This paper was converted on www.awesomepapers.org from LaTeX by an anonymous user.
Want to know more? Visit the Converter page.

Spectral theory of padicp-adic Hermite operator

Zhao tianhong
(2022/10/20)
Abstract

We give the definition of padicp-adic Hermite operator and set up the padicp-adic spectral measure. We compare the Archimedean case with non-Archimedean case. The structure of Hermite conjugate in CC^{*}-Algebra corresponds to three canonical structures of padicp-adic ultrametric Banach algebra: 1. mod pp reduction  2. Frobenius map  3. Teichmüller lift. There is a nature connection between Galois theory and Hermite operator spectral decomposition. The Galois group Gal(𝔽¯p|𝔽p)\mathrm{Gal}(\bar{\mathbb{F}}_{p}|\mathbb{F}_{p}) generate the padicp-adic spectral measure. We point out some relationships with padicp-adic quantum mechanics: 1. creation operator and annihilation operator  2. padicp-adic uncertainty principle.

1 Introduction

The study of Hermite operator is in the center of quantum mechanics. The usual Hermite operator corresponds to the observable theory in physics. In mathematics we have the spectral decomposition theorem of usual Hermite operator. Is it possible to build up a padicp-adic quantum theory by padicp-adic Hermite operator?

The positive property of a square number allows us to define positive definite quadrics in \mathbb{R}-linear space. As a result, we can define Hilbert spaces over \mathbb{R}. If we consider the Hilbert space \mathcal{H} over \mathbb{C}, we need complex conjugate in Galois group Gal(|)\mathrm{Gal}(\mathbb{C}|\mathbb{R}) to ensure the positive definiteness of inner product. The positive definiteness of inner product allows a probability annotation of quantum mechanics. The Riesz representation theorem induces the Hermite conjugate \dagger structure of bounded operator over \mathcal{H}.

When it comes to padicp-adic fields(e.g.p,p\mathbb{Q}_{p},\mathbb{C}_{p}), the same argument does not hold. The reason is that quadratic form over p\mathbb{Q}_{p} always have a non-trivial zero when the rank of quadratic form is large enough(rank5)rank\geq 5)([14],p36,Theorem 6). So if we have a quadratic form over a p\mathbb{Q}_{p}-Linear space, in general it cannot induce a norm. There are some papers to consider ultrametric Hilbert space but we will not use that definition[9].

The reasonable analogy of Hermite conjugate is the padicp-adic continuity of Frobenius map σ\sigma[11, 2, 10]. Fermat’s little theorem tells us:

x,σ(x)=xpx(modp)\forall x\in\mathbb{Z},\sigma(x)=x^{p}\equiv x\pmod{p}

The same statement holds for p\mathbb{Z}_{p}, since \mathbb{Z} is dense in p\mathbb{Z}_{p}, we have:

xp,σ(x)=xpx(modp)\forall x\in\mathbb{Z}_{p},\sigma(x)=x^{p}\equiv x\pmod{p}

In general, mod pp reduction is a ring homomorphism:ϕ:p𝔽p\phi:\mathbb{Z}_{p}\longrightarrow\mathbb{F}_{p}. There exists a canonical section named Teichmüller lift:i:𝔽ppi:\mathbb{F}_{p}\longrightarrow\mathbb{Z}_{p} such that:ϕi=id𝔽p\phi\circ i=id_{\mathbb{F}_{p}}. The section ii is multiplicative. We say xpx\in\mathbb{Z}_{p} is a Teichmüller element if xx is a lift element of 𝔽p\mathbb{F}_{p}. Let T(p)T(\mathbb{Z}_{p}) be the set of Teichmüller element. Let σ\sigma be the Frobenius map:σ:pp,xxp\sigma:\mathbb{Z}_{p}\to\mathbb{Z}_{p},x\mapsto x^{p}. We can show that:

Proposition 1.1.
T(p)={xp,σ(x)=x},Card(T(p))=pT(\mathbb{Z}_{p})=\left\{x\in\mathbb{Z}_{p},\sigma(x)=x\right\},Card(T(\mathbb{Z}_{p}))=p

We refer to [12, 3] for theory of Teichmüller lift and Witt ring.

Recall that the canonical filtration pp2\mathbb{Z}\supset p\mathbb{Z}\supset p^{2}\mathbb{Z}... induces a padicp-adic topology of \mathbb{Z} which is Hausdorff. p\mathbb{Z}_{p} is the completion of \mathbb{Z} under this topology. Equivalently, we can define a padicp-adic norm:

|x|p=pvp(x),vp(x)=sup{k,pkx}\left|x\right|_{p}=p^{-v_{p}(x)},v_{p}(x)=\sup\left\{k,p^{k}\mid x\right\}
Theorem 1.2.

Let AA be a commutative integral ring with unit,CharA\mathrm{Char}A=0,equipped with padicp-adic topology(induce by the filtration ApAp2AA\supset pA\supset p^{2}A... ),suppose AA is complete.Let A~=A/pA\tilde{A}=A/pA.The following conditions are equivalent:

  1. 1.
    xA,σ(x)=xpx(modp)\forall x\in A,\sigma(x)=x^{p}\equiv x\pmod{p}
  2. 2.
    x¯A~,σ(x¯)=x¯p=x¯\forall\overline{x}\in\tilde{A},\sigma(\overline{x})=\overline{x}^{p}=\overline{x}
  3. 3.
    xA,x=i=0xipi,σ(xi)=xi\forall x\in A,x=\sum_{i=0}^{\infty}x_{i}p^{i},\sigma(x_{i})=x_{i}
Proof.

We will only show that 131\Rightarrow 3. Suppose x,yAx,y\in A satisfy:|xy|p1p\left|x-y\right|_{p}\leq\frac{1}{p}. We have:

|σ(x)σ(y)||xy|p\left|\sigma(x)-\sigma(y)\right|\leq\frac{\left|x-y\right|}{p}

We can check {σk(x)}k=0,1,2\left\{\sigma^{k}(x)\right\}_{k=0,1,2...} is a Cauchy sequence, hence it has a limit. We call it x0x_{0}. It’s easy to show:

xx0(modp)x\equiv x_{0}\pmod{p}

So we have:

x=x0+py,yAx=x_{0}+py,y\in A

Repeating the same argument to yy, we have done. ∎

Remark 1.1.

\mathbb{Z} satisfy the property:x,σ(x)=xpx(modp)\forall x\in\mathbb{Z},\sigma(x)=x^{p}\equiv x\pmod{p} but \mathbb{Z} is not complete in padicp-adic topology. p,pn\mathbb{Z}_{p},\mathbb{Z}_{p}^{n}(Equipped with supremum norm) satisfy this property above. Let C(p,p)C(\mathbb{Z}_{p},\mathbb{Z}_{p}) be the set of continous function from p\mathbb{Z}_{p} to p\mathbb{Z}_{p}, equipped with supremum norm. C(p,p)C(\mathbb{Z}_{p},\mathbb{Z}_{p}) has this property.

Definition 1.1.

Let AA be a ring(with a unit,not necessary commutative), CharA\mathrm{Char}A=0, equipped with padicp-adic topology which is given by filtration:ApAp2AA\supset pA\supset p^{2}A.... Suppose AA is complete. Let σ:AA,xxp\sigma:A\to A,x\mapsto x^{p} be the Frobenius map. For any positive integer NN, we define:

TN(A)={xAσN(x)=x}T_{N}(A)=\left\{x\in A\mid\sigma^{N}(x)=x\right\}
T(A)=N=1TN!(A)T(A)=\bigcup_{N=1}^{\infty}T_{N!}(A)

We call set TN(A)T_{N}(A) Teichmüller element with period NN.

Remark 1.2.

We have:

TN(A)TN(A)NNT_{N}(A)\subseteq T_{N^{*}}(A)\Longleftrightarrow N\mid N^{*}

Let KK be the completion of maximal unramified extension of p\mathbb{Q}_{p} with a multiplicative norm:||:K+|\ \ |:K\to\mathbb{R}_{+}. Such norm exists and is unique. Let XX be an ultrametric Banach space over KK such that the range of norm ||:X+|\ \ |:X\to\mathbb{R}_{+} on XX satisfy: |X|=|K||X|=|K|

The following result ([1],Corollary 9.2.7) is useful for our discussion.

Theorem 1.3.

(Berkovich).Let A be a uniform commutative Banach algebra over KK with identity, such that all the characters of A take values in KK. Then the space M(A) is totally disconnected, and the Gelfand transform gives an isomorphism AC(M(A),K)A\overset{\sim}{\to}C(M(A),K)

We refer to SGA for the inspired result of Galois descent. The following is a generalization of Galois theory. Let kk be a field. Let 𝔊k\mathfrak{G}_{k} be the profinite group Gal(ks|k)\textup{Gal}(k_{s}|k).

Theorem 1.4.

(Grothendieck).The functors

{finite 𝔊ksets}opp{\left\{\textup{finite }\mathfrak{G}_{k}-sets\right\}^{opp}}{étale k-algebras}{\left\{\textup{\'{e}tale $k$-algebras}\right\}}
S{\ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ \ S}Hom𝔊ksets(S,ks)=Homsets(S,ks)𝔊k{\textup{Hom}_{\mathfrak{G}_{k}-sets}(S,k_{s})=\textup{Hom}_{sets}(S,k_{s})^{\mathfrak{G}_{k}}}
Homkalgebras(L,ks){\textup{Hom}_{k-\textup{algebras}}(L,k_{s})}L{L}

are inverse equivalences of categories.

Definition 1.2.

Let xx be a linear operator on XX,We call xx is a Teichmüller element of period \infty, if:

slimNσN!(x)=x,|x|=1s-\lim_{N\to\infty}\sigma^{N!}(x)=x,|x|=1

Let T(X)T_{\infty}(X) be the set of Teichmüller element of period \infty.

Remark 1.3.

Recall the Galois theory of finite fields:

Gal(𝔽¯p|𝔽p)limN1/N!\mathrm{Gal}(\bar{\mathbb{F}}_{p}|\mathbb{F}_{p})\simeq\lim_{\underset{N\geq 1}{\longleftarrow}}\mathbb{Z}/N!\mathbb{Z}

Teichmüller element will be used to build up orthogonal spectral measure.

Let KK be an unramified extension of p\mathbb{Q}_{p} such that ωpN=ω\omega^{p^{N}}=\omega has exactly pNp^{N} solutions. Let AA be a commutative KK-Banach algebra with unit, σ\sigma be the Frobenius map:σ:AA,xxp\sigma:A\to A,x\mapsto x^{p}. We assume the norm on AA satisfy:|A|=|K|\left|A\right|=\left|K\right|. Let A0A_{0} be the unit ball: A0={xA,|x|1}A_{0}=\left\{x\in A,\left|x\right|\leq 1\right\}. A0A_{0} is a natural 𝒪K\mathcal{O}_{K}-algebra. Moreover, A0A_{0} is complete with padicp-adic topology. Let aa be an arbitrary element in KK. By using the theory of Witt ring, we have a taylor expansion:

a=i=kaipi,aiT(𝒪K)a=\sum_{i=k}^{\infty}a_{i}p^{i},a_{i}\in T(\mathcal{O}_{K})
Definition 1.3.

