Skip to main content

Section 3.2 \(p\)–adic \(L\)–functions

Historically, Number Theorists have primarily considered two approaches to construct \(p\)-adic \(L\)–functions. The first and relatively more natural construction was put forward by Kubota-Leopoldt by interpolating special values of Dirichlet \(L\)–functions, as mentioned at the onset of this chapter.
The second approach–Iwasawa’s groundbreaking work [23] showing deep connections between \(L\)–functions and algebraic properties of cyclotomic fields–uses the Iwasawa algebra \(\Lambda(\Z_p)\) to construct \(p\)-adic \(L\)–functions. This chapter will prove that the \(p\)-adic \(L\)–functions constructed in this algebraic method are the same as the Kubota-Leopoldt \(p\)-adic \(L\)–functions obtained by analytic methods.

Subsection 3.2.1 Motivation

As explained by the creator himself in his beautiful article Analogies Between Number Fields and Function Fields [22], Iwasawa was partly motivated by the Weil Conjectures. This section describes the motivation behind Iwasawa’s construction of \(p\)-adic \(L\)–function following [22].
Algebraic Geometry is among the most important sources of Zeta functions. The Hasse–Weil Zeta function (see [44, Chapter V, §2] for the definition) attached to a projective nonsingular curve \(\mathscr{C}\text{,}\) defined over a finite field of characteristic \(p\text{,}\) stores the point-counts of \(\mathscr{C}\) over all finite fields of characteristic \(p\text{.}\) The Jacobian variety \(\mathbf{J}(\mathscr{C})\) of \(\mathscr{C}\) is a \(g\) dimensional abelian variety, where \(g\) is the genus of \(\mathscr{C}\text{.}\) The abelian group structure of \(\mathbf{J}(\mathscr{C})\) allows us to consider its \(p\)-primary part \(\mathbf{J}(\mathscr{C})[p]\text{,}\) i.e., the subgroup of \(\mathbf{J}(\mathscr{C})\) consisting of elements of \(p\)-power order. It turns out that as a abelian group \(\mathbf{J}(\mathscr{C})[p]\) has the following simple structure:
\begin{gather*} \mathbf{J}(\mathscr{C})[p]\cong (\Q_p/\Z_p)^{2g}=\underbrace{\Q_p/\Z_p\oplus\cdots\oplus\Q_p/\Z_p}_{\text{$2g$ copies}}. \end{gather*}
Therefore, the ring of endomorphisms of \(\mathbf{J}(\mathscr{C})[p]\) is isomorphic to \(M_{2g}(\Z_p)\text{.}\) Any algebraic map \(\tau:\mathscr{C}\longrightarrow\mathscr{C}\) induces an endomorphism of \(\mathbf{J}(\mathscr{C})[p]\) and hence we may attach to \(\tau\text{,}\) a matrix \(M(\tau)\in M_{2g}(\Z_p)\text{.}\) Consequently, we may attach a matrix \(M(\varphi)\) to the Frobenius endomorphism of \(\mathscr{C}\text{.}\) A special case of the Weil Conjectures [44, Chapter V, Theorem 2.2 in §2], proved by Weil himself, satisfies the property that the quotient of Hasse–Weil Zeta functions of \(\mathscr{C}\) and a genus zero curve is equal to the characteristic polynomial of \(M(\varphi)\text{.}\)
Analogies between the arithmetic of number fields and function fields (for an exposition on the arithmetic over function fields we refer to [40]) naturally point to the question: is there an analogue of this construction in the number field case? To answer this question, one needs to find an appropriate analogue of \(\mathbf{J}(\mathscr{C})\) in the number field case, i.e., when \(\mathscr{C}\) is replaced by a number field \(K\text{.}\) Note that \(\mathbf{J}(\mathscr{C})[p]\) is the group of degree 0 divisor classes in the field of rational functions of \(\mathscr{C}\text{.}\) So, the proper analogue of \(\mathbf{J}(\mathscr{C})\) in the realm of number fields is the \(p\)-primary part of the class group \(\Cl{K}\) of \(K\text{.}\) Recall the ideal class group \(\Cl{K}\) is defined to be the quotient group \(J_K/P_K\text{,}\) where \(P_K\) is the group of fractional ideals of \(K\) and \(I_K\) is its normal subgroup of principal fractional ideals [34, Chapter I, p. 22]. This analogy motivated Iwasawa to build on his work [21] on the growth of the \(p\)-primary part of the class groups in the \(\Z_p\)-tower.

