21.03.04 · number-theory / l-functions

Dirichlet density

shipped3 tiersLean: none

Anchor (Master): Serre 1973 *A Course in Arithmetic* Ch. VI §§3-4 (originator anchor for the Dirichlet-density viewpoint, with the cyclotomic / Galois-theoretic packaging carried forward in Ch. VI §3 Theorem 2); Dirichlet 1837 *Abhandl. Königl. Preuss. Akad.* 45-81 (originator of primes-in-AP in density-counting form); Chebotarev 1922 *Math. Ann.* 95, 191-228 (Chebotarev density theorem; Galois-theoretic generalisation); Hecke 1918 *Math. Z.* 1, 357-376 + 1920 *Math. Z.* 6, 11-51 (Hecke $L$-functions for number fields, the analytic-continuation engine driving Chebotarev); Tate 1950 Princeton PhD thesis (published in Cassels-Fröhlich 1967 *Algebraic Number Theory* Ch. XV, 305-347; adelic / Fourier-analytic reformulation, idele-class-group zeta integrals); Lang 1994 *Algebraic Number Theory* (Springer GTM 110, 2nd ed.) Ch. VIII (modern proof of the Chebotarev density theorem from Hecke / Artin $L$); Neukirch 1999 *Algebraic Number Theory* (Springer Grundlehren 322) Ch. VII §13 (Chebotarev + density formula in modern form); Iwaniec-Kowalski 2004 *Analytic Number Theory* (AMS Colloquium 53) Ch. 5 §5.6 (quantitative versions, the PNT in arithmetic progressions); Davenport 2000 *Multiplicative Number Theory* (Springer GTM 74) Ch. 7; Bump 1997 *Automorphic Forms and Representations* (Cambridge Studies in Advanced Mathematics 55) Ch. 1, 3 (Tate-thesis density viewpoint)

Intuition Beginner

A long-standing question in number theory is not just whether there are infinitely many primes, but how the primes distribute across residue classes. Among the integers up to a million, roughly a quarter are congruent to $1$ modulo $4$ and roughly another quarter to $3$ modulo $4$ . Are the primes inside those classes distributed in the same proportion? Dirichlet proved in 1837 that yes, every residue class coprime to the modulus contains infinitely many primes, and the share each class receives is exactly $1/ φ (m)$ .

To make this share precise one needs a notion of density on the primes. Counting "how many primes up to $x$ lie in the class" gives natural density. Dirichlet introduced a different gadget — weight each prime $p$ by $p^{- s}$ , sum the weights over the class, and compare with the sum over all primes, then let $s$ approach $1$ from above. The weighted ratio has a limit, and that limit is the Dirichlet density of the class.

Why bother with this weighted version when the unweighted count is simpler to picture? Because the weighted sum connects directly to the logarithm of an $L$ -function, and the $L$ -function lets analytic tools settle a question about whole numbers. The natural density is what you would measure with a sieve; the Dirichlet density is what analysis hands you.

Visual Beginner

The picture is the simplest informative case: primes mod $4$ . Below the modulus $4$ there are two coprime residue classes, namely $1$ and $3$ . List the primes in each class up to some bound and the lists fill up at roughly equal rates: $5, 13, 17, 29, 37, \dots$ on the $1$ -side and $3, 7, 11, 19, 23, 31, \dots$ on the $3$ -side. Both lists grow without bound, and Dirichlet's density formula predicts that the long-run share each side receives is one half.

The diagram is a finite snapshot. The density formula is a statement about the asymptotic share each colour receives as the snapshot grows. The weighting $p^{- s}$ is what makes the limit computable: as $s \to 1^{+}$ the denominator diverges like $lo g (1/ (s - 1))$ , the numerator on each side diverges like the same logarithm times $1/2$ , and the ratio approaches $1/2$ in the limit.

Worked example Beginner

Compute the Dirichlet density of the primes congruent to $1 mod 4$ and verify it equals $1/2$ .

Step 1. Set up the character. Use the non-principal character $χ$ mod $4$ with $χ (1) = + 1$ , $χ (3) = - 1$ , and $χ (n) = 0$ for $n$ even. Then the indicator of "is $p$ congruent to $1$ mod $4$ ?" on primes larger than $2$ equals $(1 + χ (p)) /2$ : when $p \equiv 1$ the character value is $+ 1$ and the indicator is $1$ ; when $p \equiv 3$ the character value is $- 1$ and the indicator is $0$ .

Step 2. Weight each prime by $p^{- s}$ and add. For each odd prime $p$ , the $p^{- s}$ contribution from the class $p \equiv 1$ equals $(1 + χ (p)) /2$ times $p^{- s}$ . So the total weighted contribution from the class is one half of the weighted total over all odd primes, plus one half of the character-twisted weighted total over odd primes.

Step 3. Compare the two pieces at $s \to 1^{+}$ . The unweighted prime-sum $1/ 2^{s} + 1/ 3^{s} + 1/ 5^{s} + \dots$ diverges like $lo g (1/ (s - 1))$ as $s$ decreases to $1$ (an asymptotic that follows from the logarithm of the Euler product for the Riemann zeta function). The character-twisted version $χ (2) / 2^{s} + χ (3) / 3^{s} + χ (5) / 5^{s} + \dots$ stays bounded as $s$ approaches $1$ , because the Dirichlet $L$ -function $L (s, χ)$ is continuous near $s = 1$ with $L (1, χ) = π /4 \neq = 0$ (the Leibniz identity $1 - 1/3 + 1/5 - 1/7 + \dots = π /4$ ).

Step 4. Take the ratio. The Dirichlet density of the class is the limit as $s \to 1^{+}$ of the ratio of the class-restricted weighted total to the all-primes weighted total. The class-restricted weighted total grows like $\frac{1}{2} lo g (1/ (s - 1))$ plus a bounded piece, the all-primes weighted total grows like $lo g (1/ (s - 1))$ plus a bounded piece, and the ratio approaches $1/2$ in the limit. The bounded pieces wash out; the orthogonality factor $1/2$ survives.