Let xAx\in A, We say xx is a padicp-adic hermite operator with period NN, if xx has a taylor expansion:

x=i=kxipi,xiTN(A0)x=\sum_{i=k}^{\infty}x_{i}p^{i},x_{i}\in T_{N}(A_{0})

Such xx has a padicp-adic orthogonal spectral measure. This is our main theorem:

Theorem 1.5.

(padicp-adic Hermite operator spectral decomposition theorem)
Let xx be a padicp-adic Hermite operator with period 1, then there exists a finitely additive, orthogonal projection valued spectral measure over p\mathbb{Q}_{p}. There exists a spectral integral:

I=p𝑑Eλx=pλ𝑑EλI=\int_{\mathbb{Q}_{p}}dE_{\lambda}\ \ \ x=\int_{\mathbb{Q}_{p}}\lambda dE_{\lambda}

Let xx be a padicp-adic Hermite operator with period N, then there exists a finitely additive, orthogonal projection valued spectral measure over KK. There exists a spectral integral:

I=K𝑑Eλx=Kλ𝑑EλI=\int_{K}dE_{\lambda}\ \ \ x=\int_{K}\lambda dE_{\lambda}
Theorem 1.6.

(Teichmüller element spectral decomposition theorem)
Let xA0x\in A_{0} be a element. xx is a Teichmüller element of period NN if and only if there exists a spectral decomposition:

λTN(K)πλ=1,λTN(K)λπλ=x,πλ2=πλ,πλπλ=0(λλ)\sum_{\lambda\in T_{N}(K)}\pi_{\lambda}=1,\ \ \sum_{\lambda\in T_{N}(K)}\lambda\pi_{\lambda}=x,\ \ \pi_{\lambda}^{2}=\pi_{\lambda},\ \ \pi_{\lambda}\pi_{\lambda^{*}}=0(\forall\lambda\neq\lambda^{*})

KK is an unramified extension of p\mathbb{Q}_{p} such that ωpN=ω\omega^{p^{N}}=\omega has exactly pNp^{N} solutions. πλ\pi_{\lambda} is non-Archimedean orthogonal projection.

Let XX be an ultrametric Banach space over KK, K=punram^K=\widehat{\mathbb{Q}_{p}^{unram}},punram^\widehat{\mathbb{Q}_{p}^{unram}} is the completion of maximal unramified extension of p\mathbb{Q}_{p}. Let xx be a linear operator on XX. xx is a Teichmüller element of period \infty if and only if there exists a spectral decomposition:

λT(K)πλ=1,λT(K)λπλ=x,πλ2=πλ,πλπλ=0(λλ)\sum_{\lambda\in T(K)}\pi_{\lambda}=1,\ \ \sum_{\lambda\in T(K)}\lambda\pi_{\lambda}=x,\ \ \pi_{\lambda}^{2}=\pi_{\lambda},\ \ \pi_{\lambda}\pi_{\lambda^{*}}=0(\forall\lambda\neq\lambda^{*})

T(K)=N=1TN!(K)T(K)=\bigcup_{N=1}^{\infty}T_{N!}(K) is the union of Teichmüller element in KK.K=punram^K=\widehat{\mathbb{Q}_{p}^{unram}},punram^\widehat{\mathbb{Q}_{p}^{unram}} is the completion of maximal unramified extension of p\mathbb{Q}_{p}.
The sum converges in strong operator topology.

Let 𝒪K={xK,|x|1}\mathcal{O}_{K}=\left\{x\in K,|x|\leq 1\right\}. Let Mn(𝒪K)M_{n}(\mathcal{O}_{K}) be the set of n×nn\times n matrix of 𝒪K\mathcal{O}_{K} coefficients. We can define a norm: A=(aij)1i,jn,|A|=sup1i,jn|aij|A=(a_{ij})_{1\leq i,j\leq n},|A|=\underset{1\leq i,j\leq n}{\sup}|a_{ij}|.

Theorem 1.7.

Let AMn(𝒪K)A\in M_{n}(\mathcal{O}_{K}) be a matrix. Suppose |A|=1|A|=1. There exists a canonical Jordan decomposition:

A=As+AnA=A_{s}+A_{n}

AsT(Mn(𝒪K))A_{s}\in T(M_{n}(\mathcal{O}_{K})) is a Teichmüller element with finite period, AnA_{n} is a topological nilpotent element such that:

|An|1,limkσk!(An)=0|A_{n}|\leq 1,\lim_{k\to\infty}\sigma^{k!}(A_{n})=0

Let A,BA,B be padicp-adic Hermite operators on XX. Let ψX,|ψ|=1\psi\in X,|\psi|=1 be the normalized wave function. Let diam(A),diam(B)diam(A),diam(B) be the spectrum diameter of A,BA,B:

diam(A)=supλ,λSpecA|λλ|diam(A)=\underset{\lambda,\lambda^{*}\in SpecA}{\sup}|\lambda-\lambda^{*}|
Theorem 1.8.

(padicp-adic uncertainty principle) We have:

|[A,B]ψ|diam(A)diam(B)\left|[A,B]\psi\right|\leq diam(A)*diam(B)

A.N.Kochubei developed a specral theory for non-Archimedean normal operator[7, 8]. We consider three canonical structures: mod pp reduction, Frobenius map,  Teichmüller lift. There are some papers about padicp-adic quantum mechanics[17, 4]. It is conjectured that the space-time is non-Archimedean, or even padicp-adic in Planck length[17]. The padicp-adic AdS-CFT is an important work[6]. We refer to [13] for the basic concepts of padicp-adic analysis, [1, 5] for the theory of non-Archimedean Banach algebra.

In section 2 we will build up the spectral measure of Teichmüller element.

In section 3 we will set up the spectral decomposition theorem of padicp-adic Hermite operator.

In section 4 we will give some examples of padicp-adic Hermite operator. We will discuss the non-Archimedean creation and annihilation operators and give a simplified proof of some examples given by A.N.Kochubei [7].

In section 5 we will give a proof for the padicp-adic uncertainty principle for padicp-adic Hermite operator.

In section 6 we will give some further discussions.

Here is a table comparing the Archimedean case with the non-Archimedean case.

Archimedean Non-Archimedean
base field ,\mathbb{R},\mathbb{C} p,K\mathbb{Q}_{p},K,KK is an unramified extension of p\mathbb{Q}_{p}
Canonical Norm(base field) usual norm norm induced by extension K|pK|\mathbb{Q}_{p}
Banach space Hilbert space ultrametric Banach space
Canonical Norm |x+y|2+|xy|2=2(|x|2+|y|2)|x+y|^{2}+|x-y|^{2}=2(|x|^{2}+|y|^{2}) |x+y|max(|x|,|y|)|x+y|\leq max(|x|,|y|)
Orthogonal projection π2=π,|x|2=|π(x)|2+|π(x)|2\pi^{2}=\pi,|x|^{2}=|\pi(x)|^{2}+|\pi^{\perp}(x)|^{2} π2=π,|x|=max(|π(x)|,|π(x)|)\pi^{2}=\pi,|x|=max(|\pi(x)|,|\pi^{\perp}(x)|)
Banach algebra CC^{*}-Algebra ultrametric Banach algebra
Galois action Hermite conjugate \dagger Frobenius map σ\sigma

2 Orthogonal projection and Teichmüller element

Let FF be a local field(,\mathbb{R},\mathbb{C} or p,K\mathbb{Q}_{p},K, KK is a finite field extension of p\mathbb{Q}_{p}) with a non-trival multiplicative norm: ||+|\ \ |\to\mathbb{R}_{+}, Suppose CharK\mathrm{Char}K=0. We assume the norm is discrete when KK is a finite field extension of p\mathbb{Q}_{p}). Let (X,||)(X,|\ \ |) be a FF-Banach space such that the range of norm satisfy: |X|=|F|+|X|=|F|\subseteq\mathbb{R}_{+}. In the Archimedean case we assume XX is a Hilbert space. In the Non-Archimedean case we assume XX is a ultrametric Banach space, i.e.the norm |||\ \ | over XX satisfies:

|x+y|max(|x|,|y|)x,yX|x+y|\leq max(|x|,|y|)\ \ \forall x,y\in X

For simplicity, we assume F=pF=\mathbb{Q}_{p}. It’s not hard to generalize this case to finite field extension of p\mathbb{Q}_{p} (because the theory of reduction and Teichmüller lift are similar). Let X0X_{0} be the set of unit ball of XX:X0={xX,|x|1}X_{0}=\left\{x\in X,|x|\leq 1\right\} Let AA be the set of continous linear map over XX: A={f:XXcontious,Klinear}A=\left\{f:X\to X\ contious,K-linear\right\} which is a canonical KK-Banach algebra. Let A0A_{0} be the unit ball of AA:A0={xA,|x|1}A_{0}=\left\{x\in A,|x|\leq 1\right\}. When KK is non-Archimedean, A0A_{0} is a nature p\mathbb{Z}_{p}-Algebra which can be reduced to A~=A0/pA0\tilde{A}=A_{0}/pA_{0}. A~\tilde{A} has a nature 𝔽p\mathbb{F}_{p}-Algebra structure.

Definition 2.1.

A projection π\pi of AA is a element satisfty:π2=π\pi^{2}=\pi ,we can also define coprojection:π=Iπ\pi^{\perp}=I-\pi . They have exact relationship:kerπ=imπ,kerπ=imπker\pi=im\pi^{\perp},ker\pi^{\perp}=im\pi.

Theorem 2.1.

For a projection π\pi,the following are equivalent:

  • 1.

    |π|=1\left|\pi\right|=1

  • 2.

    π:X0X0,X0={xX,|x|1}\pi:X_{0}\to X_{0},X_{0}=\left\{x\in X,\left|x\right|\leq 1\right\}

  • 3.

    X=π(X)^π(X)X=\pi(X)\hat{\oplus}\pi^{\perp}(X)

  • 4.

    (Archimedean)π=π\pi^{\dagger}=\pi

  • 4*.

    (non-Archimedean)π¯A~,π¯2=π¯,ππ¯(modp)\exists\bar{\pi}\in\tilde{A},\bar{\pi}^{2}=\bar{\pi},\pi\equiv\bar{\pi}\pmod{p}

Let’s explain these conditions.Condition 1 says the operator norm is 1. Condition 2 says π\pi maps the unit ball to the unit ball.Condition 3 says XX has a direct sum decomposition, which is corresponding to the decomposition of norm. Actually we have:

  • 3.

    (Archimedean)|x|2=|π(x)|2+|π(x)|2xX|x|^{2}=|\pi(x)|^{2}+|\pi^{\perp}(x)|^{2}\forall x\in X

  • 3*.

    (non-Archimedean)|x|=max(|π(x)|,|π(x)|)xX|x|=max(|\pi(x)|,|\pi^{\perp}(x)|)\forall x\in X

The Archimedean orthogonality property corresponds to the additivity of probability. The non-Archimedean orthogonality property means the probability of event is equal to the maximum of every event. Does it make sense to measure some padicp-adic or non-Archimedean probabilities or observables?

Condition 4 says the structure of Hermite conjugate is corresponding to the structure of mod pp reduction and lifting. We will focus on Condition 1 and Condition 4.

Definition 2.2.

We say a projection π\pi of AA is a Orthogonal projection if one of these condition is satisfied.