Subsection 3.2.2 Iwasawa’s construction

Fix an odd prime \(p\) (arbitrary but fixed). We start by recalling some standard notations we will use from this section onwards: for an integer \(n\geq0\)
\begin{align*} K_n \amp \colonequals\Q(\mu_{p^{n+1}})\\ K_\infty \amp \colonequals\bigcup_{n\geq0}\Q(\mu_{p^{n+1}}) \end{align*}
Therefore,
\begin{align*} \Gal(K_n/\Q) \amp \cong(\Z/p^{n+1}\Z)^\times\\ \Gal(K_\infty/\Q) \amp \cong\varprojlim_{n\geq0}(\Z/p^{n+1}\Z)^\times\cong\Z_p^\times. \end{align*}
To be more precise, the second isomorphism is given via \(a\mapsto\sigma_a\text{,}\) for \(a\in\Z_p^\times\text{,}\) where, for \(a=\sum_{k=0}^\infty a_kp^k\text{,}\) and a \(p^n\)-th root of unity \(\zeta\text{,}\) we have,
\begin{gather} \sigma_a(\zeta)\colonequals\prod_{k=0}^\infty\zeta^{a_kp^k}.\tag{3.7} \end{gather}
The product in equation (3.7) is finite as \(\zeta^{p^k}=1\) for all \(k\geq n\text{.}\)
Since \(p\) is an odd prime, \((\Z/p^{n+1}\Z)^\times\) is cyclic [19, Proposition 4.1.3] of order \(\varphi(p^{n+1})=p^n(p-1)\) and has a unique [39, Theorem 2.34] subgroup \(H_n\subseteq(\Z/p^{n+1}\Z)^\times\) which is cyclic of order \(p-1\) (and hence normal). By the Fundamental Theorem of Galois Theory(FTGT) [28, Chapter VI, Theorem 1.1], there exists a unique subfield of \(K_n\) corresponding to \(H_n\text{,}\) which we denote by \(F_n\text{.}\) Therefore, once again by FTGT,
\begin{gather} \Gal(F_n/\Q)\cong(\Z/p^{n+1}\Z)^\times/H_n\cong\Z/p^n\Z.\tag{3.8} \end{gather}
Let \(F_\infty\) be the compositum of all \(F_n\text{,}\) \(n\geq0\text{.}\) Then
\begin{gather*} \Gal(F_\infty/\Q)\cong\varprojlim_{n\geq1}\Gal(F_n/\Q)\cong\varprojlim_{n\geq1}\Z/p^n\Z\cong\Z_p. \end{gather*}
The infinite extension \(F_\infty\) of \(\Q\) is an example of a cyclotomic \(\Z_p\)-extension. It is well-known [47, p. 118] that
\begin{gather} \Z_p^\times\cong(\Z/p\Z)^\times\times(1+p\Z_p)\cong(\Z/p\Z)^\times\times\Z_p,\tag{3.9} \end{gather}
Therefore, for any \(a\in\Z_p^\times\text{,}\) there exists a unique \(\varphi(p)=(p-1)^{\rm th}\) of unity \(\omega(a)\in\Z_p\) and \([a]\in1+p\Z_p\text{,}\) such that \(\omega(a)\equiv a\pmod{p}\) and \([a]\equiv1\pmod{p}\text{.}\) The map \(a\mapsto\omega(a)\) is a \(p\)-periodic completely multiplicative function; hence, can be thought of as a Dirichlet character \(\omega:(\Z/p\Z)^\times\longrightarrow\C^\times\) modulo \(p\) (by identifying the roots of unity in \(\Z_p\) with the corresponding complex roots of unity). The character \(\omega\) is known as the Teichmüller character. The isomorphism (3.9) is given as follows
\begin{gather*} a\mapsto(\omega(a)\pmod{p},[a])\mapsto\bparen{\omega(a)\pmod{p},\frac{\log_p([a])}{\log_p(1+p)}}. \end{gather*}
For more more details refer to [47, Chapter 7] and [34, Chapter II]. Note that \(\omega\) and \(F_n\) are crucial for this thesis.
Let \(\Delta\colonequals\Gal(K_0/\Q)\text{.}\) Therefore, \(\Delta\cong(\Z/p\Z)^\times\text{.}\) Take \(\Gamma=\Gal(K_\infty/K)\text{.}\) By FTGT,
\begin{gather*} \Gamma=\Gal(K_\infty/K_0)=\varprojlim_{n\geq1}\Gal(K_n/K_0)\cong\varprojlim_{n\geq1}\Z/p^n\Z\cong\Z_p. \end{gather*}
Therefore, by the isomorphism (3.9),
\begin{gather} \Gal(K_\infty/\Q)\cong\Delta\times\Gamma.\tag{3.10} \end{gather}
To be more precise, \(\Gamma\) is the multiplicative group \(1+p\Z_p\) of principal units in \(\Z_p\text{,}\) which is isomorphic to the additive group \(\Z_p\text{.}\) We can take \(\gamma=1+p\) to be a topological generator of \(\Gamma\text{,}\) i.e., the cyclic multiplicative subgroup \((1+p)^{\Z}\) is dense in \(1+p\Z_p\text{.}\) We sometimes write \(1+p\Z_p\) as \((1+p)^{\Z_p}\) to emphasize the fact that \(1+p\Z_p\) is a multiplicative topological cyclic group generated by \(\gamma=1+p\text{.}\) The closed subgroups of \(1+p\Z_p\) are the subgroups of elements that fix \(K_n\) and hence, are \((1+p)^{p^n\Z_p}\cong\Gamma^{p^n}\text{.}\) Consequently,
\begin{gather*} \Gal(K_n/K_0)\cong\Gamma/\Gamma^{p^n}=\Gamma_n. \end{gather*}
Therefore,
\begin{gather} \Gal(K_n/\Q)\cong\Delta\times\Gamma_n.\tag{3.11} \end{gather}
The following definition divides certain Dirichlet characters into categories relevant to Iwasawa’s construction and the main result of this thesis:

Definition 3.2. Characters of first and second kind.