Step 5. Compare with natural density. The prime-number theorem in arithmetic progressions (de la Vallée Poussin 1896) gives the count $∣ {p \leq x : p \equiv 1 mod 4} ∣ \sim x / (2 lo g x)$ as $x$ grows, matching $π (x) /2$ in the limit. So the natural density of the class is also $1/2$ , equal to the Dirichlet density. For $x = 100$ , eleven primes lie in the class (compared with the prediction of $π (100) /2 = 25/2 = 12.5$ , close to the actual count).

What this tells us: the share of primes in a residue class modulo $4$ is exactly one half, both in the analytic Dirichlet-density sense and in the counting natural-density sense. The same argument with modulus $m$ and any residue $a$ coprime to $m$ replaces the factor $1/2$ with $1/ φ (m)$ .

Check your understanding Beginner

Formal definition Intermediate+

Throughout, $p$ ranges over primes in $Z_{> 0}$ and $s$ denotes a real variable approaching $1$ from above (or, when convenient, a complex variable with $Re (s) > 1$ ).

Definition (Dirichlet density). Let $A$ be a set of primes. The Dirichlet density of $A$ is $$ d(A) := \lim_{s \to 1^+} \frac{\sum_{p \in A} p^{-s}}{\sum_p p^{-s}}, $$ provided the limit exists. The denominator is a divergent series at $s = 1$ , so the ratio is interpreted by the prior assertion that the limit makes sense; equivalent normalisations replace the denominator by $lo g (1/ (s - 1))$ via the asymptotic recorded below.

Proposition (singular asymptotic; Serre Ch. VI §3 Proposition 1). As $s \to 1^{+}$ , $$ \sum_p p^{-s} = \log \frac{1}{s - 1} + O(1). $$

Sketch. Take the logarithm of the Euler product for $ζ (s) = \prod_{p} (1 - p^{- s})^{- 1}$ on $Re (s) > 1$ : $$ \log \zeta(s) = \sum_p -\log(1 - p^{-s}) = \sum_p \sum_{n \geq 1} \frac{p^{-ns}}{n} = \sum_p p^{-s} + R(s), $$ where the remainder $R (s) = \sum_{p} \sum_{n \geq 2} p^{- n s} / n$ is bounded as $s \to 1^{+}$ : the inner sum is dominated by $\sum_{n \geq 2} p^{- n} = p^{- 2} / (1 - p^{- 1}) \leq 2 p^{- 2}$ for $p \geq 2$ , so $∣ R (s) ∣ \leq 2 \sum_{p} p^{- 2} \leq 2 ζ (2)$ uniformly on $s \geq 1$ . The singular term $ζ (s) \sim 1/ (s - 1)$ at $s = 1$ (simple pole, residue $1$ ) gives $lo g ζ (s) = lo g (1/ (s - 1)) + O (1)$ , and isolating the $n = 1$ prime sum produces the claimed asymptotic. $□$

Definition (equivalent form). Using the proposition, the Dirichlet density is equivalently $$ d(A) = \lim_{s \to 1^+} \frac{\sum_{p \in A} p^{-s}}{\log \frac{1}{s - 1}}, $$ when the limit exists, since the denominator differs from $\sum_{p} p^{- s}$ by $O (1)$ and the numerator-side bounded perturbations are absorbed by the divergence of $lo g (1/ (s - 1))$ .

Definition (natural density). Let $A$ be a set of primes. The natural density of $A$ is $$ \delta(A) := \lim_{x \to \infty} \frac{|{p \in A : p \leq x}|}{\pi(x)}, $$ provided the limit exists, where $π (x) = ∣ {p \leq x} ∣$ is the prime counting function.

Proposition (natural density implies Dirichlet density). If $δ (A)$ exists, then $d (A)$ exists and $d (A) = δ (A)$ .

Proof. Write $π_{A} (x) = ∣ {p \in A : p \leq x} ∣$ . By Abel summation, $$ \sum_{p \in A} p^{-s} = s \int_2^\infty \pi_A(x) x^{-s - 1} dx $$ for $s > 1$ (the boundary term vanishes since $π_{A} (x) / x^{s} \to 0$ as $x \to \infty$ for $s > 1$ ). Hypothesis $π_{A} (x) = (δ (A) + o (1)) π (x)$ as $x \to \infty$ gives $$ \sum_{p \in A} p^{-s} = \delta(A) \cdot s \int_2^\infty \pi(x) x^{-s-1} dx + s \int_2^\infty o(\pi(x)) x^{-s-1} dx. $$ The first integral, again by Abel summation, equals $\sum_{p} p^{- s}$ . The error integral is $o (\sum_{p} p^{- s})$ as $s \to 1^{+}$ by the dominated convergence theorem applied to the ratio $o (π (x)) / π (x)$ . Dividing by $\sum_{p} p^{- s}$ and taking $s \to 1^{+}$ yields $d (A) = δ (A)$ . $□$

Non-example (the converse fails). Order the primes $p_{1} < p_{2} < p_{3} < \dots$ and let $A = {p_{n} : 2^{k} \leq n < 2^{k + 1}, k even}$ . The set $A$ contains the first half of every block of length $2^{k}$ for even $k$ and the empty half for odd $k$ . The counts $π_{A} (x) / π (x)$ oscillate between $1/3$ and $2/3$ as $x$ traverses successive blocks, so natural density does not exist. By contrast, the Dirichlet density of $A$ does exist and equals $1/2$ : the weighted sum $\sum_{p \in A} p^{- s}$ averages over the oscillation as $s \to 1^{+}$ , since the contribution of each block to the weighted sum is roughly its share of the total prime mass within the block, and the geometric weight $p^{- s}$ smooths out the block-wise oscillation. Constructions of this kind (Pólya, Knapowski-Turán) show that Dirichlet density is strictly weaker than natural density.

Sign convention. Dirichlet density takes values in $[0, 1]$ for any subset of the primes, with $d (\emptyset) = 0$ , $d (all primes) = 1$ , and finite additivity: $d (A \cup B) = d (A) + d (B)$ when $A \cap B = \emptyset$ and the two individual densities exist. Sub-additivity in general: $d (A \cup B) \leq d (A) + d (B)$ when both exist.