Now we consider the Frobenius map σ\sigma action over A0A_{0}, A0={xA,|x|1}A_{0}=\left\{x\in A,|x|\leq 1\right\},σ:A0A0,xxp\sigma:A_{0}\longrightarrow A_{0},x\mapsto x^{p}. Assume AA is a p\mathbb{Q}_{p}-commutative Banach algebra with a unit, such that |A|=|p||A|=|\mathbb{Q}_{p}| and |p|=1p|p|=\frac{1}{p}. This map reduce to A~\tilde{A} would be the usual Frobenius map of 𝔽p\mathbb{F}_{p}-Algebra. The orbit of Frobenius map σ\sigma action has several classification:
1.topological nilpotent

Nil(A)={xA0,|x|=1,limnσn(x)=0}Nil(A)=\left\{x\in A_{0}\ ,|x|=1,\lim_{n\to\infty}\sigma^{n}(x)=0\right\}

2.periodic element

T(A)={xA0,k,σk(x)=x}{0}T(A)=\left\{x\in A_{0},\exists k,\sigma^{k}(x)=x\right\}\bigcup\left\{0\right\}

3.chaos element

Chaos(A)={xA0,k,σk(x)x,|σk(x)|=1}Chaos(A)=\left\{x\in A_{0},\forall k,\sigma^{k}(x)\neq x,|\sigma^{k}(x)|=1\right\}
Theorem 2.2.

The mod pp reduction induce a correspondence:
1.topological nilpotent of AA corresponds to nilpotent elements in A~\tilde{A}

Nil(A)surjectiveNil(A~)Nil(A)\overset{surjective}{\longrightarrow}Nil(\tilde{A})

2.periodic element of AA corresponds to periodic element in A~\tilde{A}

T(A)1:1T(A~)T(A)\overset{1:1}{\longleftrightarrow}T(\tilde{A})

3.chaos element in AA remove the quasi-periodic element corresponds to chaos element in A~\tilde{A}

The proof of theorem is a standard method in theory of Witt ring[12]. If x,yT(A~)x,y\in T(\tilde{A}) commutes,it’s easy to check:x+y,xyT(A~)x+y,xy\in T(\tilde{A}).

Definition 2.3.

We call xx is a Teichmüller element if xT(A)x\in T(A). If xx acts on a ultrametric Banach space,we call xx has \infty period if:

slimNσN!(x)=x,|x|=1s-\lim_{N\to\infty}\sigma^{N!}(x)=x,|x|=1
Remark 2.1.

The importance of Teichmüller element is that Teichmüller element has a unique lifting property and it has a nature orthogonal spectrum decomposition which can be used to build up padicp-adic Hermite elements.

When we are considering a spectrum theory of operator, we should based on an algebraic closed field. At least we should consider 𝔽¯p\bar{\mathbb{F}}_{p}. From Galois theory we have:

σN(λ)=λpN=λλ𝔽pN\sigma^{N}(\lambda)=\lambda^{p^{N}}=\lambda\Leftrightarrow\lambda\in\mathbb{F}_{p^{N}}

So σN(x)=x\sigma^{N}(x)=x infers that the spectrum of xx lies in 𝔽pN\mathbb{F}_{p^{N}} (We have jumped the step of mod pp reduction because the correspondence of periodic element is 1:1). If we want to get spectral theory of algebra over 𝔽p\mathbb{F}_{p}, we should think about the action of Galois group on the spectrum. We have:

Gal(𝔽pN|𝔽p)/N\mathrm{Gal}({\mathbb{F}_{p^{N}}}|\mathbb{F}_{p})\simeq\mathbb{Z}/N\mathbb{Z}
Gal(𝔽¯p|𝔽p)limN1/N!^\mathrm{Gal}(\bar{\mathbb{F}}_{p}|\mathbb{F}_{p})\simeq\lim_{\underset{N\geq 1}{\longleftarrow}}\mathbb{Z}/N!\mathbb{Z}\simeq\widehat{\mathbb{Z}}

^\widehat{\mathbb{Z}} is the pro-finite completion of \mathbb{Z}, \mathbb{Z} is dense in ^\widehat{\mathbb{Z}}. Frobenius map σ\sigma is the topological generator of ^\widehat{\mathbb{Z}}.

Remark 2.2.

We will introduce the simplest case of spectrum decomposition. The base field kk is arbitrary.
Assume a linear transformation AA acts on finite dimensional linear space VV. Then AA generates a kk-Algebra k[A]k[A]. k[A]k[A] is isomorphic to k[X]/(f(X))k[X]/(f(X)), f(X)f(X) is the minimal polynomial of AA. f(X)f(X) has a irreducible polynomial decomposition:

f(X)=p1(X)n1pk(X)nkf(X)=p_{1}(X)^{n_{1}}...p_{k}(X)^{n_{k}}

Let hi(X)=f(X)/pi(X)nih_{i}(X)=f(X)/p_{i}(X)^{n_{i}}, hence h1(X),h2(X)hk(X)h_{1}(X),h_{2}(X)...h_{k}(X) are coprime.
So there exists ci(X)c_{i}(X) such that:

i=1kci(X)hi(X)=1\sum_{i=1}^{k}c_{i}(X)h_{i}(X)=1

Let:

ci(A)hi(A)=πic_{i}(A)h_{i}(A)=\pi_{i}

It’s easy to show that:

i=1kπi=1,πi2=πi,πiπj=0(ij)\sum_{i=1}^{k}\pi_{i}=1,\ \ \pi_{i}^{2}=\pi_{i},\ \ \pi_{i}\pi_{j}=0(\forall i\neq j)

So all πi\pi_{i} are projections and the product between them are 0. Let:

Ai=AπiVi=πi(V)A_{i}=A\pi_{i}\ \ V_{i}=\pi_{i}(V)

then we have:

1=i=1kπi,V=i=1kVi1=\sum_{i=1}^{k}\pi_{i},\ \ V=\oplus_{i=1}^{k}V_{i}
A=i=1kAi,AVi=AiA=\sum_{i=1}^{k}A_{i},\ \ A\mid_{V_{i}}=A_{i}

The first equation says the linear space is decomposed by πi\pi_{i}. The second equation says the linear transform is decomposed by πi\pi_{i}. There is a one to one correspondence between πi\pi_{i} and points in SpecA\mathrm{Spec}A. SpecA\mathrm{Spec}A is the prime spectrum of k[A]k[A]. In general we have:

V=pSpecAVpV=\oplus_{p\in\mathrm{Spec}A}V_{p}
k[A]=pSpecAk[Ap]k[A]=\oplus_{p\in\mathrm{Spec}A}k[A_{p}]

We can see that the space will be decomposed by the parameterization of spectrum. So we call this formula spectrum decomposition.
The statements above can be generalized to Euclidean domain, Principle ideal domain, Dedekind domain. If we consider the operator with metric and measure structure, then we will get theorems like Hermite operator spectrum decomposition theorem, unitary operator spectrum decomposition theorem.

To simplify issues,now we let xx be a Teichmüller element with period 1. So we have:

xp=xx^{p}=x

If the period is not 1, the theory also works. Because we can tensor a unramified field extension KK of p\mathbb{Q}_{p} such that the polynomial f(x)=xpNxf(x)=x^{p^{N}}-x splits.
Let ωi,i=1,2p\omega_{i},i=1,2...p be the root of λp=λ\lambda^{p}=\lambda in p\mathbb{Z}_{p}. Such roots exists and the amount of roots is pp, which can be given by Teichmüller lift. Now we have:

i=1p(xωi)=xpx\prod_{i=1}^{p}(x-\omega_{i})=x^{p}-x

So we have projections and spectrum decomposition:

πk=ikp(xωi)ikp(ωkωi),k=1,2,p\pi_{k}=\frac{\prod_{i\neq k}^{p}(x-\omega_{i})}{\prod_{i\neq k}^{p}(\omega_{k}-\omega_{i})},k=1,2,...p
i=1pπi=1,i=1pωiπi=x,πi2=πi,πiπj=0(ij)\sum_{i=1}^{p}\pi_{i}=1,\ \ \sum_{i=1}^{p}\omega_{i}\pi_{i}=x,\ \ \pi_{i}^{2}=\pi_{i},\ \ \pi_{i}\pi_{j}=0(\forall i\neq j)

The projections may be 0. If one of them is not 0, then it must be an orthogonal projection.

Proof.
|πk|=|ikp(xωi)ikp(ωkωi)|1|ikp(ωkωi)|\left|\pi_{k}\right|=\left|\frac{\prod_{i\neq k}^{p}(x-\omega_{i})}{\prod_{i\neq k}^{p}(\omega_{k}-\omega_{i})}\right|\leq\frac{1}{|\prod_{i\neq k}^{p}(\omega_{k}-\omega_{i})|}

The product of denominator can do mod pp reduction. We can view as a product in finite field.

ikp(ωkωi)a𝔽pa1(modp)\prod_{i\neq k}^{p}(\omega_{k}-\omega_{i})\equiv\prod_{a\in\mathbb{F}_{p}^{*}}a\equiv-1(\bmod p)

On the one hand, we know:

|πk|1|\pi_{k}|\leq 1

On the other hand, we know:

πk2=πk\pi_{k}^{2}=\pi_{k}

So if πk0\pi_{k}\neq 0, then we have:

|πk|2|πk|1|πk|=1|\pi_{k}|^{2}\geq|\pi_{k}|\geq 1\Longrightarrow|\pi_{k}|=1

Now we want to show that Teichmüller element of period \infty has a spectral decomposition, parameterized by 𝔽¯p\bar{\mathbb{F}}_{p}. Suppose XX is ultrametric Banach space over punram^\widehat{\mathbb{Q}_{p}^{unram}}, punram^\widehat{\mathbb{Q}_{p}^{unram}} is the completion of maximal unramified extension of p\mathbb{Q}_{p}. Let 𝒪={xK,|x|1}\mathcal{O}=\left\{x\in K,\left|x\right|\leq 1\right\} be the integer ring of punram^\widehat{\mathbb{Q}_{p}^{unram}} . Suppose the range of norm satisfy:|X|=|punram^|=p{0}|X|=|\widehat{\mathbb{Q}_{p}^{unram}}|=p^{\mathbb{Z}}\cup\left\{0\right\}.

Theorem 2.3.

Let xx be a linear operator on XX. xx is a Teichmüller element of period \infty if and only if there exists a spectral decomposition:

λT(K)πλ=1,λT(K)λπλ=x,πλ2=πλ,πλπλ=0(λλ)\sum_{\lambda\in T(K)}\pi_{\lambda}=1,\ \ \sum_{\lambda\in T(K)}\lambda\pi_{\lambda}=x,\ \ \pi_{\lambda}^{2}=\pi_{\lambda},\ \ \pi_{\lambda}\pi_{\lambda^{*}}=0(\forall\lambda\neq\lambda^{*})

T(K)=N=1TN!(K)T(K)=\bigcup_{N=1}^{\infty}T_{N!}(K) is the union of Teichmüller element in KK. The sum converges in strong operator topology.

Proof.