Let \(p\) be an odd prime and \(\chi\) be a Dirichlet character modulo \(p^{n+1}\) for some \(n\geq0\) (in other words \(\chi\in\mathcal{X}(p^{n+1})\text{,}\) according to Lemma 2.7). Then \(\chi\) can be considered a character of \(\Gal(K_n/\Q)\text{.}\) By the decomposition (3.11), \(\chi\) can be uniquely written as
\begin{gather} \chi=\theta\psi,\tag{3.12} \end{gather}
where \(\theta\in\widehat{\Delta}\) and \(\psi\in\widehat{\Gamma_n}\text{.}\) We call \(\theta\) a character of the first kind and \(\psi\) a character of the second kind.
The decomposition (3.12) corresponds to the decomposition of the character group
\begin{gather*} \widehat{\Delta\times\Gamma_n}\cong\widehat{\Delta}\times\widehat{\Gamma_n}. \end{gather*}
By [47, Lemma 3.1], \(\widehat\Delta\) has order \(p-1\text{.}\) Therefore, \(\theta\) is either trivial or of order \(p-1\text{,}\) i.e., \(\theta^{p-1}=1\text{,}\) and has conductor \(p\text{.}\) Observe that \(\widehat{\Gamma_n}\subset\widehat{(\Z/p^{n+1}\Z)^\times}\text{.}\) Therefore, \(\psi\in\widehat{\Gamma_n}\) is either trivial or its conductor divides \(p^{n+1}\text{.}\) But note that \(f_\psi\neq p\text{.}\) Otherwise, \(\psi\) would be a character of \((\Z/p\Z)^\times\) and hence we would get \(\chi^{p-1}=1\text{.}\) But, by [47, Lemma 3.1] again, the order of \(\widehat{\Gamma_n}\) is \(p^n\) and hence \(\psi^{p^n}=1\text{.}\) This is absurd since \(\gcd(p-1,p^n)=1\text{,}\) which forces \(\psi=1\text{;}\) a contradiction. Consequently, the conductor of \(\psi\) is of the form \(p^j\) for \(j\geq2\text{.}\) The characters of the first kind are associated with \(K_0\) (i.e., they may be regarded as characters of \(\Gal(K_0/\Q)\)) and those of the second kind are associated with \(F_n\)–the unique cyclotomic extension of degree \(p^n\) over \(\Q\text{.}\)
We follow the approach outlined in [47, Chapter 7, §7.2]. The idea is to consider certain elements of the group rings \(\mathscr{O}[\Gamma_n]\) (where \(\mathscr{O}\) is an appropriate ring extension of \(\Z_p\) as mentioned at the beginning of this chapter), called the Stickelberger elements [47, Chapter 6, §6.2]–elements that annihilate the ideal class group (Stickelberger’s Theorem [47, Theorem 6.10]). It turns out they form a compatible system and hence define an element in the inverse limit
\begin{gather*} \varprojlim_{n\geq1}\mathscr{O}[\Gamma_n]\cong\Lambda(\Z_p)\cong\mathscr{O}\llbracket T\rrbracket. \end{gather*}
The corresponding power series in \(\mathscr{O}\llbracket T\rrbracket\) will give us the \(p\)-adic \(L\)–functions.
Let \(\chi=\theta\psi\) be an even character, i.e., \(\chi(-1)=1\text{.}\) We require the following simple Lemma to proceed, which seemed to be silently used many times in standard texts but we couldn’t find a specific reference:

Proof.