Counterexamples to common slips

"Dirichlet density of ${p : p \leq N}$ for finite $N$ is $N$ -dependent." A finite set of primes has Dirichlet density $0$ , since $\sum_{p \in A} p^{- s}$ is bounded as $s \to 1^{+}$ while $\sum_{p} p^{- s}$ diverges. Only infinite sets can have non-zero Dirichlet density.
"Every set of primes has a Dirichlet density." There are sets of primes for which the limit does not exist — the construction above is one example, but one can sharpen it so that the Dirichlet-density limit itself oscillates. The notion is partial. In the situations of interest (cosets of $(Z / m)^{\times}$ , Frobenius classes in Galois groups), the limit exists by character analysis.
"The natural density is more refined than the Dirichlet density." The natural density is strictly stronger as a hypothesis (it has more existence-failure modes), but both notions, when they exist, return the same value. Dirichlet density is the broader notion: it returns answers in situations where natural density does not converge.

Key theorem with proof Intermediate+

The signature theorem of this unit is Dirichlet's theorem on primes in arithmetic progressions, in density form, as packaged in Serre's A Course in Arithmetic Ch. VI ^{[Ch. VI §3 Theorem 2]}. The non-vanishing $L (1, χ) \neq = 0$ for non-principal Dirichlet $χ$ is taken as the analytic input — it is the signature theorem of the sibling unit 21.03.02 (Dirichlet $L$ -functions), proved there.

Theorem (Dirichlet density of an arithmetic progression; Dirichlet 1837, Serre Ch. VI Theorem 2). Let $m \geq 1$ and $a \in Z$ with $g cd (a, m) = 1$ . Then the set of primes $P_{a, m} := {p : p \equiv a (mod m)}$ has Dirichlet density $$ d(P_{a, m}) = \frac{1}{\varphi(m)}, $$ where $φ$ is the Euler totient. In particular, $P_{a, m}$ is infinite.

Proof. Let $X$ denote the set of Dirichlet characters $χ$ modulo $m$ , a finite abelian group of order $φ (m)$ under pointwise multiplication. The orthogonality relation for characters gives, for any prime $p$ coprime to $m$ , $$ \frac{1}{\varphi(m)} \sum_{\chi \in X} \overline{\chi(a)} \chi(p) = \begin{cases} 1 & p \equiv a \pmod m, \ 0 & p \not\equiv a \pmod m. \end{cases} $$ The identity follows from the duality $\sum_{χ} χ (b) = φ (m) \cdot 1_{b \equiv 1 mod m}$ applied to $b = p a^{- 1}$ in $(Z / m)^{\times}$ .

Sum $p^{- s}$ against this indicator. For $s > 1$ , $$ \sum_{p \in P_{a, m}} p^{-s} = \frac{1}{\varphi(m)} \sum_{\chi \in X} \overline{\chi(a)} \sum_{p \nmid m} \chi(p) p^{-s}. $$ The inner sum is taken only over primes $p ∤ m$ , since $χ (p) = 0$ for $p ∣ m$ . The contribution of primes $p ∣ m$ to the left-hand side is a finite sum (the set of primes dividing $m$ is finite), bounded as $s \to 1^{+}$ .

The character-twisted prime sum on the right has two regimes. Principal character $χ_{0}$ : $χ_{0} (p) = 1$ for $p ∤ m$ , so $\sum_{p ∤ m} χ_{0} (p) p^{- s} = \sum_{p} p^{- s} - \sum_{p ∣ m} p^{- s} = lo g (1/ (s - 1)) + O (1)$ by the singular asymptotic, since $\sum_{p ∣ m} p^{- s}$ is a bounded finite sum.

Non-principal character $χ \neq = χ_{0}$ : the Euler product $$ L(s, \chi) = \prod_{p \nmid m} (1 - \chi(p) p^{-s})^{-1} $$ holds for $Re (s) > 1$ and gives the logarithmic expansion $$ \log L(s, \chi) = \sum_{p \nmid m} \sum_{n \geq 1} \frac{\chi(p)^n}{n p^{ns}} = \sum_{p \nmid m} \chi(p) p^{-s} + R_\chi(s), $$ with the same $n \geq 2$ tail bound $∣ R_{χ} (s) ∣ \leq 2 ζ (2)$ as in the singular asymptotic. The non-vanishing theorem $L (1, χ) \neq = 0$ for non-principal $χ$ (proved in 21.03.02 using the Landau theorem in the complex-character case and the class-number formula in the real-character case) plus the fact that $L (s, χ)$ extends to a function holomorphic at $s = 1$ for non-principal $χ$ (also proved there) imply that $lo g L (s, χ)$ is bounded as $s \to 1^{+}$ — a value $lo g L (1, χ)$ at $s = 1$ exists and the function is continuous near $s = 1$ . Therefore $$ \sum_{p \nmid m} \chi(p) p^{-s} = O(1) \qquad \text{as } s \to 1^+, \text{ for non-principal } \chi. $$

Assemble the two regimes. The principal-character term contributes $$ \frac{1}{\varphi(m)} \overline{\chi_0(a)} \sum_{p \nmid m} \chi_0(p) p^{-s} = \frac{1}{\varphi(m)} \log \frac{1}{s - 1} + O(1), $$ since $χ_{0} (a) = 1$ (as $g cd (a, m) = 1$ ). Each non-principal character contributes $O (1)$ , and there are $φ (m) - 1$ such characters. Adding, $$ \sum_{p \in P_{a, m}} p^{-s} = \frac{1}{\varphi(m)} \log \frac{1}{s - 1} + O(1) \qquad \text{as } s \to 1^+. $$ Dividing by $\sum_{p} p^{- s} = lo g (1/ (s - 1)) + O (1)$ and taking $s \to 1^{+}$ yields $$ d(P_{a, m}) = \lim_{s \to 1^+} \frac{\frac{1}{\varphi(m)} \log(1/(s - 1)) + O(1)}{\log(1/(s - 1)) + O(1)} = \frac{1}{\varphi(m)}, $$ which is the density formula. Since the density is strictly positive, the set $P_{a, m}$ is infinite — the original statement of Dirichlet 1837. $□$

