Formal Groups - Honda

J. Math. Soc.
Japan
Vol. 22, No. 2, 1970
On the theory of commutative formal groups
By Taira HONDA
(Received Nov. 10, 1969)
The theory of (commutative) formal groups was initiated by M. Lazard

and J. Dieudonne around 1954. Lazard [11], [12] studied commutative formal
groups over an arbitrary commutative ring by treating the coefficients of
power series explicitly. Whereas Dieudonn\’e investigated formal groups over
a field of characteristic $p>0$ exclusively. He reduced in [4] the study of com-
mutative formal groups over a perfect field of characteristic $p>0$ to that of
modules over a certain non-commutative ring, so-called Dieudonn\’e modules,
and obtained in [5] a complete classification of isogeny classes of commutative
formal groups over an algebraically closed field of characteristic $p>0$ . Later
Manin [16] studied isomorphism classes of simple formal groups. The study
of Xone-dimensional formal groups over -adic integer rings was begun by
$\mathfrak{p}$
Lubin [13] and a number of interesting results were obtained by him and
Tate.
In this paper we first construct a certain general family of commutative
formal groups of arbitrary dimension over a p-adic integer ring. Over the ring
$W(k)$ of Witt vectors over a perfect field of characteristic $p>0$ , this exhausts
all the commutative formal groups. These are attached to a certain type of
matrices with elements in the ring $W(k).[[T]]$ of non-commutative power
series, where is the Frobenius of $W(k)$ , and homomorphisms of these formal
$\sigma$
groups are described in terms of matrices over $W(k)_{\sigma}[[T]]$ . By reducing

the coefficients of formal groups over $W(k)mod pW(k)$ we get formal groups
over . lt is shown that all the commutative formal groups over are ob-
$k$ $k$
tained in this manner. Moreover homomorphisms of commutative formal

groups over are also described in terms of $W(k)_{\sigma}[[T]]$ -modules by lifting
$k$
these homomorphisms to power series over $W(k)$ . Thus we get the main
results of Dieudonn\’e [4] again by the method quite different from his. In
[4] he used tools peculiar to characteristic $p>0$ and his construction of formal
groups was indirect, whereas in our method the relation between formal
groups over $W(k)$ and those over is transparent and the construction of
$k$
formal groups is explicit and elementary.

We now explain briefly how to construct commutative formal groups over
$W(k)$ in case of dimension one. Take an element of $W(k)_{\sigma}[[T]]$ of the
$u$
214 T. HONDA
form $p+\sum_{\nu=1}^{\infty}c_{\nu}T(c_{\nu}\in W(k))$ and put $ pu- 1=\sum_{\nu-0}^{\infty}b_{\nu}T^{\prime}\lrcorner$

The $b_{v}$
are elements of
the fraction field of $W(k)$ and $b_{0}=1$ . Form $f(x)=\sum_{\nu=0}^{\infty}b_{\nu}x^{p^{\mu}}$ and $F(x, y)=$
$f^{-1}(f(x)+f(y))$ is a formal group over $W(k)$ . In some special case

. Then $F$
this fact can be proved by using the basic lemma of Lubin-Tate [14] (cf. [10]).
In general case we have to adopt another idea. Any formal group over $ W(k\rangle$
is isomorphic to one obtained in this manner. Let be another element of $v$
$W(k)_{\sigma}[[T]]$ of the form mentioned above and let $g(x)$ and $G(x, y)$ be the
corresponding power series and the formal group, respectively. It is known
that any homomorphism of $F$ to is of the form $g^{-1}(cf(x))$ with $c\in W(k)$ . $G$
We assert that $g^{-1}(cf(x))$ is in reality a homomorphism over $W(k)$ , if and

only if there is $t\in W(k)_{\sigma}[[T]]$ such that $vc=tu$ . All these results will be
generalized and proved for an arbitrary dimension and for more general
coefficient rings of characteristic with discrete valuation. $0$
Our results can be applied to construct and characterize formal groups

over $Z$ corresponding to a certain type of Dirichlet series with matrix co-
efficients, thus generalizing the results of the last half of our previous paper
[10]. In particular we get an interesting interpretation of the Dirichlet series
obtained from a representation of Hecke operators in the space of cusp forms
of dimension $-2$ with respect to a congruence unit group of a maximal $\Gamma_{N}$
order of an indefinite quaternion algebra over $Q$ (Shimura [19]). There is

an intimate connection between this Dirichlet series and a formal completion
of the Jacobian . $J_{N}$
\S 1. Invariant differential forms on a formal group.

1.1. Let be a ring. We denote by $S$
the module consisting of all the $S^{m}$
column vectors of dimension $m$ with components in and by $M_{m}(S)$ the full $S$
matrix ring of order $m$ with elements in S. denotes the indentity matrix $I_{m}$
of order $m$ . For $a={}^{t}(a_{1}, \cdots , a_{m})\in S^{m}$ we write for . $a^{\nu}$ ${}^{t}(a_{1}^{\nu}, \cdots , a_{m}^{\nu})$
Let $R$
be a commutative ring with the identity. Let be the set of $x$ $n$
variables , . We denote by $R[[x]]$ the ring of formal power series

$\chi_{)},$
$\cdots$
$x_{n}$
on $x_{1},$, . For basic properties of $R[[x]]$ we refer to Bourbaki [3]. We

$\cdots$
$x_{n}$
shall often regard as the column vector , .. ,

$x$
in $R[[x]]^{n}$ . Let and ${}^{t}(x_{1}$ $\cdot$
$x_{n})$ $f$
$g$be power series in $R[[x]]$ . We shall say that is congruent to modulo $f$ $g$
degree mod $\deg r$ , if and differ only in terms of total degree

$r,$ $f\equiv g$ $f$ $g$ . $\geqq r$
Let be a submodule of R.
$I$
is said to be congruent to modulo
$f$ $g$ $I,$ $f\equiv g$
$mod I$ , if all the coefficients of $f-g$ belong to . We shall write mod $\deg r$ , $I$ $f\equiv g$
$mod I$, if there are $\psi\in R[[x]]$ such that $f-g=\varphi+\psi,$

$\varphi,$
mod $\deg r$ and $\varphi\equiv 0$
$\psi\equiv 0mod I$. These definitions extend to $R[[x]]^{m}$ . If and $f={}^{t}(f_{1}, \cdots , f_{m})$
The theory of commutative formal groups 215
$g={}^{t}(g_{1}$ , $\cdot$
.. , are elements of $R[[x]]^{m},$
$g_{m})$ mod $*will$ mean $f_{i}\equiv g_{i}mod *$ $f\equiv g$
for $1\leqq i\leqq n$ . We write $R[[x]]_{0}^{m}=$ { $f\in R[[x]]^{m}|f\equiv 0$ mod deg l}.
Let be another set of variables. If
$x^{\prime}={}^{t}(x_{1}^{\prime}, \cdots, x_{m}^{\prime})$ $f(x^{\prime})={}^{t}(f_{1}(x^{\prime}), \cdots, f_{\iota}(x^{\prime}))$
is in and
$(f_{i}(x^{\prime})=f_{i}(x_{1}^{\prime}, \cdots , \chi_{m}^{\prime}))$
is in $R[[x]]_{0}^{m}$ , $R[[x^{\prime}]]^{l}$ $\varphi(x)={}^{t}(\varphi_{1}(x), \cdots , \varphi_{m}(x))$
the power series is well-defined and ,

$f_{i}(\varphi(x))=f_{i}(\varphi_{1}(x), \cdots , \varphi_{n},(x))$ ${}^{t}(f_{1}(\varphi(x)),$ $\cdots$
is an element of
$f_{t}(\varphi(x)))$ $R[[x]]^{l}$ . We denote it by or simply by , $f(\varphi(x))$ $ f\circ\varphi$
if there is no fear of ambiguity. Define the identity function of $R[[x]]_{0^{n}}$ $i$
by $i(x)=x$ . If is an element of $R[[x]]_{0}^{n}$ such that

$\varphi(x)$
mod $\deg 2$ $\varphi(x)\equiv Px$
with an invertible matrix $P$ in $M_{n}(R)$ , there is a unique element in $\psi(x)$
$R[[x]]0$ satisfying . We shall call this the inverse function

$\varphi\circ\psi=\psi\circ\varphi=i$ $\psi$
of and denote it by
$\varphi$
. $\varphi^{-1}$
We adopt the classical definition of formal group.

DEFINITION. Let and be sets (or vectors) of variables. An n-
$x$ $y$ $n$
dimensional formal group over is an element $F(x, y)$ of $R[[x, y]]_{0}^{n}$ satisfying:
$R$
i) $F(x, y)\equiv x+y$ mod $\deg 2$ ,
ii) $F(F(x, y),$ ) $=F(x, F(y, z))$

$z$
.
If satisfies $F(x, y)=F(y, x)$ moreover, $F$ is said to be commutative.
$F$
It follows from (i) that there is a unique $i_{F}(x)\in R[[x]]_{0}^{n}$ such that
$F(x, i_{F}(x))=F(i_{F}(x), x)=0$ . Part (ii) shows that $F(x, 0)=x$ and $F(O, y)=y$ .
DEFINITION. Let $F$ and $G$ be formal groups over $R$ , of dimension and $n$
$m$ , respectively. An element of $R[[x]]_{0}^{m}$ , where , is said to $\varphi$ $\chi={}^{t}(x_{1}, \cdots , x_{n})$
be a homomorphism of $F$ to , if satisfies , where $G$ $\varphi$ $\varphi\circ F=G\circ\varphi$ $(G\circ\varphi)(x, y)$
stands for . If $m=n$ and is invertible,

$G(\varphi(x), \varphi(y))$ is also a homo- $\varphi$
$\varphi^{-1}$
morphism of $G$
to . Such
$F$
is called an isomorphism and is said to be $\varphi$
$G$
(weakly) isomorphic to : $F\sim G$ over $R$ . If there is an isomorphism

$F,$ of
$\varphi$ $\varphi$
$F$ to such that

$G$
mod $\deg 2$ , we shall say that $G$ is strongly isomorphic
$\varphi(x)\equiv x$
to $F$ and
write : over $R$ . $\varphi$
$F\approx G$
If $G$ is commutative, the set $Hom_{R}(F, G)$ of all homomorphisms of $F$ to

$G$
over $R$ forms a module by defining for $(\varphi_{1}+\varphi_{2})(x)=G(\varphi_{1}(x), \varphi_{2}(x))$ $\varphi_{1},$ $\varphi_{z}$
$\in Hom_{R}(F, G)$ . In particular $End_{R}G(=Hom_{R}(G, G))$ becomes a ring by

defining the multiplication by composition of functions.
1.2. Let $A=R[[x]]$ be as in 1.1. We denote by the space of
.
$\mathfrak{D}(A;R)$
derivations of $A$
over $R$
. It is a free left A-module with a base , $D_{1},$ $\cdots$ $D_{n}$
where (cf. [3]). Denote by

$D_{i}=\partial/\partial x_{i}$
the dual A-module of , $\mathfrak{D}^{*}(A;R)$ $\mathfrak{D}(A;R)$
the space of differentials of $A$ over $R$ . For $f\in A$ the map $D\mapsto Df$ of $\mathfrak{D}(A;R\rangle$
into $A$ defines a differential, which we denote by $df$. A differential of this

form is called exact. It is well-known that $dx={}^{t}(dx_{1}, \cdots , dx_{n})$ is an A-base
of $\mathfrak{D}^{*}(A;R)$ and $df=\sum_{\iota=1}^{n}(D_{i}f)dx_{i}$ for any $f\in A$ .
216 T. HONDA
Let $B=R[[x^{\prime}]]$ be another ring of power series on $m$ variables and let
to $=\sum_{j=1}^{m}\psi_{j}(x^{\prime})dx_{j^{\prime}}$ be a differential in $\mathfrak{D}^{*}(B;R)$
. If $\varphi\in R[[x]]_{0}^{m},\sum_{j=1}^{m}\psi_{j}(\varphi(x))d\varphi_{j}(x)$
is a differential in . We denote it by . is an R-homo- $\mathfrak{D}^{*}(A;R)$ $\varphi^{*}(\omega)$ $\varphi^{*}$
morphism of into . $\mathfrak{D}^{*}(B;R)$ $\mathfrak{D}^{*}(A;R)$
Let be an n-dimensional formal group over . Introducing a new set

$F$ $R$
$l=$ of variables we may consider that

$(t_{1}, \cdots , t_{n})$
is also defined over $F$ $R_{t}$
$=R[[t]]$ .
DEFINITION. The right translation on $F$ is an element of $R_{t}[[x]]^{m}$ $T_{t}$
defined by $T_{t}(x)=F(x, t)$ . A differential ru in is said to be a right $\mathfrak{D}^{*}(A;R)$
invariant differential on $F$ if

. $ T_{t}^{*}(\omega)=\omega$
We denote by the space consisting of all right invariant dif-$\mathfrak{D}^{*}(F;R)$
ferentials on . As in the case of a Lie group or an algebraic group, we

$F$
have:
PROPOSITION 1.1. If $F$ is an n-dimensional formal group over $R,$ $\mathfrak{D}^{*}(F;R)$
is a free R-module of rank . More precisely, denoting the inverse $n$ $(\psi_{ij}(z))$
matrix of , we have and

$((\partial/\partial x_{j})F_{i}(0, z))$ $\psi_{ij}(0)=\delta_{ij}$ $\omega_{i}=\sum_{j=1}^{n}\psi_{ij}(x)dx_{j}(1\leqq i\leqq n)$
form an R-basis of . Moreover the base is characterized$\mathfrak{D}^{*}(F;R)$ $\{\omega_{1}, \cdots , \omega_{n}\}$
by these two properties.

PROOF. Differentiating $F_{i}(u, F(v, w))=F_{i}(F(u, v),$ ) relative to , we get $w$ $u_{j}$
$(\partial/\partial x_{j})F_{i}(u, F(v, w))=\sum_{k=1}^{n}(\partial/\partial x_{k})F_{i}(F(u, v),$ $w$ ) $(\partial/\partial x_{j})F_{k}(u, v)$ ,
so that
$(\partial/\partial x_{j})F_{i}(0, F(v, w))=\sum_{k=1}^{n}(\partial/\partial x_{k})F_{i}(v, w)(\partial/\partial x_{j})F_{k}(0, v)$
or by matrix notation
(1.1) $((\partial/\partial x_{j})F_{i}(0, F(v, w)))=((\partial/\partial x_{j})F_{i}(v, w))((\partial/\partial x_{j})F_{i}(0, v))$ .
Since $(\partial/\partial x_{j})F_{i}(0, z)\equiv\delta_{ij}$
mod $\deg 1$ , the matrix $((\partial/\partial x_{j})F_{i}(0, z))$
is invertible,
$\psi_{ij}(z)\in R[[z]]$ and $\psi_{ij}(0)=\delta_{ij}$ . Hence (1.1) is equivalent to
(1.2) $(T_{t}\psi_{ij}(z))((\partial/\partial x_{j})F_{i}(z, t))=(\psi_{ij}(z))$ .
Now a differential $\omega=\sum_{i=1}^{n}\psi_{i}(x)dx_{i}$
in $\mathfrak{D}^{*}(A ; R)$ is right invariant on $F$ , if and
only if
(1.3) $\psi_{j}(x)=\sum_{k=1}^{n}\psi_{k}(F(x, t))(\partial/\partial x_{j})F_{k}(x, t)$ .
This shows $\omega_{1}$ , $\cdot$
.. , $\omega_{n}\in \mathfrak{D}^{*}(F;R)$
by (1.2). On the other hand we get from
(1.3)
$\psi_{j}(0)=\sum_{k=1}^{n}\psi_{k}(t)(\partial/\partial x_{j})F_{k}(0, t)$ ,
which implies that, if $\omega\in \mathfrak{D}^{*}(F;R),$

for $1\leqq i\leqq n$ . Therefore
$\omega=0\Leftrightarrow\psi_{i}(0)=0$
the map $\Phi:\omega->{}^{t}(\psi_{1}(0), \cdot..

defines an R-isomorphism of
, \psi_{n}(0))$ into $\mathfrak{D}^{*}(F;R)$
. Since the
$R^{n}$ $\Phi(\omega_{i})(1\leqq i\leqq n)$
are the unit vectors of , the map is sur- $R^{n}$ $\Phi$
jective and $\{\omega_{1}, \cdots , \omega_{n}\}$

is a base of . $\mathfrak{D}^{*}(F;R)$
We shall call this , .. ,

$\{\omega_{1}$
the canonical base of
$\cdot$
$\omega_{n}\}$
. $\mathfrak{D}^{*}(F;R)$
PROPOSITION 1.2. Let $F,$ $G$ be formal groups over $R$ and $\varphi\in Hom_{R}(F, G)$ .
If $\eta\in \mathfrak{D}^{*}(G;R)$ , then $\varphi^{*}(\eta)\in \mathfrak{D}^{*}(F;R)$ .
PROOF. Write $\eta=\sum_{i=1}^{m}\psi_{i}(x^{\prime})dx_{i}^{\prime}$
where $m$ is the dimension of $G$
. Then
$T_{t}(\varphi^{*}(\eta))=T_{t}(\sum_{i=1}^{m}\psi_{i}(\varphi(x))d\varphi_{i}(x))$
$=\sum_{i=1}^{m}\psi_{i}(\varphi(F(x, t)))d\varphi_{i}(F(x, t))$
$=\sum_{i=1}^{m}\psi_{i}(G(\varphi(x), \varphi(t)))dG(\varphi_{i}(x), \varphi_{i}(t))$
$=\sum_{i=1}^{m}\psi_{i}(\varphi(x))d\varphi_{i}(x)$
$=\varphi^{*}(\eta)$
.
1.3. We now study invariant differential forms on a commutative formal
group.
PROPOSITION 1.3. Let $F$ be a commutative formal group over R. Then
every differential in is closed. $\mathfrak{D}^{*}(F;R)$
PROOF. Let $\omega_{i}=\sum_{j=1}\psi_{ij}(x)dx_{j}n(1\leqq i\leqq n)$ be the canonical base of $\mathfrak{D}^{*}(F;R)$