Let X0={tX,|t|1}X_{0}=\left\{t\in X,\left|t\right|\leq 1\right\},Xk=X0/pkX0,k=1,2X_{k}=X_{0}/p^{k}X_{0},k=1,2....Let ϕk:X0Xk\phi_{k}:X_{0}\to X_{k} be the canonical modpkp^{k} reduction,ϕk\phi_{k} contracts the ball of radius less than pkp^{-k} to a point in X0X_{0},ϕk\phi_{k} is surjective.Let ϕjk:XjXk,jk\phi_{jk}:X_{j}\to X_{k},j\geq k be the transition morphism.
X0X_{0} is padicp-adic complete.So we have:

limk1Xk=X0\lim_{\underset{k\geq 1}{\longleftarrow}}X_{k}=X_{0}

Equivalently, we can write the elements in X0X_{0} as:t=(t1,t2,),tj=ϕj(t)t=(t_{1},t_{2},...),t_{j}=\phi_{j}(t), ϕjk(tj)=tk\phi_{jk}(t_{j})=t_{k}, Let xx be a Teichmüller element of period \infty. T(𝔽pn!)T(\mathbb{F}_{p^{n!}}) is defined as the Teichmüller lift of 𝔽pn!\mathbb{F}_{p^{n!}} in 𝒪/pk𝒪,k=1,2\mathcal{O}/p^{k}\mathcal{O},k=1,2... We have:

σN!(x)slimx,|x|=1\sigma^{N!}(x)\overset{s-lim}{\longrightarrow}x,|x|=1

By the definition of strong limit, we have:

limNσN!(x)(t)=x(t),tX0\lim_{N\to\infty}\sigma^{N!}(x)(t)=x(t),\forall t\in X_{0}

Let xkx_{k} be the kthk-th reduction of xx. We have:xk:XkXkx_{k}:X_{k}\to X_{k}

limNσN!(xk)(tk)=xk(tk),tkXk\lim_{N\to\infty}\sigma^{N!}(x_{k})(t_{k})=x_{k}(t_{k}),\forall t_{k}\in X_{k}

Since the ball of radius less than pkp^{-k} have contracted to a point in XkX_{k}, we have:

tkXk,n,σn!(x)(tk)=x(tk)\forall t_{k}\in X_{k},\exists n,\sigma^{n!}(x)(t_{k})=x(t_{k})

So we have:Xk=n1Xk,nX_{k}=\bigcup_{n\geq 1}X_{k,n},Xk,n={tkXk,σn!(xk)(tk)=xk(tk)}X_{k,n}=\left\{t_{k}\in X_{k},\sigma^{n!}(x_{k})(t_{k})=x_{k}(t_{k})\right\},Xk,nXk,nX_{k,n}\subseteq X_{k,n^{*}} when nnn\leq n^{*}. We define:

πk,λ,n=λλ,λT(𝔽pn!)xkλλλ\pi_{k,\lambda,n}=\prod_{\lambda^{*}\neq\lambda,\lambda^{*}\in T(\mathbb{F}_{p^{n!}})}\frac{x_{k}-\lambda^{*}}{\lambda-\lambda^{*}}

πk,λ,n\pi_{k,\lambda,n} is the projection of Xk,nX_{k,n}. In general we have:

λT(𝔽pn!)λπk,λ,n=1Xk,λ,n,λT(𝔽pn!)λπk,λ,n=xkXk,λ,n,πk,λ,n2=πk,λ,n,πk,λ,nπk,λ,n=0(λλ)\sum_{\lambda\in T(\mathbb{F}_{p^{n!}})}\lambda\pi_{k,\lambda,n}=1\mid_{X_{k,\lambda,n}},\sum_{\lambda\in T(\mathbb{F}_{p^{n!}})}\lambda\pi_{k,\lambda,n}=x_{k}\mid_{X_{k,\lambda,n}},\ \ \pi_{k,\lambda,n}^{2}=\pi_{k,\lambda,n},\pi_{k,\lambda,n}\pi_{k,\lambda^{*},n}=0(\forall\lambda\neq\lambda^{*})\ \

We can get a direct sum:

Xk,n=λT(𝔽pn!)Xk,λ,n1X_{k,n}=\underset{\lambda\in T(\mathbb{F}_{p^{n!}})}{\oplus}X_{k,\lambda},\forall n\geq 1

Since Xk=n1Xk,nX_{k}=\bigcup_{n\geq 1}X_{k,n}, We have:

Xk=λT(𝔽¯p)Xk,λX_{k}=\underset{\lambda\in T(\bar{\mathbb{F}}_{p})}{\oplus}X_{k,\lambda}

We get the spectral decomposition of xkx_{k} over XkX_{k}. Let πk,λ\pi_{k,\lambda} be the projection: πk,λ:XkXk,λ\pi_{k,\lambda}:X_{k}\to X_{k,\lambda}. It’s no hard to see that:

xk=λT(𝔽¯p)λπk,λx_{k}=\sum_{\lambda\in T(\bar{\mathbb{F}}_{p})}\lambda\pi_{k,\lambda}

We want to show:

limk1Xk,λ=Xλ,X0=λT(K)^Xλ\lim_{\underset{k\geq 1}{\longleftarrow}}X_{k,\lambda}=X_{\lambda},X_{0}=\underset{\lambda\in T(K)}{\hat{\oplus}}X_{\lambda}

Here we define Xλ={tX0,x(t)=λt}X_{\lambda}=\left\{t\in X_{0},x(t)=\lambda t\right\}. λ\lambda is a element in T(K)T(K). We get two commutative diagrams:
Xj{X_{j}}Xk{X_{k}}Xλ{X_{\lambda}}limk1Xk,λ{\underset{\underset{k\geq 1}{\longleftarrow}}{\lim}X_{k,\lambda}}Xj,λ{X_{j,\lambda}}Xk,λ{X_{k,\lambda}}Xj,λ{X_{j,\lambda}}Xk,λ{X_{k,\lambda}}πj,λ\scriptstyle{\pi_{j,\lambda}}ϕjk\scriptstyle{\phi_{jk}}πk,λ\scriptstyle{\pi_{k,\lambda}}!θ\scriptstyle{\exists!\theta}ϕj\scriptstyle{\phi_{j}}ϕk\scriptstyle{\phi_{k}}ϕjk\scriptstyle{\phi_{jk}}ϕjk\scriptstyle{\phi_{jk}}
We want to show θ\theta is isomorphism. Let:

λ=(λ1,λ2,)𝒪=limk1𝒪/pk𝒪,t=(t1,t2,)limk1Xk,λ\lambda=(\lambda_{1},\lambda_{2},...)\in\mathcal{O}=\underset{\underset{k\geq 1}{\longleftarrow}}{\lim}\mathcal{O}/p^{k}\mathcal{O},t=(t_{1},t_{2},...)\in\underset{\underset{k\geq 1}{\longleftarrow}}{\lim}X_{k,\lambda}
xk(tk)=λktk,ϕk(tj)=tk,ϕk(λj)=λk,jkx_{k}(t_{k})=\lambda_{k}t_{k},\phi_{k}(t_{j})=t_{k},\phi_{k}(\lambda_{j})=\lambda_{k},j\geq k

We have:

x(t)=(λ1t1,λ2t2,),ϕk(λjtj)=λktkx(t)=(\lambda_{1}t_{1},\lambda_{2}t_{2},...),\phi_{k}(\lambda_{j}t_{j})=\lambda_{k}t_{k}

Hence it has a limit:x(t)=λtx(t)=\lambda t. So πλ=limk1πk,λ:X0Xλ\pi_{\lambda}=\underset{\underset{k\geq 1}{\longleftarrow}}{\lim}\pi_{k,\lambda}:X_{0}\to X_{\lambda} is a well-defined orthogonal projection. We only need to check the completeness of {πλ}\left\{\pi_{\lambda}\right\},i.e:

X0=λT(K)^XλX_{0}=\underset{\lambda\in T(K)}{\hat{\oplus}}X_{\lambda}

Considering the modpkp^{k} reduction:

ϕk:X0Xk=λT(𝔽¯p)Xk,λ\phi_{k}:X_{0}\to X_{k}=\underset{\lambda\in T(\bar{\mathbb{F}}_{p})}{\oplus}X_{k,\lambda}
ϕk:XλXk,λ\phi_{k}:X_{\lambda}\to X_{k,\lambda}

Since the direct sum is always finite, modpkp^{k} reduction is surjective. We get:λXλ\oplus_{\lambda}X_{\lambda} is dense in X0X_{0}. We decompose the Teichmüller element xx of period \infty:

λT(K)πλ=1,λT(K)λπλ=x,πλ2=πλ,πλπλ=0(λλ)\sum_{\lambda\in T(K)}\pi_{\lambda}=1,\ \ \sum_{\lambda\in T(K)}\lambda\pi_{\lambda}=x,\ \ \pi_{\lambda}^{2}=\pi_{\lambda},\ \ \pi_{\lambda}\pi_{\lambda^{*}}=0(\forall\lambda\neq\lambda^{*})

Finally, if xx has such a spectral decomposition(in strong operator topology), it’s easy to check xx is a Teichmüller element. ∎

Now we will think about the reduction of orthogonal projection family and the lifting of projection family. In fact there exists 1:1 correspondence. Recall our notation:AA is a KK-ultrametric Banach algebra such that: |A|=p{0}\left|A\right|=p^{\mathbb{Z}}\cup\left\{0\right\}, KK is a finite unramified extension of p\mathbb{Q}_{p}. Let A0A_{0} be the unit ball of AA:A0={xA,|x|1}A_{0}=\left\{x\in A,|x|\leq 1\right\}. A0A_{0} is a nature p\mathbb{Z}_{p}-Algebra which can be reduced to A~=A0/pA0\tilde{A}=A_{0}/pA_{0}. A~\tilde{A} has a nature 𝔽p\mathbb{F}_{p}-Algebra structure.

Lemma 2.4.

(Unique Lifting Lemma)
Suppose π¯\bar{\pi} is a projection in A~\tilde{A}. There exists a unique lifting of π¯\bar{\pi}, we call it π\pi. π\pi is orthogonal projection, σ(π)=π\sigma(\pi)=\pi.
Suppose π¯k,k=1,2p\bar{\pi}_{k},k=1,2...p is projections in A~\tilde{A}. We have:

i=1pπ¯i=1,π¯i2=π¯i,π¯iπ¯j=0(ij)\sum_{i=1}^{p}\bar{\pi}_{i}=1,\ \ \bar{\pi}_{i}^{2}=\bar{\pi}_{i},\ \ \bar{\pi}_{i}\bar{\pi}_{j}=0(\forall i\neq j)

Suppose πk,k=1,2,p\pi_{k},k=1,2,...p is the unique lifting of π¯k\bar{\pi}_{k}, then the lifting also satisfy the relationship:

i=1pπi=1,πi2=πi,πiπj=0(ij)\sum_{i=1}^{p}\pi_{i}=1,\ \ \pi_{i}^{2}=\pi_{i},\ \ \pi_{i}\pi_{j}=0(\forall i\neq j)
Proof.

