Remark 1.1
NKVRAs an amusing sidenote, the other famous mathematical object named for August Ferdinand Möbius, the Möbius strip, is also a configuration space: \ensuremath{\operatorname{Conf}}^{2}(S^{1}).
In this expository note, we revisit several classical arithmetic functions - namely Euler’s totient function, the divisor sum functions and Dedekind’s \psi-function - within a unifying algebraic framework that highlights their connections to geometry. This framework builds on prior work involving zeta functions and Möbius inversion. While our main goal is to provide a clear context for similar constructions in the future, we also make an original observation regarding Dedekind’s \psi-function.
As an amusing sidenote, the other famous mathematical object named for August Ferdinand Möbius, the Möbius strip, is also a configuration space: \ensuremath{\operatorname{Conf}}^{2}(S^{1}).
As formal Dirichlet series,
By Dirichlet convolution,
The totient function of a number field K/\mathbb{Q} is the function \phi_{K} : I_{K}^{+}\rightarrow\mathbb{Z} defined by
For any ideal \frak{a}\in I_{K}^{+},
where the sum runs over all ideals \frak{d} dividing \frak{a}. In other words, N = \phi_{K}*\zeta where \zeta(\frak{a}) = 1 is the zeta function for I_{K}^{+}.
By formula (?),
This can be rewritten
i.e. \phi_{K} = N*\mu. Then Möbius inversion gives the desired formula.
We could have proved formula (?) for the classical totient function this way too, i.e. by first proving formula (?) directly using the above argument and then applying Möbius inversion.
As formal Dirichlet series,
Since \phi_{K} = N*\mu, we get
The totient function of a variety X/\mathbb{F}_{q} is the function \phi_{X} : Z_{0}^{\ensuremath{\operatorname{eff}}}(X)\rightarrow\mathbb{Z} defined by
For any effective 0-cycle \alpha on X,
In other words, \pi = \phi_{X}*\zeta, where \zeta(\alpha) = 1 is the zeta function for Z_{0}^{\ensuremath{\operatorname{eff}}}(X) and \pi = q^{\deg(-)}.
Write \alpha = \sum_{i = 1}^{r} a_{i}x_{i}. By formula (?),
where \mu is the Möbius function for Z_{0}^{\ensuremath{\operatorname{eff}}}(X). This can be rewritten
Or in other words, \phi_{X} = \pi*\mu. Then Möbius inversion for Z_{0}^{\ensuremath{\operatorname{eff}}}(X) implies that \pi = \phi_{X}*\zeta as claimed.
For any variety X over a finite field,
Translating the identity \phi_{X} = \pi*\mu to power series produces
It is well-known that Z(X,qt) = Z(X\times\mathbb{A}^{1},t), so the claimed formula holds.
Let K/\mathbb{Q} be a number field with ring of integers \mathcal{O}_{K}. For any \alpha\in\mathcal{O}_{K} and ideal \frak{b} such that \alpha\mathcal{O}_{K} and \frak{b} are relatively prime,
Let X/\mathbb{F}_{q} be a variety, U\subseteq X an open subset and f\in\mathcal{O}_{X}(U). Then for any effective 0-cycle \beta\in Z_{0}^{\ensuremath{\operatorname{eff}}}(X) with support outside V(f) = \{x\in U\mid f(x) = 0\},
in the semilocal ring \mathcal{O}_{X,\beta}.
Let k be a finite field, g\in k[t] and D = \ensuremath{\operatorname{div}}_{0}(g) the effective divisor of zeroes of g. Then
\phi_{\mathbb{P}^{1}}(D) = \#\{f\in k[t]\mid \deg(f) < \deg(g) \text{ and } \gcd(f,g) = 1\}.
For any such f\in k[t],
holds in k[t].
Both statements follow from setting X = \mathbb{P}^{1}, U = \mathbb{P}^{1}\smallsetminus\{\infty\} \cong \mathbb{A}^{1}, identifying the rings \mathcal{O}_{\mathbb{P}^{1},D} \cong k[t]/(g) and taking cardinalities of multiplicative groups of units.
The first divisor sum of a number field K/\mathbb{Q} is the arithmetic function \sigma_{1,K} : I_{K}^{\times}\rightarrow\mathbb{Z} defined by
As formal Dirichlet series,
The first divisor sum of a variety X/\mathbb{F}_{q} is the function \sigma_{1,X} : Z_{0}^{\ensuremath{\operatorname{eff}}}(X)\rightarrow\mathbb{Z} defined by
For any variety X/\mathbb{F}_{q},
For an arithmetic scheme X, the global first divisor sum of X is the formal power series
where for each prime p,
For any arithmetic scheme X, we have
As in Remark 2.9, we can view the formulas above as shadows of the formula [\mathbb{P}^{1}] = [\mathbb{A}^{1}] + [\text{point}] in the appropriate Grothendieck ring. Further, formula (?) also suggests generalizing the divisor sum functions \sigma_{m}(n) by \sigma_{m,X}(z) = \#(\mathcal{O}/z)^{m} + 1 for each choice of object X, ring \mathcal{O} and monoid Z with z\in Z. That is, a divisor sum function is a shadow of a class of the form [\mathbb{A}^{m}] + [\text{point}] in some Grothendieck ring.
Dedekind’s \psi-function is the arithmetic function \psi : \mathbb{N}\rightarrow\mathbb{C} defined by
\#G(p^{k}) = \psi(p^{k}).
This follows from standard formulas for \#\mathbb{W}_{k}(\mathbb{F}_{q}) and \#\mathbb{W}_{k}(\mathbb{F}_{q})^{\times}. For more details, see [kzb].
\#G(n) = \psi(n).