Bridge. Dirichlet's density theorem builds toward 21.03.03 (Dedekind, Hecke, and Artin $L$ -functions), where the same logarithmic-singularity analysis applies to the Dedekind zeta of an extension number field and produces the Chebotarev density theorem — the foundational reason the prime-Frobenius distribution is governed by Galois-group conjugacy classes. The bridge is the recognition that orthogonality of characters of $(Z / m)^{\times}$ packages the indicator of a residue class as a sum that the $L$ -function analysis can probe, and the central insight is that the non-vanishing $L (1, χ) \neq = 0$ controls everything else: every non-principal character contributes a bounded perturbation, the principal character carries the full logarithmic divergence, and the share-by-character-orthogonality count $1/ φ (m)$ pops out. This is exactly the structural pattern that appears again in 21.03.03 under the heading of Chebotarev density, generalises through the Hecke and Artin $L$ -function frameworks (with the conjugacy-class share $∣ C ∣/∣ G ∣$ replacing $1/ φ (m)$ ), and the foundational reason this works is that an $L$ -function value at $s = 1$ is a residue computation that survives orthogonality averaging. Putting these together, one analytic input — the non-vanishing of $L$ at $s = 1$ — drives an entire family of density theorems for primes in residue classes, Frobenius conjugacy classes, and idele-class-group cosets.

Exercises Intermediate+

Exercise 4 (medium, symbolic).

Show that the Dirichlet density of the set of primes $p \equiv 1 (mod 4)$ equals $\frac{1}{2}$ directly from the formula $\sum_{p \in P_{1, 4}} p^{- s} = \frac{1}{2} (\sum_{p > 2} p^{- s} + \sum_{p > 2} χ_{1} (p) p^{- s})$ for the non-principal character $χ_{1}$ mod $4$ .

Hint

The character $χ_{1}$ has $L (1, χ_{1}) = π /4 \neq = 0$ (Leibniz). The singular term comes from the principal piece only.

Answer

The first sum is $\sum_{p > 2} p^{- s} = lo g (1/ (s - 1)) + O (1)$ as $s \to 1^{+}$ (singular asymptotic minus the bounded $p = 2$ term). The second sum is $\sum_{p} χ_{1} (p) p^{- s} = lo g L (s, χ_{1}) + O (1)$ , and since $L (1, χ_{1}) = π /4 \neq = 0$ with $L (s, χ_{1})$ continuous near $s = 1$ , the second sum is bounded as $s \to 1^{+}$ . Therefore $$ \sum_{p \in P_{1, 4}} p^{-s} = \tfrac{1}{2} \log(1/(s - 1)) + O(1), $$ and dividing by $\sum_{p} p^{- s} = lo g (1/ (s - 1)) + O (1)$ gives $d (P_{1, 4}) = 1/2$ . Rubric: full credit for separating the principal and non-principal contributions, plus the explicit use of $L (1, χ_{1}) \neq = 0$ .

Exercise 5 (medium, symbolic).

Define upper Dirichlet density $\overline{d} (A) := lim sup_{s \to 1^{+}} \frac{\sum _{p \in A} p ^{- s}}{\sum _{p} p ^{- s}}$ and lower Dirichlet density $\underline{d} (A) := lim inf_{s \to 1^{+}} \frac{\sum _{p \in A} p ^{- s}}{\sum _{p} p ^{- s}}$ . Show that $\overline{d} (A) = \underline{d} (A) = d (A)$ when $d (A)$ exists, and that both upper and lower densities lie in $[0, 1]$ .

Hint

When a limit exists, the limit superior and limit inferior coincide with it. For the bounded-range claim, use that the ratio is itself bounded in $[0, 1]$ for every $s > 1$ .

Answer

If $d (A)$ exists, then by the definition of limit, both $lim sup$ and $lim inf$ of the ratio equal $d (A)$ . For the bounded-range claim, the ratio satisfies $0 \leq \sum_{p \in A} p^{- s} / \sum_{p} p^{- s} \leq 1$ for every $s > 1$ (the numerator is a non-negative sub-sum of the denominator), so the $lim sup$ and $lim inf$ both lie in $[0, 1]$ . Therefore $\overline{d} (A), \underline{d} (A) \in [0, 1]$ always, and they coincide with $d (A)$ when the limit exists. Rubric: full credit for the inclusion-of-bounds argument and the limit-superior / limit-inferior identity.

Exercise 6 (medium, symbolic).

Compute the Dirichlet density of the set of primes $p$ that split in $Q (i)$ (equivalently, primes $p$ with $p \equiv 1 (mod 4)$ together with the exceptional $p = 2$ ramified case).

Hint

A prime $p$ splits in $Q (i)$ if and only if $- 1$ is a square mod $p$ , which by quadratic reciprocity is the condition $p \equiv 1 (mod 4)$ .

Answer

A prime $p$ splits in $Q (i)$ if and only if the polynomial $X^{2} + 1$ has a root mod $p$ , equivalently $- 1$ is a quadratic residue mod $p$ , equivalently $p \equiv 1 (mod 4)$ (Euler's criterion). The exceptional case $p = 2$ ramifies and contributes a single term, irrelevant to density. So the splitting set has Dirichlet density equal to $d (P_{1, 4}) = 1/2$ . This is the simplest instance of the Chebotarev density theorem: the splitting primes of a quadratic extension form a set of density $1/2 = 1/∣ Gal (Q (i) / Q) ∣$ . Rubric: full credit for the quadratic-reciprocity identification plus the density computation.

Exercise 7 (hard, short-answer).

Construct a set of primes $A$ with Dirichlet density $1/2$ but no natural density. (Outline the construction; a full proof is not required.)

Hint

Block-construction: include long runs of primes alternating with long gaps so the $π_{A} (x) / π (x)$ ratio oscillates, but weight by $p^{- s}$ averages over the blocks.