Let \(\psi\) be a character of the second kind. If \(\psi\) is trivial there’s nothing to prove. Otherwise, there exists \(\nu\geq1\) such that \(\psi^{p^\nu}=1\text{.}\) If \(\psi\) were odd, then \(1=\chi(-1)^{p^\nu}=(-1)^{p^\nu}=-1\) since \(p\) is odd, which is a contradiction.
Suppose, for the sake of contradiction, that \(\omega(-1)=1\text{.}\) Let \(a\) be a primitive root modulo \(p\text{,}\) i.e., the smallest positive integer \(n\) such that \(a^n\equiv1\pmod{p}\) is \(n=p-1\) (such an \(a\) exists by [19, Proposition 4.1.3]). Therefore,
\begin{gather*} a^{(p-1)/2}\equiv-1\pmod{p}\quad\text{and}\quad \omega(a)\equiv a\pmod{p}. \end{gather*}
Combining these, we have
\begin{gather*} -1\equiv a^{(p-1)/2}\equiv(\omega(a))^{(p-1)/2}=\omega(a^{(p-1)/2})\equiv\omega(-1)=1\pmod{p}, \end{gather*}
which is absurd since \(p\) is odd and hence \(p\nmid 2\text{.}\) This completes the proof.
By Lemma 3.3, \(\theta^*\coloneqq\omega\theta^{-1}\) is odd. For any \(a\in\Z_p^\times\text{,}\) the corresponding element \(\sigma_a\in\Gal(K_n/\Q)\) (as defined in equation (3.7)) decomposes as \(\sigma_a=\delta(a)\gamma_n(a)\) according to the decomosition (3.11). Therefore, \(\delta(a)\in\Delta\) and \(\gamma_n(a)\in\Gamma_n\text{.}\) Define the following elements
\begin{align} \xi_n \amp \colonequals-\frac{1}{p^{n+1}}\sum_{\substack{0\lt a\lt p^{n+1}\\\gcd(a,p)=1}}a\delta(a)^{-1}\gamma_n(a)^{-1},\tag{3.13}\\ \eta_n \amp \colonequals(1-(1+p)\gamma_n(1+p)^{-1})\xi_n.\tag{3.14} \end{align}
The values of \(\theta\) are \((p-1)^{\rm th}\) roots of unity. Since \(\Q_p\) contains \((p-1)^{\rm th}\) roots of unity, \(\xi_n\in\Q_p[\Gamma_n]\) and \(\eta_n\in\Z_p[\Delta\times\Gamma_n]\text{.}\) Note that \(a\mapsto\sigma_a=\delta(a)\gamma_n(a)\) is an isomorphism \((\Z/p^{n+1}\Z)^\times\longrightarrow\Delta\times\Gamma_n\) of multiplicative groups. Therefore, \(\sigma_{a(1+p)}=\sigma_a\sigma_{1+p}\text{.}\) Since \(1+p\equiv1\pmod{p}\) and \(\delta(1+p)\in\Delta\cong(\Z/p\Z)^\times\text{,}\) we have \(\delta(1+p)=\delta(1)=1\text{.}\) Therefore,
\begin{gather} \sigma_{a(1+p)}^{-1}=\sigma_a\gamma_n(1+p)^{-1}.\tag{3.15} \end{gather}
Using Euclid’s Division Algorithm, one can easily check that
\begin{gather} \frac{b\pmod{p^{n+1}}}{p^{n+1}}=\bbrace{\frac{b}{p^{n+1}}},\tag{3.16} \end{gather}
where \(b\) is any positive integer and \(\{x\}\) denotes the fractional part \(\{x\}=x-\lfloor x\rfloor\) of a real number \(x\text{.}\) Note that \(a\mapsto a(1+p)\) is an isomorphism since \(\gcd(1+p,p^n)=1\text{.}\) Combining all these we get
\begin{align*} \eta_n \amp =(1-(1+p)\gamma_n(1+p)^{-1})\xi_n\\ \amp =-\sum_{a\in(\Z/p^{n+1}\Z)^\times}\frac{a}{p^{n+1}}\sigma_a^{-1}+\sum_{a\in(\Z/p^{n+1}\Z)^\times}\frac{a(1+p)}{p^{n+1}}\sigma_a^{-1}\gamma_n(1+p)^{-1}\\ \amp =\sum_{a\in(\Z/p^{n+1}\Z)^\times}-\frac{a(p+1)\pmod{p^{n+1}}}{p^{n+1}}\sigma_{a(1+p)}^{-1}+\frac{a(1+p)}{p^{n+1}}\sigma_a^{-1}\gamma_n(1+p)^{-1} \end{align*}
changing variable \(a\mapsto a(1+p)\)
\begin{align*} \amp =\sum_{a\in(\Z/p^{n+1}\Z)^\times}-\bbrace{\frac{a(1+p)}{p^{n+1}}}\sigma_a^{-1}\gamma_n(1+p)^{-1}\frac{a(1+p)}{p^{n+1}}\sigma_a^{-1}\gamma_n(1+p)^{-1} \end{align*}
using identities (3.15) and (3.16)
\begin{align} \amp =-\sum_{a\in(\Z/p^{n+1}\Z)^\times}\bparen{\bbrace{\frac{a(1+p)}{p^{n+1}}}-\frac{a(p+1)}{p^{n+1}}}\sigma_a^{-1}\gamma(1+p)^{-1}.\tag{3.17} \end{align}

Definition 3.4. Idempotent of a character [47, Chapter 6, §6.3].

Let \(G\) be a finite abelian group and \(\chi\in\widehat{G}\) be a character. The associated idempotent element \(\varepsilon_\chi\) is defined as the following element in \(\overline{\Q}[G]\)
\begin{gather} \varepsilon_\chi\colonequals\frac{1}{\abs{G}}\sum_{\sigma\in G}\chi(\sigma)\sigma^{-1}.\tag{3.18} \end{gather}
Elementary computations show that \(\varepsilon_\chi\) is indeed an idempotent element in the group ring \(\overline{\Q}[G]\text{,}\) i.e., \(\varepsilon_\chi^2=\varepsilon_\chi\text{.}\) We refer to [33, p. 8] for a discussion on idempotents attached to characters of a finite group. In the following lemma, we will prove two basic properties of these idempotents:

Proof.

