证明. Note A(B(x))=x, we see we can apply B to both side and get B(A(B(x)))=B(x). On the other hand, note if B(A(x))=x then we can apply evB to both side and get B(A(B(x)))=B(x), but evB is injective, which forces us to have B(A(x))=x.
□
命题 1.1.9.A(x) 在 R[[x]] 中可逆当且仅当 a0 在 R 中可逆.
证明.(⇒): exercise.
(⇐): Consider G(x)=∑n≥0(−1)na0−n−1xn. Then we see (a0+x)G(x)=1 and now apply evA+ to both side of the equation and we get (a0+A+(x))G(A+(x))=1, where a0+A+(x)=A(x) and hence we are done, i.e. G(A+(x)) is the inverse of A(x).
Linear approximation: for any A(x)∈R[[x]], we have A(u)=A(v)+(u−v)A′(v)(mod(u−v)2).
2.
Newton’s method: we will construct f0(t)=0, then fn+1(t)=fn(t)−F′(t,fn(t))F(t,fn(t)) and we take the limit to define f(t)=limn→∞fn(t). Then we use linear approximation to prove the limit exists.
Now we give the detailed proof of Hensel’s lemma.
证明. Existence: let f0(t)=0, fn+1(t)=fn(t)−F′(t,fn(t))F(t,fn(t)). By induction, fn(t)∈R[[t]]+ for all n, hence the right hand side is defined. Now apply the linear approximation with A(x)=F(t,x), u=fn(t),v=fn−1(t) we getF(t,fn(t))=F(t,tn−1(t))+(fn(t)−fn−1(t))F′(t,fn−1(t))(mod(fn(t)−fn−1(t))2)Thus we see F(t,fn+1(t)) is divisible by (fn(t)−fn−1(t))2.
On the other hand, since F′(t,fn(t)) is invertible, from the definition of fn we seefn+1(t)−fn(t)is divisible by F(t,fn(t)). Thus we seeValt(fn(t)−fn−1(t))≤2ValtF(t,fn(t))≤2Valt(fn+1(t)−fn(t))and hence limn→∞Valt(fn+1(t)−fn(t))=∞ and thus f(t)=limn→∞fn(t)=∑n≥0(fn+1(t)−fn(t)) exists. Now take limit of the defining equation of fn, we getn→∞limfn+1(t)=n→∞limfn(t)−F′(t,fn(t))F(t,fn(t))f(t)=f(t)−F′(t,f(t))F(t,f(t))where F′(t,f(t))=0 and hence F(t,f(t))=0 as desired.
Uniqueness: Suppose F(t,f(t))=F(t,g(t))=0. Apply linear approximation with A(x)=F(t,x) and u=g(t),v=f(t) we getF(t,g(t))=F(t,f(t))+(g(t)−f(t))F′(t,f(t))(mod(g(t)−f(t))2)and hence g(t)−f(t) is divisible by (g(t)−f(t))2. Since g(t)−f(t) is not a unit in R[[t]] we must have g(t)−f(t)=0 and the proof follows.
□
如下是两个 Hensel 引理的应用.
命题 1.1.15.A(x)∈R[[x]]+ 有复合逆当且仅当 [x′]A(x) 在 R 中可逆.
证明.(⇒): Exercise.
(⇐): Let F(x,λ)=x−A(λ) and F′(x,λ)=∂λ∂F(x,λ). Then F(0,0)=0−A(0)=0 and F′(0,0)=A′(0) by definition is invertible. Now by Hensel’s lemma we can find B(λ)∈R[[x]]+ such that F(x,B(x))=0=x−A(B(x)) as desired.
(⇐): Write A(x)=a2x2m+ higher terms. Let F(x,y)=(1+y)2−a2x2mA(x). Then we see F′(x,y):=∂y∂F(x,y)=2(1+y). Then F(0,0)=0 and F′(0,0)=2 and so we can use Hensel’s lemma and get f(x)∈F[[x]]+ such that F(x,f(x))=0 and hence a2x2m(1+f(x))2=A(x), i.e. axm(1+f(x)) is a square root of A(x) as desired.
存在独特的 A(x)∈R[[x]]+ 使得 A(x)=xϕ(A(x)). This equation is called the LIFT equation.
2.
对所有 n≥1, [xn]A(x)=n1[λn−1]ϕ(λ)n
3.