Answer

Order the primes $p_{1} < p_{2} < \dots$ . Let $B_{k} = {p_{n} : 2^{2^{k}} \leq n < 2^{2^{k + 1}}}$ be the $k$ -th block. Define $A = ⋃_{k} B_{2 k}$ , taking every other block. At the right endpoint of $B_{2 k}$ , the ratio $π_{A} (x) / π (x)$ is close to $\sum_{j \leq k} ∣ B_{2 j} ∣/ \sum_{j \leq 2 k} ∣ B_{j} ∣ \approx (4^{k} - 1) / (4^{2 k} - 1) \cdot ratio shift$ , which oscillates as $k$ varies through even and odd values. So natural density does not exist. For Dirichlet density, the weighted contribution of each block to $\sum_{p \in A} p^{- s}$ smooths out via the $p^{- s}$ damping: by direct computation against the Chebyshev / Mertens estimate $\sum_{p \leq x} p^{- 1} = lo g lo g x + M + O (1/ lo g x)$ (Mertens 1874), the Dirichlet density of $A$ is exactly $1/2$ . Rubric: full credit for the block-construction outline plus the qualitative explanation of why $p^{- s}$ weighting averages the oscillation.

Exercise 8 (hard, short-answer).

State the Chebotarev density theorem and show that Dirichlet's theorem on primes in arithmetic progressions follows as the special case $K = Q$ , $L = Q (ζ_{m})$ .

Hint

The Galois group $Gal (Q (ζ_{m}) / Q)$ is isomorphic to $(Z / m)^{\times}$ , abelian, so its conjugacy classes are singletons. The Frobenius at a prime $p ∤ m$ corresponds to the residue $p mod m$ under this isomorphism.

Answer

Theorem (Chebotarev 1922). Let $L / K$ be a finite Galois extension of number fields with $G = Gal (L / K)$ , and let $C \subseteq G$ be a conjugacy class. The set ${p unramified prime of K : Frob_{p} \in C}$ has Dirichlet density $∣ C ∣/∣ G ∣$ .

Specialisation to Dirichlet's theorem. Take $K = Q$ and $L = Q (ζ_{m})$ , the $m$ -th cyclotomic field. Then $G = Gal (Q (ζ_{m}) / Q) ≅ (Z / m)^{\times}$ via the isomorphism $σ_{a} \mapsto a$ where $σ_{a} (ζ_{m}) = ζ_{m}^{a}$ . The Frobenius element at a prime $p ∤ m$ is $σ_{p}$ , the automorphism corresponding to $p mod m$ . Since $G$ is abelian, every conjugacy class $C$ is a singleton ${σ_{a}}$ , and $∣ C ∣/∣ G ∣ = 1/ φ (m)$ . The Chebotarev set ${p ∤ m : Frob_{p} = σ_{a}}$ is exactly ${p ∤ m : p \equiv a (mod m)} = P_{a, m}$ , with density $1/ φ (m)$ . This recovers Dirichlet's density theorem from the Galois-theoretic statement. Rubric: full credit for the statement, the isomorphism $G ≅ (Z / m)^{\times}$ , and the Frobenius identification.

Lean formalization Intermediate+

Mathlib has the partial Dirichlet-character infrastructure in Mathlib.NumberTheory.DirichletCharacter.Basic and a partial Riemann-zeta apparatus in Codex.NumberTheory.LFunctions.RiemannZeta, but ships no named Dirichlet-density theorem. The intended formalisation reads schematically:

import Mathlib.NumberTheory.DirichletCharacter.Basic
import Codex.NumberTheory.LFunctions.RiemannZeta
import Codex.NumberTheory.LFunctions.DirichletL

/-- Dirichlet density of a set of primes, when the limit exists. -/
noncomputable def dirichletDensity (A : Set Nat) : ℝ := sorry
  -- defined as `Filter.lim (𝓝[>] 1) (fun s => (∑ p ∈ A.filter Nat.Prime, (p : ℝ)^(-s))
  --                                            / (∑ p ∈ Nat.Prime, (p : ℝ)^(-s)))`

/-- Singular asymptotic: ∑_p p^(-s) ~ log(1/(s-1)) as s → 1^+. -/
theorem prime_sum_singular_asymptotic :
    ∀ ε > 0, ∃ δ > 0, ∀ s, 1 < s → s < 1 + δ →
      |(∑ p in (Nat.Prime.range s), (p : ℝ)^(-s)) - Real.log (1 / (s - 1))| < ε :=
  sorry

/-- Primes in arithmetic progression have Dirichlet density 1/φ(m). -/
theorem dirichlet_density_arithmetic_progression
    (m : Nat) (a : Nat) (hcop : Nat.Coprime a m) :
    dirichletDensity {p | p.Prime ∧ p ≡ a [MOD m]} = 1 / (Nat.totient m) :=
  sorry
  -- proof: orthogonality of Dirichlet characters + non-vanishing L(1, χ) ≠ 0
  -- (the latter is `L_nonvanish_at_one` in DirichletL)

The proof gap is substantial. Mathlib needs: the singular asymptotic $\sum_{p} p^{- s} = lo g (1/ (s - 1)) + O (1)$ as $s \to 1^{+}$ , derived from the Euler product for $ζ (s)$ and subdominance of the $n \geq 2$ tail; Dirichlet density itself as a limit in the filter $N_{> 1} (1)$ ; the orthogonality-average derivation of the density formula from the (also-unformalised) non-vanishing $L (1, χ) \neq = 0$ ; and the Abel-summation comparison establishing $δ (A) \Rightarrow d (A)$ direction. The Chebotarev density theorem is a further target, predicated on a complete Hecke / Artin $L$ -function infrastructure not currently in Mathlib.