Suppose we select a lifting of π¯\bar{\pi} named aa, such that:a2=a+py,yA0a^{2}=a+py,y\in A_{0}. Considering the Cauchy sequence:{σk(a),k=1,2,}\left\{\sigma^{k}(a),k=1,2,...\right\}. The σ\sigma is Frobenius map. Then there exists a unique limit π\pi such that:σ(π)=π,π2=π\sigma(\pi)=\pi,\pi^{2}=\pi. We have π\pi is orthogonal projection.
Suppose we select liftings of πk¯\bar{\pi_{k}} named πk\pi_{k}. First, we have estimate:

|πiπj|1p(ij)|\pi_{i}\pi_{j}|\leq\frac{1}{p}(\forall i\neq j)

Secondly, suppose πiπj0\pi_{i}\pi_{j}\neq 0, we have (This estimate depends on the commutative of πi\pi_{i}):

|πiπj|1(ij)|\pi_{i}\pi_{j}|\geq 1(\forall i\neq j)

which leads to a contradiction. Finally, (i=1kπi)2=i=1kπi\left(\sum_{i=1}^{k}\pi_{i}\right)^{2}=\sum_{i=1}^{k}\pi_{i} and i=1kπi\sum_{i=1}^{k}\pi_{i} is in the neighborhood of 1, so i=1kπi=1\sum_{i=1}^{k}\pi_{i}=1. ∎

3 Spectral measure of padicp-adic Hermite elements

Now let’s recall the defnition of padicp-adic Hermite elements.
Let KK be an unramified extension of p\mathbb{Q}_{p} such that ωpN=ω\omega^{p^{N}}=\omega has exactly pNp^{N} solutions. Let AA be a commutative KK-Banach algebra with unit, σ\sigma be the Frobenius map:σ:AA,xxp\sigma:A\to A,x\mapsto x^{p}. We assume the norm on AA satisfy:|A|=|K|\left|A\right|=\left|K\right|. Let A0A_{0} be the unit ball: A0={xA,|x|1}A_{0}=\left\{x\in A,\left|x\right|\leq 1\right\}.

Definition 3.1.

Let xAx\in A, We say xx is a padicp-adic hermite operator with period NN, if xx has a taylor expansion:

x=i=kxipi,xiTN(A0)x=\sum_{i=k}^{\infty}x_{i}p^{i},x_{i}\in T_{N}(A_{0})

Let k=0,N=1k=0,N=1. Then we have:xip=xi,i=0,1,2x_{i}^{p}=x_{i},i=0,1,2..., the assumption k=0k=0 is convenient for our discussion.

The following lemma is not clearly point out before. However,it is obvious.

Lemma 3.1.

If π\pi,π\pi^{*} are orthogonal projections, and they are commute. Then ππ\pi\pi^{*} are either orthogonal projection or 0.

We will show that how to define a spectral measure of xx
xix_{i} satisfies a polynomial equation:xip=xix_{i}^{p}=x_{i}. So we can define a spectral decompositon of xix_{i}.We have:

πi,j,j=1,2..pπi,j2=πi,jπi,jπi,k=0(jk)\exists\pi_{i,j},j=1,2..p\ \ \ \pi_{i,j}^{2}=\pi_{i,j}\ \ \pi_{i,j}\pi_{i,k}=0(\forall j\neq k)
j=1pπi,j=Ij=1pωjπi,j=xi\sum_{j=1}^{p}\pi_{i,j}=I\ \ \sum_{j=1}^{p}\omega_{j}\pi_{i,j}=x_{i}

ωj\omega_{j} is the root of λp=λ\lambda^{p}=\lambda. Since |xi|=1|x_{i}|=1 (if xi=0x_{i}=0, the same argument holds) and the norm over AA has the Non-Archimedean property, it’s not hard to prove that:

|πi,j|=1or 0|\pi_{i,j}|=1\ or\ 0

So the spectral decompositon of xix_{i} is an orthogonal spectral decompositon.
We can apply this spectral decompositon step by step to get the spectral decompositon of xx.
Actually we have:

I=j=1pπ0,jI=\sum_{j=1}^{p}\pi_{0,j}
I=j,k=1pπ0,jπ1,kI=\sum_{j,k=1}^{p}\pi_{0,j}\pi_{1,k}
......
I=limni0,i1,in=1pπ0,i0π1,i1πn,in=p𝑑EλI=\lim_{n\to\infty}\sum_{i_{0},i_{1},...i_{n}=1}^{p}\pi_{0,i_{0}}\pi_{1,i_{1}}...\pi_{n,i_{n}}=\int_{\mathbb{Z}_{p}}dE_{\lambda}
x0=j=1pωjπ0,jx_{0}=\sum_{j=1}^{p}\omega_{j}\pi_{0,j}
x0+px1=j,k=1p(ωj+pωk)π0,jπ1,kx_{0}+px_{1}=\sum_{j,k=1}^{p}(\omega_{j}+p\omega_{k})\pi_{0,j}\pi_{1,k}
......
x=limni0,i1,in=1p(ωi0+pωi1++pnωin)π0,i0π1,i1πn,in=pλ𝑑Eλx=\lim_{n\to\infty}\sum_{i_{0},i_{1},...i_{n}=1}^{p}(\omega_{i_{0}}+p\omega_{i_{1}}+...+p^{n}\omega_{i_{n}})\pi_{0,i_{0}}\pi_{1,i_{1}}...\pi_{n,i_{n}}=\int_{\mathbb{Z}_{p}}\lambda dE_{\lambda}
π0,1π0,2π0,pπ1,1π1,2π1,pπ2,1π2,2π2,p\begin{matrix}\pi_{0,1}&\pi_{0,2}&...&\pi_{0,p}&\\ \pi_{1,1}&\pi_{1,2}&...&\pi_{1,p}&\\ \pi_{2,1}&\pi_{2,2}&...&\pi_{2,p}&\\ ...&...&...&...&\end{matrix}

For each path from top to bottom on the diagram corresponding to a unique number in p\mathbb{Z}_{p}. Counting from top to bottom, the projections on the first floor correspond to the canonical resolution of p\mathbb{Z}_{p} to small discs of whose radius is r=1pr=\frac{1}{p}. The second floor correspond to the resolution of small discs of p\mathbb{Z}_{p} whose radius is r=1p2r=\frac{1}{p^{2}} and so on.

In this sense we build up the fractal spectral measure of Hermite operator xx.It is a finite additive Orthogonal projection valued measure over p\mathbb{Z}_{p}.
In general,without the assumption of k=0k=0,we have:

Theorem 3.2.

(padicp-adic Hermite operator Spectral decomposition theorem)
Let xx be a padicp-adic Hermite operator, there exists a spectral integral:

I=p𝑑Eλx=pλ𝑑EλI=\int_{\mathbb{Q}_{p}}dE_{\lambda}\ \ \ x=\int_{\mathbb{Q}_{p}}\lambda dE_{\lambda}

We can compare the result with the Archimedean case:

Theorem 3.3.

(Hermite operator spectral decomposition theorem)
Let xx be a Hermite operator, there exists a spectral integral:

I=𝑑Eλx=λ𝑑EλI=\int_{\mathbb{R}}dE_{\lambda}\ \ \ x=\int_{\mathbb{R}}\lambda dE_{\lambda}

There is a nature connection with the path-intergral in quantum mechanics. As we know, the original definition of the path-intergral is to apply the spectral decomposition theorem in different time again and again to calculate the propagator. Our definition is related to this statement. There is a one to one correspondence between the path in the following diagram to the filtration of balls in p\mathbb{Z}_{p}. We should sum the all paths in the diagram to give the spectral decomposition.

π0,1π0,2π0,pπ1,1π1,2π1,pπ2,1π2,2π2,p\begin{matrix}\pi_{0,1}&\pi_{0,2}&...&\pi_{0,p}&\\ \pi_{1,1}&\pi_{1,2}&...&\pi_{1,p}&\\ \pi_{2,1}&\pi_{2,2}&...&\pi_{2,p}&\\ ...&...&...&...&\end{matrix}

What happens in finite dimension case? Let XX be a p\mathbb{Q}_{p}-linear space with finite dimension, equipped with a non-Archimedean norm such that |X|=|p|=p{0}|X|=|\mathbb{Q}_{p}|=p^{\mathbb{Z}}\cup\left\{0\right\}. Let xx be a padicp-adic Hermite operator acts on XX. Let Πn={π0,i0π1,i1πn,in}ik=1,2p{0}\varPi_{n}=\left\{\pi_{0,i_{0}}\pi_{1,i_{1}}...\pi_{n,i_{n}}\right\}_{i_{k}=1,2...p}-\left\{0\right\}, Πn\varPi_{n} is the spectral decomposition of x0+px1++pnxnx_{0}+px_{1}+...+p^{n}x_{n}, which is complete and orthogonal. The element in Πn+1\varPi_{n+1} is the subprojection of Πn\varPi_{n}. It’s obvious to see:

Card(Πn)Card(Πn+1)dimXCard(\varPi_{n})\leq Card(\varPi_{n+1})\leq dimX

Hence it has a limit:

limnCard(Πn)\lim_{n\to\infty}Card(\varPi_{n})

Hence we have: Πn\varPi_{n} is stable for some n0n_{0}. Let Π\varPi be the limit set of Πn\varPi_{n}.Π\varPi is composed of a finite number of Orthogonal projections. Actually, we have:

x=i=1Card(Π)λiπi,λipx=\sum_{i=1}^{Card(\varPi)}\lambda_{i}\pi_{i},\lambda_{i}\in\mathbb{Q}_{p}
i=1Card(Π)πi=1,πi2=πi,πiπj=0(ij)\sum_{i=1}^{Card(\varPi)}\pi_{i}=1,\ \ \pi_{i}^{2}=\pi_{i},\ \ \pi_{i}\pi_{j}=0(\forall i\neq j)

We will give a more detailed description immediately.

4 Examples

In this section, We want to give some examples of padicp-adic Hermite operator.
Let pn\mathbb{Q}_{p}^{n} equipped with supremum norm:

xpn,x=(x1,x2,xn),xip,|x|=sup1in(|xi|)\forall x\in\mathbb{Q}_{p}^{n},x=(x_{1},x_{2}...,x_{n}),x_{i}\in\mathbb{Q}_{p},|x|=\sup_{1\leq i\leq n}\left(|x_{i}|\right)

Here nn is an arbitrary integer. Let AA be a linear operator acts on pn\mathbb{Q}_{p}^{n}. The following conditions are equivalent:

  1. 1.

    AA is a padicp-adic Hermite operator.

  2. 2.

    There exists a orthogonal projection spectral decomposition:

    λπλ=1,λλπλ=A,πλ2=πλ,πλπλ=0(λλ)\sum_{\lambda}\pi_{\lambda}=1,\ \ \sum_{\lambda}\lambda\pi_{\lambda}=A,\ \ \pi_{\lambda}^{2}=\pi_{\lambda},\ \ \pi_{\lambda}\pi_{\lambda^{*}}=0(\forall\lambda\neq\lambda^{*})

    λp\lambda\in\mathbb{Q}_{p} is the eigenvalue of AA.

  3. 3.

    UGLn(p),A=Udiag(λ1,λn)U1\exists U\in GL_{n}(\mathbb{Z}_{p}),A=Udiag(\lambda_{1},...\lambda_{n})U^{-1}

Proof.

121\Rightarrow 2:We have shown.

232\Rightarrow 3:The columns of elements in GLn(p)GL_{n}(\mathbb{Z}_{p}) is padicp-adic orthonormalized. We mean for any gGLn(p)g\in GL_{n}(\mathbb{Z}_{p}), write gg as a column vector combination:g=(X1,X2,Xn)g=(X_{1},X_{2}...,X_{n}). We have:

(c1,c2,cn)pn,|i=1nciXi|=sup1im|ci|\forall(c_{1},c_{2}...,c_{n})\in\mathbb{Q}_{p}^{n},\left|\sum_{i=1}^{n}c_{i}X_{i}\right|=\sup_{1\leq i\leq m}\left|c_{i}\right|

The converse proposition is also true. Actually there is a one to one correspondence between padicp-adic orthonormalized column vectors with elements in GLn(p)GL_{n}(\mathbb{Z}_{p}). GLn(p)GL_{n}(\mathbb{Z}_{p}) is the padicp-adic substitute of the Archimedean case orthogonal group:On()O_{n}(\mathbb{R}) or Un()U_{n}(\mathbb{C}). We can find a orthonormal basis of pn\mathbb{Q}_{p}^{n} corresponds with the orthogonal projection spectral decomposition to get UU.