.
We shall prove $d\omega_{i}=0$ for $1\leqq i\leqq n$ . First $d\omega_{i}$
is a right invariant 2-form,
since
$T_{t}^{*}(d\omega_{i})=T_{t}^{\star}(\sum_{=j1}^{n}d\psi_{ij}(x)$
A $dx_{j})$
$=\sum_{j}d\psi_{ij}(F(x, t))\wedge dF_{j}(x, t)$
$=d(T_{t}^{\star}(\omega_{i}))$
$=d\omega_{i}$
.
Now differentiating
$\sum_{k=1}^{n}(\partial/\partial x_{k})F_{i}(0, z)\psi_{kj}(z)=\delta_{ij}$
relative to $z_{\iota}$
and putting $z=0$ , we get
$\sum_{k}(\partial^{2}/\partial x_{k}\partial y_{\iota})F_{i}(0,0)\psi_{kj}(0)+\sum_{k}(\partial/\partial x_{k})F_{i}(0,0)(\partial/\partial x_{\iota})\psi_{kj}(0)=0$ ,
which is reduced to
218 T. HONDA
$(\partial^{2}/\partial x_{j}\partial y_{l})F_{i}(0,0)+(\partial/\partial x_{\iota})\psi_{ij}(0)=0$ ,

since
$\psi_{kj}(0)=\delta_{kj}$ and $(\partial/\partial x_{k})F_{i}(0,0)=\delta_{ik}$
.
Hence, by the commutativity of $F$
we get
$(\partial/\partial x_{\iota})\psi_{ij}(0)=-(\partial^{2}/\partial x_{j}\partial y_{l})F_{i}(0,0)$
$=-(\partial^{2}/\partial x_{\iota}\partial y_{j})F_{i}(0,0)$
$=(\partial/\partial x_{j})\psi_{il}(0)$
.
Since
$d\omega_{i}=\sum_{f.l}(\partial/\partial x_{\iota})\psi_{ij}(x)dx_{l}$ A $dx_{j}$
$=\sum_{j<l}((\partial/\partial x_{l})\psi_{ij}(x)-(\partial/\partial x_{j})\psi_{il}(x))dx_{\iota}\wedge dx_{j}$ ,
the coefficients of $dx_{\iota}$

A $dx_{j}$
in $d\omega_{i}$
have no constant term. So we have only
to prove that, if $\eta=\sum_{i<j}\lambda_{ij}(x)dx_{i}\wedge dx_{j}$ is right invariant on $F$ and $\lambda_{ij}(0)=0$
for all $1\leqq i<j\leqq n,$must be equal to $\eta$

$0$
. An easy computation shows that
$ T_{t}^{*}(\eta)=\eta$ is equivalent to
$\lambda_{kl}(x)=\sum_{i<j}\lambda_{ij}(F(x, t))\left|\begin{array}{lll}(\partial/\partial x_{k})F_{i}(x,t) & (\partial/\partial x_{l})F_{i}(x, & t)\\(\partial/\partial x_{k})F_{j}(x,t) & (\partial/\partial x_{\iota})F_{j}(x,t) & \end{array}\right|$
,
which implies
$\lambda_{kl}(0)=\sum_{Kj}\lambda_{ij}(t)\left|\begin{array}{ll}(\partial/\partial x_{k})F_{i}(0,t) & (\partial/\partial x_{\iota})F_{i}(0,t)\\(\partial/\partial x_{k})F_{j}(0,t) & (\partial/\partial x_{\iota})F_{j}(0,t)\end{array}\right|$
for $1\leqq k<l\leqq n$ . Since the matrix is regular, this shows in $((\partial/\partial x_{j})F_{i}(0, t))$
fact for all $i<j\Rightarrow\lambda_{ij}(t)=0$ for all $i<j$ .

$\lambda_{ij}(0)=0$
We now consider the case where $R$ is a Q-algebra. In this case every
power series in $R[[x]]$ is termwise integrable with respect to . The fol- $x_{i}$
lowing lemma is essentially well-known in elementary analysis and the proof

is easy.
LEMMA 1.4. If $R$ is a Q-algebra, a closed differential in is exact. $\mathfrak{D}^{*}(A;R)$
The following theorem, mentioned in [10], was also proved in [7] in a

slightly different manner.
THEOREM 1. Let $F$ be an n-dimensional commutative formal group over a
Q-algebra $R$ and let be the canonical base of . Then
$\omega={}^{t}(\omega_{1}, \cdots , \omega_{n})$ $\mathfrak{D}^{*}(F;R)$
there exists a unique element of such that $\omega=df$. This satisfies $f$ $R[[x]]_{0}^{n}$ $f$
$f(x)\equiv x$ mod $\deg 2$
and
$F(x, y)=f^{-1}(f(x)+f(y))$ .
In particular $F(x, y)\approx x+y$ over $R$
.
PROOF. The existence of follows from Proposition 1.3 and Lemma 1.4. $f$
The uniqueness follows from the fact that $d\varphi=0$ for $\varphi\in R[[x]]$ , if and only
if is a constant. Since
$\varphi$
, we have $f(x)\equiv x$ mod $\deg 2$ . Now, $df(x)$
$\psi_{ij}(0)=\delta_{ij}$
being right invariant, we have

$df(F(x, t))=df(x)$ ,
which implies
$f(F(x, t))-f(x)\in R[[t]]$ .
Writing $g(t)=f(F(x, t))-f(x)$ and putting $x=0$ we get
$g(t)=f(t)$ .
Thus we have
$f(F(x, t))=f(x)+f(t)$
or
$F(x, t)=f^{-1}(f(x)+f(t))$ .
This completes the proof of our theorem.
1.4. Let be an integral domain of characteristic
$R$ $0$
and $K$ its fraction field.
LEMMA 1.5. Let and $x={}^{t}(x_{1}, \cdots , x_{n})$ $y={}^{t}(y_{1}, \cdots , y_{n})$ be sets of variables. $n$
If $\psi\in K[[x]]^{m}$
satisfies
$\psi(x+y)=\psi(x)+\psi(y)$ ,
must be linear, . . there is an

$\psi$ $i$
$e$
matrix over $K$ such that $\psi(x)=Cx$ .
$m\times n$ $C$
PROOF. We have only to consider the case where $m=1$ and is a $\psi$
homogeneous polynomial. Then our assertion is verified by a simple com-

putation. (See the proof of Lemma 3.2)
Let $F$ be a commutative formal group over $R$ , of dimension . By Theorem $n$
1 there is $f(x)\in K[[x]]_{0}^{n}$ such that mod $\deg 2$ and $F(x, y)=f^{-1}(f(x)+f(y))$ .
$f\equiv i$
If there is another element of $K[[x]]_{0}^{n}$ satisfying $h$

mod $\deg 2$ and $h\equiv i$
$F(x, y)=h^{-1}(h(x)+h(y))$ , we have
$f\circ h^{-1}\equiv i$
mod $\deg 2$ ,
$(f\circ h^{-1})(x+y)=(f\circ h^{-1})(x)+(f\circ h^{-1})(y)$ .
Hence we get $f\circ h^{-1}=i$ or $f=h$ by Lemma 1.5.
DEFINITION. Let $R$ and $K$ be as above; let $F$ be an n-dimensional com-
mutative formal group over $R$ . The unique element of $K[[x]]_{0}^{n}$ , such that $f$
mod $\deg 2$ and $F(x, y)=f^{-1}(f(x)+f(y))$ , is called the transformer of $F$.

$f\equiv i$
Let be another commutative formal group over $R$ , of dimension $m$ and
$G$
with the transformer . If $\varphi\in Hom_{R}(F, G)$ , we have

$g$
$\varphi(f^{-1}(f(x)+f(y)))=g^{-1}(g(\varphi(x)+g(\varphi(y))))$ .
Substituting $x,$ $y$ by $f^{-1}(x),$ $f^{-1}(y)$ , respectively, we get
$(g\circ\varphi\circ f^{-1})(x+y)=(g\circ\varphi\circ f^{-1})(x)+(g\circ\varphi\circ f^{-1})(y)$ .
220 T. HONDA
Hence by Lemma 1.5 there is an $m\times n$ matrix $C$

over such that
$K$ $(g\circ\varphi\circ f^{-1})(x)$
$=Cx$ . This implies $\varphi(x)=g^{-1}(Cf(x))$ . As $\varphi(x)\equiv Cx$

mod deg2, $C$
is a matrix
with elements in . $R$
PROPOSITION 1.6. Let be as above. Every element of $Hom_{R}(F, G)$ $F,$ $f,$ $G,$ $g$ $\varphi$
has the form , where is an matrix over R. Conversely,

$g^{-1}\circ(Cf)$ $C$ $m\times n$ $C$
being an matrix over $R,$ $g^{-1}\circ(Cf)\in

$m\times n$ Hom_{R}(F, G)$ ,
if and only if $g^{-1}\circ(Cf)$
has coefficients in R. The map yields an isomorphism of $Hom_{R}(F, G)$ $\varphi-\div C$
into the module of $mXn$ matrices over R. If $F=G$ in particular, this map is
a ring isomorphism of $End_{R}F$ into $M_{n}(R)$ .
PROOF, The first assertion has already been proved. The second follows
from
$(g^{-1}\circ(Cf))\circ F=G\circ(g^{-1}\circ(Cf))$ .
The rests follow from the definitions.
\S 2. Formal groups over a -adic integer ring. $\mathfrak{p}$
Throughout the rest of this paper we exclusively deal with commutative

formal groups. By a formal group we always mean a commutative one.
Let $K$ be a discrete valuation field of characteristic and let and be the $0$ $0$ $\mathfrak{p}$
ring of integers in $K$ and the maximal ideal of , respectively. We assume $0$
that the residue class field is of characteristic $p>0$ . Consider the $k=0/\mathfrak{p}$
following condition on $K$ :

$(F)$ There are an endomorphism of $K$ and a power of such that $\sigma$
$q$ $p$
$\alpha^{\sigma}\equiv\alpha^{q}$
$mod \mathfrak{p}$
for any $\alpha\in 0$
.
We note , since sends a unit of to and . In this section
$\mathfrak{p}^{\sigma}=\mathfrak{p}$ $\sigma$ $0$ $0$ $p^{\sigma}=p$
we study formal groups over , when $K$ satisfies $(F)$ . We do not assume the $0$
completeness of $K$.
Let be a finite extension of the -adic number field
$K_{0}$
and let be $p$ $Q_{p}$ $q$
the cardinal of its residue field. Then it is well-known that an unramified

extension of (of finite or infinite degree) or its completion satisfies $(F)$
$K_{0}$
with a Frobenius . $\sigma$
2.1. Let be the non-commutative power series ring on

$K_{\sigma}[[T]]$ with $T$
the multiplication rule: for . We denote by (resp. ) $T\alpha=\alpha^{\sigma}T$ $\alpha\in K$ $\mathfrak{B}_{m,n}$ $\mathfrak{A}_{m_{J}n}$
the module consisting of all matrices over (resp. ). $m\times n$ $K_{\sigma}[[T]]$ $0_{\sigma}[[T]]$
Let be a set of variables. For $f\in K[[x]]_{0}^{m}$ and

$x={}^{t}(x_{1}, \cdots , x_{n})$ $n$
$u=\sum_{\nu-- 0}^{\infty}C_{\nu}T^{\nu}\in \mathfrak{B}_{l,m}$ (where the C. are matrices over $K$ ), we define an element
$u*f$ of $K[[x]]_{0}^{l}$ by
$(u*f)(x)=\sum_{\nu=0}^{\infty}C_{\nu}f^{\sigma^{\nu}}(x^{q^{\mathcal{V}}})$ .
This is well-defined, since $f(x)$ has no constant term. If $v=\sum_{\nu=0}^{\infty}D_{\nu}T$

“
is in $\mathfrak{B}_{k,l}$
,
we have
(2.1) $(vu)*f=v*(u*f)$ ,
since
$(v*(u*f))(x)=\sum_{\nu=0}^{\infty}D_{\nu}\sum_{\mu=0}^{\infty}C_{\mu}^{\sigma^{v}}f^{\sigma\mu+\nu}(x^{q\mu+\nu})$
$=\sum_{\lambda=0}^{\infty}\sum_{\mu+\nu=\lambda}D_{\nu}C_{\mu}^{\sigma^{\mathcal{V}}}f^{\sigma^{\lambda}}(x^{q^{\lambda}})$
$=((vu)*f)(x)$ .
From now on we fix a prime element $\pi$
of $0$
.
LEMMA 2.1. For any rational integers }) $\geqq 0,$ $a\geqq 1$ and $m\geqq 1$ we have
$\pi^{-\nu}(X+\pi Y)^{mp^{a\nu}}\equiv\pi^{-\nu}X^{mp^{av}}$ $mod \mathfrak{p}$
.
In particular we have
$m^{-1}(X+pY)^{m}\equiv m^{-1}X^{m}$ $mod pZ_{p}$
for $m\geqq 1$ .
This is Lemma 4 of [10]. As the proof is elementary and easy, we omit
it here.
We write (resp. ) for (resp.
$\mathfrak{A}_{n}$
). $\mathfrak{B}_{n}$
$\mathfrak{A}_{n,n}$ $\mathfrak{B}_{n,n}$
DEFINITION. An element of is said to be special, if mod $\deg 1$ .

$u$ $\mathfrak{A}_{n}$ $u\equiv\pi I_{n}$
Let be an invertible matrix in $M_{n}(0)$ and let

$P$
be a special element of . $u$ $\mathfrak{A}_{n}$
An element of $K[[x]]_{0}^{n}$ is said to be of type $(P;u)$ , if fsatisfies the folloN-

$f$
ing two conditions:

i) $f(x)\equiv Px$ mod $\deg 2$ ,
ii) $(u*f)(x)\equiv 0$ $mod \mathfrak{p}$
.
If $f$
is of type , we shall simply say that
$(I_{n} ; u)$ $f$ is of type $u$ .
Let be special and put
$u\in \mathfrak{A}_{n}$ $w=u^{-1}\pi(\in \mathfrak{B}_{n})$
. Then, $i$
being the identity
function,
$(u*(w*i))(x)=((uw)*i)(x)=\pi x\equiv 0$ $mod \mathfrak{p}$
.
This implies that $(u^{-1}\pi)*i$ is of type . $u$
LEMMA 2.2. Let be special and put $u\in \mathfrak{A}_{n}$ $u^{-1}\pi=I_{n}+\sum_{\nu=1}^{\infty}B_{v}T^{\nu}$ . Then we
have for .
$\pi^{\nu}B_{\nu}\in M_{n}(0)$ $\nu\geqq 0$
PROOF. Write and replace $u=\pi I_{n}+\sum_{\nu=I}^{\infty}C_{\nu}T^{\nu}$ $T$ by $\pi T$

in the equality
$(\pi I_{n}+\sum_{\nu=I}^{\infty}C_{\nu}T^{\nu})(I_{n}+\sum_{\nu=1}^{\infty}B_{\nu}T^{\nu})=\pi I_{n}$ .

Then we get
222 T. HONDA
$(’\llcorner$
This implies , since is also a prime element of .

$\pi^{v}B_{\nu}\in M_{n}(0)$ $\pi^{\sigma\mu}$
$0$
2.2. The following two lemmas play crucial roles in our further inves-
tigation and will be used repeatedly.
LEMMA 2.3. Let $f\in K[[x]]_{0}^{n}$ be of type $(P;u)$ and let be an element of $v$
. Let be an element of
$\mathfrak{A}_{m,n}$
,
$\psi$
being a finite set of variables. If $K[[x^{\prime}]]_{0}^{n}$ $x^{\prime}$
the coef7 cients (of components) of of terms of (total) degree $\leqq r-1$ , belong $\psi,$
for some
$lo\mathfrak{v}$
, we have $r\geqq 2$
$ v*(f\circ\psi)\equiv(v*f)\circ\psi$ mod $\deg(r+1)$ , $mod \mathfrak{p}$

.
If $\psi\in \mathfrak{o}[[x^{\prime}]]_{0}^{n}$
in particular, we have
$ v*(f\circ\psi)\equiv(v*f)\circ\psi$ $mod \mathfrak{p}$
.
LEMMA 2.4. If $f$ (resp. $g$ ) $\in K[[x]]_{0}^{n}$ is of type $(P;u)$ (resp. of type ( $Q$ ; )),
$u$
then $g^{-1}\circ f\in 0[[x]]_{0}^{n}$ .

