrm{d}2$ . can be written as follows: $¥chi(K^{c})=¥sum_{i=0}^{d-1}¥sum_{r=1}^{k_{i+1}-k_{i}-1}(-1)^{r-1}¥varpi_{k_{¥mathrm{i}}+r}$ where $k_{0}=k_{d}-l-1$ . With the notation as above, we define a Hamiltonian (2. 12) $h_{0}$ for our system by $h_{0}=¥sum_{K¥in¥Psi_{3}}f_{K}+¥sum_{K¥in¥Psi 1}¥chi(K^{c})f_{K}$ $=¥sum f_{k_{1}}f_{k_{2}}f_{k_{3}}+¥sum_{i=0}^{2n}(¥sum_{r=1}^{2n}(-1)^{r-1}¥varpi_{i+r})f_{i}$ when $l$ $(k_{1},k_{2},k_{3})$ (1, 0, 1) (2. 13) is taken over all triples where the summation such that $0¥leq k_{1}<k_{2}<k_{3}¥leq 2n$ and that $(k_{1},k_{2}, k_{3})¥equiv(0,1,0)$ or by . When $f=2n+1$ , we define $=2n$ , $¥sum f_{k_{1}}f_{k_{2}}f_{k_{3}}$ $¥mathrm{m}¥mathrm{o}¥mathrm{d}2$ $h_{0}$ $h_{0}=¥sum_{K¥in¥Psi_{4}}f_{K}+¥sum_{K¥in¥ovalbox{¥tt¥small REJECT}_{2}}¥chi(K^{c})f_{K}+(¥sum_{i=1}^{2n+1}(-1)^{i-1}¥varpi_{i})^{2}$ . 490 Masatoshi NOUMI and Yasuhiko YAMADA We remark that, when $f=2n+1$ , the sum appearing in the constant term of (2. 13) has an alternative expression (2. 14) $¥sum_{i=1}^{2n+1}(-1)^{i-1}¥varpi_{i}=¥frac{1}{2}¥sum_{r=0}^{n}¥alpha_{2r+1}$ by simple roots. We give in Appendix some explicit formulas of for convenience. Proposition 2.2. follows: (2. 15) When $I$ $=2n$ , the as $h_{0}$ differential system (0.1) can $f_{j}^{¥prime}=¥{h_{0},f_{j}¥}+¥delta_{j,0}k(j=0, -’ 2n)$ for small $l$ be expressed . Hence one has (2. 16) When $¥varphi^{¥prime}=¥{h_{0}, ¥varphi¥}+k¥frac{¥partial¥varphi}{¥partial f_{0}}$ $l$ $=2n+1$ , the (2. 17) with defined $g_{0}$ . (0.2) can be expressed as follows: by (2.5). , Hence one has $¥varphi^{¥prime}=¥{h_{0}, ¥varphi¥}-¥frac{k}{2}(¥sum_{i=0}^{2n+1}(-1)^{i}f_{i}¥frac{¥partial¥varphi}{¥partial f_{i}})+kg_{0}¥frac{¥partial¥varphi}{¥partial f_{0}}$ any $¥varphi¥in C(¥alpha; f)$ Proof. (2. 19) system $¥varphi¥in C(¥alpha; f)$ $f_{j}^{¥prime}=¥{h_{0},f_{j}¥}-(-1)^{j}¥frac{k}{2}f_{j}+¥delta_{j,0}kg_{0}(j=0, -’ 2n+1)$ (2. 18) for differential for any . For each $j=0$ , $X_{j}(¥varphi)=$ $f$ _’ , we define the vector field $¥{¥varphi,f_{j}¥}=(¥frac{¥partial}{¥partial f_{j-1}}-¥frac{¥partial}{¥partial f_{j+1}})¥varphi$ We now consider the case of computed as follows: $A_{2n}^{(1)}$ . $X_{j}$ by . From the definition (2.12), $X_{j}(h_{0})$ $¥sum_{K¥in¥ovalbox{¥tt¥small REJECT}_{2}(¥Gamma¥backslash ¥{j-1¥})}f_{K}-¥sum_{K¥in¥ovalbox{¥tt¥small REJECT}_{2}(¥Gamma¥backslash ¥{j+1¥})}f_{K}+¥chi(¥Gamma¥backslash ¥{j-1¥})-¥chi(¥Gamma¥backslash ¥{j+1¥})$ $=f_{j}¥sum_{r=1}^{n}f_{j+2r-1}-f_{j}¥sum_{r=1}^{n}f_{j+2r}-¥varpi_{j-1}+2¥varpi_{j}-¥varpi_{j+1}$ $=f_{j}(¥sum_{i=1}^{2n}(-1)^{i-1}f_{j+i})+¥alpha_{j}-¥delta_{j,0}k=F_{j}-¥delta_{j,0}k$ , is Higher Order Painleve Equations with the notation of of Type 491 $A_{l}^{(1)}$ extended to subdiagrams of . This $G$ proves (2.15). (For a subdiagram of , stands for the set of all $K ¥ subset G$ subsets with $|K|=d$ such that the connected components of are all chains of even nodes.) Formulas (2. 