313\Rightarrow 1: We have the Taylor expansion of diag(λ1,λn)diag(\lambda_{1},...\lambda_{n}) in pn\mathbb{Q}_{p}^{n}. So it is a padicp-adic Hermite operator. Moreover, it’s easy to see AA is a padicp-adic Hermite operator if and only if U1AUU^{-1}AU is a padicp-adic Hermite operator. ∎

Let KK be a field, CharK=0\mathrm{Char}K=0. Let K[X1,X2,Xn]K[X_{1},X_{2},...X_{n}] be the polynomials with nn variables over KK.

For homogeneous polynomials, we have Euler Theorem:

Definition 4.1.

We call f(X1,X2,Xn)K[X1,X2,Xn]f(X_{1},X_{2},...X_{n})\in K[X_{1},X_{2},...X_{n}] is a homogeneous polynomial of degree kk if:

f(αX1,αX2,αXn)=αkf(X1,X2,Xn),αK×f(\alpha X_{1},\alpha X_{2},...\alpha X_{n})=\alpha^{k}f(X_{1},X_{2},...X_{n}),\forall\alpha\in K^{\times}
Theorem 4.1.

We call Δ=i=1nXiXi\Delta=\sum_{i=1}^{n}X_{i}\frac{\partial}{\partial X_{i}} Euler operator, f(X1,X2,Xn)K[X1,X2,Xn]f(X_{1},X_{2},...X_{n})\in K[X_{1},X_{2},...X_{n}] is a homogeneous polynomial of degree kk if and only if:Δf=kf\Delta f=kf

Let K<X1,X2,Xn>K<X_{1},X_{2},...X_{n}> be the Tate Algebra:

K<X1,X2,Xn>={f=i1,i2,inai1,i2,inX1i1X2i2Xnin,ai1,i2,in0}K<X_{1},X_{2},...X_{n}>=\left\{f=\sum_{i_{1},i_{2},...i_{n}}^{\infty}a_{i_{1},i_{2},...i_{n}}X_{1}^{i_{1}}X_{2}^{i_{2}}...X_{n}^{i_{n}},a_{i_{1},i_{2},...i_{n}}\to 0\right\}

with respect to the Gauss norm:

|i1,i2,inai1,i2,inX1i1X2i2Xnin|=sup|ai1,i2,in|\left|\sum_{i_{1},i_{2},...i_{n}}^{\infty}a_{i_{1},i_{2},...i_{n}}X_{1}^{i_{1}}X_{2}^{i_{2}}...X_{n}^{i_{n}}\right|=\sup\left|a_{i_{1},i_{2},...i_{n}}\right|

Here we assume KK is a field extension of p\mathbb{Q}_{p} with non-Archimedean norm.

Proposition 4.2.

The Euler operator Δ\Delta is a padicp-adic Hermite operator over K<X1,X2,Xn>K<X_{1},X_{2},...X_{n}> with eigenvalue: k=0,1,2,k=0,1,2,... and eigenfunction: f0=1;f1,1=X1,f1,2=X2f1,n=Xn;f_{0}=1;f_{1,1}=X_{1},f_{1,2}=X_{2}...f_{1,n}=X_{n};.... Eigenfunctions are padicp-adic orthonormalized. In general we can define the creation and annihilation operators:

ai+=Xi,ai=Xi,[ai,ai+]=1,Δ=i=1nai+aia^{+}_{i}=X_{i},a^{-}_{i}=\frac{\partial}{\partial X_{i}},[a_{i}^{-},a_{i}^{+}]=1,\Delta=\sum_{i=1}^{n}a^{+}_{i}a^{-}_{i}

Now we will simplify examples given by A.N.Kochubei. We will follow on [7] and try to establish a general framework.
Let C(p,p)C(\mathbb{Z}_{p},\mathbb{C}_{p}) be the set of continous function from p\mathbb{Z}_{p} to p\mathbb{C}_{p}. We define:

(a+f)(x)=xf(x1),(af)(x)=f(x+1)f(x),xp(a^{+}f)(x)=xf(x-1),(a^{-}f)(x)=f(x+1)-f(x),x\in\mathbb{Z}_{p}

a+,aa^{+},a^{-} are bounded and satify the relation [a,a+]=1[a^{-},a^{+}]=1. Let A=a+aA=a^{+}a^{-}, then AA is padicp-adic hermite operator with eigenfunction:

Pn(x)=x(x1)(xn+1)n!,n1;P0(x)=1P_{n}(x)=\frac{x(x-1)...(x-n+1)}{n!},n\geq 1;P_{0}(x)=1

such that APn=nPnAP_{n}=nP_{n}. Let (af)(x)=f(x+1)(a^{*}f)(x)=f(x+1), We have:[a,a+]=1[a^{*},a^{+}]=1.Let’s define:

A=a+a,(Af)(x)=xf(x)A^{*}=a^{+}a^{*},(A^{*}f)(x)=xf(x)

AA^{*} is padicp-adic hermite operator because A=xA^{*}=x is a element in C(p,p)C(\mathbb{Z}_{p},\mathbb{Z}_{p}) , which has the property:

f(x)C(p,p),σ(f(x))=(f(x))pf(x)(modp)\forall f(x)\in C(\mathbb{Z}_{p},\mathbb{Z}_{p}),\sigma(f(x))=(f(x))^{p}\equiv f(x)\pmod{p}

We want to summarize the examples listed above. Let XX be a KK-ultrametric Banach space(KK is extension of p\mathbb{Q}_{p}) or Hilbert space over \mathbb{C}. In quantum mechanics and quantum field theory we know the creation operators can express particle production. Here is our definition:

Definition 4.2.

We call a triple (Ω,a^+,a^)(\Omega,\hat{a}^{+},\hat{a}^{-}) creation and annihilation operators if they satisfy the following conditions(ΩX\Omega\in X is a special element):

  • 0.

    We call Ω\Omega vacuum or ground state, a^+\hat{a}^{+} creation operator, a^\hat{a}^{-} annihilation operator.

  • 1.

    {(a^+)n(Ω)}n=0,1,\left\{(\hat{a}^{+})^{n}(\Omega)\right\}_{n=0,1,...} generates XX. They are orthogonal to each other.

  • 2.

    a^(Ω)=0\hat{a}^{-}(\Omega)=0, [a^,a^+]=1[\hat{a}^{-},\hat{a}^{+}]=1.

We will express the condition 1 more accurately.

  • 1.

    (Archimedean)

    Span(Ω,a^+(Ω),(a^+)2(Ω))¯=X\overline{Span(\Omega,\hat{a}^{+}(\Omega),(\hat{a}^{+})^{2}(\Omega)...)}=X
    nm,((a^+)n(Ω),(a^+)m(Ω))=0\forall n\neq m,((\hat{a}^{+})^{n}(\Omega),(\hat{a}^{+})^{m}(\Omega))=0
  • 1*.

    (non-Archimedean)

    Span(Ω,a^+(Ω),(a^+)2(Ω))¯=X\overline{Span(\Omega,\hat{a}^{+}(\Omega),(\hat{a}^{+})^{2}(\Omega)...)}=X
    |n=0kcn(a^+)n(Ω)|=sup0nk|cn||(a^+)n(Ω)||,k=0,1,2\left|\sum_{n=0}^{k}c_{n}(\hat{a}^{+})^{n}(\Omega)\right|=\underset{0\leq n\leq k}{\sup}\left|c_{n}\right|\left|(\hat{a}^{+})^{n}(\Omega)\right||,k=0,1,2...
Proposition 4.3.

In the conditions above, we have: N=a^+a^N=\hat{a}^{+}\hat{a}^{-} is a Hermite operator(usual hermite operator if XX is a Hilbert space, padicp-adic hermite operator if XX is a KK-ultrametric Banach space) with the spectrum:SpecN={0,1,2,}SpecN=\left\{0,1,2,...\right\}. Ω\Omega is the cyclic vector.

Remark 4.1.

In Quantum mechanics, we know the hamilton of Harmonic oscillator:

H=12mω2x212md2dx2H=\frac{1}{2}m\omega^{2}x^{2}-\frac{1}{2m}\frac{d^{2}}{dx^{2}}

It is well-known that:

a^+=12mω(mωxddx),a^=12mω(mωx+ddx)\hat{a}^{+}=\frac{1}{\sqrt{2m\omega}}(m\omega x-\frac{d}{dx}),\hat{a}^{-}=\frac{1}{\sqrt{2m\omega}}(m\omega x+\frac{d}{dx})
Ω=emωx22,H=ω(a^+a^+12)\Omega=e^{-\frac{m\omega x^{2}}{2}},H=\omega(\hat{a}^{+}\hat{a}^{-}+\frac{1}{2})

Let X=K<X>X=K<X> be the Tate Algebra, we define:

a^+=x,a^=ddx\hat{a}^{+}=x,\hat{a}^{-}=\frac{d}{dx}
Ω=1,N=a^+a^\Omega=1,N=\hat{a}^{+}\hat{a}^{-}

There is not a unique choice. We can also define:

a^+=x+h(ddx),a^=ddx;h(X)K<X>,|h(X)|1\hat{a}^{+}=x+h(\frac{d}{dx}),\hat{a}^{-}=\frac{d}{dx};h(X)\in K<X>,\left|h(X)\right|\leq 1
Ω=1,N=a^+a^\Omega=1,N=\hat{a}^{+}\hat{a}^{-}

The more accurately vacuum of the Tate Algebra is not identity function. We should view it as 1𝒪1_{\mathcal{O}} since the Tate Algebra describe the analytic geometry of 𝒪\mathcal{O}. The statement listed above coincide with Tate’s thesis[16]. We have the correspondence of fast decreasing functions:

1pex22,x1_{\mathbb{Z}_{p}}\longleftrightarrow e^{-\frac{x^{2}}{2}},x\in\mathbb{R}

We will show there exists a decomposition of n×nn\times n padicp-adic matrix AMn(𝒪K)A\in M_{n}(\mathcal{O}_{K}), K=punram^K=\widehat{\mathbb{Q}_{p}^{unram}}. Consider the modpmodp reduction: Mn(𝒪K)Mn(𝔽¯p)M_{n}(\mathcal{O}_{K})\to M_{n}(\bar{\mathbb{F}}_{p})

Mn(𝔽¯p)M_{n}(\bar{\mathbb{F}}_{p}) is a non-commutative 𝔽p\mathbb{F}_{p}-algebra. Let AMn(𝒪K)A\in M_{n}(\mathcal{O}_{K}) be a matrix. We can define a norm: A=(aij)1i,jn,|A|=sup1i,jn|aij|A=(a_{ij})_{1\leq i,j\leq n},|A|=\underset{1\leq i,j\leq n}{\sup}|a_{ij}|. Let 𝒜\mathcal{A} be the KK-algebra generated by AA. Let A~Mn(𝔽¯p)\widetilde{A}\in M_{n}(\bar{\mathbb{F}}_{p}) be the reduction of AA. The coefficient of A~\widetilde{A} can be embedded into a common finite extension of 𝔽p\mathbb{F}_{p}. So A~\tilde{A} generates a finite dimensional commutative 𝔽p\mathbb{F}_{p}-algebra 𝒜~\widetilde{\mathcal{A}}.