Put $h=(u^{-1}\pi)*i$ . First we will prove the first assertion of Lemma 2.3 for
$f=h$ . Write
$u^{-1}\pi=I_{n}+\sum_{\nu=1}^{\infty}B_{v}T^{\nu}$ , $v=\sum_{\nu=0}^{\infty}A_{\nu}T^{\nu}$
We have
(2.2) $((v*h)\circ\psi)(x^{\prime})=(((vu^{-1}\pi)*i)\circ\psi)(x^{\prime})$
$=\sum_{\mu’\nu}A_{\nu}B_{u}^{\sigma^{\nu}}\psi(x^{\prime})^{q^{\alpha+\nu}}$
Now
(2.3) $B_{\mu}^{\sigma^{\nu}}\psi(x^{\prime})^{q\mu+\nu}=\pi^{\mu}B_{\mu}^{\sigma^{\nu}}\pi^{-\mu}\psi(x^{\prime})^{q\mu+\nu}$
and $\pi^{\mu}B_{u}^{\sigma^{\nu}}\in M_{n}(0)$

by Lemma 2.2. We will prove
(2.4) $\pi^{-\mu}\psi(x^{\prime})^{q\mu+\nu}\equiv\pi^{-\mu}(\psi^{\sigma^{y}}(x^{\prime q^{\nu}}))^{q\mu}$
mod $\deg(r+1)$ , $mod \mathfrak{p}$
.
If $\mu=\nu=0,$ $(2.4)$ is trivial. If $\mu=0$ and $\nu\geqq 1$
, we have
$\psi(x^{\prime})^{q^{\mathcal{V}}}\equiv\psi^{\sigma^{\nu}}(x^{\prime q^{\nu}})$
mod $\deg(r+1)$ , $mod \mathfrak{p}$ ,
since terms of of degree $\psi$ $\geqq r$

do not affect this congruence. (Note $\psi(0)=0.$ )
Assume . Because $\mu\geqq 1$
$\psi(x^{\prime})^{q^{\nu}}\equiv\psi^{\sigma^{\nu}}(x^{\gamma q^{v}})$
mod $\deg r$ , $mod \mathfrak{p}$
,
we get (2.4) by Lemma 2.1 and by the fact $\psi(0)=0$ . This completes the proof
of (2.4). Thus we get from (2.2), (2.3) and (2.4)
$((v*h)\circ\psi)(x^{\prime})\equiv\sum_{\mu\nu}A_{\nu}B_{\mu}^{\sigma^{\nu}}(\psi^{\sigma\nu}(x^{\prime q^{\nu}}))^{q\mu}$ mod $\deg(r+1)$ , $mod \mathfrak{p}$
$=(v*(h\circ\psi))(x^{\prime})$ .
PROOF LEMMA 2.4. Since $g^{-1}\circ f=(g^{-1}\circ h)\circ(h^{-1}\circ f)=(h^{-1}\circ g)^{-1}\circ(h^{-1}\circ f)$
OF
and $(h^{-1}\circ g)(x)\equiv Qx$ mod deg2, we have only to prove . Put $h^{-1}\circ f\in 0[[\chi]]_{0}^{n}$
or
$ h^{-1}\circ f=\varphi$ . The first-degree coefficients of
$ f=h\circ\varphi$ are in . Assume $\varphi$
$0$
that the coefficients of , of (total) degree $\leqq r-1$ , are integers for some
$\varphi$ . $r\geqq 2$
By Lemma 2.3 for $f=h$ we have

$\pi\varphi=(u*h)\circ\varphi\equiv u*(h\circ\varphi)$ mod $\deg(r+1)$ , $mod \mathfrak{p}$
$=u*f\equiv 0$ $mod \mathfrak{p}$

.
This implies that the r-th degree coefficients of are also integers. This $\varphi$
completes our proof by induction.

PROOF OF LEMMA 2.3. We have only to prove the first assertion. Nota-
tions being as above,
$v*(f\circ\psi)=v*((h\circ\varphi)\circ\psi)=v*(h\circ(\varphi\circ\psi))$
$\equiv(v*h)\circ(\varphi\circ\psi)$ mod $\deg(r+1)$ , $mod \mathfrak{p}$
$=((v*h)\circ\varphi)\circ\psi$ .
Since $\varphi(x)\equiv Px$
mod $\deg 2$ , we have
$((v*h)\circ\varphi)(x)\equiv A{}_{0}Px\equiv(v*(h\circ\varphi))(x)$ mod $\deg 2$ .
Put $\lambda_{1}(x)=((v*h)\circ\varphi)(x)-A{}_{0}Px$ and $\lambda_{2}(x)=(v*(h\circ\varphi))(x)-A{}_{0}Px$ . Then $\lambda_{1}\equiv\lambda_{2}\equiv 0$
mod $\deg 2$ and $\lambda_{1}\equiv\lambda_{2}mod \mathfrak{p}$

by what we have proved. It follows from this
$\lambda_{1}\circ\psi\equiv\lambda_{2}\circ\psi$
mod $\deg(r+1)$ , $mod \mathfrak{p}$ ,
since the terms of $\psi$
of degree $r$
do not affect this congruence. Hence we get
$ v*(f\circ\psi)\equiv((v*h)\circ\varphi)\circ\psi$ mod $\deg(r+1)$ , $mod \mathfrak{p}$
$=A{}_{0}P\psi+\lambda_{1}\circ\psi$
$\equiv A{}_{0}P\psi+\lambda_{2}\circ\psi$
$=(v*(h\circ\varphi))\circ\psi$
$=(v*f)\circ\psi$ .
This completes the proof of our lemma.
2.3. The results of 2.2 first allow us to construct certain formal groups
over . $0$
THEOREM 2. Assume $K$ satisfies $(F)$ . Let be an invertible matrix in $P$
$M_{n}(0)$ and let be a special element of

$u$ . If $f\in K[[x]]_{0}^{n}$ is of type $(P;u)$ , $\mathfrak{A}_{n}$
$F(x, y)=f^{-1}(f(x)+f(y))$ is a formal group over . Let $g\in K[[x]]_{0}^{n}$ be of type $\mathfrak{o}$
( ; u) for an invertible matrix

$Q$ and put $G(x, y)=g^{-1}(g(x)+g(y))$ . Then we $Q$
have $G\sim F$ over . If $P=Q$ in particular, we have

$0$
over . $G\approx F$ $\mathfrak{o}$
PROOF. Form $h=(u^{-1}\pi)*i$ and $H(x, y)=h^{-1}(h(x)+h(y))$ . It is clear that

224 T. HONDA
$H(x, y)\equiv x+y$ mod $\deg 2$ .

Assume that the coefficients of $H$, of terms of degree $\leqq r-1$ , are integers for
some . By Lemma 2.3 we have
$r\geqq 2$
$\pi H(x, y)=((u*h)\circ H)(x, y)$
$\equiv(u*(h\circ H))(x, y)$ mod $\deg(r+1)$ , $mod \mathfrak{p}$
$=(u*h)(x)+(u*h)(y)$
$=\pi x+\pi y\equiv 0$ $mod \mathfrak{p}$
.
This implies that the r-th degree coefficients of $H$ are also integers. This
proves $H(x, y)\in 0[[x, y]]$ by induction. All the assertions of our theorem
follow from this and from Lemma 2.4, because if . $ F=\varphi^{-1}\circ H\circ\varphi$ $ f=h\circ\varphi$
As for examples, see \S 5.

PROPOSITION 2.5. Let be an invertible matrix in $M_{n}(0)$ and let $P$ be a $u$
special element of . Then $f\in K[[x]]_{0}^{n}$ is of type $(P;u)$ , if and only if

$\mathfrak{A}_{n}$
is $f$
of the form with such

$((u^{-1}\pi)*i)\circ\varphi$ that mod deg2. $\varphi\in 0[[x]]_{0}^{n}$ $\varphi(x)\equiv Px$
PROOF. Only if part is Lemma 2.4. Conversely, if

” ‘’
and $\varphi\in\circ[[x]]_{0}^{n}$
mod $\deg
$\varphi(x)\equiv Px$ 2$ , we have, writing $h=(u^{-1}\pi)*i$ ,
$(h\circ\varphi)(x)\equiv Px$ mod $\deg 2$
and by Lemma 2.3

$u*(h\circ\varphi)\equiv(u*h)\circ\varphi=\pi\varphi\equiv 0$ $mod \mathfrak{p}$
.
This completes our proof.
Dually to Proposition 2.5 we have
PROPOSITION 2.6. Let $f\in K[[x]]_{0}^{n}$ be of type $(P;u)$ for an invertible matrix
of $M_{n}(0)$ and a special element of ; Let
$P$ $u$ $\mathfrak{A}_{n}$
$v$
be a matrix in $\mathfrak{A}_{m,n}$
. Then
$v*f\equiv 0$ $mod \mathfrak{p}$ ,
if and only if there exists $t\in \mathfrak{A}_{m,n}$

such that $v=tu$ .
PROOF. If $v=tu$ with $t\in \mathfrak{A}_{m,n}$ , then
$v*f=t*(u*f)\equiv 0$ $mod \mathfrak{p}$

.
Conversely, assume for $v*f\equiv 0mod \mathfrak{p}$ $v\in \mathfrak{A}_{m,n}$
.
Put $h=(u^{-1}\pi)*i$ and .
$\varphi=h^{-1}\circ f$
Since $\varphi$
is an invertible element of $\mathfrak{v}[[x]]_{0}^{n}$
by Lemma 2.4, we have
$(\nu*h)\circ\varphi\equiv v*(h\circ\varphi)=v*f\equiv 0$ $mod \mathfrak{p}$
by Lemma 2.3, so that

(2.5) $v*h=((v*h)\circ\varphi)\circ\varphi^{-1}\equiv 0$ modp.
Put $vu^{-1}\pi=\sum_{\nu=0}^{\infty}A_{\nu}T^{\nu}$
. Since
$v*h=v*((u^{-1}\pi)*i)=(vu^{-1}\pi)*i$ ,
we have from (2.5)

$\sum_{\nu=0}^{\infty}A_{\nu}x^{q^{\nu}}\equiv 0$ $mod \mathfrak{p}$
,
which implies . This completes our proof.

$vu^{-1}=(vu^{-1}\pi)\pi^{-1}\in \mathfrak{A}_{m,n}$
2.4. We now study homomorphisms of formal groups constructed in

Theorem 2. $M_{m,n}(0)$ denotes the module of all the matrices with elements $m\times n$
in . $0$
THEOREM 3. Assume $K$ satisfies $(F)$ . Let and be special and $u\in \mathfrak{A}_{n}$ $v\in \mathfrak{A}_{m}$
let $f\in K[[x]]_{0}^{n}$ (resp. $g\in K[[x]]_{0}^{m}$ ) be of type (resp. of type ). Form $F(x,$ $u$ $v$ $ y\rangle$
$=f^{-1}(f(x)+f(y))$ and $G(x, y)=g^{-1}(g(x)+g(y))$ . Then $g^{-1}\circ(Cf)\in Hom_{0}(F, G)$

for $C\in M_{m,n}(0)$ , if and only if there exists such that $vC=tu$ . $t\in \mathfrak{A}_{m,n}$
PROOF. Put $\varphi=g^{-1}\circ(Cf)$ . By Proposition 1.6 $\varphi\in Hom_{0}(F, G)$ if and only
if . In view of Lemma 2.4 we may assume $f=(u^{-1}\pi)*i$ and
$\varphi\in \mathfrak{o}[[x]]_{0}^{m}$
$g=(v^{-1}\pi)*i$ . If , we have by Lemma 2.3

$\varphi\in 0[[x]]_{0}^{m}$
$(vC)*f=v*(Cf)=v*(g\circ\varphi)$
$\equiv(v*g)\circ\varphi=\pi\varphi\equiv 0$ modp.
Hence, by Proposition 2.6, there exists such that $vC=tu$ . Conversely, $t\in \mathfrak{A}_{m,n}$
suppose that there is such that $vC=tu$ . As mod $\deg 2$ , the

$t\in \mathfrak{A}_{m,n}$ $\varphi(x)\equiv Cx$
first-degree coefficients of are integral. Assume that i-th degree coefficients $\varphi$
of are integral for $i\leqq r-1(r\geqq 2)$ . By Lemma 2.3 we have then
$\varphi$
$\pi\varphi=(v*g)\circ\varphi$
$\equiv v*(g\circ\varphi)$ mod $\deg(r+1)$ , $mod \mathfrak{p}$
$=\nu*(Cf)=(vC)*f$
$=(tu)*f=t*(u*f)$
$\equiv 0$ $mod \mathfrak{p}$
.
This shows that the r-th degree coefficients of $\varphi$
are integral. Hence we get
by induction.
$\varphi\in \mathfrak{o}[[x]]_{0}^{m}$
$CoROLLARY$ . Let be as in Theorem 3. The module $Hom_{0}(F, G)$ is

$F,$ $G$
canonically isomorphic to . $M_{m,n}(0)\cap v^{-1}\mathfrak{A}_{m,n}u$
By Theorem 3 $g^{-1}\circ(Cf)\in Hom_{0}(F, G)$ for , if and only if $C\in M_{m,n}(\mathfrak{o})$
. Our assertion follows from this and from Proposition 1.6.

$C\in\nu^{-1}\mathfrak{A}_{m,n}u$
\S 3. The non-ramified case.

Let $K$ be as in \S 2. In \S 3 we assume moreover that:
, o, p and $k$
The valuation of $K$ is unramified and $(F)$ is satisfied with $q=p$ .

$(F_{1})$
The ring of Witt vectors over a perfect field

$W(k^{\prime})$ of characteristic $k^{\prime}$
226 T. HONDA
$p>0$ satisfies (cf. [22]). Under we can take as the fixed prime
$(F_{1})$ $(F_{1})$ $p$
element of . $0$
3.1. Let be the set of variables as usual. Let $N$ be the set of all $x$ $n$
the non-negative rational integers. For we write for $\alpha=$ $(\alpha_{1}, \cdots , \alpha_{n})\in N^{n}$ $x^{\alpha}$
. Then is the degree of . For

$x_{1}^{\alpha_{1}}\cdots x_{n}^{\alpha_{n}}$
, let $|\alpha|=\alpha_{1}+\cdots+\alpha_{n}$ $x^{\alpha}$
$1\leqq i\leqq n$ $\epsilon_{i}$
denote the vector of whose j-th component is . Then $N^{n}$ $\delta_{ij}(1\leqq j\leqq n)$ $x^{r\epsilon_{i}}=x_{i}^{r}$
for $r\in N$. Every element of $K[[x]]$ is written in the form . $\sum_{\alpha\in N^{n}}a_{\alpha}x^{\alpha}(a_{\alpha}\in K)$
LEMMA 3.1. For define the form in $Z[X, Y]$ as follows: $r\geqq 2$ $\Lambda_{r}(X, Y)$
If is not a power of a prime number, we put $\Lambda_{r}(X, Y)=(X+Y)^{r}-X^{r}-Y^{r}$ .

$r$
If is a power of a prime number , we put $\Lambda_{r}(X, Y)=l^{-1}((X+Y)^{r}-X^{-r}-Y^{r})$ .

$r$ $l$
Then is a primitive polynomial in $Z[X, Y]$ .

$\Lambda_{r}$
PROOF. Easy. See also [11], III.

For any commutative ring $R,$ is considered a polynomial in $R[X, Y]$ . $\Lambda_{r}$
LEMMA 3.2. Let be a form of degree satisfying $\lambda(x)=\sum_{|\alpha|=r}a_{\alpha}x^{\alpha}(a_{\alpha}\in K)$

$r$
(3.1) $\lambda(x+y)\equiv\lambda(x)+\lambda(y)$ $mod \mathfrak{p}$

.
Then, is not a power of
if $r$ $p,$ $a_{\alpha}\in \mathfrak{p}$
for all . If is a power of $\alpha$ $r$ $p,$ $a_{\alpha}\in 0$
for all and for

$\alpha$
$a_{\alpha}\in \mathfrak{p}$ $\alpha\neq r\epsilon_{i}(1\leqq i\leqq n)$
.
PROOF. Take such that $\alpha\in N^{n}$
. If two of , .. , , say
$|\alpha|=r$ $\alpha_{1}$
$\cdot$
$\alpha_{n}$ $\alpha_{1}$
and
$\alpha_{2}$ , are not equal to $0$
, the coefficient of $x_{1}^{a_{1}}y_{2}^{\alpha_{2}}$ $y_{n}^{\alpha_{n}}$
on the left side of (3.1)
is and no term of this form appears on the right.
$a_{\alpha}$
Hence we have $a_{\alpha}\in \mathfrak{p}$
for such . If , we have $\alpha$

$\alpha=r\epsilon_{i}$
$a_{\alpha}\{(x_{i}+y_{i})^{r}-x_{i}^{r}-y_{i^{r}}\}\equiv 0$ $mod \mathfrak{p}$
from (3.1). Then our assertion is a direct consequence of Lemma 3.1.