17) for the case of can be established in a similar way. In fact is given by $¥Gamma=¥{0,1, ¥ldots, ¥mathit{1}¥}$ $¥Psi_{d}$ $¥Gamma$ $¥ovalbox{¥tt¥small REJECT}_{d}(G)$ $G¥backslash K$ $A_{2n+1}^{(1)}$ $X_{j}(h_{0})$ $¥sum_{K¥in¥ovalbox{¥tt¥small REJECT}_{3}(¥Gamma¥backslash ¥{j-1¥})}f_{K}-¥sum_{K¥in¥Psi_{3}(¥Gamma¥backslash ¥{j+1¥})}f_{K}$ $+¥sum_{¥{i¥}¥in¥ovalbox{¥tt¥small REJECT}_{1}(¥Gamma¥backslash ¥{j-1¥})}¥chi(¥Gamma¥backslash ¥{j-1, i¥})f_{i}-¥sum_{¥{i¥}¥in¥ovalbox{¥tt¥small REJECT}_{1}(¥Gamma¥backslash ¥{j+1¥})}¥chi(¥Gamma¥backslash ¥{j+1, i¥})f_{i}$ $=f_{j}(¥sum_{K¥in¥Psi_{2}([j+1,j-2])}f_{K}-¥sum_{K¥in¥Psi_{2}([j+2,j-1])}f_{K})$ $+(-¥varpi_{j-1}+2¥varpi_{j}-¥varpi_{j+1}+(-1)^{j}2¥sum_{i=1}^{2n+1}(-1)^{i-1}¥varpi_{i})f_{j}$ $+(-¥varpi_{j-1}+2¥varpi_{j}-¥varpi_{j+1})¥sum_{r=1}^{n}f_{j+2r}$ $=f_{j}(¥sum_{K¥in¥Psi_{2}([j+1,j-2])}f_{K}-¥sum_{K¥in¥ovalbox{¥tt¥small REJECT}_{2}([j+2,j-1])}f_{K})$ $+(¥alpha_{j}-¥delta_{j,0}k+(-1)^{j}¥sum_{r=0}^{n}¥alpha_{2r+1})f_{j}+(¥alpha_{j}-¥delta_{j,0}k)¥sum_{r=1}^{n}f_{j+2r}$ $=F_{j}+(-1)^{j}¥frac{k}{2}f_{j}-¥delta_{j,0}k(¥sum_{r=0}^{n}f_{2r})$ where $[a, b]$ stands for the chain , $¥{a, a+1, ¥ldots, b-1, b¥}$ . This proves (2. 17). $¥square $ 3. Canonical coordinates and the Hamiltonian system By using Proposition 2.2, we introduce canonical coordinates for our differential system. We discuss the two cases of and separately. $A_{2n}^{(1)}$ Case (3.1) $A_{2n}^{(1)}$ $A_{2n+1}^{(1)}$ : In view of (2.3), we define a new coordinate system $(q;p;x)=(q_{1}, ¥ldots, q_{n};p_{1}, ¥ldots,p_{n};x)$ 492 Masatoshi NOUMI and Yasuhiko YAMADA for the affine space with coordinates (3.2) $q_{1}=f_{2}$ , $q_{2}=f_{4}$ $p_{1}=f_{1}$ , $p_{2}=f_{1}+f_{3}$ , , $¥ldots$ $f=(f_{0}, -’ f_{2n})$ as follows: , $q_{n}=f_{2n}$ $p_{n}=f_{1}+f_{3}+-+f_{2n-1}$ , ,?’ $x=g=f_{0}+f_{1}+¥cdots+f_{2n}$ . Note that the inverse of this coordinate transformation is given by $f_{0}=x-q_{1}-q_{2}¥_--q_{n}-p_{n}$ , (3.3) $f_{1}=p_{1}$ , , $f_{2}=q_{1}$ $f_{2n-1}=p_{n}-p_{n-1}$ $f_{3}=p_{2}-p_{1}$ , $f_{2n}=q_{n}$ , $f_{4}=q_{2}$ ,?’ . Then it is easy to show that (3.4) for $i,j$ $¥{p_{i}, q_{j}¥}=¥delta_{i,j}$ $=1$ , _’ (3.5) $n$ , $¥{q_{i}, q_{j}¥}=¥{p_{i},p_{j}¥}=¥{p_{i},x¥}=¥{q_{i}, x¥}=0$ Hence we have . $¥{¥varphi, ¥psi¥}=¥sum_{i,j=1}^{n}(¥frac{¥partial¥varphi}{¥partial p_{i}}¥frac{¥partial¥psi}{¥partial q_{i}}-¥frac{¥partial¥varphi}{¥partial q_{i}}¥frac{¥partial¥psi}{¥partial p_{i}})$ Noting that $¥partial/¥partial f_{0}=¥partial/¥partial x$ (3.6) $(¥varphi, ¥psi¥in C(¥alpha;f))$ . , from (2.16) we obtain $¥varphi^{¥prime}=¥{h_{0}, ¥varphi¥}+k¥frac{¥partial¥varphi}{¥partial x}$ for any $¥varphi¥in C(¥alpha; f)$ . Let us denote by $H=H(q;p;x)¥in C(¥alpha)[q;p;x]$ the polynomial in $(q;p)$ which in the coordinates $(q;p;x)$ . Then we see that represents our Hamiltonian the differential system (0.1) is equivalent to the Hamiltonian system $h_{0}$ (3.7) where (3.8) $¥frac{dq_{i}}{dt}=¥{H, q_{i}¥}=¥frac{¥partial H}{¥partial p_{i}}$ $i=1$ , $¥ldots,n$ . , $¥frac{dp_{i}}{dt}=¥{H,p_{i}¥}=-¥frac{¥partial H}{¥partial q_{i}}$ The Hamiltonian $H$ , $¥frac{dx}{dt}=k$ , is determined explicitly as follow: $H=(x-¥sum_{i=1}^{n}q_{i})(¥sum_{i=1}^{n}q_{i}p_{i})-¥sum_{i=1}^{n}q_{i}p_{i}^{2}-¥sum_{1¥leq i<j¥leq n}q_{i}(p_{i}-p_{j})q_{j}$ , $-¥sum_{i=1}^{n}(¥sum_{r=1}^{i}¥alpha_{2r-1})q_{i}+¥sum_{i=1}^{n}¥alpha_{2i}p_{i}+¥beta x$ Higher Order Pa ι nleve Equations of Type 493 $A_{l}^{(1)}$ where (3.9) $¥beta=¥sum_{i=1}^{2n}(-1)^{i-1}¥varpi_{i}=¥frac{1}{2n+1}¥sum_{r=1}^{n}((n+1-r)¥alpha_{2r-1}-r¥alpha_{2r})$ Case $A_{2n+1}^{(1)}$ . : Note first that (2.17) implies (3. 