By Definition 3.4, we have
\begin{align*} \varepsilon_\chi g \amp =\bparen{\frac{1}{\abs{G}}\sum_{\sigma\in G}\chi(\sigma)\sigma^{-1}}g\\ \amp =\frac{1}{\abs{G}}\sum_{\sigma\in G}\chi(\sigma)\sigma^{-1}\\ \amp =\frac{1}{\abs{G}}\sum_{h\in G}\chi(gh)h^{-1} \end{align*}
changing variable \(h^{-1}=\sigma^{-1}g\)
\begin{align*} \amp =\chi(g)\varepsilon_\chi. \end{align*}
For the second identity, observe that
\begin{align*} \varepsilon_\chi\varepsilon_\psi \amp =\varepsilon_\chi\bparen{\frac{1}{\abs{G}}\sum_{\kappa\in G}\psi(\kappa)\kappa^{-1}}\\ \amp =\frac{1}{\abs{G}}\sum_{\kappa\in G}\psi(\kappa)\varepsilon_{\chi}\kappa^{-1}\\ \amp =\frac{1}{\abs{G}}\sum_{\kappa\in G}\psi(\kappa)\chi(\kappa^{-1})\varepsilon_\chi\\ \amp =\varepsilon_\chi\bparen{\frac{1}{\abs{G}}\sum_{\kappa\in G}\psi(\kappa)\overline{\chi}(\kappa)}\\ \amp =0, \end{align*}
where the last equation follows from [42, Theorem 3, p. 15] as \(\chi\neq\psi\text{.}\)
In what follows, we attach two elements \(\xi_n(\theta),\eta_n(\theta)\in\Z_p[\Gamma_n]\) to a character \(\chi=\theta\psi\text{,}\) for all sufficiently large \(n\text{.}\) The choice of this notation [47, Chapter 7, §7.2] emphasizes on their dependence on \(\theta\text{.}\) We will see that \(\{\xi_n(\theta)\},\{\eta_n(\theta)\}\) form a compatible system and will give us the \(p\)–adic \(L\)–function. By Lemma 3.5,
\begin{align*} \varepsilon_{\theta^*}\xi_n \amp =\varepsilon_{\theta^*}\bparen{-\frac{1}{p^{n+1}}\sum_{\substack{0\lt a\lt p^{n+1}\\\gcd(a,p)=1}}a\delta(a)^{-1}\gamma_n(a)^{-1}}\\ \amp =-\frac{1}{p^{n+1}}\sum_{\substack{0\lt a\lt p^{n+1}\\\gcd(a,p)=1}}a\varepsilon_{\theta^*}\delta(a)^{-1}\gamma_n(a)^{-1}= \xi_n(\theta^*)\varepsilon_{\theta^*}, \end{align*}
where
\begin{gather} \xi_n(\theta)\colonequals-\frac{1}{p^{n+1}}\sum_{\substack{0\lt a\lt p^{n+1}\\\gcd(a,p)=1}}a\theta\omega^{-1}(a)\gamma_n(a)^{-1}.\tag{3.19} \end{gather}
Similarly, we can use Lemma 3.4 to show that \(\varepsilon_{\theta^*}\eta_n=\eta_n(\theta)\varepsilon_{\theta^*}\text{,}\) where
\begin{align} \eta_n(\theta) \amp \colonequals(1-(1+p)\gamma(1+p)^{-1})\xi_n(\theta)\notag\\ \amp =\sum_{a\in(\Z/p^{n+1}\Z)^\times}\bparen{\frac{a(p+1)}{p^{n+1}}-\bbrace{\frac{a(1+p)}{p^{n+1}}}}\times\theta^*(a)\gamma_n(a)^{-1}\gamma_n(1+p)^{-1}.\tag{3.20} \end{align}
using the formula (3.17) The fact that \(\theta^*\) is odd implies that \(\eta_n(\theta)\in\Z_p[\Gamma_n]\) [47, Proposition 7.6 (a)].
Recall that a Dirichlet character \(\chi\) is a character modulo \(N\) for any multiple \(N\) of \(f_\chi\text{,}\) the conductor of \(\chi\text{.}\) For some \(n\geq1\text{,}\) if we start with a character \(\chi\) of \(\Delta\times\Gamma_n\text{,}\) then, for all \(m\geq n\text{,}\) \(\chi\) can be regarded as a character of \(\Delta\times\Gamma_m\text{.}\) Therefore, whenever \(\xi_n(\theta),\eta_n(\theta)\) are defined, we can also define \(\xi_m(\theta),\eta_m(\theta)\) for all \(m\geq n\text{.}\) It turns out that the sequences \(\{\xi_n(\theta)\},\{\eta_n(\theta)\}\) are compatible with respect to the natural maps \(\Z_p[\Gamma_m]\longrightarrow\Z_p[\Gamma_n]\text{.}\) As a result, their limits lie in the Iwasawa Algebra \(\Lambda(\Z_p)\text{.}\) We state the following Proposition without proof, which will be used in the proof of the main theorem of this chapter.