PROPOSITION 3.3. Let $F$ be an n-dimensional formal group over and let $0$
be its transformer. Then there exists a special element

$f$
of such that $u$ $\mathfrak{A}_{n}$
$f$
is of type . $u$
PROOF. As $f(x)\equiv x$ mod $\deg 2$ , we have $pf(x)\equiv 0$ mod $\deg 2,$ . Sup- $mod \mathfrak{p}$
pose that for there are matrices , in $M_{n}(0)$ satisfying

$\mu\geqq 0$ $C_{1},$ $\cdots$ $C_{\mu}$
(3.2) $pf(x)+\sum_{\nu=1}^{\mu}C_{\nu}f^{\sigma^{\nu}}(x^{p^{\nu}})\equiv 0$
mod $\deg(p^{\mu}+1)$ , $mod \mathfrak{p}$
.
Write $f_{i}(x)=\sum_{\alpha}a_{\alpha,i}x^{\alpha}$ for $1\leqq i\leqq n$ . Since $df_{i}(x)\in \mathfrak{D}^{*}(F;0)$ by the results of
\S 1, the $(\partial/\partial x_{j})f_{i}(x)$
have integral coefficients. In particular we have $\alpha_{j}a_{\alpha,i}\in 0$
for $1\leqq j\leqq n$ . Hence by Lemma 2.1 we get
$a_{\alpha,i}(x+py)^{\alpha}=\alpha_{1}a_{\alpha,i}\alpha_{1}^{-1}(x_{1}+py_{1})^{\alpha_{1}}\prod_{j=2}^{n}(x_{j}+py_{j})^{\alpha_{j}}$
$\equiv\alpha_{1}a_{\alpha,i}\alpha_{1}^{-1}x_{1}^{a_{1}}\prod_{J=2}^{n}(x_{j}+py_{j})^{\alpha_{j}}$ $mod \mathfrak{p}$

$=x_{1^{y_{1}}}(a_{\alpha,i}\prod_{j=2}^{n}(x_{j}+py_{j})^{\alpha_{j}}$ .
By repeating the same argument we have
(3.3) $a_{\alpha,i}(x+py)^{\alpha}\equiv a_{\alpha,i}x^{\alpha}$ $mod \mathfrak{p}$

.
Put now
(3.4) $pf(x)+\sum_{\nu=1}^{\mu}f^{\sigma^{\rho}}(x^{P^{\nu}})\equiv\sum_{|\beta|\geqq p^{\mu+1}}b_{\beta}x^{\beta}$
$mod \mathfrak{p}$ $(b_{\beta}\in K^{n})$ .
Substituting $x$ by $F(x, y)$ in (3.4) we get
(3.5) $pf(F(x, y))+\sum_{\nu=I}^{\mu}f^{\sigma^{v}}(F(x, y)^{P^{\nu}})\equiv\sum_{|\beta|\geqq p^{\mu+1}}b_{\beta}F(x, y)^{\beta}$

$mod \mathfrak{p}$
.
By (3.3) the left side of (3.5) is congruent $mod \mathfrak{p}$

to
$pf(F(x, y))+\sum_{\nu=1}^{\mu}C_{\nu}f^{\sigma^{\nu}}(F^{\sigma^{\nu}}(x^{p^{\nu}}, y^{P^{\nu}}))$
$=pf(x)+\sum_{\nu=1}^{\mu}C_{\nu}f^{\sigma^{\nu}}(x^{p^{\nu}})+pf(y)+\sum_{\nu=1}^{\mu}C_{\nu}f^{\sigma^{\nu}}(y^{p^{\nu}})$
$\equiv\sum_{|\beta|\geqq p^{\mu}+1}b_{\beta}(x^{\beta}+y^{\beta})$
.
Thus, denoting by $b_{\beta,i}$

the i-th component of $b_{\beta}$
, we get
(3.6) $\sum_{|\beta|\geqq p^{\mu}+1}b_{\beta,i}\{F(x, y)^{\beta}-x^{\beta}-y^{\beta}\}\equiv 0$
$mod \mathfrak{p}$
for Let
$1\leqq i\leqq n$
. $r$
be the minimum value of $|\beta|$
such that $b_{\beta,i}\not\in \mathfrak{p}$
for some .
$i$
Then (3.6) implies

$\sum_{|\beta|=r}b_{\beta,i}\{(x+y)^{\beta}-x^{\beta}-y^{\beta}\}\equiv 0$
$mod \mathfrak{p}$
.
Applying Lemma 3.2 to this we see $r\geqq p^{\mu+1}$
. At any rate we have
$\sum_{|\beta|=p^{\mu+1}}b_{\beta,i}\{(x+y)^{\beta}-x^{\beta}-y^{\beta}\}\equiv 0$
$mod \mathfrak{p}$
.
Hence, by Lemma 3.2, for
$b_{\beta,i}\in 0$ and $\beta=p^{\mu+1}\epsilon_{j}(1\leqq j\leqq n)$ for other
$b_{\beta,i}\in \mathfrak{p}$ $\beta$
such that $|\beta|=p^{\mu+1}$ . Therefore we can find a matrix in $C_{\mu+1}$ $M_{n}(0)$ satisfying
$pf(x)+\sum_{\nu=1}^{\mu}C_{\nu}f^{\sigma^{\nu}}(x^{P^{\nu}})\equiv-C_{\mu+1}x^{p\mu+1}$
mod $\deg(p^{\mu+1}+1)$ , $mod \mathfrak{p}$
,
from which follows
(3.7) $pf(x)+\sum_{\nu=1}^{\mu+1}C_{\nu}f^{\sigma^{\nu}}(x^{p^{\nu}})\equiv 0$
mod $\deg(p^{\mu+1}+1)$ , $mod \mathfrak{p}$
.
228 T. HONDA
Thus we have been able to replace by $\mu+1$ in (3.2). This implies the $\mu$
existence of , $\cdots\in M_{n}(0)$ satisfying $C_{1},$ $C_{2},$ $\cdots$ $C_{\nu},$
(3.8) $pf(x)+\sum_{\nu\approx 1}^{\infty}$

C. $f^{\sigma^{\nu}}(x^{p\nu})\equiv 0$ $mod \mathfrak{p}$
.
This means that $f$ is of type $u$ , where $u=pI_{n}+\sum_{\nu=1}^{\infty}C_{\nu}T^{\nu}$ .
3.2. By Theorem 2 and Proposition 3.3 every n-dimensional formal group
over is obtained from a special element of . Let and be n-dimensional
$0$ $\mathfrak{A}_{n}$ $F$ $G$
formal groups over , with the transformers and . By Proposition 3.3 $0$ $f$ $g$
there exist special elements of such that (resp. g) is of type (resp. $u,$ $v$ $\mathfrak{A}_{n}$
$f$ $u$
of type ). By the uniqueness of transformer

$v$
over if and only if $F\approx G$ $\mathfrak{o}$
. By Theorem 3 this happens if and only if there is

$g^{-1}\circ f\in 0[[x]]_{0}^{n}$ $t\in \mathfrak{A}_{n}$
such that $v=tu$ . It is clear that such is a unit in . Let and be $t$ $\mathfrak{A}_{n}$
$u^{\prime}$ $v^{\prime}$
elements of . We shall say that is left associate with , if there is a

$\mathfrak{A}_{n}$ $v^{\prime}$ $u^{\gamma}$
unit in such that

$t^{\prime}$
. We have proved the following theorem :

$\mathfrak{A}_{n}$ $v^{\prime}=t^{\prime}u^{\prime}$
THEOREM 4. Assume $K$

satisfies . Then every n-dimensional formal $(F_{1})$
group over is obtained from a special element of by the method of Theorem

$\mathfrak{o}$
$\mathfrak{A}_{n}$
2. The strong isomorphism classes of n-dimensional groups over correspond $0$
bijectively to the left associate classes of special elements of . $\mathfrak{A}_{n}$
COROLLARY. Let $M$ be a complete system of representatives of . $0mod \mathfrak{p}$
Then the strong isomorphism classes of n-dimensional formal groups over $\mathfrak{v}$
correspond bijectively to the special elements of whose coeffcient matrices $\mathfrak{A}_{n}$
have elements in $M$.

PROOF. Let be a fixed special element of and let $u=pI_{n}+\sum_{\nu=1}^{\infty}C_{\nu}T^{\nu}$ $Qt_{n}$
$t=I_{n}+\sum_{\nu=1}^{\infty}A_{J}\backslash T^{\nu}$
be a unit in $\mathfrak{A}_{n}$
. Then we have
$tu=pI_{n}+\sum_{\nu=1}^{\infty}(pA_{\nu}+\sum_{\mu\backslash \nu}A_{\mu}C_{\nu-\mu}^{\sigma}\mu)T^{\nu}$ .
Therefore we can choose successively and uniquely so that the $A_{1},$ $ A_{2}\cdots$
coefficients of the $T$ “ in

have all their elements in $M$. Our assertion follows $tu$
from this and from Theorem 4.

3.3. As for the classification of (strong) isomorphism classes of n-dimen-
sional groups over , it is preferable to construct a module space over . In $0$ $0$
the following we will perform it in case $n=1$ and is complete. $\mathfrak{o}$
The following lemma is a slight modification of Lemma 2.1 of [16].

LEMMA 3.4. In addition to the condition , suppose that is complete. $(F_{1})$ $0$
Let $u=p+\sum_{\nu=1}^{\infty}c_{\nu}T^{\nu}(c_{\nu}\in \mathfrak{o})$ be a special element of $0_{\sigma}[[T]]$ . If all the $c_{\nu}$

are in
$\mathfrak{p}$
, there is a unit $t$

in $0_{\sigma}[[T]]$ such that $tu=p$ . If $c_{1},$ $\cdots$
, $c_{h-1}\in \mathfrak{p}$
but $c_{h}\not\in \mathfrak{p}$
,
then there is a unit $t$
in $0.[[T]]$ such that tu is of the form $p+\sum_{\nu=1}^{h}b_{\nu}T^{\nu}$
where
$b_{1},$ $\cdots$
, $b_{h-1}\in \mathfrak{p}$
and $b_{h}\not\in \mathfrak{p}$
.
PROOF. If all the $c_{\nu}$
are in $\mathfrak{p}$
, it suffices to put $t=pu^{-1}$ . Assume $c_{1}$ , $\cdot$

.. , $c_{h- 1}$
but
$\in \mathfrak{p}$ $c_{h}\in E\mathfrak{p}$
. We will show that for every $i\geqq 1$
we can choose $b_{1}^{(i)}$
, $\cdot$
.. , $b_{h^{f)}}^{(}\in 0$
and a unit $t_{i}$

of $0_{\sigma}[[T]]$ satisfying
$(1\leqq\nu\leqq h)$ ,
$mod \mathfrak{p}^{l}mod \mathfrak{p}$
(3.9) $\left\{\begin{array}{l}b_{\nu}^{(i+1)}\equiv b_{\nu}^{(i)} mod\mathfrak{p}^{i}, b_{\nu}^{(1)}\equiv c_{\nu}\\t_{i}\equiv l moddeg1, t_{i+1}\equiv t_{i}\\t_{i}u\equiv p+\Sigma b_{\nu}^{(i)}Th mod\mathfrak{p}^{i}.\end{array}\right.$
First put $b_{1}^{(1)}=\ldots=b_{h-1}^{(1)}=0,$ $b_{h}^{(1)}=c_{h}$

and $t_{1}=c_{h}(.\sum_{=,Jn}^{\infty}c_{h}T^{\nu-h})^{-1}$
As $c_{h}$
is a unit,
$t_{1}\in 0_{\sigma}[[T]]$ . Since
$t_{1}u\equiv c_{h}T^{h}$ $mod \mathfrak{p}$
,
(3.9) is $satisfi^{\mathfrak{Q}}.d$
by $\{b_{\nu}^{(1)} ; t_{1}\}$
with $i=1$ . Suppose that we have already found
$\{b_{\nu^{j)}}^{(} ; t_{j}\}$
for $1\leqq j\leqq i$
satisfying (3.9). We try to determine $b_{\nu}^{(i+1)}=b_{\nu^{i)}}^{\dot{\iota}}+p^{i}d_{\nu}^{(i)}$
$(1\leqq\nu\leqq h)$
and $t_{i-\vdash 1}=r_{i}+p^{i}v_{i}$
so that
(3.10) $(t_{i}+p^{i}v_{i})u\equiv p+\sum_{\nu\propto 1}(b_{\nu}^{(i)}+p{}^{t}d_{\nu}^{(7)})T^{\nu}$
$mod \mathfrak{p}^{i+1}$
Put $w_{i}=p^{-i}\{t_{i}u-(p+\sum_{\nu=1}^{h}b_{\nu}^{(j)}T^{\nu})\}(\in 0_{\sigma}[[T]])$ . Since $p^{i}u\equiv p^{i}(\sum_{\nu=h}c_{\nu}T^{\nu})mod \mathfrak{p}^{i+1}$ ,

(3.10) is reduced to
(3.11) $v_{i}\sum_{\nu--h}^{\infty}c_{\nu}T^{\nu}\equiv\sum_{/=1}^{h}d_{\nu}^{(?)}T‘‘-w_{i}$ $mod \mathfrak{p}$

.
As has no constant term, we can choose
$w_{i}$ , .. , so that the right $d_{1}^{(i)}$ $\cdot$
$d_{h}^{(i)}\in 0$
hand side of (3.11) has no term of degree . Hence we can find a series $\leqq h$
, without constant term and satisfying (3.11). By induction this

$v_{i}\in 0_{\sigma}[[T]]$
proves the existence of for all . Put and for $\{b_{\nu}^{(i)} ; t_{i}\}$
$i$
$f=\varliminf t_{i}$ $b_{\nu}=\varliminf_{i}b_{\nu}^{(i)}$
$1\leqq v\leqq h$ . Then

satisfy the requirement of our lemma. $\{b_{v} ; t\}$
be a l-dimensional formal group over . We shall say that $F$ is of

Let $F$ $0$
height if the reduction of $F$ modulo is of height

$h$
(cf. [11]). $\mathfrak{p}$
$h$
PROPOSITION 3.5. Let $K$ be a complete discrete valuation field satisfying

. The strong isomorphism classes of l-dimensional formal groups over ,
$(F_{1})$ $0$
of height $h(1\leqq h<\infty)$ , correspond bijectively to the special elements of the form
$u=p+\sum_{\nu=1}^{h}b_{\nu}T^{\nu}$
where $b_{1},$ $\cdots$
, $b_{h-1}\in \mathfrak{p}$
but $b_{h}$
is a unit of . Let $0$ $v=p+\sum_{\nu=1}^{h}c_{\nu}T^{\nu}$
be another special element of this form. Then the formal group obtained from
$u$is weakly isomorphic to the one obtained from , if and only if there exists $v$
a unit of such that $c$

for . $0$ $c_{\nu}=c^{1-\sigma^{\nu}}b_{\nu}$ $1\leqq\nu\leqq h$
PROOF. Let $F$ be a l-dimensional formal group over . Then its trans- $\mathfrak{o}$
former is of type for a s-oecial element . If all the coefficients of

$f$ $u^{\prime}$ $u^{\prime}$ $u^{\prime}$
230 T. HONDA
are in $\mathfrak{p}$
, then $F(x, y)\approx x+y$ by Lemma 3.4 and Theorem 2. If not, $f$ is also
of type $u$ , where $u$ is a special element of the form $p+\sum_{\nu-I}^{h}b_{\nu}T^{\nu}(b_{1}$ , $\cdot$
.. $b_{h- 1}\in \mathfrak{p}$ ,
$b_{h}\not\in \mathfrak{p})$
. We will prove that $F$ is of height $h$
. Since
$(1+p^{-1}\sum_{\nu=1}^{h-1}b_{\nu}T^{\nu})^{-1}u=p+b_{h}T^{h}+\cdots$ ,
it suffices to prove that a formal group obtained from a special element $u^{\prime/\prime}$
of the form is of height . Put $(pu^{r\gamma-1})*i=h$ . Then

$p+b_{h}T^{h}+\cdots(b_{h}\not\in \mathfrak{p})$ $h$
$ h(x)=\chi-p^{-1}b_{h}x^{p^{h}}+\cdots$
and so
$ h^{-1}(ph(x))=px-b_{h}x^{P^{h}}+\cdots+p^{-1}b_{h}(px-\cdots)^{ph}+\cdots$
$\equiv-b_{h}x^{P^{h}}+\cdots$ $mod \mathfrak{p}$

,
which prove that $h^{-1}(h(x)+h(y))$ is of height . $h$
Now suppose that there exist a unit in and a unit $c$ $0$ $t=\sum_{\nu=0}^{\infty}a_{\nu}T^{\nu}$
in $ 0_{\sigma}[[T]\sum$
such that $vc=tu$ . Comparing the $(\nu+h)$ -th degree coefficients of both mem-
bers of
$(\sum_{\psi=0}^{\infty}a_{\nu}T^{\nu})(p+\sum_{\nu=1}^{h}b_{\nu}T^{\nu})=(p+\sum_{\nu=1}^{h}c_{v}T^{\nu})c$
for $v>0$ , we get
(3.12) $a_{\nu}b_{h}^{\sigma^{\nu}}+\sum_{\mu=1}^{h-1}a_{\nu+\mu}b_{h-\mu}^{\sigma^{v+\mu}}+pa_{\nu+h}=0$
.
Since is a unit, it follows from (3.12) that
$b_{h}$
for . Hence we get $a_{\nu}\in \mathfrak{p}$ $\nu\geqq 1$
for again by
$a_{\nu}\in \mathfrak{p}^{2}$
(3.12). Repeating the same argument we see
$\nu\geqq 1$ $a_{\nu}\in \mathfrak{p}^{1}$
for every and for every . This implies $a_{\nu}=0$ for

$\nu\geqq 1$
, and $t=a_{0}=c_{-}$ $i\geqq 1$ $\nu\geqq 1$
Our proposition follows from this, from Theorem 3 and from Theorem 4.
In the above proof we proved that $vc=tu$ implied $t=c$ . Thereby we dick
not use the fact that (resp. t) is a unit. Therefore we get by Theorem $c$ $3_{j}$
PROPOSITION 3.6. Let be as in Proposition 3.5 and let be format $u,$ $v$ $F,$ $G$
groups attached to them. Then the module $Hom_{0}(F, G)$ is canonically isomorphic
to $\{c\in 0|vc=cu\}$ .
\S 4. Formal groups over a field of characteristic $p>0$ .

be a discrete valuation field satisfying $(F)$ of \S 2. For a power
Let $K$
series $f\in 0[[x]]^{m},$ $f*denotes$ the power series in $k[[x]]^{m}$ obtained by reducing
the coefficients of modulo . In \S 4 we will study the reductions of formal $f$ $\mathfrak{p}$
groups over and their homomorphisms. $0$
4.1. Our first task is to prove two lemmas.