10) $g_{0}^{¥prime}=¥frac{k}{2}g_{0}$ , $g_{1}^{¥prime}=¥frac{k}{2}g_{1}$ . Hence, by setting (3.11) $¥tilde{f_{2r}}=g_{0}f_{2r}$ , $¥tilde{f_{2r+1}}=g_{0}^{-1}f_{2r+1}(r=0,1, -’ n)$ , we obtain (3.12) $(j =0,1, ¥_’ 2n+1)$ . $¥tilde{f_{j}}^{¥prime}=¥{h_{0},¥tilde{f_{j}}¥}+¥delta_{j,0}kg_{0}^{2}$ We now introduce a new coordinate system (3.13) $(q;p;x)=(q_{1}, -’ q_{n}; p_{1}-’ p_{n}; x_{0},x_{1})$ as follows: $q_{1}=g_{0}f_{2}$ (3.14) , $p_{1}=g_{0}^{-1}f_{1}$ $q_{2}=g_{0}f_{4}$ , , $¥ldots$ , $q_{n}=g_{0}f_{2n}$ $p_{2}=g_{0}^{-1}(f_{1}+f_{3})$ $x_{0}=g_{0}=f_{0}+f_{2}+-+f_{2n}$ , , , $¥ldots,p_{n}=g_{0}^{-1}(f_{1}+f_{3}+ --+f_{2n-1})$ , $x_{1}=g_{1}=f_{1}+f_{3}+¥cdots+f_{2n+1}$ . The inverse transformation is then given by $f_{0}=x_{0}-x_{0}^{-1}(q_{1}+¥cdots+q_{n})$ (3.15) $f_{1}=x_{0}p_{1}$ , $f_{2}=x_{0}^{-1}q_{1}$ $f_{2n-1}=x_{0}(p_{n}-p_{n-1})$ , , , $f_{3}=x_{0}(p_{2}-p_{1})$ $f_{2n}=x_{0}^{-1}q_{n}$ , , $f_{4}=x_{0}^{-1}q_{2}$ $f_{2n+1}=x_{1}-x_{0}p_{n}$ . Since $¥{p_{i}, q_{j}¥}=¥delta_{i,j}$ (3.16) , $¥{q_{i}, q_{j}¥}=¥{p_{i}, p_{j}¥}=0$ , $¥{p_{i},x_{0}¥}=¥{q_{i},x_{0}¥}=¥{p_{i}, x_{1}¥}=¥{q_{i},x_{1}¥}=¥{x_{0},x_{1}¥}=0$ for $i,j$ (3.17) $=1$ , $¥ldots,n$ , we have the same formula as (3.5) and $¥varphi^{¥prime}=¥{h_{0}, ¥varphi¥}+¥frac{k}{2}(x_{0}¥frac{¥partial¥varphi}{¥partial x_{0}}+x_{1}¥frac{¥partial¥varphi}{¥partial x_{1}})$ ,?’ for any $¥varphi¥in C(¥alpha; f)$ 494 Masatoshi NOUMI and Yasuhiko YAMADA by (2.18). Let us denote by $H=H(q;p;x)¥in C(¥alpha)[q;p;x^{¥pm 1}]$ the polynomial in $(q;p)$ which represents our Hamiltonian in the coordinates $(q;p;x)$ . Then we see that the differential system (0.2) is equivalent to the Hamiltonian system $h_{0}$ $¥frac{dq_{i}}{dt}=¥{H, q_{i}¥}=¥frac{¥partial H}{¥partial p_{i}}$ , $¥frac{dp_{i}}{dt}=¥{H,p_{i}¥}=-¥frac{¥partial H}{¥partial q_{i}}$ , (3.18) $¥frac{dx_{0}}{dt}=¥frac{k}{2}x_{0}$ where $i=1$ , (3.18) _’ $n$ , $¥frac{dx_{1}}{dt}=¥frac{k}{2}x_{1}$ The Hamiltonian . , $H$ is given in the form $H=(x_{0}^{2}-¥sum_{i=1}^{n}q_{i})(¥sum_{i=1}^{n}q_{i}p_{i}(¥frac{x_{1}}{x_{0}}-p_{i}))$ ? $¥sum_{1¥leq i<j¥leq n}q_{i}q_{j}(p_{i}-p_{j})(¥frac{x_{1}}{x_{0}}+p_{i}-p_{j})+2¥gamma¥sum_{i=1}^{n}q_{i}p_{i}$ $-¥frac{x_{1}}{x_{0}}¥sum_{i=1}^{n}¥beta_{i}q_{i}+¥sum_{i=1}^{n}¥alpha_{2i}p_{i}+(¥gamma-¥varpi_{2n+1})x_{0}x_{1}+¥gamma^{2}$ , where (3.20) $¥beta_{i}=¥sum_{r=1}^{i}¥alpha_{2r1}¥_(i=1, -’ n)$ , $¥gamma=¥sum_{i=1}^{2n+1}(-1)^{i-1}¥varpi_{i}=¥frac{1}{2}¥sum_{r=0}^{n}¥alpha_{2r+1}$ . Summarizing the results of this section, we have Theorem 3.1. symmetry (3.20) of type (1) The differential system (0.1) with affine Weyl group $A_{2n}^{(1)}$ is equivalent to the Hamiftonian system $¥frac{dq_{i}}{dt}=¥frac{¥partial H}{¥partial p_{i}}$ , $¥frac{dp_{i}}{dt}=-¥frac{¥partial H}{¥partial q_{i}}$ $(i =1, ¥ldots,n)$ , with an auxiliary variable such that $dx/dt$ $=k$ , where the $H=H(q;p;x)$ is the polynomial (3.8). is equivalent to the (2) The differential system (0.2) of type Hamiltonian system $x$ $A_{2n+1}^{(1)}$ (3.20) $¥frac{dq_{i}}{dt}=¥frac{¥partial H}{¥partial p_{i}}$ , $¥frac{dp_{i}}{dt}=-¥frac{¥partial H}{¥partial q_{i}}$ $(i =1, ¥ldots,n)$ , Higher Order Painleve Equations of Type 495 $A_{l}^{(1)}$ such that $dx_{0}/dt=kx_{0}/2$ , with two auxiliary variables where $H=H(q;p;x)$ is the polynomial (3.19). $dx_{1}/dt=kx_{1}/2$ $x_{0},x_{1}$ See Appendix for explicit formulas of the Hamiltonians $H$ , for small /. Remark 3.2. When we regard our systems as Hamiltonian systems, it , would be more convenient to use constant parameters as $x=kt+c$ for $x_{1}=c_{1}e^{kt/2}$ for $x_{0}=c_{0}e^{kt/2}$ , rather than the auxiliary variables as in , and Theorem 3. 