Advanced results Master

Chebotarev density theorem and the Galois generalisation

Theorem (Chebotarev density theorem; Chebotarev 1922 Math. Ann. 95). Let $L / K$ be a finite Galois extension of number fields with Galois group $G = Gal (L / K)$ , and let $C \subseteq G$ be a conjugacy class. The set of unramified primes $p$ of $K$ whose Frobenius element $Frob_{p} \in C$ (well-defined as a conjugacy class) has Dirichlet density $$ d\big({\mathfrak{p} : \mathrm{Frob}_\mathfrak{p} \in C}\big) = \frac{|C|}{|G|}. $$

The theorem extends Dirichlet's theorem from cyclotomic extensions to arbitrary finite Galois extensions of number fields. The proof, modulo Hecke / Artin $L$ -function infrastructure, follows the same orthogonality-average pattern: the indicator of the conjugacy class $C$ is expressed as a sum over irreducible characters $χ$ of $G$ via the orthogonality of characters of finite groups, and the analytic input is the holomorphy and non-vanishing of $L (s, χ, L / K)$ at $s = 1$ for $χ$ different from the principal character. The non-principal irreducible characters are handled via Brauer induction (Brauer 1947), reducing to one-dimensional characters which are themselves Hecke / Dirichlet characters by class field theory (Artin reciprocity, Artin 1927).

The Chebotarev density theorem is the structural reason every density question about primes in number fields can be reduced to a Galois-theoretic question about the distribution of Frobenius elements. Applications include: density of primes representing a given binary quadratic form (Lagrange-Gauss), density of primes splitting in a given extension (equivalently, density of primes whose Frobenius is the identity, which by Chebotarev is $1/ [L : K]$ ), and density of supersingular primes for an elliptic curve (CM case gives density $1/2$ by Chebotarev applied to the Galois closure of the CM order).

Connection to the analytic class number formula

The Dirichlet-density theorem records the leading-order behaviour of an $s = 1$ residue analysis, but the same residue analysis carries deeper information. The analytic class number formula of Dirichlet (1839/40 Crelle's Journal 19, 21) ^{[source pending]} relates the residue of the Dedekind zeta function to the class number and regulator: for a number field $K$ of degree $n = r_{1} + 2 r_{2}$ over $Q$ , $$ \mathrm{res}_{s = 1} \zeta_K(s) = \frac{2^{r_1} (2\pi)^{r_2} h_K R_K}{w_K \sqrt{|d_K|}}, $$ with $h_{K}$ the class number, $R_{K}$ the regulator, $w_{K}$ the number of roots of unity, and $d_{K}$ the discriminant. The formula refines the leading $lo g (1/ (s - 1))$ singularity from a " $1/ φ (m)$ share of unity" to an arithmetic identity coupling five invariants of the number field to the analytic residue. Dirichlet originally proved both his density theorem and this formula together — the non-vanishing $L (1, χ) \neq = 0$ in the real-character case follows from the strict positivity of the right-hand side via Artin's factorisation $ζ_{K} (s) = ζ (s) L (s, χ)$ for the quadratic field $K = Q (D)$ .

Tate's thesis and the adelic viewpoint

Tate's thesis (Tate 1950, published in Cassels-Fröhlich 1967, Ch. XV) ^{[source pending]} recasts the entire Dirichlet / Hecke $L$ -function machinery as a Fourier-analytic computation on the idele class group $A_{K}^{\times} / K^{\times}$ of a number field. The starting point is the global zeta integral $$ Z(s, \chi, f) = \int_{\mathbb{A}_K^\times} f(x) \chi(x) |x|^s d^\times x, $$ where $f$ is a Schwartz-Bruhat function on $A_{K}$ and $χ$ is a Hecke character. For an appropriate choice of $f$ , the zeta integral specialises to a Hecke $L$ -function with archimedean Gamma factors; the functional equation $Z (s, χ, f) = Z (1 - s, χ^{- 1}, \hat{f})$ follows from Pontryagin self-duality of the adeles $A_{K}$ and the Poisson summation formula on the discrete subgroup $K \subset A_{K}$ .

The Dirichlet density of a set of primes $A$ , in this framework, is the residue at $s = 1$ of the partial zeta integral against the indicator function of $A$ on the finite-place part of the idele class group, divided by the residue at $s = 1$ of the full zeta integral. The orthogonality-average derivation of $d (P_{a, m}) = 1/ φ (m)$ becomes a Fourier-inversion on the finite quotient $A_{Q}^{\times} / (Q^{\times} \prod_{p} Z_{p}^{\times}) ≅ (Z / m)^{\times}$ at finite level. The Chebotarev density theorem becomes the Plancherel theorem on the Galois-extension idele class group, with the Frobenius-class indicator decomposed across irreducible representations of $G$ . The adelic viewpoint is the foundational reason every density theorem in number theory is part of one structural framework.

The Tate-thesis reformulation appears again in 21.03.03 (Dedekind, Hecke, Artin $L$ -functions) and forms the $GL_{1}$ -prototype of the Langlands programme; it builds toward the higher-rank automorphic $L$ -functions where density theorems of Sato-Tate type (density of Frobenius eigenvalue distribution in $SU (2)$ -conjugacy classes) generalise Chebotarev to non-abelian Galois representations.

Quantitative density: the PNT in arithmetic progressions

The Dirichlet-density theorem is qualitative — it says $d (P_{a, m}) = 1/ φ (m)$ exists. The quantitative refinement is the prime-number theorem in arithmetic progressions (de la Vallée Poussin 1896 in the modulus-fixed case; Page 1935 and Siegel 1935 uniformly in modulus, with the ineffective Siegel bound): $$ \pi(x; m, a) := |{p \leq x : p \equiv a \pmod m}| = \frac{1}{\varphi(m)} \frac{x}{\log x} + O!\left(\frac{x}{(\log x)^2}\right) \qquad (x \to \infty, , m \text{ fixed}). $$

The error term sharpens with the zero-free region of $L (s, χ)$ via the explicit formula. Siegel's theorem (Siegel 1935) gives an effective error $O_{ε} (x e^{- c l o g x})$ uniformly for $m \leq (lo g x)^{A}$ for any $A$ , but only ineffectively when allowed to depend on $ε$ . Linnik's theorem (Linnik 1944 Mat. Sb. 15) ^{[source pending]} bounds the least prime $p \equiv a (mod m)$ by $≪ m^{L}$ for an absolute constant $L$ , now $L \leq 5$ (Xylouris 2011 Acta Arith. 150); conjecturally $L = 1 + ε$ under the generalised Riemann hypothesis.