LEMMA 4.1. Let $f\in K[[x]_{0}^{n}$ be of type $(P;u)$ and let $\psi(x^{\prime})\in 0[[x^{\prime}]]_{0}^{n}$ where
$x^{\prime}$
is a finite set of variables. Then we have

$f^{-1}(\pi\psi(x^{\prime}))\equiv 0$ $mod \mathfrak{p}$
.
PROOF. Put $h=(u^{-1}\pi)*i$ . By Lemma 2.4 it suffices to prove
$h^{- 1}(\pi x)\equiv 0$ $mod \mathfrak{p}$
.
Write $h(x)=\sum_{\nu}B_{\nu}x^{q^{\nu}}$ and $h^{-1}(\pi x)=l(x)$ . Since $1(x)\equiv\pi x$ mod deg2, the first-
degree coefficients of are in . Assume for that the i-th degree
$l$
$\mathfrak{p}$ $r\geqq 2$
coefficients of are in for all $i\leqq r-1$ . Write

$l$
$1(x)=\pi l^{(r)}(x)+\Delta^{(\gamma)}(x)$
where
$\mathfrak{p}$
and mod $\deg r$ . Then it follows from $h(l(x))=\pi x$

$l^{(\gamma)}(x)\in 0[[x]]_{0}^{n}$ $\Delta^{(r)}(x)\equiv 0$
(4.1) $l(x)+\sum_{\nu=1}^{r-1}\pi^{q^{\nu}}B_{\nu}l^{(r)}(x)^{q^{\nu}}\equiv\pi x$
mod $\deg(r+1)$ .
Since $\pi^{q^{v}}B_{\nu}\in\pi M_{n}(0)$ for $\nu\geqq 1$
by Lemma 2.2, it follows from (4.1)
1 $(x)\equiv 0$ mod $\deg(r+1)$ , $mod \mathfrak{p}$
.
Hence the r-th degree coefficients of 1 are also in . Thus we get $\mathfrak{p}$
$l\equiv 0mod \mathfrak{p}$
by induction.
LEMMA 4.2. Let be special and let $f\in K[[x]]_{0}^{n}$ be of type . Let
$u\in \mathfrak{A}_{n}$ $u$
and . Then
$\psi_{1}\in K[[x^{\prime}]]_{0}^{n}$ mod p, if and only if
$\psi_{2}\in 0[[x^{\prime}]]_{0}^{n}$ $f\circ\psi_{1}\equiv f\circ\psi_{2}$
.
$\psi_{1}\equiv\psi_{2}mod \mathfrak{p}$
PROOF. Suppose . Then we have clearly . Put

$\psi_{1}\equiv\psi_{2}mod \mathfrak{p}$ $\psi_{1}\in 0[[x]]_{0}^{n}$
$h=(u^{-1}\pi)*i$ and . Since by Lemma 2.4 and

$ h^{-1}\circ f=\varphi$ $\varphi\in 0[[x]]_{0}^{n}$ $\varphi\circ\psi_{1}\equiv\varphi\circ\psi_{2}$
mod p, we obtain by Lemma 2.1 and 2.2

$h\circ(\varphi\circ\psi_{1})\equiv h\circ(\varphi\circ\psi_{2})$ $mod \mathfrak{p}$
i. e. Conversely assume
$f\circ\psi_{1}\equiv f\circ\psi_{2}$
mod p. modp and put $f\circ\psi_{1}\equiv f\circ\psi_{2}$
. Then
$\pi\lambda=f^{-1}(f\circ\psi_{1}-f\circ\psi_{2})$ by Lemma 4.1. Since $F(x, y)=\Gamma^{1}(f(x)$ $\lambda\in \mathfrak{o}[[x]]_{0}^{n}$
$+f(y))$ has coefficients in , it follows from $\mathfrak{v}$
$f\circ\psi_{1}=f\circ\psi_{2}+f\circ(\pi\lambda)$
$i$
. . $e$ $\psi_{1}=F(\psi_{2}, \pi\lambda)$ that $\psi_{1}\equiv\psi_{2}mod \mathfrak{p}$

.
4.2. We now study a certain type of homomorphisms of to $c*$ for $F^{*}$
formal groups over . $F,$ $G$ $0$
THEOREM 5. Suppose $K$ satisfies $(F)$ . Let $F$ and be formal groups over $G$
, of dimension
$0$
and $m$ and with transformers and , respectively. Suppose
$n$ $f$ $g$
that (resp. g) is of type

$f$ (resp. of type v) for special elements and $u$ $u\in \mathfrak{A}_{n}$
$v\in \mathfrak{A}_{m}$
.
(i) Put $\varphi=\varphi_{w}=g^{-1}\circ(w*f)$ for $w\in \mathfrak{A}_{m,n}$
. Then $\varphi(x)\in 0[[x]]_{0}^{m}$
if and only
if there exists such that $vw=tu$ .
$t\in \mathfrak{A}_{m,n}$
(ii) If , then
$\varphi_{w}\in 0[[\chi]]_{0}^{m}$ $\varphi_{w}^{\star}\in Hom_{k}(F^{*}, G^{*})$ .
232 T. HONDA
Let
(iii) be of type for a special element
$h$
. If $v^{\prime}$
$v^{\prime}\in \mathfrak{A}_{l}$ $\varphi_{w^{\prime}}=h^{-1}\circ(w^{\prime}*g)$
has integral coefficients for , then $w^{\prime}\in \mathfrak{A}_{l,m}$ $\varphi_{w}^{*},$ $\circ\varphi_{w}^{\star}=\varphi_{w’ w}^{*}$
.
PROOF. In order to prove (i) we may assume $g=(v^{-1}\pi)*i$ . Suppose there
is such that $vw=tu$ . Clearly the first-degree coefficients of
$t\in \mathfrak{A}_{m,n}$
are $\varphi$
integers. Assume for that the i-th degree coefficients of are integers
$r\geqq 2$ $\varphi$
for $i\leqq r-1$ . By Lemma 2.3 we have

$\pi\varphi=(v*g)\circ\varphi\equiv v*(g\circ\varphi)$ mod $\deg(r+1)$ , $mod \mathfrak{p}$
$=v*(w*f)=(vw)*f=(tu)*f$
$=t*(u*f)\equiv 0$ modp.
This implies that the r-th degree coefficients of are also integers. This $\varphi$
shows by induction. Conversely, suppose

$\varphi(x)\in 0[[\chi]]_{0}^{m}$ . By $\varphi=\varphi_{w}\in 0[[x]]_{0}^{m}$
Lemma 2.3 we get

$(vw)*f=v*(w*f)=v*(g\circ\varphi)$
$\equiv(v*g)\circ\varphi=\pi\varphi\equiv 0$ modp.
Hence, by Proposition 2.6 we can find $t\in \mathfrak{A}_{m,n}$
such that $vw=tu$ . This proves
(i). Now we have
go $(\varphi\circ F)=(g\circ\varphi)\circ F=(w*f)\circ F$
and by Lemma 2.3
$((w*f)\circ F)(x, y)\equiv(w*(f\circ F))(x, y)$ $mod \mathfrak{p}$
$=(w*f)(x)+(w*f)(y)$
$=(g\circ\varphi)(x)+(g\circ\varphi)(y)$
$=g(G(\varphi(x), \varphi(y)))$ .
Thus we get mod p. By Lemma 4.2 it follows from this
$g\circ(\varphi\circ F)\equiv g\circ(G\circ\varphi)$
that . This implies $\varphi^{*}\in Hom_{k}(F^{*}, G^{*})$ . Let us prove (iii).

$\varphi\circ F\equiv G\circ\varphi mod \mathfrak{p}$
By Lemma 2.3 we have

$h\circ(\varphi_{w’}\circ\varphi_{w})=(h\circ\varphi_{w^{\prime}})\circ\varphi_{w}=(w^{\prime}*g)\circ\varphi_{w}$
$\equiv w^{\prime}*(g\circ\varphi_{w})$
mod p
$=w^{\prime}*(w*f)=(w^{\prime}w)*f$ .
By (i) there is such that . Since
$t^{\prime}\in \mathfrak{A}_{l,m}$ $v^{\prime}w^{\prime}=t^{\prime}v$ $v^{\prime}w^{\prime}w=t^{\prime}vw=t^{\prime}tu,$ $\varphi_{w^{\prime}w}$
$=h^{-1}\circ((w^{\prime}w)*f)$ has integral coefficients by (i). Since

$h\circ(\varphi_{w^{\prime}}\circ\varphi_{w})\equiv h\circ\varphi_{w’ w}$ $mod \mathfrak{p}$
as we have shown, it follows from Lemma 4.2 that

$\varphi_{w},$ $\circ\varphi_{w}\equiv\varphi_{ww}$ modp.
This proves (iii).
COROLLARY. Put $E=0.[[T]]$ . The submodule of $Hom_{k}(F^{*}, G^{*})$ , consisting

of homomorphisms of the form , is canonically isomorphic to the $\varphi_{w}^{*}(w\in \mathfrak{A}_{m,n})$
module of all right E-homomorphisms of $E^{n}/uE^{n}$ into $E^{m}/vE^{m}$ . In particular

the subring of $End_{k}F^{*}$ , consisting of homomorphisms of the form $(f^{-1}\circ(w*f))^{*}$
, is canonically isomorphic to the right E-endomorphism ring of $E^{n}/uE^{n}$ .
$(w\in \mathfrak{A}_{n})$
PROOF. If $tu=vw$ , then

$t(uE^{n})=vwE^{n}\subset vE^{m}$ .
Thus induces a right E-homomorphism
$t$
of $E^{n}/uE^{n}$ into $E^{m}/vE^{m}$ . Con- $\Phi_{t}$
versely, as is easily verified, every right E-homomorphism of $E^{n}/uE^{n}$ into

$E^{m}/t)E^{m}$ is of the form with such that . We will show
$\Phi_{t}$
$t\in \mathfrak{A}_{m,n}$ $tu\in v^{\backslash }?I_{m,n}$
that if and only if

$\varphi_{w}^{\star}=0$ $\Phi_{t}=0:\varphi_{w}^{*}=0\Leftrightarrow g^{-1}\circ(w*f)\equiv 0mod \mathfrak{p}\Leftrightarrow w*f\equiv 0mod \mathfrak{p}$
(by Lemma $4.2$) (by Proposition $2.6$)

$\Leftrightarrow w\in \mathfrak{A}_{m,n}u$ $\Leftrightarrow tu\in v_{\backslash }^{\backslash )}Y_{m,n}u\Leftrightarrow t\in v\mathfrak{A}_{m,n}\Leftrightarrow tE^{n}$
. This implies that

$\subset vE^{m}\{\Rightarrow\Phi_{t}=0$
and correspond bijectively. The $\varphi_{w}^{*}$ $\Phi_{t}$
second assertion follows from this and from Theorem 5, (iii).

4.3. If $K$ satisfies , every element of $Hom_{k}(F^{*}, G^{*})$ is of the form
$(F_{1})$ $\varphi_{w}^{*}$
with . To prove it we need the following lemma.

$w\in \mathfrak{A}_{m,n}$
LEMMA 4.3. Suppose $K$ satisfies . Let $F$ be an n-dimensional formal $(F_{1})$
group over and let be its transformer. Put $M=\{\psi\in K[[x]]|(\psi\circ F)(x, y)$
$0$ $f$
$\equiv\psi(x)+\psi(y)$ mod p}. Then $M$ is topologically generated by and by $\mathfrak{p}[[x]]$
as o-module. (We define the topology of $K[[x]]$ by

$\{f_{t}^{\sigma^{\nu}}(x^{v^{\nu}})|1\leqq i\leqq n, \nu\geqq 0\}$
taking { $f\in K[[x]]|f\equiv

$I_{\nu}=$ 0$ mod $\deg(\nu+1)(\nu\geqq 1)$ } as a base of neigh-
borhoods of ) $0.$
PROOF. It is clear that . By Lemma 2.3 and by Proposition $\mathfrak{p}[[x]]\subset M$
3.3 we have
$f^{\sigma^{\nu}}(F(x, y)^{p^{\nu}})=((T^{\nu}*f)\circ F)(x, y)$
$=(T^{\nu}*(f\circ F))(x, y)$ mod $p$
$=(T^{\nu}*f)(x)+(T^{\nu}*f)(y)$
$=f^{\sigma^{\nu}}(x^{p^{\lrcorner}}\backslash )+f^{\sigma^{\nu}}(y^{p^{\nu}})$
.
This implies for $1\leqq i\leqq n,$ . Let be any element of Mand
$f_{i}^{\sigma^{\nu}}(x^{p\nu})\in M$ $\nu\geqq 0$ $\psi$
let be the lowest degree such that

$r$
mod $\deg(r+1),$ . Then $\psi\in M$ $\psi\not\equiv 0$ $mod \mathfrak{p}$
implies that the r-th degree homogeneous part of satisfies $\psi^{(r)}$ $\psi$
(4.2) $\psi^{(r)}(x+y)\equiv\psi^{(r)}(x)+\psi^{(r)}(y)$ mod $\deg(r+1)$ , $mod \mathfrak{p}$

.
By Lemma 3.2 (4.2) implies that is a power of $r$ $p,$ $sayp^{h}$ (if $ r<\infty$ ) and that
there exist , satisfying $c_{1},$ $\cdots$
$c_{n}\in 0$
$\psi(x)-\sum_{i=1}^{n}c_{i}x\mathscr{S}^{\hslash}\equiv 0$
.
Hence we get
234 $\prime r$
. HONDA
(4.3) $\psi(x)-\sum_{i=1}^{n}c_{i}f_{i}^{\sigma^{h}}(x^{p^{h}})\equiv 0$
.
Applying the same argument to the left side of (4.3) in place of and re- $\psi$
peating this procedure we see in fact that $p[[x]]$ and the $f_{i}^{\sigma^{\nu}}(x^{p\nu})(1\leqq i\leqq n, \nu\geqq 0)$
generate a dense o-submodule of $M$.

THEOREM 6. Suppose $K$ satisfies . The map: , defined in $(F_{1})$ $\Phi_{t}-\rangle$ $\varphi_{w}^{*}$
Theorem 5, is a bijection of $Hom_{E}(E^{n}/uE^{n}, E^{m}/vE^{m})$ onto $Hom_{k}(F^{*}, G^{*})$ . In

particular $End_{k}F^{*}$ is canonically isomorphic to $End_{E}(E^{n}/uE^{n})$ .
PROOF. It suffices to prove the surjectivity. We may assume $f=(u^{-1}\pi)*i$
and $g=(v^{-1}\pi)*i$ . For $\varphi_{*}\in Hom_{k}(F^{*}, G^{*})$ , take such that . $\varphi\in 0[[x]]_{0}^{m}$ $\varphi^{*}=\varphi_{*}$
Since , we get by Lemma 4.2

$\varphi\circ F\equiv G\circ\varphi mod \mathfrak{p}$
(4.4) $ g\circ\varphi\circ F\equiv g\circ G\circ\varphi$ $mod \mathfrak{p}$

.
Put $\psi=g\circ\varphi$ . Then (4.4) implies
(4.5) $\psi(F(x, y))\equiv\psi(x)+\psi(y)$ $mod \mathfrak{p}$
.
By Lemma 4.3 it follows from (4.5) that there exists $w\in \mathfrak{A}_{m,n}$ satisfying
$\psi\equiv w*f$ modp,
or
$g\circ\varphi\equiv w*f$ $mod \mathfrak{p}$
.
By Lemma 4.2 this implies that $g^{-1}\circ(w*f)\in 0[[x]]_{0}^{m}$ and $\varphi\equiv g^{-1}\circ(w*f)mod \mathfrak{p}$ .
Thus we have , which was $\varphi_{w}^{*}=\varphi^{*}=\varphi_{*}$
to be proved.
4.4. Now we will show that, if $K$ satisfies , any formal group over $(F_{1})$ $k$
is obtained by reducing a formal group over . $0$
The following lemma is due to [12].

LEMMA 4.4. Let be a commutative ring and let $X=$ $R$ and $(X_{1}, \cdots , X_{n})$
$Y=$ be systems of
$(Y_{1}, \cdots , Y_{n})$ variables. Suppose that a form $\Delta(X, Y)$ of $n$
degree in $R[X, Y]$ is a commutative 2-cocycle, . .

$r$
$i$
$e$
$\Delta(X, Y)=\Delta(Y, X)$ ,
(4.6) $\Delta(Y, Z)-\Delta(X+Y, Z)+\Delta(X, Y+Z)-\Delta(X, Y)=0$ .