1. $A_{2n}^{(1)}$ $A_{2n+1}^{(1)}$ 4. $¥tau$ -Functions and In the following, we introduce a family of Hamiltonians , of Section 2, we the Hamiltonian From system. for our , functions , define the other Hamiltonians , _’ by the diagram rotation: $h_{0}$ $¥tau_{0}$ $¥ldots$ $¥ldots,h_{l}$ $¥tau-$ $h_{0}$ $¥tau_{l}$ $h_{1}$ (4. 1) $h_{l}$ $(i =1, ¥ldots, l)$ $h_{i}=¥pi(h_{i-1})$ . is invariant under of highest degree in , Note that the component of degree comhighest a have common Hence , the diagram rotation. _’ ponent, and have different coefficients in lower degrees. We also introduce the -functions , _’ for our system to be the dependent variables such that by the linear , , namely as functions determined from differential equations $h_{0}$ $f_{0}$ $h_{0}$ $¥tau$ $¥tau_{0}$ $¥ldots,f_{l}$ $h_{l}$ $¥tau_{l}$ $h_{j}=k(¥log¥tau_{j})^{¥prime}$ $f_{0}$ (4.2) $k¥tau_{j}^{¥prime}=h_{j}¥tau_{j}$ $¥ldots,f_{l}$ $(j =0,1, --’ l)$ . In this section, we show that the affine Weyl group symmetry lifts to the level of $¥tau$ -functions as well. We first remark that our Hamiltonians have some remarkable properties in relation to the action of $W$. Proposition 4.1. With respect to the action Hamiltonians have the following invariance: $A_{2n}^{(1)}$ : $s_{i}(h_{j})=h_{j}+¥delta_{i,j}k¥frac{¥alpha_{j}}{f_{j}}$ $A_{2n+1}^{(1)}$ : $s_{i}(h_{j})=h_{j}+¥delta_{i,j}k¥frac{¥alpha_{j}}{f_{j}}g_{j}$ of the Weyl group, the affine $(i,j =0, ¥_ , f)$ , (4.3) where $g_{j}$ stands for $g_{0}$ or $g_{1}$ according as $j¥equiv 0$ $(i,j =0, ¥ldots, ¥mathit{1})$ or 1 $¥mathrm{m}¥mathrm{o}¥mathrm{d}2$ We have only to show the case where $j=0$ . we compute as follows: Proof. $¥Delta_{i}(h_{0})$ , . In the case of $A_{2n}^{(1)}$ , 496 Masatoshi NOUMI and Yasuhiko YAMADA $¥Delta_{i}(f_{i-1}f_{i}f_{i+1})+¥Delta_{i}(f_{i-1}f_{i})¥sum_{K¥in¥Psi_{1}([i+3,i-2])}f_{K}+¥Delta_{i}(f_{i}f_{i+1})¥sum_{K¥in ¥mathit{9}_{1}^{7}([i+2,i-3])}f_{K}$ $+¥Delta_{i}(f_{i-1})¥sum_{K¥in¥ovalbox{¥tt¥small REJECT}_{1}([i+2,i-2])}f_{K}+¥Delta_{i}(f_{i+1})¥sum_{K¥in¥ovalbox{¥tt¥small REJECT}_{1}([i+2,i-2])}f_{K}$ $+¥sum_{j=0}^{2n}¥Delta_{i}(¥chi(¥Gamma¥backslash ¥{j¥}))f_{j}+s_{i}(¥chi(¥Gamma¥backslash ¥{i-1¥}))¥Delta_{i}(f_{i-1})+s_{i}(¥chi(¥Gamma¥backslash ¥{i+1¥}))¥Delta_{i}(f_{+1})$ $=(f_{i-1}-f_{i+1}-¥frac{¥alpha_{i}}{f_{i}})-¥sum_{K¥in¥Psi_{1}([i+3,i-2])}f_{K}+¥sum_{K¥in¥ovalbox{¥tt¥small REJECT}_{1}([i+2,i-3])}f_{K}$ $-¥frac{1}{f_{i}}(¥sum_{K¥in¥Psi_{1}([i+2,i-2])}f_{K})+¥frac{1}{f_{i}}(¥sum_{K¥in¥Psi_{1}([i+2,i-2])}f_{K})$ $+¥sum_{K¥in¥Psi_{1}([i+1,i-2])}f_{K}-¥sum_{K¥in¥ovalbox{¥tt¥small REJECT}_{1}([i+2,i-1])}f_{K}$ ? $(¥chi(¥Gamma¥backslash ¥{i-1¥})-¥alpha_{i})¥frac{1}{f_{i}}+(¥chi(¥Gamma¥backslash ¥{i+1¥})+¥alpha_{i})¥frac{1}{f_{i}}$ $=(¥alpha_{i}+¥varpi_{i-1}-2¥varpi_{i}+¥varpi_{i+1})¥frac{1}{f_{i}}=¥delta_{i,0}¥frac{k}{f_{i}}$ which proves (4.3) for $A_{2n}^{(1)}$ . The case of , $A_{2n+1}^{(1)}$ can be verified similarly. We need also to know how adjacent pairs of the Hamiltonians related. We propose some lemmas for this purpose. Lemma 4.2. For each $i=0,1$ , $¥ldots$ $h_{j}$ $¥square $ are , 1, one has $¥pi(¥chi([i-1, i]))-¥chi([i, i+1])=¥left¥{¥begin{array}{l}-¥frac{1}{f+1}k¥¥¥frac{f}{f+1}k¥end{array}¥right.$ $((i=0)i¥neq 0),$ . (4.4) $¥pi^{-1}(¥chi([i+1, i+2]))-¥chi([i, i+1])=¥left¥{¥begin{array}{l}¥frac{1}{l+1}k¥¥-¥frac{l}{l+1}k¥end{array}¥right.