Proof.

Definition 3.7.

For \(\theta=1\text{,}\) define
\begin{gather} f(T,1)\colonequals\frac{g(T,1)}{1-\frac{1+p}{1+T}}\tag{3.21} \end{gather}

Proof.

When \(\theta=1\text{,}\) it follows from Definition 3.7. Let \(\theta\neq1\text{.}\) Recall that \(1+p\) is a topological generator of the multiplicative group \(1+p\Z_p\) of principal units in \(\Z_p\text{.}\) The image of \(1+p\) in the \(\Gamma/\Gamma^{p^n}=\Gamma_n\) corresponds to \(\gamma_n(1+p)\text{;}\) hence \(\lim\gamma_n(1+p)=\gamma\text{,}\) which corresponds to \(1+T\text{.}\) Therefore,
\begin{gather*} \lim\bparen{1-(1+p)\gamma_n(1+p)^{-1}}\longleftrightarrow1-\frac{1+p}{1+T}=h(T) \end{gather*}
as defined in equation (3.23). Recall the definitions of \(\xi_n(\theta)\) and \(\eta_n(\theta)\) in the equations (3.19) and (3.17), which imply, by Proposition 3.6
\begin{align*} f(T,\theta) \amp \longleftrightarrow \lim\xi_n(\theta)\\ \amp =\lim\frac{\eta_n(\theta)}{\bparen{1-(1+p)\gamma_n(1+p)^{-1}}}\\ \amp \longleftrightarrow\frac{g(T,\theta)}{h(T)}. \end{align*}
This completes the proof as the correspondence \(\gamma\longleftrightarrow1+T\) induces an isomorphism \(\Lambda(\Z_p)\cong\Z_p\llbracket T\rrbracket\) by Theorem E.
Now, we are ready to state and prove the main theorem of this chapter, which is one of the groundbreaking contributions of Iwasawa in the study of cyclotomic fields. It provides an alternate realization of the \(p\)–adic Dirichlet \(L\)–functions, denoted by \(L_p(s,\chi)\text{,}\) which were first constructed by Kubota-Leopoldt (see [47, Chapter 5, §5.2]). They proved that for \(\chi\neq1\text{,}\) \(L_p(s,\chi)\) is analytic on \(\{s\in\C_p\colon\abs{s}_p\lt p^{1-1/(p-1)}\}\text{,}\) and for \(\chi=1\text{,}\) \(L_p(s,1)=\zeta_p(s)\) is meromorphic with a simple pole at \(s=1\) [47, Theorem 5.11]. We decided to include the proof, taken from [47], due to its significance, in general, and in the proof of the main theorem of this thesis.

Proof.