Proposition 4.4.

Let 𝒜~\widetilde{\mathcal{A}} be a finite dimensional commutative 𝔽p\mathbb{F}_{p}-algebra. Let X~𝒜~\widetilde{X}\in\widetilde{\mathcal{A}} be a arbitrary element. Let σ\sigma be the Frobenius morphism. There exists a canonical Jordan decomposition:

X~=Xs~+Xn~,Xs~=limkσk!(X~)\widetilde{X}=\widetilde{X_{s}}+\widetilde{X_{n}},\widetilde{X_{s}}=\lim_{k\to\infty}\sigma^{k!}(\widetilde{X})

Xs~\widetilde{X_{s}} is a Teichmüller element:

limkσk!(Xs~)=Xs~\lim_{k\to\infty}\sigma^{k!}(\widetilde{X_{s}})=\widetilde{X_{s}}

Moveover, Xs~\widetilde{X_{s}} has finite period.
Xn~\widetilde{X_{n}} is a nilpotent element:

limkσk!(Xn~)=0\lim_{k\to\infty}\sigma^{k!}(\widetilde{X_{n}})=0
Proof.

The Frobenius morphism σ\sigma is a 𝔽p\mathbb{F}_{p}-linear map on 𝒜~\widetilde{\mathcal{A}}, since 𝒜~\widetilde{\mathcal{A}} is finite dimensional commutative 𝔽p\mathbb{F}_{p}-algebra. From linear algebra, 𝒜~\widetilde{\mathcal{A}} has a Jordan decomposition:

𝒜~=𝒜~s𝒜~n\widetilde{\mathcal{A}}=\widetilde{\mathcal{A}}_{s}\oplus\widetilde{\mathcal{A}}_{n}

We have: σ𝒜~s\sigma\mid_{\widetilde{\mathcal{A}}_{s}} is invertible, σ𝒜~n\sigma\mid_{\widetilde{\mathcal{A}}_{n}} is nilpotent. Consider the cardinality of 𝒜~\widetilde{\mathcal{A}} is finite, we have:

x𝒜~s,k,σk(x)=x\forall x\in\widetilde{\mathcal{A}}_{s},\exists k,\sigma^{k}(x)=x

Theorem 4.5.

Let AMn(𝒪K)A\in M_{n}(\mathcal{O}_{K}) be a matrix. Suppose |A|=1|A|=1. There exists a canonical Jordan decomposition:

A=As+AnA=A_{s}+A_{n}

AsT(Mn(𝒪K))A_{s}\in T(M_{n}(\mathcal{O}_{K})) is a Teichmüller element with finite period, AnA_{n} is a topological nilpotent element such that:

|An|1,limkσk!(An)=0|A_{n}|\leq 1,\lim_{k\to\infty}\sigma^{k!}(A_{n})=0
Proof.

Let A~\widetilde{A} be the reduction of AA, A~=As~+An~\widetilde{A}=\widetilde{A_{s}}+\widetilde{A_{n}}. Let AsA_{s}^{*} be a lift of As~\widetilde{A_{s}} in 𝒜\mathcal{A}, which is the K-algebra generated by AA. 𝒜\mathcal{A} is commutative, so we have a limit independent of the choice of AsA_{s}^{*}:

limkσk!(As)=As\lim_{k\to\infty}\sigma^{k!}(A_{s}^{*})=A_{s}

From the unique lift lemma, AsA_{s} is a Teichmüller element. So we have a canonical Jordan decomposition:

A=As+An,An=AAsA=A_{s}+A_{n},A_{n}=A-A_{s}

Finally, the reduction of AnA_{n} is An~\widetilde{A_{n}}. So we have:

limkσk!(An~)=0|An|1,limkσk!(An)=0\lim_{k\to\infty}\sigma^{k!}(\widetilde{A_{n}})=0\Rightarrow|A_{n}|\leq 1,\lim_{k\to\infty}\sigma^{k!}(A_{n})=0

5 padicp-adic uncertainty principle

In this section, we want to set up uncertainty principle for padicp-adic Hermite operator. Let KK be the completion of maximal unramified extension of p\mathbb{Q}_{p}. Let XX be an ultrametric Banach space over KK such that the range of norm ||:X+|\ \ |:X\to\mathbb{R}_{+} on XX satisfy: |X|=|K||X|=|K|. In a sense it is a formal proof in physics.

Lemma 5.1.

Let AA be a padicp-adic Hermite operator. Let SpecASpecA be the spectrum of AA. We have:

|A|=supλSpecA|λ|\left|A\right|=\underset{\lambda\in SpecA}{\sup}|\lambda|
Definition 5.1.

We define the spectrum diameter of AA:

diam(A)=supλ,λSpecA|λλ||A|diam(A)=\underset{\lambda,\lambda^{*}\in SpecA}{\sup}|\lambda-\lambda^{*}|\leq|A|
Lemma 5.2.

For any translation: A~=Aμ\widetilde{A}=A-\mu, μSpecA\mu\in SpecA, we have:

|A~|=diam(A)|\widetilde{A}|=diam(A)
Proof.

We only prove the Lemma 5.2. The ultrametric property shows that:

|A~|=supλSpecA|λμ|supλ,λSpecA|λλ|=diam(A)|\widetilde{A}|=\underset{\lambda\in SpecA}{\sup}|\lambda-\mu|\leq\underset{\lambda,\lambda^{*}\in SpecA}{\sup}|\lambda-\lambda^{*}|=diam(A)
diam(A)=diam(A~)|A~|diam(A)=diam(\widetilde{A})\leq|\widetilde{A}|

Let A,BA,B be padicp-adic Hermite operators on XX. Let ψX,|ψ|=1\psi\in X,|\psi|=1 be the normalized wave function.

Theorem 5.3.

(padicp-adic uncertainty principle) We have:

|[A,B]ψ|diam(A)diam(B)\left|[A,B]\psi\right|\leq diam(A)*diam(B)
Proof.

We have:

|[A,B]ψ|diam(A)diam(B)\left|[A,B]\psi\right|\leq diam(A)*diam(B)

Let A~,B~\widetilde{A},\widetilde{B} be the translation of A,BA,B such that:

|A~|=diam(A),|B~|=diam(B)|\widetilde{A}|=diam(A),|\widetilde{B}|=diam(B)

We have:

|[A,B]ψ|=|[A~,B~]ψ|=|(A~B~B~A~)ψ|sup(|A~B~ψ|,|B~A~ψ|)|A~||B~||ψ|\left|[A,B]\psi\right|=\left|[\widetilde{A},\widetilde{B}]\psi\right|=\left|(\widetilde{A}\widetilde{B}-\widetilde{B}\widetilde{A})\psi\right|\leq\sup(|\widetilde{A}\widetilde{B}\psi|,|\widetilde{B}\widetilde{A}\psi|)\leq|\widetilde{A}||\widetilde{B}||\psi|
|A~||B~||ψ|=diam(A)diam(B)1=diam(A)diam(B)|\widetilde{A}||\widetilde{B}||\psi|=diam(A)*diam(B)*1=diam(A)*diam(B)

Remark 5.1.

Recall the classical uncertainty principle:

12|[A,B]|AB\frac{1}{2}\left|\left\langle[A,B]\right\rangle\right|\leq\bigtriangleup A*\bigtriangleup B

Here A,BA,B is the classical Archimedean Hermite operator, A,B\bigtriangleup A,\bigtriangleup B is the variance of A,BA,B.

6 Further discussion

We refer to [15] for theory of formal group scheme.

Notation.

We assume KK is a field extension of p\mathbb{Q}_{p} with a non-Archimedean norm. Let A,BA,B be the ultrametric KK-Banach algebra(not necessary commutative) with unit. A0,B0A_{0},B_{0} be the unit ball of A,BA,B. Let HomnonArch(A,B)\textup{Hom}_{non-Arch}(A,B) be the set of Norm decreasing KK-Banach algebra morphism:

HomnonArch(A,B)={f is K-Banach Algebra morphism,xA,|f(x)||x|}\textup{Hom}_{non-Arch}(A,B)=\left\{f\textup{ is $K$-Banach Algebra morphism,}\forall x\in A,\left|f(x)\right|\leq\left|x\right|\right\}

It’s obvious to see the norm decreasing morphism ff satisfy: f(A0)B0f(A_{0})\subseteq B_{0}. Let σ\sigma be the Frobenius map σ:xxp\sigma:x\mapsto x^{p}
We can define the set of non-Archimedean Unitary operator of AA:

U(A)={uA;|u|=|u1|=1}U(A)=\left\{u\in A;\left|u\right|=\left|u^{-1}\right|=1\right\}

the set of non-Archimedean orthogonal projection of AA:

Π(A)={πA;|π|=1,π2=π}\Pi(A)=\left\{\pi\in A;\left|\pi\right|=1,\pi^{2}=\pi\right\}

the set of Teichmüller element of AA:

T(A)={xA;k,σk(x)=x,|x|=1}T(A)=\left\{x\in A;\exists k,\sigma^{k}(x)=x,\left|x\right|=1\right\}

the set of Hermite operator of AA:

H(A)={xA;x=i=0xipi,σ(xi)=xi,|xi|1,xixj=xjxi}H(A)=\left\{x\in A;x=\sum_{i=0}^{\infty}x_{i}p^{i},\sigma(x_{i})=x_{i},\left|x_{i}\right|\leq 1,x_{i}x_{j}=x_{j}x_{i}\right\}

For simplifying the issues, we assume the Hermite operator have a common period 11 and bounded.

Proposition 6.1.

The norm decreasing morphism fHomnonArch(A,B)f\in Hom_{non-Arch}(A,B) preserves all the set above. We have:

f(U(A))U(B),f(Π(A))Π(B),f(T(A))T(B),f(H(A))H(B)f(U(A))\subseteq U(B),f(\Pi(A))\subseteq\Pi(B),f(T(A))\subseteq T(B),f(H(A))\subseteq H(B)

What about the Archimedean case?

Notation.

Let A,BA,B be the CC^{*}-Algebra(not necessary commutative) with unit.
Let HomArch(A,B)\textup{Hom}_{Arch}(A,B) be the set of CC^{*}-Algebra morphism. We mean:

HomArch(A,B)={f is -Banach algebra morphism,xA,f(x)=f(x)}\textup{Hom}_{Arch}(A,B)=\left\{f\textup{ is $\mathbb{C}$-Banach algebra morphism,}\forall x\in A,f(x^{\dagger})=f(x)^{\dagger}\right\}

It is known that the morphism between CC^{*}-Algebra are all Norm decreasing.
We can define the set of Unitary operator of AA:

U(A)={uA;u=u1}U(A)=\left\{u\in A;u^{\dagger}=u^{-1}\right\}

the set of Orthogonal projection of AA:

Π(A)={πA;π=π,π2=π}\Pi(A)=\left\{\pi\in A;\pi^{\dagger}=\pi,\pi^{2}=\pi\right\}

the set of Hermite operator of AA:

H(A)={xA;x=x}H(A)=\left\{x\in A;x^{\dagger}=x\right\}
Proposition 6.2.