$ $((i¥neq l)i=f)’$ . Let $L$ be a subset of $¥Gamma=¥{0,1, ¥ldots, f¥}$ with $|L|=2m$ and suppose that $L$ is a disjoint union of chains of even nodes. We say that $L$ contains $[/, i+1]$ evenly, if $[i, i+1]¥subset L$ and the complement splits into chains of even nodes. With this terminology, we have $L¥backslash [i, i+1]$ of Higher Order Painleve Equations Lemma 4.3. (4.5) Under the assumption on $L$ $L$ (4.6) contains [0, $¥pi^{-1}$ according as $L$ 1] evenly, or not. or $[l, 0]$ $-¥frac{m}{l+1}k$ , Similarly, $(¥chi(¥pi L))-¥chi(L)=-¥frac{l-m+1}{l+1}k$ contains 497 $A_{l}^{(1)}$ above, one has $¥pi(¥chi(¥pi^{-1}L))-¥chi(L)=¥frac{l-m+1}{l+1}k$ according as Type or $¥frac{m}{l+1}k$ , evenly, or not. We omit the proof of Lemmas 4.2 and 4.3 since they can be proved by direct calculations from the definitions. Proposition 4.4. { $¥mathrm{I})$ In the case has (4.7) of $¥mathrm{A}_{2n}^{(1)}(f =2n)$ , for each $j=0$ , $h_{j+1}-h_{j}=k¥sum_{r=1}^{n}f_{j+2r}-¥frac{nk}{2n+1}x$ where $x=¥sum_{i=0}^{2n}f_{i}$ . $¥mathit{2}n$ _’ , Hence $-h_{j-1}+2h_{j}-h_{j+1}=k¥sum_{r=1}^{n}(f_{j+2r-1}-f_{j+2r})$ , (4.8) $h_{j-1}-h_{j+1}=k(f_{j}-¥frac{1}{2n+1}x)$ . (2) In the case of $A_{2n+1}^{(1)}(l =2n+1)$ , one has (4.9) $h_{j+1}-h_{j}=k¥sum_{1¥leq r¥leq s¥leq n}f_{j+2r}f_{j+2s+1}$ $-¥frac{nk}{2n+2}¥sum_{K¥in¥Psi_{2}}f_{K}+(-1)^{j}¥frac{k}{4}¥sum_{i=0}^{2n+1}(-1)^{i}¥alpha_{i}$ for each $j=0$ , _’ $2n+1$ . , Hence $-h_{j-1}+2h_{j}-h_{j+1}=k¥sum_{1¥leq r¥leq s¥leq n}(f_{j+2r-1}f_{j+2s}-f_{j+2r}f_{j+2s+1})$ (4. 10) $+k(¥frac{k}{2}-¥sum_{r=0}^{n}¥alpha_{j+2r})$ , $h_{j-1}-h_{j+1}=kx_{j+1}(f_{j}-¥frac{1}{n+1}x_{j})$ , , one 498 Masatoshi NOUMI and Yasuhiko YAMADA where $i¥equiv 0$ $x_{0}=¥sum_{r=0}^{2n}f_{2r}$ or 1 Proof. $¥mathrm{m}¥mathrm{o}¥mathrm{d}2$ and $x_{1}=¥sum_{r=0}^{2n}f_{2r+1}$ , and $x_{i}$ stands for $x_{0}$ or $x_{1}$ according as . When $¥mathit{1}=2n$ , we have $h_{0}=¥sum_{K¥in¥ovalbox{¥tt¥small REJECT}_{3}}f_{K}+¥sum_{i=0}^{2n}¥chi([i+1, i-1])f_{i}$ , (4. 11) $h_{1}=¥sum_{K¥in¥ovalbox{¥tt¥small REJECT}_{3}}f_{K}+¥sum_{i=0}^{2n}¥pi(¥chi([i, i-2]))f_{i}$ , hence (4. 12) $h_{1}-h_{0}=¥sum_{i=0}^{2n}(¥pi(¥chi([i, i-2]))-¥chi([i+1, i-1]))f_{i}$ . By Lemma 4.3, we compute (4. 13) $h_{1}-h_{0}=-¥frac{nk}{2n+1}(f_{0}+¥sum_{r=1}^{n}f_{2r-1})+¥frac{(n+1)k}{2n+1}¥sum_{r=1}^{n}f_{2r}$ $=k¥sum_{r=1}^{n}f_{2r}-¥frac{nk}{2n+1}¥sum_{i=0}^{2n}f_{i}$ , which gives (4.7) for $j=0$ . Formulas (4.7) for $j=1$ , _’ are obtained by applying the diagram rotation . Formulas (4.8) follow directly from (4.7). When $=2n+1$ , we have $¥mathit{2}n$ $¥pi$ $l$ $h_{0}=¥sum_{K¥in¥Psi_{4}}f_{K}+¥sum_{K¥in¥ovalbox{¥tt¥small REJECT}_{2}}¥chi(K^{c}))f_{K}+¥gamma_{1}^{2}$ , (4. 14) $h_{1}=¥sum_{K¥in¥Psi_{4}}f_{K}+¥sum_{K¥in¥ovalbox{¥tt¥small REJECT}_{2}}¥pi(¥chi(¥pi^{-1}K^{c}))f_{K}+¥gamma_{0}^{2}$ where $¥gamma_{1}=(¥sum_{r=0}^{n}¥alpha_{2r+1})/2$ and $¥gamma_{0}=¥pi(¥gamma_{1})=$ $(¥sum_{r=0}^{n}¥alpha_{2r})/2$ . , Hence we have $h_{1}-h_{0}=¥sum_{K¥in¥Psi_{2}}(¥pi(¥chi(¥pi^{-1}K^{c}))-¥chi(K^{c}))f_{K}+¥frac{k}{2}(¥gamma_{0}-¥gamma_{1})$ . ¥ since ¥ . The coefficients of are computed by Lemma 4.3 to obtain (4.9) for $j=0$ . Formulas (4.9) for the other are obtained by the diagram rotation, and formulas (4.10) follow directly from (4.9). $ gamma_{0}+ gamma_{1}=k/2$ $f_{K}$ $j$ $¥square $ Higher Order Painleve Equations of Type 499 $A_{l}^{(1)}$ By combining Propositions 4.1 and 4.4, we obtain Proposition 4.5. The differential system (0.1) (resp. (0.2)) is expressed as : follows in terms of Hamiltonians , $h_{0}$ (4. 