Then, if is not a power of a prime number,
$r$
$\Delta$
is a 2-coboundary, . . there $i$

$e$
is a form of degree such that

$\Gamma(X)$ $r$
$\Delta(X, Y)=\Gamma(X)-\Gamma(X+Y)+\Gamma(Y)$ .
If $r$
is a power of a prime, is cohomologous to a linear combination of
$\Delta$
$\Lambda_{r}(X_{i}, Y_{i})(1\leqq i\leqq n)$ with

coefficients in $R$ .
PROOF. In case $n=1$ this is Lemma 3 of [11]. (For the proof of this
case see also [7], p. 62.) In general we can reduce the case $n=m$ to the
case $n=m-1$ by making use of the result of Lyndon [15] on normal co-
homology groups. (See also [12]). For the convenience of the reader we will
perform this reduction in the following. We first note $\Delta(X, 0)=0=\Delta(0, X)$ .
(Put $Y=Z=0$ in (4.6)). Let us write $X^{\prime}=(X_{1}, \cdots , X_{m- 1}),$ $Y^{\prime}=(Y_{1}, \cdots , Y_{m- 1})$ ,
. . $X=(X^{\prime}, X_{m}),$ $Y=(Y^{\prime}, Y_{m})$ and $\Delta(X, Y)=\Delta(X^{\prime}, X_{m}, Y^{\prime}, Y_{m})$ . Define
$i$
$e$
by $\Delta_{1}$
(4.7) $\Delta_{1}(X, Y)=\Delta(X, Y)$
$-\{\Delta(0, X_{m}, X^{\prime}, 0)-\Delta(0, X_{m}+Y_{m}, X^{\prime}+Y^{\prime}, 0)+\Delta(0, Y_{m}, Y^{\prime}, 0)\}$ .
Then $\Delta_{1}$
is also a commutative 2-cocycle cohomologous to $\Delta$
. Putting $X^{\prime}=0$
,
$Y_{m}=0$ in (4.7) we get
(4.8) $\Delta_{1}(0, X_{m}, Y^{\prime}, 0)=0$
and by commutativity
(4.8) $\Delta_{1}(X^{\prime}, 0,0, Y_{m})=0$ .
Now putting $X^{\prime}=0,$ $Y_{m}=Z_{m}=0$ in (4.6) for $\Delta=\Delta_{1}$
we get
$\Delta_{1}(Y^{\prime}, 0, Z^{\prime}, 0)-\Delta_{1}(Y^{\prime}, X_{m}, Z^{\prime}, 0)+\Delta_{1}(0, X_{m}, Y^{\prime}+Z^{\prime}, 0)-\Delta_{1}(0, X_{m}, Y^{\prime}, 0)=0$ .
By (4.8) this implies
(4.9) $\Delta_{1}(Y^{\prime}, X_{m}, Z^{\prime}, 0)=\Delta_{1}(Y^{\prime}, 0, Z^{\prime}, 0)$
.
In the same way we obtain
(4.10) $\Delta_{1}(X^{\prime}, Y_{m}, 0, Z_{m})=\Delta_{1}(0, Y_{m}, 0, Z_{m})$ .
Putting $Y^{\prime}=Z_{m}=0$ in (4.6) for $\Delta_{1}=\Delta$
we get
$\Delta_{1}(0, Y_{m}, Z^{\prime}, 0)-\Delta_{1}(X^{\prime}, X_{m}+Y_{m}, Z^{\prime}, 0)$
$+\Delta_{1}(X^{\prime}, X_{m}, Z^{\prime}, Y_{m})-\Delta_{1}(X^{\prime}, X_{m}, 0, Y_{m})=0$ .

By (4.8), (4.9) and (4.10) this implies
$\Delta_{1}(X^{\prime}, X_{m}, Z^{\prime}, Y_{m})=\Delta_{1}(X^{\prime}, 0, Z^{\prime}, 0)+\Delta_{1}(0, X_{m}, 0, Y_{m})$ ,
which completes the reduction: the case the case $n=m-1$ . $ n=m\Rightarrow$
THEOREM 7. Suppose $K$ satisfies of \S 3. For any formal group $(F_{1})$ $F_{*}$
over there exists a formal group over such that $F^{*}=F_{*}$ .

$k$ $F$ $\mathfrak{o}$
PROOF. Let be the dimension of . Take $n$

such that $F_{*}$ $\varphi(x)\in 0[[x]]_{0}^{n}$
$\varphi(x)\equiv x$ mod deg2 and $u(T)=pI_{n}+\sum_{\nu=1}^{\infty}C_{\nu}T^{\nu}\in \mathfrak{A}_{n}$ and form $ f=((pu^{-1})*i)\circ\varphi$ .

Then is a formal group over . We will prove that
$F(x, y)=f^{-1}(f(x)+f(y))$ $0$
we can choose the coefficients of and , successively so that $F^{*}=F_{*}$ . $\varphi$

$C_{1}$ $C_{2},$ $\cdots$
Suppose that we have already chosen the i-th degree coefficients of for $\varphi$
$i\leqq r-1$ and the for $p^{\nu}<r$ so that $C_{\nu}$
(4.11) $F^{*}\equiv F_{*}$

mod $\deg r$ .
236 T. HONDA
Letting the other coefficients of be equal to and the for be equal $\varphi$
$0$ $C_{\nu}$ $p^{\nu}\geqq r$
to O-matrix for example, form $ g=((pu^{-1})*i)\circ\varphi$ and $G(x, y)=g^{-1}(g(x)+g(y))$ .

Then $G$ is a formal group over and we have $\mathfrak{o}$
(4.12) $G^{*}\equiv F_{*}$

mod $\deg r$ .
It follows from (4.12) and from the associative law of formal group that the
r-th degree homogeneous part of $G^{*}-F_{*}$ is a commutative 2-cocycle in $\Delta$
$k[x]^{n}$ (cf. [11], [12]). If is not a power of , we can find by Lemma 4.4
$r$ $p$
whose components are forms of degree and satisfy

$\psi\in \mathfrak{o}[x]^{n}$ $r$
(4.13) $G^{*}(x, y)-F_{*}(x, y)\equiv\psi^{*}(x)-\psi^{*}(x+y)+\psi^{*}(y)$ mod $\deg(r+1)$ .

Let be the element of
$h$ $\mathfrak{o}[[x]]_{0}^{n}$
, obtained by replacing $\varphi$
by $\varphi-\psi$
in the
definition of and put $H(x, $g$ y)=h^{-1}(h(x)+h(y))$ . Since $ h\equiv g-\psi$ mod $\deg(r+1)$ ,
we get
$H(x, y)=h^{-1}(h(x)+h(y))$
$\equiv g^{-1}(g(x)+g(y))-\{\psi(x)+\psi(y)-\psi(x+y)\}$ mod $\deg(r+1)$ .

This implies
$H^{*}(x, y)\equiv G^{*}(x, y)-\{\psi^{*}(x)+\psi^{*}(y)-\psi^{*}(x+y)\}$ mod $\deg(r+1)$
$\equiv F_{*}(x, y)$ mod $\deg(r+1)$ .

Thus we have been able to replace by $r$ $r+1$ in (4.11). If is a power of , $r$ $p$
say $r=p^{h}$ , we can find by Lemma 4.4 $\psi\in \mathfrak{o}[x]^{n}$

whose components are forms
of degree and $D\in M_{n}(0)$ such that
$r$
(4.14) $G^{*}(x, y)-F_{*}(x, y)\equiv\psi^{*}(x)-\psi^{*}(x+y)+\psi^{*}(y)-D^{*}\Lambda_{r}(x, y)$
mod $\deg(r+1)$ ,
$wh^{\alpha}.re$
we have written $\Lambda_{r}(x, y)={}^{t}(\Lambda_{r}(x_{1}, y_{1})$ , $\cdot$
.,
, $\Lambda_{r}(x_{n}, y_{n}))$
. Replacing $\varphi$
by
$\varphi-\psi$
and $u$ by $u+DT^{h}$ in the definition of $g$ , we get an element $h$
of $0[[x]]_{0}^{n}$ .
Since
$p(pI_{n}+\sum_{\nu=\iota}^{h-1}C_{\nu}T^{\nu}+DT^{h})^{-1}\equiv p(pI_{n}+\sum_{v=1}^{h-1}$
C. $T^{\nu})^{-1}-p^{-1}DT^{h}$
mod $\deg(h+1)$ ,
we have
(4.15) $h(x)\equiv g(x)-\psi(x)-p^{-1}Dx^{r}$ mod $\deg(r+1)$ .
Put $H(x, y)=h^{-1}(h(x)+h(y))$ . Then we get from (4.15)
(4.16) $H(x, y)\equiv G(x, y)-\{\psi(x)+\psi(y)-\psi(x+y)\}+D\Lambda_{r}(x, y)$ mod $\deg(r+1)$ .
$lt$
follows from (4.14) and (4.16) that
$H^{*}(x, y)\equiv G^{*}(x, y)-\{\psi^{*}(x)+\psi^{*}(y)-\psi^{*}(x+y)\}+D^{*}\Lambda_{r}(x, y)$
$\equiv F_{*}(x, y)$ mod $\deg(r+1)$ .

Thus we have been able to replace by $r+1$ in (4.11) in this case too. This $r$
proves the existence of and satisfying $F^{*}=F_{*}$ . $u$ $\varphi$
When $K$ satisfies , all the formal groups over are obtained from $(F_{1})$ $k$
special elements by Theorem 7 and homomorphisms of these groups are

described in Theorem 6 and its corollary. In case where is the ring of Witt $0$
vectors over a perfect field of characteristic $p>0$ , these results are nothing $k^{\prime}$
other than the main results of Dieudonn\’e [4]. Using these results Dieudonn\’e
[5] gave a complete classification of isogeny classes of formal groups over
when
$k^{\prime}$
is algebraically closed. For this see also [2], [8] and [16].
$k^{\prime}$
\S 5. Examples and applications.

5.1. The group of Witt vectors of length . $n$
Let be a perfect field of characteristic $p>0$ and let

$k$
$0=W(k)$ be the ring of
Witt vectors over . Put $u=pI_{n}-C_{1}T$ where $k$
. Then
$C_{1}=\left(\begin{array}{lll}0 & 1. & 0\\\vdots & & 1\\0 & \cdots & .0\end{array}\right)\in M_{n}(0)$
it is easily verified that the reduction of the formal group with the transformer
$(pu^{-1})*i$ is the group of Witt vectors of length (cf. [5], p. 120). $n$
5.2. The group for $n\geqq 2,$ $m\geqq 1$ . $G_{n,m}$
Let and be as in 5.1. Put $u=pI_{n}-C_{1}T-C_{m+1}T^{m+1}$ with

$k,$ $0$ $C_{1}$
and form $h=(pu^{-1})*i$ and $H(x, y)=h^{-1}(h(x)+h(y))$ .

$C_{m+1}=\left(\begin{array}{llll}0 & \cdots & \cdots & 0\\\vdots & & & \vdots\\ 0 & \cdots & \cdots & 0\\1 & 0 & \cdots & 0\end{array}\right)$
Then, as is seen from

is the group $G_{n,m}(=G_{n,0,m}$ by the notation of [5], $H^{*}$
[5]). Suppose that ) contains a primitive $(p^{m+n}-1)$ -th root of unity. Put $\zeta$
$w$
. Then as , we have $WC_{1}=C_{1}W^{\sigma}$ and $WC_{m+1}=$ $w^{\sigma}=w^{p}$
$W=\left(\begin{array}{ll}w^{p^{n- 1}} & 0\\0 & w^{p}w\end{array}\right)$
$C_{m+1}W^{\sigma^{m+1}}$
, so that $Wu=uW$. By Theorem 3 this implies End. $H$.
$ h^{-1}(Wh(x))\in$
On the other hand $(T*i)(x)=x^{p}\in End_{k}H^{*}$ , since $H$ is defined over . Let $Z_{p}$
$E$
be the -subalgebra of generated by and
$Z_{p}$ . The $End_{k}H^{*}$ $(h^{-1}\circ(Wh))^{*}$ $T*i$
coefficients of components of $h^{-1}\circ(Wh)$ are polynomials in $Q_{p}[w]$ . Since

, these polynomials belong to $Z_{p}[w]$ , the ring of integers
$h^{-1_{O}}(Wh)\in \mathfrak{o}[[x]]_{0}^{n}$
in $Q_{p}(w)$ . Therefore we have

(5.1) $(T*i)\circ(h^{-1}\circ(Wh))^{*}=(h^{-1}\circ(W^{\sigma}h))^{*}\circ(T*i)$ .
If $(m, n)=1$ and is algebraically closed, $End_{k}H^{*}$ is isomorphic to the (unique) $k$
maximal order in the central division algebra of rank $(m+n)^{2}$ over , and $Q_{p}$
238 T. HONDA
invariant $n/(m+n)$ ([5], p. 129-130). Since $Q_{p}(w)$ is the unramified extension

of degree $m+n$ of and $T*i$ is clearly a prime element in $End_{k}H^{*},$ $(5.1)$
$Q_{p}$
implies $E=End_{k}H^{*}$ when $(m, n)=1$ .

5.3. The Lubin-Tate group $(n=1)$ .
Suppose $K$ satisfies $(F)$ of \S 2. For $\alpha\in \mathfrak{o},$ $\alpha\neq 0,$ $u_{\alpha}=\pi-\alpha^{\sigma- 1}T$
is a special
element. Put . An easy computation shows
$f_{a}=((u_{\alpha}^{-1}\pi)*i)$
(5.2) $f_{\alpha}(x)=\sum_{\nu=0}^{\infty}\pi^{-(1+\sigma+\cdots+\sigma^{\nu- 1})}\alpha^{\sigma^{\nu}- 1}x^{q^{\nu}}$
By Theorem 2, is a formal group over . Since

$F_{\alpha}(x, y)=f_{\overline{\alpha}^{1}}(f_{\alpha}(x)+f_{\alpha}(y))$ $0$ $\alpha u_{\alpha}$
$=u_{1}\alpha,$ has integral coefficients by Theorem 3. When

$f_{1}^{-1}(\alpha f_{\alpha}(x))$
and $\pi^{\sigma}=\pi$
$\alpha=1,$
coincides with the group constructed in [10], Theorem 2. (Theorem
$F_{\alpha}$
2 of [10] can be reduced to the case $a=1$ by replacing $K$ by its unramified
extension of degree ) $a.$
5.4. Interpretation of the Artin-Hasse function.

Suppose $K$ satisfies of \S 3. Put $g(x)=-\log(1-x)=\sum_{?n=1}^{\infty}m^{-1}x^{m}$ . It is
$(F_{1})$
easily verified that is of type $ p-\tau$ . Put now $g$
$L(\alpha, x)=\sum_{\nu=0}^{\infty}p^{-\nu}\alpha^{\sigma^{\nu}}x^{p}$
“
for $\alpha\in 0$
.
Then $g^{-1}(L(\alpha, x))$ has integral coefficients by the result of 5.3. This is a
$homomorphismofF_{\alpha}$ tog $(g(x)+g(y))=x+y-xy$ . Sinceg $(x)=1-\exp(-x)$ ,
$\exp(-L(\alpha, x))$ has coefficients in . This is nothing other than the Artin-Hasse
$0$
exponential function ([1]).

5.5. The characteristic equation for the Frobenius endomorphism.
Suppose $K$ satisfies $(F)$ . Assume and let be a special element $\pi^{\sigma}=\pi$ $u$
of such that $uT=Tu$ . This implies that all coefficients of are -invariant.
$\mathfrak{A}_{n}$
$u$ $\sigma$
Since the elements of and $T$ generate a commutative subring of , $u$ $\mathfrak{o}_{\sigma}[[T]]$
we can consider the cofactor matrix of : $w$ $u$
(5.3) $uw=wu=(\det u)I_{n}$ .

Form $f=(u^{-1}\pi)*i$ and $F(x, y)=f^{-1}(f(x)+f(y))$ . By (5.3) and by Theorem 5,
(i) $(f^{-1}\circ(w*f))^{*}\in End_{k}F^{*}$ . Then by Theorem 5, (iii) and by Lemma 4.1,
(5.4) $f^{-1}\circ((\det u)*f)\equiv(f^{-1}\circ(u*f))\circ(f^{-1}\circ(w*f))$ $mod \mathfrak{p}$
$\equiv 0$
.
Write $\det u=\pi^{n}+\sum_{\nu=1}^{\infty}c_{\nu}T^{\nu},$ $c_{\nu}\in 0$ . Since $c_{\nu}^{\sigma}=c_{\nu},$ $f^{-1}\circ(c_{\nu}f)\in End_{o}F$ for $\nu\geqq 1$
by
Theorem 3. Put $[c_{\nu}]^{*}=(f^{-1}\circ(c_{\nu}f))^{*}$ and $\xi(x)=x^{q}$ . Since $f^{\sigma}=f,$ $(5.4)$ implies
that satisfies the equation
$\xi$
$[\pi^{n}]^{*}+\sum_{\nu=1}^{\infty}[c_{\nu}]^{*}\xi^{\nu}=0$
in $End_{k}F^{*}$ .
\S 6. Formal groups over $Z$ . Applications to zeta functions.
6.1. Suppose that for every prime number and for every there $p$ $\nu\geqq 1$
is given a matrix in $M_{n}(Z)$ and that commutes with if and

$C_{p^{\nu}}$ $C_{p^{\nu}}$ $C_{l^{\mu}}$ $p$ $l$
are distinct primes. Let be a complex variable and consider the (formal) $s$
Dirichlet series
$(I_{n}+C_{p}p^{-s}+\cdots+c_{p^{\nu}}p^{\nu-1-\nu S}+ )^{-1}=\sum_{p=0}^{\infty}A_{p^{\nu}}p^{-\nu s}$ .
Since $A_{p^{\nu}}$
is expressed by , with coefficients in $Z,$ commutes
$C_{p},$ $\cdots$ $C_{p^{\nu}}$ $A_{p^{\nu}}$
with $A_{\iota^{\mu}}$
if $p\neq l$ . Hence we can consider the global Dirichlet series
(6.1) $\prod_{p}(I_{n}+C_{p}p^{-s}+\cdots+c_{p^{\nu}}p^{\nu-1-\nu s}+ )^{-1}=\sum_{m=1}^{\infty}A_{m}m^{-s}$ ,
where $A_{mm},$ $=A_{m}A_{m^{\prime}}=A_{m},A_{m}$ if $(m, m^{\prime})=1$ .