The CC^{*}-Algebra morphism fHomArch(A,B)f\in Hom_{Arch}(A,B) preserves all the set above. We have:

f(U(A))U(B),f(Π(A))Π(B),f(H(A))H(B)f(U(A))\subseteq U(B),f(\Pi(A))\subseteq\Pi(B),f(H(A))\subseteq H(B)

In Archimedean case(AA is CC^{*}-Algebra) we have x,yH(A)x+yH(A)\forall x,y\in H(A)\Rightarrow x+y\in H(A). In non-Archimedean case we will not get the same result. However it can show that:

Proposition 6.3.

Suppose x,yH(A)x,y\in H(A) such that xy=yxxy=yx, we have:x+y,xyH(A)x+y,xy\in H(A), AA is CC^{*}-Algebra or ultrametric Banach algebra.

We define:h=K[X0,X1,]/(X0pX0,X1pX1,)h=K\left[X_{0},X_{1},...\right]/(X_{0}^{p}-X_{0},X_{1}^{p}-X_{1},...) with a norm:

|iai0,i1,X0i0X1i1|=supωiT(p)|ai0,i1,ω0i0ω1i1|\left|\sum_{i}a_{i_{0},i_{1},...}X_{0}^{i_{0}}X_{1}^{i_{1}}...\right|=\underset{\omega_{i}\in T(\mathbb{Z}_{p})}{\sup}\left|a_{i_{0},i_{1},...}{\omega}_{0}^{i_{0}}{\omega}_{1}^{i_{1}}...\right|

Since f=iai0,i1,X0i0X1i1hf=\sum_{i}a_{i_{0},i_{1},...}X_{0}^{i_{0}}X_{1}^{i_{1}}...\in h is a finite sum, |f|\left|f\right| is well-defined. So we can do completion of Banach algebra: hh\to\mathcal{H}, \mathcal{H} is the ”home of all padicp-adic Hermite operator”. We have:

HomnonArch(,A)1:1H(A)\textup{Hom}_{non-Arch}(\mathcal{H},A)\overset{1:1}{\longleftrightarrow}H(A)

In Archimedean case, the ”home of all Unitary operator” or ”home of all Hermite operator” actually is a scheme. We know the Gelfand representation shows that the unitary operator can be realized as the continous function on S1\mathrm{S}^{1}, the Hermite operator can be realized as the continous function on \mathbb{R}. We have:

HomArch(C(S1),A)1:1U(A)\textup{Hom}_{Arch}(C(\mathrm{S}^{1}),A)\overset{1:1}{\longleftrightarrow}U(A)
HomArch(C(I),A)1:1H1(A),I=[1,1]\textup{Hom}_{Arch}(C(I),A)\overset{1:1}{\longleftrightarrow}H_{\leq 1}(A),I=[-1,1]\subset\mathbb{R}

Which is the classical result. In the final, we define:

K<t,t1>={f;f=i=aiti,|ai|0,i,|f|=sup|ai|}K<t,t^{-1}>=\left\{f;f=\sum_{i=-\infty}^{\infty}a_{i}t^{i},\left|a_{i}\right|\to 0,i\to\infty,\left|f\right|=\sup\left|a_{i}\right|\right\}

We have:

HomnonArch(K<t,t1>,A)1:1U(A)\textup{Hom}_{non-Arch}(K<t,t^{-1}>,A)\overset{1:1}{\longleftrightarrow}U(A)


When we talk about the canonical norm on pn\mathbb{Q}_{p}^{n}, the nature observation is to use the Galois theory. The usual norm on \mathbb{C} can be defined as:

|x|=xx¯=(|detx|)1|:|x|x|=\sqrt{x\bar{x}}=\left(|detx|\right)^{\frac{1}{|\mathbb{C}:\mathbb{R}|}}\ \ \forall x\in\mathbb{C}

We view xx as a \mathbb{R}-Linear transform on \mathbb{C}, the determinant of zz is multipliable and we can check the triangle inequality of zz¯\sqrt{z\bar{z}}. Similarly, suppose K is a finite Galois field extension of p\mathbb{Q}_{p} with degree nn. There exists a canonical multipliable norm:

|x|=(σGal(K|p)σ(x))1|K:p|=(|detx|)1|K:p|xK|x|=(\prod_{\sigma\in\mathrm{Gal}(K|\mathbb{Q}_{p})}\sigma(x))^{\frac{1}{|K:\mathbb{Q}_{p}|}}=\left(|detx|\right)^{\frac{1}{|K:\mathbb{Q}_{p}|}}\ \ \forall x\in K

This multipliable norm is non-Archimedean and unique. In field extension theory of p\mathbb{Q}_{p}, there exists a kind of isotropy Galois extension of p\mathbb{Q}_{p} named unramified extension. Such extension is unique. There exists a p\mathbb{Q}_{p}-base(not unique) of KK named {e1,e2en}\left\{e_{1},e_{2}...e_{n}\right\} such that the norm can be written as a supremum norm:

x=i=1nxiei,xip,xKx=\sum_{i=1}^{n}x_{i}e_{i},x_{i}\in\mathbb{Q}_{p},\forall x\in K
|x|=sup1in(|xi|)|x|=\sup_{1\leq i\leq n}\left(|x_{i}|\right)

There is an another way to show the supremum norm on pn\mathbb{Q}_{p}^{n} is canonical. Let GLn()GL_{n}(\mathbb{R}) be the group of linear invertible transform on n\mathbb{R}^{n}, GLn(p)GL_{n}(\mathbb{Q}_{p}) be the group of linear invertible transform on pn\mathbb{Q}_{p}^{n}. Then we have:

Theorem 6.4.

On()O_{n}(\mathbb{R}) is the maximal compact subgroup of GLn()GL_{n}(\mathbb{R}) .GLn(p)GL_{n}(\mathbb{Z}_{p}) is the maximal compact subgroup of GLn(p)GL_{n}(\mathbb{Q}_{p}). The maximal compact subgroup is unique up to conjugate.

We can prove that On()O_{n}(\mathbb{R}) is the isometric group of n\mathbb{R}^{n} (equiped with the usual quadric norm). GLn(p)GL_{n}(\mathbb{Z}_{p}) is the isometric group of pn\mathbb{Q}_{p}^{n} (equiped with the Non-Archimedean supremum norm). In this sense, the norm above is canonical.

Let Norm(pn)Norm(\mathbb{Q}_{p}^{n}) be the set of Non-Archimedean norms such that they take values in pp^{\mathbb{Z}}. Let Norm(n)Norm(\mathbb{R}^{n}) be the set of Archimedean norms such that they can induce a inner product.

Theorem 6.5.

There is a one to one correspendence:

Norm(pn)1:1GLn(p)/GLn(p)1:1Lat(pn)Norm(\mathbb{Q}_{p}^{n})\overset{1:1}{\longleftrightarrow}GL_{n}(\mathbb{Q}_{p})/GL_{n}(\mathbb{Z}_{p})\overset{1:1}{\longleftrightarrow}Lat(\mathbb{Q}_{p}^{n})
Norm(n)1:1GLn(n)/On()1:1Sym+(n)Norm(\mathbb{R}^{n})\overset{1:1}{\longleftrightarrow}GL_{n}(\mathbb{R}^{n})/O_{n}(\mathbb{R})\overset{1:1}{\longleftrightarrow}Sym_{+}(\mathbb{R}^{n})

Lat(pn)Lat(\mathbb{Q}_{p}^{n}) is the set of lattice in pn\mathbb{Q}_{p}^{n}. A lattice in pn\mathbb{Q}_{p}^{n} is a p\mathbb{Z}_{p}- submodule LL such that:ppL=pn\mathbb{Q}_{p}\otimes_{\mathbb{Z}_{p}}L=\mathbb{Q}_{p}^{n} and LL is compact. (So it is isomorphic to pn\mathbb{Z}_{p}^{n}) Sym+(n)Sym_{+}(\mathbb{R}^{n}) is the set of positive definite symmetric matrix of GLn(n)GL_{n}(\mathbb{R}^{n}).

References

  • [1] Vladimir G. Berkovich. Spectral theory and analytic geometry over non-Archimedean fields, volume 33 of Mathematical Surveys and Monographs. American Mathematical Society, Providence, RI, 1990.
  • [2] Alexandru Buium, Claire Ralph, and Santiago Simanca. Arithmetic differential operators on p\mathbb{Z}_{p}. Journal of Number Theory, 131, 09 2008.
  • [3] Alain Connes. The Witt construction in characteristic one and quantization. In Noncommutative geometry and global analysis, volume 546 of Contemp. Math., pages 83–113. Amer. Math. Soc., Providence, RI, 2011.
  • [4] B. Dragovich, A. Yu. Khrennikov, S. V. Kozyrev, and I. V. Volovich. On pp-adic mathematical physics. p-Adic Numbers Ultrametric Anal. Appl., 1(1):1–17, 2009.
  • [5] Alain Escassut. Analytic elements in pp-adic analysis. World Scientific Publishing Co., Inc., River Edge, NJ, 1995.
  • [6] Steven S. Gubser, Johannes Knaute, Sarthak Parikh, Andreas Samberg, and Przemek Witaszczyk. pp-adic AdS/CFT. Comm. Math. Phys., 352(3):1019–1059, 2017.
  • [7] Anatoly N. Kochubei. Non-Archimedean normal operators. J. Math. Phys., 51(2):023526, 15, 2010.
  • [8] Anatoly N. Kochubei. On some classes of non-Archimedean operator algebras. In Advances in ultrametric analysis, volume 596 of Contemp. Math., pages 133–148. Amer. Math. Soc., Providence, RI, 2013.
  • [9] S. Ludkovsky and B. Diarra. Spectral integration and spectral theory for non-Archimedean Banach spaces. Int. J. Math. Math. Sci., 31(7):421–442, 2002.
  • [10] Lance Miller and Alexandru Buium. Arithmetic differential geometry in the arithmetic pde setting, i: connections. 02 2022.
  • [11] Bjorn Poonen. pp-adic interpolation of iterates. Bull. Lond. Math. Soc., 46(3):525–527, 2014.
  • [12] Joseph Rabinoff. The Theory of Witt Vectors. arXiv e-prints, page arXiv:1409.7445, September 2014.
  • [13] Alain M. Robert. A course in pp-adic analysis, volume 198 of Graduate Texts in Mathematics. Springer-Verlag, New York, 2000.
  • [14] J.-P. Serre. A course in arithmetic. Graduate Texts in Mathematics, No. 7. Springer-Verlag, New York-Heidelberg, 1973. Translated from the French.
  • [15] Neil P. Strickland. Formal schemes and formal groups. In Homotopy invariant algebraic structures (Baltimore, MD, 1998), volume 239 of Contemp. Math., pages 263–352. Amer. Math. Soc., Providence, RI, 1999.
  • [16] J. T. Tate. Fourier analysis in number fields, and Hecke’s zeta-functions. In Algebraic Number Theory (Proc. Instructional Conf., Brighton, 1965), pages 305–347. Thompson, Washington, D.C., 1967.
  • [17] V. S. Vladimirov and I. V. Volovich. pp-adic quantum mechanics. Comm. Math. Phys., 123(4):659–676, 1989.