The L-function of CM elliptic curve $E$ over an imaginary quadratic field can be written as the product of the Hecke L-functions (for simplicity, I assume that the base field of the elliptic curve and the CM field is the same field $K$). This Hecke character's target is $\mathbb{C}^{*}$ and not $S^1$.

The Hecke character is often written as a character whose values are in the unit circle. If we write

$L(E,s):=L(\chi,s)L(\overline{\chi},s)$ $\quad$ ($\chi$ is the Hecke character whose values are not in $S^1$)

$\mathbb{L}(E,s):=L(\chi',s)L(\overline{\chi'},s)$ $\quad$ ($\chi'$ is the normalize Hecke character whose values are in $S^1$),

my question is if one can relate $L(E,1)$ and $\mathbb{L}(E,1)$. Please tell me.

  • 2
    $\begingroup$ Do you know how to express the $L$-function of a non-normalized Hecke character $\chi$ in terms of the $L$-function of its normalized form $\chi'$? $\endgroup$ – KConrad Jan 3 '16 at 15:44

The two $L$-functions are shifts of each other. That is, $\mathbb{L}(E,s)=L(E,s+s_0)$ for some fixed $s_0$. BTW some people do not appreciate the importance to shift normalize every automorphic $L$-function so that its center is $s=1/2$. This normalization creates order in the world and pays respect to Riemann as well.

  • 6
    $\begingroup$ Indeed, specifically, for elliptic curves there is a "competing" tradition of making the critical strip $0\le \sigma\le 2$, and critical line $\sigma=1$... $\endgroup$ – paul garrett Jan 3 '16 at 17:41
  • 3
    $\begingroup$ It makes sense to use the other normalization for algebraicity issues; for instance, the field generated over $\mathbf{Q}$ by the coefficients of the Hecke-normalized $L$-function of an elliptic curve (i.e., functional equation between $s$ and $1-s$) is of infinite degree over $\mathbf{Q}$, and depends on the elliptic curve (through the set of supersingular primes, for instance). $\endgroup$ – Denis Chaperon de Lauzières Jan 3 '16 at 19:05
  • 1
    $\begingroup$ @DenisChaperondeLauzières: No doubt it makes sense, but I don't like it, especially that most automorphic $L$-functions seem highly transcendental. Algebraicity is a rare feature. $\endgroup$ – GH from MO Jan 3 '16 at 19:27
  • 1
    $\begingroup$ Somewhat related to @DenisChaperondeLauzières's comment, although I myself do like the $s\leftrightarrow 1-s$ convention, it is true that Deligne's conjectures, e.g., proven in some cases by Shimura and others for holomorphic elliptic or Hilbert modular forms' L-functions, about special values, do seem to "want" a wider critical strip, etc. And, yes, these results are highly anomalous. Once again, I think there's no single best convention... so one always must think about context. In a different universe things might have been otherwise... $\endgroup$ – paul garrett Jan 3 '16 at 22:53
  • $\begingroup$ @paulgarrett: I think the symbol $L(s,\pi)$ should be reserved for the finite $L$-function $\prod_p L_p(s,\pi)$ with functional equation $s\leftrightarrow 1-s$ and $\pi\leftrightarrow\tilde\pi$. Anything else, e.g. $L(s,\pi)$ with the gamma factors attached, or $L(s,\pi)$ shifted by some constant or rescaled in any way, should be denoted with a different symbol such as $\Lambda(s,\pi)$ or $D(s,\pi)$. This could be carried out in our universe. $\endgroup$ – GH from MO Jan 4 '16 at 1:48

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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