THEOREM 8. Let $\{C_{p^{\mu}}\}$
and $\{A_{m}\}$ be as above and form $f(x)=\sum_{m=1}^{\infty}m^{-1}A_{m}x^{m}$
. Then
$\in Q[[x]]_{0}^{n}$
(6.2) $pf(x)+\sum_{\nu=1}^{\infty}$
C. $f(x^{p^{\nu}})\equiv 0$ $mod pz_{p}$
for every and $F(x, $p$ y)=f^{-1}(f(x)+f(y))$ is a formal group over . $Z$
PROOF. Put
(6.3) $p(pI_{n}+\sum_{\nu=1}^{\infty}C_{p^{\nu}}T^{\nu})^{-1}=\sum_{\nu=0}^{\infty}B_{p^{\nu}}T^{\nu}$
.
Replacing $T$ by $ p\tau$

in (6.3) we get $B_{p^{\nu}}=p^{-\nu}A_{p^{\nu}}$ . Now
(6.4) $pf(x)+\sum_{\nu=1}^{\infty}C_{p^{\nu}}f(x^{p^{\nu}})=p\sum_{m=1}^{\infty}m^{-1}A_{m}x^{m}+\sum_{\nu=1}^{\infty}C_{p^{\nu}}\sum_{m=1}^{\infty}m^{-1}A_{m}x^{mp^{\nu}}$
For $p+k$ let $D_{kp^{\nu}}$

be the coefficient of $x^{kp^{\nu}}$
on the right side of (6.4). If $\nu=0$ ,
then
$D_{kp^{\nu}}=pk^{-1}A_{k}\equiv 0$ $mod pz_{p}$ .
If $\nu\geqq 1$
, then
$D_{kp^{\nu}}=pk^{- 1}p^{-\nu}A_{kp^{\nu}}+\sum_{\mu=1}^{\nu}C_{p^{\mu}}(kp^{\nu-f^{J}})^{-1}A_{kp^{\nu-\mu}}$
$=k^{- 1}A_{k}(p^{-(\nu-1)}A_{p^{\nu}}+\sum_{\mu=1}^{\nu}C_{p^{\mu}}p^{-(\nu-\mu)}A_{p^{\nu-\alpha}})$
$=k^{-1}A_{k}(pB_{p^{\nu}}+\sum_{\mu=1}^{\nu}C_{p^{\mu}}B_{p^{\nu-\mu}})$
$=0$ .
Thus (6.2) is proved. Moreover, by Theorem 2 the coefficients of $F$ are
-integral for every . Hence $F(x, y)\in Z[[x, y]]$ . This completes our proof.
$p$ $p$
COROLLARY 1. Any l-dimensional formal group over $Z$ is strongly iso-

240 T. HONDA
morphic to one obtained in Theorem 8. The strong isomorphism classes cor-

respond bijectively to Dirichlet series of the form (6.1) with $n=1$ such that
$0\leqq C_{p^{\nu}}<p$ .
PROOF. Let be a l-dimensional formal group
$F$ over and let
$Z$ be its $f$
transformer. By Theorem 4 we can find , , $C_{p}$ $C_{p^{2}}$ $\cdot..\in Z$

for every satisfying $p$
$pf(x)+\sum_{\nu=1}^{\infty}C_{\nu}f(x^{p\nu})\equiv 0$ $mod pz_{p}$ .
Let $G$
be the formal group over $Z$ obtained from the Dirichlet series
$\prod_{p}(1+\sum_{\nu=1}^{\infty}C_{p^{\nu}}p^{\nu-1-\nu s})^{-I}$
By Theorem 8 and Theorem 2 $F\approx G$
over $Z_{p}$
for every
. Since the strong isomorphism of to is unique, this implies
$p$ over $F$ $G$ $F\approx G$
$Z$. The second assertion is a consequence of the Corollary of Theorem 4.

COROLLARY 2. Notations and assumptions being as in Theorem 8, assume
moreover that the commute with each other for a fixed prime . Put
$C_{p^{\nu}}$ $p$ $[C_{p^{\nu}}]$
and $\xi(x)=x^{p}$ . Then $[C_{p^{\nu}}]\in End_{Z}F$ for

$=f^{-1}\circ(C_{p^{\nu}}f)$ and satisfies the $\nu\geqq 1$ $\xi$
equation
(6.5) $[pI_{n}]^{*}+\sum_{\nu=1}^{\infty}[C_{p^{\nu}}]^{*}\xi^{\nu}=0$
in $End_{k}F^{*}$ , where $k=Z/pZ$.

PROOF. Since $C_{p^{\nu}}$
commutes with $lI_{n}+\sum_{\mu=1}^{\infty}C_{\iota^{\mu}}T^{\mu}$ for any $l,$ $[C_{p^{\nu}}]$ is 1-
integral by Theorem 3. Hence $[C_{p^{\nu}}]\in End_{Z}F$ by Proposition 1.6. The equa-
tion (6.5) is a direct consequence of (6.2) and of Lemma 4.1.
6.2. The results of 6.1 can be applied to zeta functions of the following
types:
(a) Dirichlet L-functions.
(b) Zeta functions of elliptic curves over $Q$ .
(c) Dirichlet series obtained from a rational representation of Hecke
operators in the space of cusp forms of dimension $-2$ with respect to a
congruence unit group of an indefinite quaternion algebra over $Q$ (cf. [19]).
We have already studied (a) and (b) in [10]. We note that we can remove
the assumption on in [10], Theorem 5: $S$
THEOREM 9. Let be a l-dimensional abelian variety over $Q$ and let $F$ be
$C$
a formal minimal model for over $Z$ (cf. [10]). Let $L_{p}(s)$ be the p-factor of $C$
the function of and put

$L$
for any set of prime numbers.
$C$
$L_{s}(s)=\prod_{p\in S}L_{p}(s)$
$S$
Then the formal group obtained from $L_{s}(s)$ is strongly isomorphic to $F$
over
.
$\bigcap_{p\frac{\prime}{\sim}S}(Z_{p}\cap Q)$
PROOF. Let be the formal group obtained from $G$

. Since $L_{p}(s)=1$ , $L_{s}(s)$
or of the form $(1-a_{p}p^{-S}+p^{1-ZS})^{-1},$ is a formal group over $Z$ by

$(1\pm p^{-S})^{-1}$ $G$
Theorem 8. As a strong isomorphism of to $F$ is unique if it exists, it $G$

suffices to prove over for every $p\in S$ . Let be the reduction of

$F\approx G$ $Z_{p}$ $C_{p}$
$C$
modulo . The cases where $p$has a singular point were treated in [10]. $C_{p}$
Suppose that is an abelian variety with $L_{p}(s)=(1-a_{p}p^{-s}+p^{1-2S})^{-1}$ . Since

$C_{p}$
the Frobenius of satisfies $\xi$ $C_{p}$
$\xi^{2}-a_{p}\xi+p=0$ ,
the transformer $f$ of $F$ satisfies

(6.6) $f^{-1}(pf(x)-a_{p}f(x^{p})+f(x^{P^{2}}))\equiv 0$ $mod pZ_{p}$ .
By Lemma 4.2 it follows from (6.6)
(6.7) $pf(x)-a_{p}f(x^{p})+f(x^{p^{2}})\equiv 0$ $mod pz_{p}$ .
The fact $F\approx G$
over $Z_{p}$
follows from (6.7), Theorem 8 and Theorem 2. This
completes the proof of our theorem.
Notations being as above, put $L_{c}(s)=\prod_{p}L_{p}(s)$ and let $G$
be the formal
group attached to it. Then there is $\varphi(x)\in Z[[x]]$ such that mod $\deg 2$ $\varphi(x)\equiv\chi$
and . If the conjecture of Weil [21] on

$F\circ\varphi=\varphi\circ G$
is true, the power $L_{c}(s)$
series would be the q-expansion of a suitable automorphic function with

$\varphi$
“ “
respect to where $N$ is the conductor of .

$\Gamma_{0}(N)$ $C$
It would be interesting to see that our results yield a simple proof of a

special case of the main result of Eichler [6] and Shimura [18]. Let $j(z)$ be
the elliptic modular function and put $L=Q(j(z), j(Nz))$ for $N\geqq 2$ . Then is $L$
a field of algebraic function over and $LC$ is the field of automorphic func- $Q$
tions with respect to the subgroup of SL $(2, Z)$ . We shall consider $\Gamma_{0}(N)$
the case where the genus of is equal to 1. Let be a complete non-singular $L$ $C$
model for over . Since $j(z)$ has q-expansion

$L$ $Q$
(6.8) $ j(z)=q^{-1}+744+\cdots$
with coefficients in $Z$ where $q=\exp(2\pi\sqrt{-1}z)$ , the infinite point $ z=i\infty$

corresponds to a rational point on and can be considered an abelian $\mathfrak{B}$ $C$ $C$
variety over , with the origin . Expanding the group law of by means
$Q$ $\mathfrak{P}$ $C$
of the local parameter at we get a formal group $F$ over . By the

$j(z)^{-1}$ $\mathfrak{P}$
$Q$
theory of reduction there exists a finite set of prime numbers such that $S^{\prime}$
for the reduction

$p\not\in S^{\prime}$
of $Cmod p$ is non-singular and is a local
$C_{p}$ $j(z)^{-1}$
parameter at the origin of . Then, for has -integral coefficients

$C_{p}$ $p\not\in S^{\prime}F$ $p$
and the p-th power endomorphism of the reduction of $Fmod p$ satisfies $F_{p}$
the same characteristic equation as that of . Let be the transformer of $C_{p}$ $f$
$F$ . Then $df(x)$ is the canonical invariant differential on . . the $F,$ $i$
$e$ $j(z)^{-1}-$
expansion of a differential of the first kind on . Let be the q-expansion $C$ $\varphi(q)$
of . Then $\varphi(x)\in Z[[x]]$ and

$j(z)^{-1}$ mod $\deg 2$ by (6.8). Put $\varphi(x)\equiv\chi$
242 T. HONDA
$df(\varphi(x))=\sum_{m=1}^{\infty}a_{m}x^{m- 1}dx$ $(a_{1}=1)$ .

Then, as is well-known, $\sum_{m=1}^{\infty}a_{m}q^{m}$
is the q-expansion of a cusp form of dimen-
sion $-2$ with respect to and by Hecke [9] the Dirichlet series $\Gamma_{0}(N)$ $\sum_{m=1}^{\infty}a_{m}m^{-s}$
has an Euler product of the form
$\prod_{p|N}(1-a_{p}p^{-s})^{-1}\prod_{v+N}(1-a_{p}p^{-s}+p^{1- 2s})^{-1}$ , $a_{p}\in Z$ .

Form $G(x, y)=g^{-1}(g(x)+g(y))$ with . By Theorem 8 is a formal $ g=f\circ\varphi$ $G$
group over , so that

$Z$ $F$ is also a formal group over $Z$. Let be a prime $p$
number such that and $p+N$. Then, by Corollary 2 of Theorem 8 the $p\not\in S^{\prime}$
Frobenius of is a root of the equation $G_{p}$
(6.9) $p-a_{p}X+X^{2}=0$ .
Since $F\approx G$
over
is also the characteristic equation for the Frobenius $Z,$ $(6.9)$
of , and then of
$F_{p}$
. Therefore $(1-a_{p}p^{-S}+p^{1-2s})^{-1}$ coincides with the $C_{p}$ $L$
function of . This proves the principal theorem of [18] in this case.

$C_{p}$
REMARK. By considering N\’eron’s minimal model for , we can prove $L$
that the -factor of the Hecke Dirichlet series coincides with that of the zeta
$p$
function of , assuming only that is a local parameter at the origin of

$L$ $j(z)^{-1}$
. See [10] as for the case

$C_{p}$
is singular. In view of the conjecture of $C_{p}$
Weil it is plausible that is a formal minimal model for . $F$ $C$
6.3. We now deal with (c). We use the terminology, notations and results
of Shimura [19]. Let be an indefinite quaternion algebra over and let $\Phi$
$Q$
$0$
be a maximal order in . For a natural number $N$ prime to the discriminant $\Phi$
of denotes the group consisting of units in such that $N(\gamma)=1$ and

$\Phi,$ $\Gamma_{N}$
$\gamma$
$0$
$\gamma\equiv 1mod$ No. is a discontinuous group operating on the upper half plane.
$\Gamma_{N}$
Let be the field of automorphic functions relative to

$ff_{N}$
and let be its $\Gamma_{N}$ $n$
genus. Take and JN as in [19]. is a function field over

$\mathfrak{L}_{N},$
such $\mathfrak{C}_{N}$ $\mathfrak{L}_{N}$
$Q$
that is its complete non-singular model and JN is a Jacobian of

$\mathfrak{L}{}_{N}C=ff_{N},$ $\mathfrak{C}_{N}$
, each defined over

$\mathfrak{C}_{N}$
. Let and be the spaces of differentials $Q$ $\mathfrak{D}_{0}(\mathfrak{C}_{N})$ $\mathfrak{D}_{0}(J_{N})$
of the first kind on and , respectively. For if $\mathfrak{C}_{N}$ $J_{N}$ $f,$ $g\in \mathfrak{L}_{N},$ $gdf\in \mathfrak{D}_{0}(\mathfrak{C}_{N})$
and only if . Let be a base of , defined

$gf^{\prime}\in S_{2}(\Gamma_{N})$ $\omega=\{\omega_{1}, \cdots , \omega_{n}\}$ $\mathfrak{D}_{0}(\mathfrak{C}_{N})$
over . Fixing a canonical map $Q$ (which may not be defined over ), $\mathfrak{C}_{N}\rightarrow J_{N}$ $Q$
let ) and be the

$tt$
corresponding bases of
$\eta$ and , respectively. $S_{2}(\Gamma_{N})$ $\mathfrak{D}_{0}(J_{N})$
For such that $N\alpha>0,$ $(N, \alpha)=1,$

$\alpha\in \mathfrak{o}$
operates on on the one $\Gamma_{N}\alpha\Gamma_{N}$ $S_{2}(\Gamma_{N})$
hand. Let denote its representation matrix relative to . On the

$\mathfrak{T}_{2}(\Gamma_{N}\alpha\Gamma_{N})$ $\mathfrak{w}$
other hand yields a correspondence of over $Q$ where

$\Gamma_{N}\alpha\Gamma_{N}$ $X_{\eta}$ $\mathfrak{C}_{N}$
$q=\alpha 0$
and then induces an endomorphism of . This is defined over ([19], $\xi$ $J_{N}$ $\xi$ $Q$
p. 325). Denoting by the representation matrix of with respect to , $M^{a}(\xi)$ $\xi$

$\eta$
we have
(6.10) $M^{a}(\xi)=\mathfrak{T}_{2}(\Gamma_{N}\alpha\Gamma_{N})$
([19], p. 327), where $M^{a}(\xi)\in M_{n}(Q)$ . By [19] the $\mathfrak{T}_{2}(\Gamma_{N}\alpha\Gamma_{N})$

are semi-simple
and commute with each other, and their eigenvalues are algebraic integers.
Hence there is a regular matrix $P$ in $M_{n}(Q)$ such that the $P^{-1}\mathfrak{T}_{2}(\Gamma_{N}\alpha\Gamma_{N})P$
are all in $M_{n}(Z)$ . By changing the bases if necessary, we may assume that
the are already in $M_{n}(Z)$ .
$\mathfrak{T}_{2}(\Gamma_{N}\alpha\Gamma_{N})$
Let be the set of prime numbers which fail to satisfy at least one of
$S_{1}$
P. ) ) in [19]. Then $1$

is a finite set. Let be the set of prime divisors
$\sim 10$ $S_{1}$ $S_{2}$
of . By Theorem 4 of [19] we have for $peS_{1}VS_{2}$

$d(\Phi)$
(6.11) $\tilde{X}_{\mathfrak{q}}=\Pi+\Pi_{0}^{\prime}\tilde{Y}_{p}$
,
where is an integral left o-ideal such that $N(q)=p,$

$q$
is the Frobenius of $\Pi$
$\tilde{\mathfrak{C}}_{N}$
and is defined on p. 315 of [19]. Correspondingly we have

$Y_{p}$
(6.12) $\tilde{\xi}_{p}=\pi+\pi^{\prime}\circ\tilde{\eta}_{p}$
.
be a system of local parameters $(\in Q(J_{N}))$ at the
Now let $t=\{t_{1}$ , $\cdot$
.. , $t_{n}\}$
origin of . Expanding the group law of JN into power series relative to ,

$J_{N}$ $t$
we get an n-dimensional formal group $F$ over $Q$ . We shall call this formal
group a formal model for . (A formal model is also obtained from the t- $J_{N}$
expansion of a base of , defined over $Q$). By the theory of reduction $\mathfrak{D}_{0}(J_{N})$
([20], Chapter III) there is a finite set of prime numbers such that for $S_{3}$
$p\not\in S_{3}$
:
(i) is a system of local parameters at the origin of
$t$ $\tilde{J}_{N}=the$
reduction
of JN $mod p$ .
(ii) The differentials , have good reductions $mod $\eta_{1},$
$\cdots$
$\eta_{n}$ p$ and yield a
base of $\mathfrak{D}_{0}(\tilde{J}_{N})$
.
Assume has coefficients in and an endomorphism
$p\not\in S_{1}US_{2}US_{3}$ . Then $F$ $Z_{p}$
of of , corresponding to some
$\xi$
, induces an endomorphism of $F$
$J_{N}$ $\Gamma_{N}\alpha\Gamma_{N}$
over . Let be the transformer of and let $f^{-1}\circ(C(\xi)f)(C(\xi)\in M_{n}(Z_{p}))$

$Z_{p}$ $f$ $F$
denote this endomorphism of . Since is also defined over , it induces $F$ $\xi^{\prime}$
$Q$
the endomorphism of over . Now it follows from (6.12) that $f^{-1}\circ(C(\xi^{\prime})f)$ $F$ $Z_{p}$
$\tilde{\xi}_{p}^{\prime}=\pi^{\prime}+\tilde{\eta}_{p}^{\prime}\circ\pi$
and then
\langle 6.13) $p-\tilde{\xi}_{p}^{\prime}\circ\pi+\tilde{\eta}_{p}^{\prime}\circ\pi^{2}=0$
.
This implies
$f^{-1}(pf(x)-C(\xi_{p}^{\prime})f(x^{p})+C(\eta_{p}^{\prime})f(x^{p^{2}}))\equiv 0$ $mod pz_{p}$ ,
or by Lemma 4.2
244 T. HONDA
(6.14) $pf(x)-C(\xi_{p}^{\prime})f(x^{p})+C(\eta_{p}^{\prime})f(x^{p^{2}})\equiv 0$ $mod pZ_{p}$ .