15) _’ $h_{l}$ $k¥frac{f_{j}^{¥prime}}{f_{j}}=s_{j}(h_{j})+h_{j}-h_{j1}¥_-h_{j+1}$ $(j =0, ¥_’ l)$ , or equivalently, (4. 16) where $k¥frac{f_{j}^{¥prime}}{f_{j}}=s_{j}(h_{j})-h_{j}+¥sum_{i=0}^{l}h_{i}a_{ij}$ $A$ $=(a_{ij})_{0¥leq i,j¥leq l}$ is the generalized Carton matrix of (1.5). Formulas (4. 15) can be verified by rewriting the right-hand side as the sum , and then by applying Proposition 4.1 and $(s_{j}(h_{j})-h_{j})+(-h_{j+1}¥_+2h_{j}-h_{j+1})$ 4.4. ¥ the field of rational functions in In what follows, we denote by ¥ $C( ¥ alpha; f)$ . We define the structure of differential with coefficients in , ¥ ¥ and extend the automorphisms , now by field of (4.2). We of $C(¥alpha; f)$ to $C(¥alpha,f;¥tau)$ by setting $C( alpha; f; tau)$ $¥tau_{0}$ _’ $¥tau_{l}$ $C( alpha; f; tau)$ (4.17) where $s_{0}$ $s_{i}(¥tau_{j})=¥tau_{j}(i¥neq j)$ $i,j$ $=0$ , $¥ldots$ , $f$ , $s_{j}(¥tau_{j})=¥frac{¥tau_{j-1}¥tau_{j+1}}{¥tau_{j}}f_{j}$ , $¥pi(¥tau_{j})=¥tau_{j+1}$ . and Theorem 4.6 ([8]). The automorphisms , of ¥ ¥ scribed above define a representation of the extended affine Weyl group $s_{0}$ _’ $¥pi$ $s_{l}$ Proposition 4.7. -functions (4. 18) The variables $f_{j}$ $(j =0,1, ¥_’ l)$ as follows: $f_{j}$ $¥tilde{W}$ . by $¥tau$ are expressed in terms of $f_{j}=¥frac{1}{k}(h_{j1}¥_-h_{j+1})+¥frac{x}{2n+1}$ $=¥frac{¥tau_{j-1}^{¥prime}}{¥tau_{j-1}}-¥frac{¥tau_{j+1}^{¥prime}}{¥tau_{j+1}}+¥frac{x}{2n+1}$ de- $C( alpha; f; tau)$ Note that, by Proposition 4.4, one obtains an expression of functions. $¥tau$ $¥pi$ $¥ldots,s_{l}$ $(j =0, ¥ldots, ¥mathit{2}n)$ 500 Masatoshi NOUMI and Yasuhiko YAMADA when $f=2n$ , and (4. 19) $f_{j}=¥frac{1}{kx_{j+1}}(h_{j-1}-h_{j+1})+¥frac{x_{j}}{n+1}$ $=¥frac{1}{x_{j+1}}(¥frac{¥tau_{j-1}^{¥prime}}{¥tau_{j-1}}-¥frac{¥tau_{j+1}^{¥prime}}{¥tau_{j+1}})+¥frac{x_{j}}{n+1}$ when $l$ $(j =0, --’ 2n+1)$ $=2n+1$ . Hence we have Proposition 4.8. For each $j=0$ , 1, the action of on the is given by the following bilinear operators of Hirota type: _’ (4.20) $s_{j}$ $z$ -functions $¥tau_{j}$ $s_{j}(¥tau_{j})=¥frac{1}{¥tau_{j}}(D_{t}+¥frac{x}{2n+1})¥tau_{j-1}¥cdot¥tau_{j+1}$ $=¥frac{1}{¥tau_{j}}(¥tau_{j-1}^{¥prime}¥tau_{j+1}-¥tau_{j-1}¥tau_{j+1}^{¥prime}+¥frac{x}{2n+1}¥tau_{j-1}¥tau_{j+1})$ when $¥mathit{1}=2n$ , and (4.21) $s_{j}(¥tau_{j})=¥frac{1}{¥tau_{j}}(¥frac{1}{x_{j+1}}D_{t}+¥frac{x_{j}}{n+1})¥tau_{j-1}¥cdot¥tau_{j+1}$ $=¥frac{1}{¥tau_{j}}(¥frac{1}{x_{j+1}}(¥tau_{j-1}^{¥prime}¥tau_{j+1}-¥tau_{j-1}¥tau_{j+1}^{¥prime})+¥frac{x_{j}}{n+1}¥tau_{j-1}¥tau_{j+1})$ when as ¥ $¥mathit{1}=2n+1$ $j equiv 0$ or 1 , where $¥mathrm{m}¥mathrm{o}¥mathrm{d}2$ $x_{j}$ stands for $x_{0}=¥sum_{r=0}^{n}f_{2r}$ or $x_{1}=¥sum_{r=0}^{n}f_{2r+1}$ according . From Proposition 4.5, we obtain Theorem 4.9. The action of the extended with the derivation of the differential fiefd ¥ affine $C( alpha; f;¥tau)$ Proof. Since case when (4.22) $i=j$ . Weyl group $s_{i}(¥tau_{j})^{¥prime}=s_{i}(¥tau_{j}^{¥prime})$ $¥underline{s_{j}(¥tau_{j})^{¥prime}}=+-+¥underline{¥tau_{j-1}^{¥prime}}¥underline{¥tau_{j+1}^{¥prime}}¥underline{¥tau_{j}^{¥prime}}¥underline{f_{j}^{¥prime}}$ $¥tau_{j-1}$ $¥tau_{j+1}$ $¥tau_{j}$ commutes . the equality is obvious if By Proposition 4.5, we compute $s_{j}(¥tau_{j})$ $¥tilde{W}$ $f_{j}$ $=¥frac{1}{k}(h_{j-1}+h_{j+1}-h_{j}+k¥frac{f_{j}^{¥prime}}{f_{j}})$ $=¥frac{1}{k}s_{j}(h_{j})=s_{j}(¥frac{¥tau_{j}^{¥prime}}{¥tau_{j}})=¥frac{s_{j}(¥tau_{j}^{¥prime})}{s_{j}(¥tau_{j})}$ , $i¥neq j$ , we consider the Higher Order Painleve Equations which implies commutes with ’. $s_{j}(¥tau_{j})^{¥prime}=s_{j}(¥tau_{j}^{¥prime})$ of Type 501 $A_{l}^{(1)}$ It is clear that the action of the diagram rotation . $¥square $ $¥pi$ Theorem 4.9 means that one can lift the Backlund transformations of our are to the level of -functions so that each system for the variables invariant with respect to the subgroup $ W_{j}=¥langle s0, ¥_’ s_{j1}¥_’ s_{j1,-}¥_’ s_{l}¥rangle$ of $W$ $(j =0, ¥ldots, f)$ , and that one has the multiplicative formulas $¥mathrm{r}$ $f_{j}$ (4.23) $¥tau_{j}$ $(j =0, ¥ldots, l)$ $f_{j}=¥frac{¥tau_{j}s_{j}(¥tau_{j})}{¥tau_{j-1}¥tau_{j+1}}$ in terms of -functions. We have thus for the dependent variables , _’ Backlund transformations of our differential of the structure guaranteed that system is consistent with the general scheme of our previous paper [8]. As a consequence we see that the Backlund transformations of our system provides in the sense of [8]. the discrete dynamical systems of type $f_{0}$ $f_{l}$ $¥mathrm{r}$ $A_{l}^{(1)}$ Appendix A. 1. Explicit formulas of $f_{0}^{¥prime}$ for small $l$ . , $A_{2}^{(1)}$ : $f_{0}^{¥prime}=f_{0}(f_{1}-f_{2})+¥alpha_{0}$ $A_{4}^{(1)}$ : $f_{0}^{¥prime}=f_{0}(f_{1}-f_{2}+f_{3}-f_{4})+¥alpha_{0}$ $A_{3}^{(1)}$ : $f_{0}^{¥prime}=f_{0}(f_{1}f_{2}-f_{2}f_{3})+(¥frac{k}{2}-¥alpha_{2})f_{0}+¥alpha_{0}f_{2}$ $A_{5}^{(1)}$ : $f_{0}^{¥prime}=f_{0}(f_{1}f_{2}+f_{1}f_{4}+f_{3}f_{4}-f_{2}f_{3}-f_{2}f_{5}-f_{4}f_{5})$ , $+(¥frac{k}{2}-¥alpha_{2}-¥alpha_{4})f_{0}+¥alpha_{0}(f_{2}+f_{4})$ A.2. Explicit formulas of $h_{0}$ for small $f$ , . . $A_{2}^{(1)}$ : $h_{0}=f_{0}f_{1}f_{2}+¥frac{1}{3}(¥alpha_{1}-¥alpha_{2})f_{0}+¥frac{1}{3}(¥alpha_{1}+2¥alpha_{2})f_{1}-¥frac{1}{3}(2¥alpha_{1}+¥alpha_{2})f_{2}$ $A_{4}^{(1)}$ : $h_{0}=f_{0}f_{1}f_{2}+f_{1}f_{2}f_{3}+f_{2}f_{3}f_{4}+f_{3}f_{4}f_{0}+f_{4}f_{0}f_{1}$ $+¥frac{1}{5}(2¥alpha_{1}-¥alpha_{2}+¥alpha_{3}-2¥alpha_{4})f_{0}+¥frac{1}{5}(2¥alpha_{1}+4¥alpha_{2}+¥alpha_{3}+3¥alpha_{4})f_{1}$ $-¥frac{1}{5}(3¥alpha_{1}+¥alpha_{2}-¥alpha_{3}+2¥alpha_{4})f_{2}+¥frac{1}{5}(2¥alpha_{1}-¥alpha_{2}+¥alpha_{3}+3¥alpha_{4})f_{3}$ $-¥frac{1}{5}(3¥alpha_{1}+¥alpha_{2}+4¥alpha_{3}+2¥alpha_{4})f_{4}$ 502 Masatoshi NOUMI and Yasuhiko YAMADA : $A_{3}^{(1)}$ $h_{0}=f_{0}f_{1}f_{2}f_{3}+¥frac{1}{4}(¥alpha_{1}+2¥alpha_{2}-¥alpha_{3})f_{0}f_{1}+¥frac{1}{4}(¥alpha_{1}+2¥alpha_{2}+3¥alpha_{3})f_{1}f_{2}$ $-¥frac{1}{4}(3¥alpha_{1}+2¥alpha_{2}+¥alpha_{3})f_{2}f_{3}+¥frac{1}{4}(¥alpha_{1}-2¥alpha_{2}-¥alpha_{3})f_{3}f_{0}+¥frac{1}{4}(¥alpha_{1}+¥alpha_{3})^{2}$ : $A_{5}^{(1)}$ $h_{0}=f_{0}f_{1}f_{2}f_{3}+f_{1}f_{2}f_{3}f_{4}+f_{2}f_{3}f_{4}f_{5}+f_{3}f_{4}f_{5}f_{0}+f_{4}f_{5}f_{0}f_{1}+f_{5}f_{0}f_{1}f_{2}$ $+¥frac{1}{3}(¥alpha_{1}+2¥alpha_{2}+¥alpha_{4}-¥alpha_{5})f_{0}f_{1}+¥frac{1}{3}(¥alpha_{1}+2¥alpha_{2}+3¥alpha_{3}+¥alpha_{4}+2¥alpha_{5})f_{1}f_{2}$ $-¥frac{1}{3}(2¥alpha_{1}+¥alpha_{2}-¥alpha_{4}+¥alpha_{5})f_{2}f_{3}+¥frac{1}{3}(¥alpha_{1}-¥alpha_{2}+¥alpha_{4}+2¥alpha_{5})f_{3}f_{4}$ $-¥frac{1}{3}(2¥alpha_{1}+¥alpha_{2}+3¥alpha_{3}+2¥alpha_{4}+¥alpha_{5})f_{4}f_{5}+¥frac{1}{3}(¥alpha_{1}-¥alpha_{2}-2¥alpha_{4}-¥alpha_{5})f_{5}f_{0}$ $+¥frac{1}{3}(¥alpha_{1}-¥alpha_{2}+¥alpha_{4}-¥alpha_{5})f_{0}f_{3}+¥frac{1}{3}(¥alpha_{1}+2¥alpha_{2}+¥alpha_{4}+2¥alpha_{5})f_{1}f_{4}$ $-¥frac{1}{3}(2¥alpha_{1}+¥alpha_{2}+2¥alpha_{4}+¥alpha_{5})f_{2}f_{5}+¥frac{1}{4}(¥alpha_{1}+¥alpha_{3}+-¥alpha_{5})^{2}$ A.3. Explicit formulas of $H$ for small . $l$ $A_{2}^{(1)}$ : $H=(x-q_{1})q_{1}p_{1}-q_{1}p_{1}^{2}-¥alpha_{1}q_{1}+¥alpha_{2}p_{1}+¥frac{1}{3}(¥alpha_{1}-¥alpha_{2})x$ $A_{4}^{(1)}$ : $H=(x-q_{1}-q_{2})(q_{1}p_{1}+q_{2}p_{2})-q_{1}p_{1}^{2}-q_{2}p_{2}^{2}-q_{1}(p_{1}-p_{2})q_{2}$ $-¥alpha_{1}q_{1}-(¥alpha_{1}+¥alpha_{3})q_{2}+¥alpha_{2}p_{1}+¥alpha_{4}p_{2}+¥frac{1}{5}(2¥alpha_{1}-¥alpha_{2}+¥alpha_{3}-2¥alpha_{4})x$ $A_{3}^{(1)}$ : $H=(x_{0}^{2}-q_{1})p_{1}q_{1}(¥frac{x_{1}}{x_{0}}-p1)+(¥alpha_{1}+¥alpha_{3})q_{1}p_{1}-¥alpha_{1}¥frac{x_{1}}{x_{0}}q_{1}+¥alpha_{2}x_{0}^{2}p_{1}$ $+¥frac{1}{4}$ $A_{5}^{(1)}$ : $(¥alpha_{1}-2¥alpha_{2}-¥alpha_{3})x_{0}x_{1}+¥frac{1}{4}$ $(¥alpha_{1}+¥alpha_{3})^{2}$ $H=(x_{0}^{2}-q_{1}-q_{2})(q_{1}p_{1}(¥frac{x_{1}}{x_{0}}-p_{1})+q_{2}p_{2}(¥frac{x_{1}}{x_{0}}-p_{2}))$ ? $q_{1}q_{2}(p_{1}-p_{2})(¥frac{x_{1}}{x_{0}}+p_{1}-p_{2})+(¥alpha_{1}+¥alpha_{3}+¥alpha_{5})(q_{1}p_{1}+q_{2}p_{2})$ $-¥frac{x_{1}}{x_{0}}(¥alpha 1q1+(¥alpha]+¥alpha 3)q2)+x^{2}¥mathrm{o}(¥alpha 2p1+¥alpha 4p2)$ $+¥frac{1}{3}(¥alpha_{1}-¥alpha_{2}-2¥alpha_{4}-¥alpha_{5})x_{0}x_{1}+¥frac{1}{4}(¥alpha_{1}+¥alpha_{3}+¥alpha_{5})^{2}$ References Groupes et Algebres de Lie, Chapitres 4, 5 et 6, Elements de Mathematique, Masson, Paris, 1981. Kac, V. G., Infinite dimensional Lie algebras, Third edition, Cambridge University Press, 1990. [1] Bourbaki, N., [2] Higher Order Painleve Equations of Type $A_{l}^{(1)}$ 503 [3] Macdonald, I. G., Affine root systems and Dedekind’s -function, Inv. Math, 15 (1972), 91-143. Noumi, M., Okada, S., Okamoto, K. and Umemura, H., Special polynomials associated with [4] the Painleve equations II, to appear in the Proceedings of the Taniguchi Symposium $¥eta$ “Integrable Systems and Algebraic Geometry”, RIMS, Kyoto University, Japan, 1997. [5] Noumi, M. and Okamoto, K., Irreducibility of the second and the fourth Painleve equations, Funkcial. Ekvac., 40 (1997), 139-163. [6] Noumi, M. and Yamada, Y., Symmetries in the fourth Painleve equation and Okamoto polynomials, to appear in Nagoya Math. J. (q-alg/9708018). [7] Noumi, M. and Yamada, Y., Umemura polynomials for Painleve V equation, to appear in Phys. Lett. A. [8] Noumi, M. and Yamada, Y., Affine Weyl groups, discrete dynamical systems and Painleve equations, to appear in Comm. Math. Phys. (math.QA/9804132). [9] Noumi, M. and Yamada, Y., Symmetric forms of the Painleve equations, in preparation. [10] Okamoto, K., Studies of the Painleve equations, I. Ann. Math. Pura Appl., 146 (1987), 337-381; II. Jap. J. Math., 13 (1987), 47-76; III. Math. Ann., 275 (1986), 221-255; IV. Funkcial. Ekvac., 30 (1987), 305-332. [11] Umemura, H., On the irreducibility of the first differential equation of Painleve, in “Algebraic Geometry and Commutative Algebra in honor of Masayoshi Nagata”, pp. 101119, Kinokuniya-North-Holland, 1987. [12] Umemura, H., Special polynomials associated with the Painleve equations I, to appear in the Proceedings of the Workshop on “Painleve Transcendents”, CRM, Montreal, Canada, 1996. nuna adreso: Masatoshi Noumi Department of Mathematics Graduate School of Science and Technology Kobe University Rokko, Kobe 657-8501 Japan E-mail: noumicmath.kobe-u.ac.jp Yasuhiko Yamada Department of Mathematics Graduate School of Science and Technology Kobe University Rokko, Kobe 657-8501 Japan E-mail: yamadaycmath.kobe-u.ac.jp (Ricevita la 5-an de junio, 1998)