The qualitative-quantitative gap is real. Dirichlet density tells you the share is $1/ φ (m)$ in the limit; PNT in AP tells you the share is approached at a definite rate, with an error controlled by the zero-free region. Sieve methods (Bombieri-Vinogradov 1965, the large sieve) extend the uniform-modulus regime, and the equidistribution of primes in arithmetic progressions is one of the central open questions of analytic number theory under the Elliott-Halberstam conjecture and the Generalised Riemann Hypothesis for Dirichlet $L$ -functions.

Synthesis. Dirichlet density is the foundational tool that builds toward every density theorem for primes in number theory, and the central insight is that the singular behaviour of $lo g ζ (s)$ at $s = 1$ — namely $lo g ζ (s) = lo g (1/ (s - 1)) + O (1)$ — combines with orthogonality of characters and the non-vanishing of $L (s, χ)$ at $s = 1$ to extract the share $1/ φ (m)$ that an arithmetic progression captures of the total prime mass. This is exactly the structural pattern that appears again in 21.03.03 (Dedekind, Hecke, Artin $L$ -functions) under the heading of the Chebotarev density theorem, generalises through the analytic class number formula whose residue analysis refines the leading singularity to a full arithmetic identity, and is dual to the Tate-thesis viewpoint that recasts the entire density framework as a Plancherel computation on the idele class group.

Putting these together, the Dirichlet-density / Chebotarev / Tate-thesis chain identifies the analytic kernel of every number-theoretic density question with the residue of a $GL_{1}$ zeta integral at $s = 1$ , and the central insight is that this residue is governed by a single non-vanishing statement: $L (1, χ) \neq = 0$ for non-principal Hecke characters. The foundational reason this works is that an $L$ -function is the analytic continuation of a Dirichlet series whose Euler factors encode multiplicative structure on the primes, and the bridge is the recognition that the share of primes captured by a conjugacy class $C \subseteq G$ is the average of the irreducible characters of $G$ evaluated against $C$ , weighted by the analytic input from the corresponding Hecke / Artin $L$ -functions. This pattern recurs in higher rank: the Sato-Tate density (Frobenius eigenvalue distribution in $SU (2)$ -conjugacy classes for elliptic curves over $Q$ , proved by Clozel-Harris-Shepherd-Barron-Taylor 2008), the Langlands density questions for automorphic $L$ -functions, and the equidistribution of Hecke eigenvalues on $GL_{n}$ -automorphic forms all share the same residue-at- $s = 1$ analytic engine.

Full proof set Master

Theorem (Dirichlet's density formula), proof. Given in the Intermediate-tier section above: orthogonality of characters reduces the residue-class indicator to a sum over the character group of $(Z / m)^{\times}$ , the principal-character contribution gives the full $lo g (1/ (s - 1))$ divergence at $s = 1^{+}$ , every non-principal-character contribution is bounded by the non-vanishing $L (1, χ) \neq = 0$ , and division by $\sum_{p} p^{- s} = lo g (1/ (s - 1)) + O (1)$ extracts the share $1/ φ (m)$ . $□$

Theorem (singular asymptotic $\sum_{p} p^{- s} = lo g (1/ (s - 1)) + O (1)$ ), full proof. Begin with the Euler product $ζ (s) = \prod_{p} (1 - p^{- s})^{- 1}$ valid for $Re (s) > 1$ . Take the logarithm: $$ \log \zeta(s) = \sum_p -\log(1 - p^{-s}) = \sum_p \sum_{n \geq 1} \frac{p^{-ns}}{n}. $$ Split the inner sum at $n = 1$ : $$ \log \zeta(s) = \sum_p p^{-s} + \sum_p \sum_{n \geq 2} \frac{p^{-ns}}{n} =: \sum_p p^{-s} + R(s). $$ Bound the remainder. For each prime $p \geq 2$ , the inner sum is geometric-dominated: $$ \sum_{n \geq 2} \frac{p^{-ns}}{n} \leq \sum_{n \geq 2} p^{-ns} = \frac{p^{-2s}}{1 - p^{-s}} \leq \frac{p^{-2}}{1 - 2^{-1}} = 2 p^{-2} $$ for $s \geq 1$ . Summing over primes, $$ |R(s)| \leq 2 \sum_p p^{-2} \leq 2 \sum_{n \geq 2} n^{-2} = 2(\zeta(2) - 1) < 2. $$ So $R (s) = O (1)$ as $s \to 1^{+}$ .

On the other side, $ζ (s) = \frac{1}{s - 1} + γ + O (s - 1)$ as $s \to 1$ (simple pole with residue $1$ , Euler-Mascheroni constant $γ$ as the next term). Taking logarithms, $$ \log \zeta(s) = \log \frac{1}{s - 1} + \log(1 + \gamma(s - 1) + O((s - 1)^2)) = \log \frac{1}{s - 1} + O(1) $$ as $s \to 1^{+}$ . Subtracting the bounded $R (s)$ , $$ \sum_p p^{-s} = \log \zeta(s) - R(s) = \log \frac{1}{s - 1} + O(1). \quad \square $$

Theorem (natural density implies Dirichlet density), full proof. Suppose $δ (A) = δ$ exists. Set $π_{A} (x) = ∣ {p \in A : p \leq x} ∣$ and $π (x) = ∣ {p : p \leq x} ∣$ . Hypothesis: $π_{A} (x) = δ π (x) + o (π (x))$ as $x \to \infty$ .