对于任意 f(λ)∈R((λ)) 和 n>0, 我们有[xn]f(A(x))=n1[λn−1]f′(λ)ϕ(λ)n对于 n=0 我们有[x0]f(A(x))=[λ0]f(λ)+[λ−1]f′(λ)log(ϕ(0)ϕ(λ))where the second term is zero when Valλf(λ)≥0.
Suppose we have polynomial p(z0,z1,z2,...)∈R[z0,z1,z2,...]. For example, p(z0,z1,z2,...)=z1z42+z999∈Q[z0,z1,z2,...]. Given such a polynomial, we get a function ϕp:R[[x]]→R given byU(x)=n≥0∑unxn↦p(u0,u1,u2,...)We call such ϕp an algebraic function.
For example, given p(z0,z1,z2,...)=z1z42+z999 then p(∑n≥0nxn)=1⋅42+999.
Similarly given p(z1,z2,...)∈R[z1,z2,...], we get a function ϕp:R[[x]]+→R given by ∑n≥1unxn↦p(u1,u2,...), which is called an algebraic function as well.
For example, if p(z0,z1,...)=zn then ϕp=[xn].
定义 1.3.1. An algebraic transformation of FPS is a function f:A→B with A,B∈{R[[x]],R[[x]]+} such that [xn]∘f is an algebraic function for all n≥0.
例子 1.3.2.
1.
Squaring map: sq:R[[x]]→R[[x]] is given by U(x)↦U(x)2. We see [xn]∘sq:U(x)↦[xn]U(x)2=∑k=0nukun−k, which is a polynomial in u0,u1,... and hence it is an algebraic transformation.
2.
Differentiation map: dxd:R[[x]]→R[[x]] is given by U(x)↦U′(x). This is algebraic transformation since [xn]U′(x)=(n+1)un+1, which is a polynomial in u0,u1,... as desired.
3.
Evaluation map: Given B(x)∈R[[x]]+, the evaluation map evB:R[[x]]→R[[x]] is given by U↦U(B(x)). We have [xn]∘evB(u(x)) is a linear function of u0,u1,u2,..., as one can show.
4.
Application map: given A(x)∈R[[x]], we get apA:R[[x]]+→R[[x]] is given by U(x)↦A(U(X)). This time [xn]∘apA(U(x)) is a degree n polynomial of u1,u2,u3,..., as one can show.
定义 1.3.3. Let Tf=Tfull=Tfull(R[[x]]) be the set of algebraic transformations R[[x]]→R[[x]].
We use T be the set of algebraic transformations R[[x]]+→R[[x]].
We use T+ be the set of algebraic transformations R[[x]]+→R[[x]]+.
In particular we haveT+↪T↞Tf
We also note g∈T, h∈T+ then g∘h∈T. To see why composition is again algebraic trans, just note composition of polynomials is still polynomial.
On the other hand, f∈Tf and g∈T then f∘g∈T.
Note we get a ring structure on T because the codomain of T is a ring, hence we can define f,g∈T then we define f+g:U(x)↦f(U)+g(U(x)). Define the similar point-wise subtraction and multiplications, we get a ring.
The 0 element will be the map U(x)↦0 and the 1 element will be U(x)↦1.
We note Tf is a ring, but T+ is not a ring as it is missing the multiplicative identity.
备注 1.3.4. We can also take limits of algebraic transformations. Given f0,f1,f2,...∈T, or in Tf,T+. Then we sayn→∞limfn=f∈Tif and only if limn→∞fn[U(x)]=f(U(x)) for all FPS U(x)∈R[[x]]+. Hence we can also talk about infinite sums and products of algebraic transformations.
备注 1.3.5. In particular we can define composition with FPS. If A(x)=∑n≥0anxn∈R[[x]], and f∈T+(R[[x]]). Then we can define A(f)=∑n≥0anfn∈T. As an exercise, show that A(f)=apA∘f.
例子 1.3.6. Let f=evx2 and g=dxd. Then what does exp(f2g)∘f do?
Apply this to U(x), we see exp(f2g)∘f[U(x)]=exp(f2g)[U(x2)]=exp(f(U(x2))2⋅g(U(x2)))=exp(U(x4)2⋅2xU′(x2))
备注 1.3.7. Caution: note algebraic transformations are not formal power series. For example, if we have A(x),B(x)∈R[[x]],U(x)∈R[[x]]+. If A(U(x))=B(U(x)) we know in fact A=B.
However, on the other hand, if we have f,g∈T, then f[U(x)]=g[U(x)] does not imply f=g.