Observe that \(\abs{s}_p\lt p^{1-1/(p-1)}\) implies
\begin{align} \abs{\zeta_\psi(1+p)^s-1}_p \amp =\abs{\zeta_\psi\exp(s\log_p(1+p))-1}_p\tag{3.25}\\ \amp =\abs{\zeta_\psi\sum_{n=0}^\infty\frac{(s\log(1+p))^n}{n!}-1}_p\tag{3.26}\\ \amp =\abs{\zeta_\psi-1+s\sum_{n=1}\frac{s^{n-1}\log_p^n(1+p)}{n!}}_p.\tag{3.27} \end{align}
Since \(\zeta_\psi=\psi(1+p)^{-1}\) is a root of unity of \(p\)-power order, \(\zeta_\psi-1\) is a uniformizer associated with a local field \(\Q_p(\mu_{p^\nu})\) for some \(\nu\geq1\text{.}\) Therefore, using equation (3.27), one can easily verify that \(|\zeta_\psi(1+p)^s-1|\lt 1\text{.}\) Therefore, the right-hand side of equation (3.24) converges and is an analytic function of \(s\text{.}\) Since \(\Z\hookrightarrow\Z_p\) is dense, it suffices to derive the formula (3.24) for \(s=1-m\text{,}\) where \(m\in\N\text{.}\)
We will work with \(\eta_n(\theta)\) and \(g(T,\theta)\) as \(f(T,\theta)\) is merely a normalization of \(g(T,\theta)\) and in all cases, \(\eta_n(\theta)\) and \(g(T,\theta)\) have coefficients in \(\Z_p\text{.}\) Recall the isomorphism \(\vartheta:1+p\Z_p\xrightarrow{\sim}\Z_p\)
\begin{gather*} a\mapsto\vartheta(a)=\frac{\log_p[a]}{\log_p(1+p)}. \end{gather*}
Therefore, \(\log_p(1+p)^{\vartheta(a)}=\log_p[a]\) and hence \(\gamma_n(a)=\gamma(1+p)^{\vartheta(a)}\text{.}\) Let \(\Z_p[T]\) be the ring of polynomials with coefficients in \(\Z_p\text{.}\) Then
\begin{gather*} \Z_p[\Gamma_n]\cong\Z_p[T]/\ideal{(1+T)^{p^n}-1}, \end{gather*}
where the isomorphism is induced by \(\gamma\pmod{\Gamma^{p^n}}\mapsto 1+T\pmod{(1+T)^{p^n}-1}\text{.}\) Since \(\gamma_n(1+p)\in\Gamma/\Gamma^{p^n}\) corresponds to \(1+T\pmod{(1+T)^{p^n}-1}\text{,}\) \(\gamma_n(a)\) corresponds to \((1+T)^{\vartheta(a)}\pmod{(1+T)^{p^n}-1}\text{.}\) As a result, by Proposition 3.6 and the formula (3.20), we get that
\begin{gather} g(T,\theta)\equiv\sum_{a\in(\Z/p^{n+1}\Z)^\times}\bparen{\frac{a(p+1)}{p^{n+1}}-\bbrace{\frac{a(1+p)}{p^{n+1}}}}\times\theta^*(a)(1+T)^{-\vartheta(a)-1}\notag\\ \pmod{(1+T)^{p^n}-1}.\tag{3.28} \end{gather}
Applying Euclid’s Division Algorithm to \(a(1+p)\) when divided by \(p^{n+1}\text{,}\) let
\begin{gather} a(1+p)=b_ap^{n+1}+r_a,\quad 0\leq r_a\lt p^{n+1}.\tag{3.29} \end{gather}
Therefore,
\begin{gather*} \frac{a(p+1)}{p^{n+1}}-\bbrace{\frac{a(1+p)}{p^{n+1}}}=b_a \end{gather*}
for all \(a\in(\Z/p^{n+1}\Z)^\times\text{.}\) Also,
\begin{gather*} \vartheta(a(1+p))=\frac{\log_p[a(1+p)]}{\log_p(1+p)}=\frac{\log_p[a]}{\log_p(1+p)}+1=\vartheta(a)+1 \end{gather*}
and \(a(1+p)\equiv r_a\pmod{p^{n+1}}\) implies \(\vartheta(a(1+p))\equiv\vartheta(r_a)\pmod{p^{n+1}}\text{.}\) Therefore,
\begin{gather*} -\vartheta(a)-1\equiv-\vartheta(r_a)\pmod{p^{n}}. \end{gather*}
Consequently, the congruence (3.28) transforms to
\begin{gather} g(T,\theta)\equiv\sum_{a\in(\Z/p^{n+1}\Z)^\times}b_a\theta^*(r_a)(1+T)^{-\vartheta(r_a)}\pmod{(1+T)^{p^n}-1}.\tag{3.30} \end{gather}
For two positive integers \(m\) and \(n\text{,}\) where \(n\) is sufficiently large (we need to choose \(n\) large enough so that the character \(\chi\) we started with may be regarded as a character of \(\Gal(K_n/\Q)\text{,}\) in other words, \(f_\chi\mid p^{n+1}\)). Plugging in \(T=\zeta_\psi(1+p)^{1-m}-1\) in equation (3.30), we get
\begin{gather} g(\zeta_\psi(1+p)^{1-m}-1,\theta)\equiv\sum_{a\in(\Z/p^{n+1}\Z)^\times}b_a\theta^*(r_a)(\zeta_{\psi}(1+p)^{1-m})^{-\vartheta(r_a)}\notag\\ \pmod{(\zeta_\psi(1+p)^{1-m})-1}.\tag{3.31} \end{gather}
Since \(\psi\in\widehat{\Gamma_n}\) and \(\#\widehat{\Gamma_n}=p^n\text{,}\)
