I am reading about $p$-adic $L$-functions and I have one question in mind.

To start with, I will write a proof I've learned of a congruence of $L$-values:

Theorem: Let $p\geq5$ be a prime, $\alpha\geq1$ be any integer, and $F_{\alpha}$ be the subextension of $\Bbb Q(p^{\alpha+1})$ such that ${\rm Gal}(F_{\alpha}/\Bbb Q)=\Bbb Z/p^{\alpha}\Bbb Z$, which exists by Galois theory since the Galois group of $\Bbb Q(\zeta_{p^{\alpha+1}})$ is cyclic of order $p^{\alpha}(p-1)$. Then $\zeta_{F_{\alpha}}(-1)\equiv\zeta_{F_{\alpha+1}}(-1)\pmod{p}$.

Proof: Let $\Lambda=\Bbb Z_p[\![T]\!]$ be the Iwasawa algebra, $\omega:(\Bbb Z/p\Bbb Z)^{\times}\longrightarrow\Bbb Q(\zeta_p)\subset\Bbb C_p$ be the $p$-adic cyclotomic character. There exists a function $g\in\Lambda$ (sometimes denoted as $g(T,\omega^{-1})$) such that \begin{align*} g((1+p)^{-1}-1)=(1-p)\zeta(-1)=\frac{1}{12}(p-1) \end{align*}For every Dirichlet character $\chi\neq1$, the conductor $p^{\alpha}$ is the $p$-primary order and hence associated to $F_{\alpha}$ such that \begin{align*} g(\zeta(1+p)^{-1}-1)=L(-1,\chi) \end{align*} where $\zeta$ is the value of $\chi$ on the topological generator $(1+p)$ of $1+p\Bbb Z_p$. We can also write in the following way: put \begin{align*} T=(1+p)^{-1}(1+X)-1=-\frac{p}{p+1}+\frac{X}{p+1} \end{align*} Therefore, $g(T)=f(X)$ for $f\in\Lambda$ now verifies \begin{align*} f(\zeta-1)=g((1+p^{-1})\zeta-1)=L(-1,\chi) \end{align*} For $\chi=1$, we can write $f(0)=L(-1,1)=\zeta(-1)$. Therefore, we have \begin{align*} \prod_{\zeta}f(\zeta-1)=\prod_{\chi}L(-1,\chi)=(1-p)\zeta_{F_{\alpha}}(-1) \end{align*} Here we have used the well-known Artin formalism for complex $L$-functions.(Example: For $K=\Bbb Q(\zeta_n)$ a cyclotomic field, we have $\zeta_K(s)=\prod_{\chi}L(s,\chi)$, product taken over the Dirichlet characters modulo $n$.) Therefore, \begin{align*} \zeta_{F_{\alpha}}(-1)-\zeta_{F_{\alpha+1}}(-1) &=(1-p)\zeta_{F_{\alpha}}(-1)H_{\alpha+1} \end{align*} where \begin{align*} H_{\alpha+1}=\prod_{\substack{\zeta\\\zeta\text{ primitive}\\\text{of order $p^{\alpha+1}$}}}f(\zeta-1)-1 \end{align*} There are total $\varphi(p^{\alpha})=(p-1)p^{\alpha}$ such primitive roots of unity. Also, there is a unit $a_0$ such that \begin{align*} f(\zeta-1)\equiv a_0\pmod{\zeta-1} \end{align*} It is a well-known fact that \begin{align*} \prod_{\substack{\zeta\\\zeta\text{ primitive}\\\text{of order $p^{\alpha+1}$}}}(\zeta-1)=\Phi_{p^{\alpha+1}}(1)=p \end{align*} Therefore, \begin{align*} H_{\alpha+1} &=\prod_{\substack{\zeta\\\zeta\text{ primitive}\\\text{of order $p^{\alpha+1}$}}}f(\zeta-1)\\ &\equiv a_0^{(p-1)p^{\alpha}}-1\pmod{p}\\\implies H_{\alpha+1} &\equiv 0\pmod{p}\\\tag{since $a_0$ is a unit in $\Bbb Z/p\Bbb Z$} \end{align*} This completes the proof.

Please let me know if this proof is correct!

Here as we can see the proof uses the factorization formula in a crucial way. I am thinking about similar factorization for $p$-adic $L$-functions. I know that in his "Lectures on $p$-adic $L$-functions" Iwasawa defined: if $K$ is a real abelian extension then the $p$-adic zeta function for $K$ is given by $\zeta_{K,p}(s)=\prod_{\chi\in\widehat{{\rm Gal}(K/\Bbb Q)}}L_p(s,\chi)$. This is clearly motivated by the Artin formalism for complex $L$-functions.