Let $E$ be the subring of generated by endomorphisms corresponding
$End_{Q}J_{N}$
to $\{\Gamma_{N}\alpha\Gamma_{N}|\alpha\in 0, N(\alpha)>0, (\alpha, N)=1\}$ . Then, as $E\otimes Q$ is a commutative

semi-simple algebra over $Q$ , the map yields an isomorphism of $E$ into $\xi-\rangle$ $\xi^{\prime}$
$End_{Q}J_{N}$ . Now is self-dual and

$J_{N}$
is the transposed matrix of ,
$M^{a}({}^{t}\xi)$ $M^{(}{}^{t}(\xi)$
since $M^{l}((\xi)\in M_{n}(Q)$ . (For example see [20], p. 25). As is conjugate $M^{a}(\xi^{\prime})$
with $M^{a}({}^{t}\xi),$ $M^{a}(\xi)$ and $M^{a}(\xi^{\prime})$

have the same trace. Therefore there is an
invertible matrix $P_{1}\in M_{n}(Q)$ such that
(6.15) $M^{a}(\xi^{\prime})=P_{1}^{-1}M^{a}(\xi)P_{1}$
for all $\xi\in E$ .
Now since the t-expansion of is a base of and $\eta$
$\mathfrak{D}^{*}(F;Q)$ $C(\xi^{\prime})(\xi\in E)$ is the
representation matrix of relative to the canonical base $df(x)$ of $\xi^{\prime}$
, we
$\mathfrak{D}^{*}(F ; Q)$
can find an invertible matrix $P_{2}\in M_{n}(Q)$ such that

(6.16) $C(\xi^{\prime})=P_{2}^{-1}M^{a}(\xi^{\prime})P_{2}$
for all $\xi\in E$ .
Putting $P_{8}=P_{1}P_{2}$ , we get from (6.15), (6.16)
(6.17) $C(\xi^{\prime})=P_{3}^{-1}M^{a}(\xi)P_{3}$ for all $\xi\in E$ .
Let $S_{4}$
be the set of prime numbers $p$
such that $P_{3}$
or $P_{3}^{-1}$
is not $p$ -integral,
and put $S=\bigcup_{i=1}^{4}S_{i}$
. $S$
is a finite set. For $p\in ES$ we get from (6.14) and (6.17)
(6.18) $pP_{\theta}f(x)-M^{a}(\xi_{p})P_{3}f(x^{p})+M^{a}(\eta_{p})P_{8}f(x^{p^{2}})\equiv 0$ $mod pZ_{p}$ .

Now replacing the parameters $t={}^{t}(t_{1}, \cdots , t_{n})$ by $u=P_{3}t$ , we obtain the
formal model $H(x, y)=P_{3}F(P_{3}^{-1}x, P_{3}^{-1}y)$ of $J_{N}$ , with the transformer $h(x)$
$=P_{3}f(P_{3}^{-1}x)$ . For $p\not\in S$

we have
$(P_{3}^{-1}x)^{p^{\nu}}\equiv P_{3^{-1}}x^{p^{\nu}}$ $mod pz_{p}$
and then by Lemma 4.2

(6.19) $f((P_{3}^{-1}x)^{p^{\nu}})\equiv f(P_{3}^{-1}x^{p^{v}})$ $mod pz_{p}$ .
By (6.18) and (6.19) we get finally
(6.20) $ph(x)-M^{l}(\xi_{p})h(x^{P})+M^{tl}(\eta_{p})h(x^{p^{2}})\equiv 0$ $mod pz_{p}$
for $peS$ .
Now we have
(6.21) $M^{a}(\xi_{p})=\mathfrak{T}_{2}$ ( $p$ ; No) and $M^{a}(\eta_{p})=R_{2}$ ( $p$ ; No)
([19], p. 327). Let $M$ be the product of all primes in $S$

and put $Z_{s}^{\prime}=\bigcap_{pfS}(Z_{p}\cap Q)_{-}$
The Dirichlet series

$\prod_{pIMN}[I_{n}-\mathfrak{T}_{2}(p;No)p^{-s}+R_{2}(p;N\mathfrak{o})p^{1- 2s}]^{-1}=\sum_{(m,MI\eta=1}\mathfrak{T}_{2}(m;No)m^{-S}$
is a main part of the one defined in [19]. Let be the formal group over $G$
$Z$ corresponding to it by Theorem 8. By Theorem 2 it follows from (6.20)

and (6.21) that over for every $peS$ . Hence
$G\approx H$
over by the $Z_{p}$ $G\approx H$ $Z_{s}^{\prime}$
uniqueness of strong isomorphism. We have proved the following theorem:

THEOREM 10. Let notations be as in [19] and let be an integral rep- $\mathfrak{T}_{2}$
resentation as above. Then there is a finite set of prime numbers such that $S$
the formal group obtained from the Dirichlet series is $\sum_{(m,\ovalbox{\tt\small REJECT} N)=1}\mathfrak{T}_{2}(m;No)m^{-s}$
strongly isomorphic over to a formal model for . $Z_{S}^{\prime}$ $J_{N}$
Thus the matrix Dirichlet series itself (not only its $\sum \mathfrak{T}_{2}(m;No)m^{-\$}$
determinant) has important significance for What kind of curve over $Q$ $J_{N}$
.
has a Jacobian whose formal completion is isomorphic to a formal group
corresponding to a matrix Dirichlet series with Euler product ?
6.4. All zeta functions, which we studied in 6.2 and 6.3, are of the form
. Do there exist number-theoretic Dirichlet series of
$I_{p}I(I_{n}+C_{p}p^{-s}+C_{\mathcal{D}^{2}}p^{1-2S})^{-1}$
the form (6.1) such that not all are equal to for ? If such ones $C_{p^{\nu}}$
$0$ $\nu\geqq 3$
exist, formal groups over $Z$ obtained from them would be non-algebroid.
Their transformers would be obtained from analytic functions, perhaps satis-
fying suitable kinds of differential equations.
Osaka University
References
[1] E. Artin and H. Hasse, Die beiden Erg\"anzungss\"atze zum Reziprozit\"atsgesetz der
-ten Potenzreste im K\"orper der -ten Einheitswurzeln, Abh. Math. Sem. Univ.
$l^{n}$ $l^{n}$
Hamburg, 6 (1928), 146-162.

[2] I. Barsotti, Moduli canonici e gruppi analitici commutativi, Ann. Scuola Norm.
Sup. Pisa, 13 (1959), 303-372.
[3] N. Bourbaki, Alg\‘ebre, Chapitre IV., 2e \’ed., Hermann, Paris, 1959.
[4] J. Dieudonne, Lie groups and Lie hyperalgebras over a field of characteristic
$p>0$ (IV), Amer. J. Math., 77 (1955), 429-452.
[5] J. Dieudonn\’e, Groupes de Lie et hyperalg\‘ebres de Lie sur un corps de caract\’eris-
tique $p>0$ (VII), Math. Ann., 134 (1957), 114-133.
[6] M. Eichler, Quatern\"are quadratische Formen und die Riemannsche Vermutung
f\"ur die Kongruenzzetafunktion, Arch. Math., 5 (1954), 355-366.
[7] A. Frohlich, Formal groups, Lecture Notes in Mathematics, Springer, Berlin-
Heidelberg.New York, 1968.
[8] P. Gabriel, Sur les cat\’egories localement noeth\’eriennes et leurs applications aux
alg\‘ebres \’etudi\’ees par Dieudonn\’e, S\’eminaire J. P. Serre, (1960).
[9] E. Hecke, \"Uber Modulfunktionen und die Dirichletscher Reihen mit Eulersher
Produktentwickelung II, Math. Ann., 114 (1937), 316-351; Mathematische Werke,
672-707.
[10] T. Honda, Formal groups and zeta-functions, Osaka J. Math., 5 (1968), 199-213.
[11] M. Lazard, Sur les groupes de Lie formels \‘a un param\‘etre, Bull. Soc. Math.
246 T. HONDA
France, 83 (1955), 251-274.

[12] M. Lazard, Lois de groupes et analyseurs, Ann. Sci. \’Ecole Norm. Sup., (3) 72
(1955), 299-400.
[13] J. Lubin, One-parameter formal Lie groups over p-adic integer rings, Ann. of
Math., 80 (1964), 464-484.
[14] J. Lubin and J. Tate, Formal complex multiplication in local fields, Ann. of
Math., 81 (1965), 380-387.
[15] R. Lyndon, The cohomology theory of group extensions, Duke Math. J., 15 (1948),
271-292.
[16] Y. Manin, The theory of commutative formal groups over fields of finite char.
acteristic, Russian Math. Surveys, 18 (1963), 1-81.
[17] A. N\’eron, Mod\‘eles minimaux des vari\’et\’es ab61iennes sur les corps locaux et
globaux, Publ. Math. I. H. E. S., 21 (1964).
[18] G. Shimura, Correspondances modulaires et les fonctions de courbes alg\’ebriques,
$\zeta$
J. Math. Soc. Japan, 10 (1958), 1-28.

[19] G. Shimura, On the zeta-functions of the algebraic curves uniformized by certain
automorphic functions, J. Math. Soc. Japan, 13 (1961), 275-331.
[20] G. Shimura and Y. Taniyama, Complex multiplication of abelian varieties and
its applications to number theory, Publ. Math. Soc. Japan, No. 6, 1961.
[21] A. Weil, \"Uber die Bestimmung Dirichletscher Reihen durch Funktionalgleich-
ungen, Math. Ann., 168 (1967), 149-156.
[22] E. Witt, Zyklische K\"orper und Algebren der Charakteristik vom Grad , J.
$p$ $p^{n}$
Reine Angew. Math., 176 (1936), 126-140.

Formal Groups - Honda

Uploaded by

Copyright:

Available Formats

Formal Groups - Honda

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Formal Groups - Honda

Uploaded by

Copyright:

Available Formats

J. Math. Soc.

On the theory of commutative formal groups

(Received Nov. 10, 1969)

The theory of (commutative) formal groups was initiated by M. Lazard

groups are described in terms of matrices over $W(k)_{\sigma}[[T]]$ . By reducing

tained in this manner. Moreover homomorphisms of commutative formal

formal groups is explicit and elementary.

form $p+\sum_{\nu=1}^{\infty}c_{\nu}T(c_{\nu}\in W(k))$ and put $ pu- 1=\sum_{\nu-0}^{\infty}b_{\nu}T^{\prime}\lrcorner$

$f^{-1}(f(x)+f(y))$ is a formal group over $W(k)$ . In some special case

is isomorphic to one obtained in this manner. Let be another element of $v$

We assert that $g^{-1}(cf(x))$ is in reality a homomorphism over $W(k)$ , if and

Our results can be applied to construct and characterize formal groups

order of an indefinite quaternion algebra over $Q$ (Shimura [19]). There is

\S 1. Invariant differential forms on a formal group.

variables , . We denote by $R[[x]]$ the ring of formal power series

on $x_{1},$, . For basic properties of $R[[x]]$ we refer to Bourbaki [3]. We

shall often regard as the column vector , .. ,

degree mod $\deg r$ , if and differ only in terms of total degree

$mod I$, if there are $\psi\in R[[x]]$ such that $f-g=\varphi+\psi,$

the power series is well-defined and ,

if there is no fear of ambiguity. Define the identity function of $R[[x]]_{0^{n}}$ $i$

by $i(x)=x$ . If is an element of $R[[x]]_{0}^{n}$ such that

with an invertible matrix $P$ in $M_{n}(R)$ , there is a unique element in $\psi(x)$

$R[[x]]0$ satisfying . We shall call this the inverse function

We adopt the classical definition of formal group.

i) $F(x, y)\equiv x+y$ mod $\deg 2$ ,

ii) $F(F(x, y),$ ) $=F(x, F(y, z))$

stands for . If $m=n$ and is invertible,

(weakly) isomorphic to : $F\sim G$ over $R$ . If there is an isomorphism

$F$ to such that

If $G$ is commutative, the set $Hom_{R}(F, G)$ of all homomorphisms of $F$ to

$\in Hom_{R}(F, G)$ . In particular $End_{R}G(=Hom_{R}(G, G))$ becomes a ring by

where (cf. [3]). Denote by

into $A$ defines a differential, which we denote by $df$. A differential of this

is a differential in . We denote it by . is an R-homo- $\mathfrak{D}^{*}(A;R)$ $\varphi^{*}(\omega)$ $\varphi^{*}$

morphism of into . $\mathfrak{D}^{*}(B;R)$ $\mathfrak{D}^{*}(A;R)$

Let be an n-dimensional formal group over . Introducing a new set

$l=$ of variables we may consider that

defined by $T_{t}(x)=F(x, t)$ . A differential ru in is said to be a right $\mathfrak{D}^{*}(A;R)$

invariant differential on $F$ if

We denote by the space consisting of all right invariant dif-$\mathfrak{D}^{*}(F;R)$

ferentials on . As in the case of a Lie group or an algebraic group, we

matrix of , we have and

form an R-basis of . Moreover the base is characterized$\mathfrak{D}^{*}(F;R)$ $\{\omega_{1}, \cdots , \omega_{n}\}$

by these two properties.

which implies that, if $\omega\in \mathfrak{D}^{*}(F;R),$

the map $\Phi:\omega->{}^{t}(\psi_{1}(0), \cdot..

jective and $\{\omega_{1}, \cdots , \omega_{n}\}$

We shall call this , .. ,

$=\sum_{i=1}^{m}\psi_{i}(\varphi(F(x, t)))d\varphi_{i}(F(x, t))$

$=\sum_{i=1}^{m}\psi_{i}(G(\varphi(x), \varphi(t)))dG(\varphi_{i}(x), \varphi_{i}(t))$

PROOF. Let $\omega_{i}=\sum_{j=1}\psi_{ij}(x)dx_{j}n(1\leqq i\leqq n)$ be the canonical base of $\mathfrak{D}^{*}(F;R)$

$=\sum_{j}d\psi_{ij}(F(x, t))\wedge dF_{j}(x, t)$

$\sum_{k}(\partial^{2}/\partial x_{k}\partial y_{\iota})F_{i}(0,0)\psi_{kj}(0)+\sum_{k}(\partial/\partial x_{k})F_{i}(0,0)(\partial/\partial x_{\iota})\psi_{kj}(0)=0$ ,

$(\partial^{2}/\partial x_{j}\partial y_{l})F_{i}(0,0)+(\partial/\partial x_{\iota})\psi_{ij}(0)=0$ ,

$=-(\partial^{2}/\partial x_{\iota}\partial y_{j})F_{i}(0,0)$

$=\sum_{j<l}((\partial/\partial x_{l})\psi_{ij}(x)-(\partial/\partial x_{j})\psi_{il}(x))dx_{\iota}\wedge dx_{j}$ ,

the coefficients of $dx_{\iota}$

for all $1\leqq i<j\leqq n,$must be equal to $\eta$

is a differential in . We denote it by . is an R-homo- $\mathfrak{D}^{}(A;R)$ $\varphi^{}(\omega)$ $\varphi^{*}$

morphism of into . $\mathfrak{D}^{}(B;R)$ $\mathfrak{D}^{}(A;R)$

$ v(f\circ\psi)\equiv(vf)\circ\psi$ mod $\deg(r+1)$ , $mod \mathfrak{p}$

$vf=t(u*f)\equiv 0$ $mod \mathfrak{p}$