\begin{gather*} \zeta_\psi^{p^n}=\psi(1+p)^{-p^n}=1. \end{gather*}
Therefore,
\begin{align*} (\zeta_\psi(1+p)^{1-m})^{p^n}-1 \amp =(1+p)^{(1-m)p^n}-1\\ \amp \equiv0\pmod{p^{n+1}}, \end{align*}
which implies that
\begin{gather} g(\zeta_\psi(1+p)^{1-m}-1,\theta)\equiv\sum_{a\in(\Z/p^{n+1}\Z)^\times}b_a\theta^*(r_a)(\zeta_{\psi}(1+p)^{1-m})^{-\vartheta(r_a)}\notag\\ \pmod{p^{n+1}}.\tag{3.32} \end{gather}
Now, \(\log_p(1+p)^{\vartheta(r_a)}=[r_a]\) and hence \(\zeta_\psi^{-\vartheta(r_a)}=\psi(1+p)^{\vartheta(r_a)}=\psi(r_a)\text{.}\) Therefore,
\begin{gather} g(\zeta_\psi(1+p)^{1-m}-1,\theta)\equiv\sum_{a\in(\Z/p^{n+1}\Z)^\times}b_a\theta\omega^{-1}(r_a)\psi(r_a)[r_a]^{m-1}\notag\\ \pmod{p^{n+1}}.\tag{3.33} \end{gather}
Recall that for any \(\alpha\in\Z_p^\times\text{,}\) \(\alpha=\omega(\alpha)[\alpha]\text{.}\) Therefore,
\begin{align*} \theta\omega^{-1}(r_a)\psi(r_a)[r_a]^{m-1} \amp =\theta\psi(r_a)\times\omega^{-m}(r_a)\times(\omega(r_a)[r_a])^{m-1}\\ \amp =\chi\omega^{-m}(r_a)r_a^{m-1}. \end{align*}
Combining all these, we may write congruence (3.33) as
\begin{gather} g(\zeta_\psi(1+p)^{1-m}-1,\theta)\equiv\sum_{a\in(\Z/p^{n+1}\Z)^\times}b_a\chi\omega^{-m}(r_a)r_a^{m-1}\pmod{p^{n+1}}.\tag{3.34} \end{gather}
We’ve chosen \(n\) large enough so that \(f_\chi\mid n\text{.}\) Thus
\begin{gather} \chi\omega^{-m}(a(1+p))=\chi\omega^{-m}(r_a)\tag{3.35} \end{gather}
since \(a(1+p)\equiv r_a\pmod{p^{n+1}}\text{.}\) Also,
\begin{align*} (a(1+p))^{m} \amp =(b_ap^{n+1}+r_a)^m\\ \amp \equiv r_a^m+mr_a^{m-1}b_ap^{n+1}\pmod{p^{2(n+1)}}. \end{align*}
Therefore,
\begin{align} \chi\omega^{-m} \amp (1+p)(1+p)^m\sum_{a\in(\Z/p^{n+1}\Z)^\times}\chi\omega^{-m}(a)a^m\notag\\ \amp =\sum_{a\in(\Z/p^{n+1}\Z)^\times}\chi\omega^{-m}(a(1+p))(a(1+p))^m\notag\\ \amp \equiv \sum_{a\in(\Z/p^{n+1}\Z)^\times}\chi\omega^{-m}(r_a)r_a^m+mp^{n+1}\sum_{a\in(\Z/p^{n+1}\Z)^\times}b_a\chi\omega^{-m}(r_a)r_a^{m-1}\notag\\ \amp \hfill\qquad\qquad\qquad\qquad\qquad\qquad\qquad\qquad\qquad\qquad\pmod{p^{2(n+1)}}.\tag{3.36} \end{align}
It is easy to see that as \(a\) runs over \((\Z/p^{n+1}\Z)^\times\) then so does \(r_a\text{,}\) (recall the definition of \(r_a\) from equation (3.29)). As a result,
\begin{gather*} \sum_{a\in(\Z/p^{n+1}\Z)^\times}\chi\omega^{-m}(a)a^m=\sum_{a\in(\Z/p^{n+1}\Z)^\times}\chi\omega^{-m}(r_a)r_a^m. \end{gather*}
Since \(\omega\) is a character modulo \(p\text{,}\) \(\omega^{-m}(1+p)=1\text{.}\) Using congruences (3.34) and (3.36), we obtain,
\begin{align*} g(\zeta_\psi(1+p)^{1-m}-1,\theta) \amp \equiv\sum_{a\in(\Z/p^{n+1}\Z)^\times}b_a\chi\omega^{-m}(r_a)r_a^{m-1}\\ \amp \equiv\frac{\bparen{(1+p)^m\chi(1+p)-1}}{mp^{n+1}}\sum_{a\in(\Z/p^{n+1}\Z)^\times}\chi\omega^{-m}(a)a^m\\ \amp \qquad\qquad\qquad\qquad\qquad\qquad\qquad\qquad\qquad\pmod{p^{n+1}}. \end{align*}
Therefore, by definition (3.23) and Proposition 3.6,
\begin{align*} g(\zeta_\psi \amp (1+p)^{1-m}-1,\theta)\\ \amp =\frac{\bparen{(1+p)^m\chi(1+p)-1}}{m}\lim_{n\to\infty}\frac{1}{p^{n+1}}\sum_{a\in(\Z/p^{n+1}\Z)^\times}\chi\omega^{-m}(a)a^m\\ \amp =-\frac{h(\zeta_\psi(1+p)^{1-m}-1)}{m}(1-\chi\omega^{-m}(p)p^{m-1})B_{n,\chi\omega^{-m}}, \end{align*}
where the last equality follows from [47, Lemma 7.11]. Therefore, an appeal to equation (3.5) shows that
\begin{align*} f(\zeta_\psi(1+p)^{1-m}-1,\theta) \amp =\frac{g(\zeta_\psi(1+p)^{1-m}-1,\theta)}{h(\zeta_\psi(1+p)^{1-m}-1)}\\ \amp =-(1-\chi\omega^{-m}(p)p^{m-1})\frac{B_{n,\chi\omega^{-m}}}{m}=L_p(1-m,\chi). \end{align*}
This completes the proof.