My question is: Can we define a $p$-adic analog of $\zeta_{F_{\alpha}}$ and rephrase the proof of the Theorem above in terms of a factorization of $p$-adic zeta function $\zeta_{F_{\alpha},p}$ into $p$-adic Dirichlet $L$-functions?

Can we obtain simpler proof of other congruences of complex $L$-values using $p$-adic $L$-function/$p$-adic Artin formalism?

Edit/Addition: The Galois group of the subextension $F_{\alpha}$ is inside the totally real subfield of $\Bbb Q(\zeta_{\alpha})$. Hence following Iwasawa/Leopoldt, we may define their $p$-adic $L$-function as \begin{align*} \zeta_{F_{\alpha},p}(1-n) &=\prod_{\chi}L_p(1-n,\chi) \end{align*} where the product is taken over the characters $\chi$ of ${\rm Gal}(F_{\alpha}/\Bbb Q)\cong\Bbb Z/p^{\alpha}\Bbb Z$.

Also, we know by the interpolation formula that \begin{align*} L_p(1-n,\chi)=(1-\chi(p)\omega^{-n}(p)p^{n-1})L(1-n,\chi\omega^{-n})\equiv L(1-n,\chi\omega^{-n})\pmod{p} \end{align*} ($\omega$ is the $p$-adic cyclotomic character)which gives us \begin{align*} \zeta_{F_{\alpha},p}(-1)\equiv\prod_{\chi}L(-1,\chi\omega^{-2})\pmod{p} \end{align*}

From this can we say the that the Theorem above gives us a congruence for $p$-adic L-functions as follows?

\begin{align*} \zeta_{F_{\alpha},p}(-1)\equiv\zeta_{F_{\alpha+1},p}(-1)\pmod{p} \end{align*}

Thanks in advance for any helpful comments/answers!

PS. This is not so well-thought or well-written, so apologies for posting here.

  • 2
    $\begingroup$ Please define $F_{\alpha}$: specifying its Galois group is far from sufficient. $\endgroup$ Feb 13, 2023 at 21:35
  • $\begingroup$ @HenriCohen Since the Galois group of $\Bbb Q(p^{\alpha+1})$ is cyclic of order $p^{\alpha}(p-1)$, it has a subextension whose Galois group is cyclic of order $p^{\alpha}$. So, I meant that subextension. $\endgroup$
    – ShBh
    Feb 13, 2023 at 22:00
  • 1
    $\begingroup$ I thought as much. You should edit accordingly. $\endgroup$ Feb 13, 2023 at 22:03
  • $\begingroup$ @HenriCohen I have edited! Thanks for the suggestion! $\endgroup$
    – ShBh
    Feb 13, 2023 at 22:11

1 Answer 1


I have not checked your proof, but did some numerical experiments, which should be easy to prove:

For $p=3$ your claim is definitely false: for $\alpha=1$, $2$, $3$ the values of the zeta function are -1/9, -373312/27, -bignumber/81, so probably always integer/$3^{\alpha+1}$.

For $p=5$ and $p=7$, the zeta values are integers/3, and for $\alpha=1$ they are indeed congruent modulo $p$, and even modulo $p^2$, so perhaps modulo $p^{\alpha+1}$ in general.

  • $\begingroup$ I am really sorry! My proof actually works for primes $p\geq 5$. I have edited. Do you mean this congruence is true for the $p$-adic Zeta functions as well that I mentioned at the end? $\endgroup$
    – ShBh
    Feb 14, 2023 at 11:27
  • 2
    $\begingroup$ There is a good reason why $p = 3$ does not work: for $p = 3$, you have $-1 = 1 \bmod {p-1}$. Throughout this theory, one needs to split into cases according to the value of $s$ modulo $p-1$ (or mod 4 if $p = 2$); and the values $s = 1 \bmod {p-1}$ are bad, because you are in the same congruence class as $s = 1$ where the zeta function has a pole -- meaning you get growing denominators, just as in this example. $\endgroup$ Feb 14, 2023 at 21:24
  • $\begingroup$ @DavidLoeffler Yes! That's very illuminating! Thanks for the comment. Do you think that the p-adic L-function congruence is true at the end? $\endgroup$
    – ShBh
    Feb 15, 2023 at 2:03

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.