Is there a group scheme \mathcal{G}/\ensuremath{\operatorname{Spec}}\mathbb{Z} such that G(n) \cong \mathcal{G}(\mathbb{Z}/n\mathbb{Z}) for all n? For example, is there an interpretation of the groups G(n) in terms of the ring of big Witt vectors?
The \psi-function of a number field K/\mathbb{Q} is the function \psi_{K} : I_{K}^{\times}\rightarrow\mathbb{Z} defined by
Let \frak{a},\frak{b},\frak{p}\subset\mathcal{O}_{K} be ideals, with \frak{p} prime. Then
If \frak{a} and \frak{b} are relatively prime, then \psi_{K}(\frak{a}\frak{b}) = \psi_{K}(\frak{a})\psi_{K}(\frak{b}).
\psi_{K}(\frak{p}) = \sigma_{1,K}(\frak{p}) = N(\frak{p}) + 1. In particular, if \frak{a} is squarefree then \psi_{K}(\frak{a}) = \sigma_{1,K}(\frak{a}).
As arithmetic functions on I_{K}^{\times}, \psi_{K} = N*|\mu_{K}|, where \mu_{K} is the Möbius function in I(I_{K}^{\times}).
\psi(\frak{a}) = [SL_{2}(\mathcal{O}_{K}) : \Gamma_{0}(\frak{a})], where \Gamma_{0}(\frak{a}) is the subgroup of matrices in SL_{2}(\mathcal{O}_{K}) which are upper triangular mod \frak{a}.
\psi_{K}(\frak{a}) = \#G(\frak{a}) where
(a) – (c) follow from the definition. (d) is proven in the usual way, i.e. by counting the number of cyclic submodules of rank N(\frak{a}) in E[\frak{a}], where E is an elliptic curve over K (equivalently, cyclic submodules in E_{tors} which are isomorphic as \mathcal{O}_{K}-modules to \mathcal{O}_{K}/\frak{a}). Finally, (e) follows from the same standard formulas as above.
The Dirichlet series for \psi_{K} satisfies the formula
where \zeta_{K}(s) is the Dedekind zeta function of K.
This follows from (c) and the identity
which is a straightforward generalization of the same formula over \mathbb{Q}.
The \psi-function of a variety X/\mathbb{F}_{q} is the function \psi_{X} : Z_{0}^{\ensuremath{\operatorname{eff}}}(X)\rightarrow\mathbb{Z} defined by
Let \alpha,\beta\in Z_{0}^{\ensuremath{\operatorname{eff}}}(X) and let x\in|X|. Then
If \alpha and \beta are relatively prime, then \psi_{X}(\alpha + \beta) = \psi_{X}(\alpha)\psi_{X}(\beta).
\psi_{X}(x) = \sigma_{1,X}(x) = q^{\deg(x)} + 1. In particular, if \alpha is a log effective 0-cycle(A log cycle is one for which each irreducible cycle appears with coefficient 1.) then \psi_{X}(\alpha) = \sigma_{1,X}(\alpha).
As arithmetic functions on Z_{0}^{\ensuremath{\operatorname{eff}}}(X), \psi_{X} = \pi*|\mu_{X}|, where \mu_{X} is the Möbius function in I(Z_{0}^{\ensuremath{\operatorname{eff}}}(X)).
\psi(\alpha) = [SL_{2}(\mathcal{O}_{X,\alpha}) : \Gamma_{0}(I_{\alpha})], where \Gamma_{0}(I_{\alpha}) is the subgroup of matrices in SL_{2}(\mathcal{O}_{X,\alpha}) which are upper triangular mod I_{\alpha}.
\psi_{X}(\alpha) = \#G(\alpha) where
(a) – (c) and (e) are the same as in the proof of Proposition 4.6. For (e), counting sublattices in (\mathcal{O}_{X,\alpha}/I_{\alpha})^{2} isomorphic to \mathcal{O}_{X,\alpha}/I_{\alpha} gives the result.
The power series for \psi_{X} satisfies
where X_{\mathbb{F}_{q^{2}}} = X\times_{\ensuremath{\operatorname{Spec}}\mathbb{F}_{q}}\ensuremath{\operatorname{Spec}}\mathbb{F}_{q^{2}}.
For any arithmetic scheme X, we have
Liouville’s \lambda-function is the arithmetic function \lambda : \mathbb{N}\rightarrow\mathbb{C} defined by \lambda(1) = 1 and
That is, \lambda(n) = (-1)^{\Omega(n)} where \Omega\left (\prod p_{i}^{a_{i}}\right ) = \sum a_{i}.
Extend \lambda to a number field K/\mathbb{Q} by setting
where \Omega\left (\prod \frak{p}_{i}^{a_{i}}\right ) = \sum a_{i}. Show \lambda_{K}*|\mu_{K}| = \delta in the incidence algebra I(I_{K}^{\times}), which implies
Repeat for a variety X/\mathbb{F}_{q} and for a scheme X/\ensuremath{\operatorname{Spec}}\mathbb{Z}.
If possible, find a class \xi in the Grothendieck ring of schemes such that \#\xi(\mathbb{Z}/n\mathbb{Z}) = \lambda(n) and similarly for each generalization of \psi in Problem 5.2.
Does this identity generalize to (i) number fields? (ii) higher dimensional geometric objects? (iii) a formula in some Grothendieck ring? What about formulas involving the sum of squares functions r_{k}(n) for k\geq 3?
The text on this page is from arXiv:2504.13328v2 by Andrew Kobin, under the licence CC BY 4.0. The corpus made these changes: it read the items out of the LaTeX source, it expanded the macros of the paper, it put each proof after its statement, and it converted the text to HTML. The formulas are the TeX of the paper, and KaTeX renders them. A formula or a reference can look different from the PDF. The page is published under CC BY 4.0.
The record of each item is in analysis/objects/2504/2504.13328.jsonl.