Apply Abel summation. For $s > 1$ , $$ \sum_{p \in A} p^{-s} = \sum_{p \in A} \int_p^\infty s t^{-s - 1} dt = s \int_2^\infty \pi_A(t) t^{-s - 1} dt, $$ where we have used $\int_{p}^{\infty} s t^{- s - 1} d t = p^{- s}$ and interchanged sum and integral via Fubini (justified by absolute convergence for $s > 1$ ). Substituting the asymptotic $π_{A} (t) = δ π (t) + o (π (t))$ , $$ \sum_{p \in A} p^{-s} = \delta \cdot s \int_2^\infty \pi(t) t^{-s - 1} dt + s \int_2^\infty o(\pi(t)) t^{-s - 1} dt. $$ The first integral is exactly $\sum_{p} p^{- s}$ by the same Abel summation applied to the full prime set. The error integral: write $o (π (t)) = ε (t) π (t)$ with $ε (t) \to 0$ . Then $$ \frac{s \int_2^\infty \varepsilon(t) \pi(t) t^{-s - 1} dt}{\sum_p p^{-s}} = \frac{\int_2^\infty \varepsilon(t) \pi(t) t^{-s - 1} dt}{\int_2^\infty \pi(t) t^{-s - 1} dt} \to 0 \quad (s \to 1^+), $$ since the integrands are dominated by $π (t) t^{- s - 1}$ (integrable for $s > 1$ ), the ratio $ε (t) \to 0$ as $t \to \infty$ , and as $s \to 1^{+}$ the measure $π (t) t^{- s - 1} d t$ concentrates on $t \to \infty$ . (Apply dominated convergence after normalising.)

Therefore $$ \frac{\sum_{p \in A} p^{-s}}{\sum_p p^{-s}} = \delta + o(1) \quad (s \to 1^+), $$ and $d (A) = δ = δ (A)$ . $□$

Theorem (worked example: $d (P_{1, 4}) = 1/2$ ), proof. Apply the Dirichlet-density formula with $m = 4$ , $a = 1$ , $φ (4) = 2$ : the density is $1/ φ (4) = 1/2$ . Explicitly, the two Dirichlet characters mod $4$ are $χ_{0}$ (principal, value $1$ on ${1, 3}$ and $0$ elsewhere) and $χ_{1}$ (non-principal real, $χ_{1} (1) = + 1$ , $χ_{1} (3) = - 1$ , $χ_{1} (0) = χ_{1} (2) = 0$ ). The orthogonality average for the indicator of $1 mod 4$ is $1_{p \equiv 1} = \frac{1}{2} (χ_{0} (p) + χ_{1} (p))$ . Summing $p^{- s}$ : $$ \sum_{p \equiv 1 \bmod 4} p^{-s} = \tfrac{1}{2} \sum_p \chi_0(p) p^{-s} + \tfrac{1}{2} \sum_p \chi_1(p) p^{-s}. $$ First sum: $\sum_{p} χ_{0} (p) p^{- s} = \sum_{p ∤ 4} p^{- s} = \sum_{p} p^{- s} - 2^{- s} = lo g (1/ (s - 1)) + O (1)$ . Second sum: $\sum_{p} χ_{1} (p) p^{- s} = lo g L (s, χ_{1}) + O (1)$ from the Euler-product log, and $L (s, χ_{1})$ extends to a function holomorphic on $Re (s) > 0$ with $L (1, χ_{1}) = \sum_{n \geq 1} χ_{1} (n) / n = 1 - 1/3 + 1/5 - 1/7 + \dots = π /4 \neq = 0$ (Leibniz 1674 / Madhava ca. 1400). So $lo g L (s, χ_{1})$ stays bounded as $s \to 1^{+}$ , and the second sum contributes $O (1)$ . Adding, $$ \sum_{p \equiv 1 \bmod 4} p^{-s} = \tfrac{1}{2} \log \frac{1}{s - 1} + O(1). $$ Dividing by $\sum_{p} p^{- s} = lo g (1/ (s - 1)) + O (1)$ and taking $s \to 1^{+}$ gives $d (P_{1, 4}) = 1/2$ . $□$

Theorem (Chebotarev density theorem), stated without proof — see Lang 1994 Algebraic Number Theory (GTM 110, 2nd ed.) Ch. VIII or Neukirch 1999 Ch. VII §13 ^{[source pending]}. The full proof requires the analytic continuation, holomorphy at $s = 1$ , and non-vanishing $L (1, χ, L / K) \neq = 0$ of Hecke and Artin $L$ -functions for non-principal irreducible characters of $G = Gal (L / K)$ , plus Brauer induction to reduce higher-dimensional characters to one-dimensional ones (which are Hecke characters by Artin reciprocity, Artin 1927). The orthogonality-average pattern then runs identically to the Dirichlet case, with $∣ C ∣/∣ G ∣$ replacing $1/ φ (m)$ as the principal-character share.

Connections Master

Dirichlet $L$ -functions 21.03.02. The Dirichlet-density theorem is downstream of the non-vanishing $L (1, χ) \neq = 0$ for non-principal characters; that non-vanishing is the only analytic input the density argument requires. The sibling unit proves the non-vanishing in two cases (complex characters via the Landau positive-coefficient theorem, real characters via the analytic class number formula); this unit consumes the result and extracts the density formula via orthogonality of characters. The bridge between the two units is the recognition that the analytic behaviour of $L (s, χ)$ at $s = 1$ controls the distribution of primes across residue classes.
Riemann zeta function 21.03.01. The singular asymptotic $\sum_{p} p^{- s} = lo g (1/ (s - 1)) + O (1)$ as $s \to 1^{+}$ is the foundational input that makes Dirichlet density well-defined. The asymptotic is derived from $lo g ζ (s) = lo g (1/ (s - 1)) + O (1)$ via the Euler product and subdominance of the $n \geq 2$ tail. Every density theorem in number theory begins with this singular-behaviour analysis, and the Riemann zeta function is the universal denominator against which other prime sums are compared.
Dedekind / Hecke / Artin $L$ -functions 21.03.03. The Chebotarev density theorem is the Galois generalisation of Dirichlet's density formula, and the analytic input is the holomorphy and non-vanishing at $s = 1$ of Hecke and Artin $L$ -functions for non-principal characters of the Galois group. The sibling unit develops the Hecke / Artin $L$ -function framework that the Chebotarev proof consumes; this unit consumes a slice of the framework at the level of cyclotomic extensions, and the connection between the two is the recognition that the Galois group $Gal (Q (ζ_{m}) / Q) ≅ (Z / m)^{\times}$ identifies Dirichlet characters with one-dimensional Galois characters.
Modular forms on $SL_{2} (Z)$ 21.04.01. The same residue-at- $s = 1$ analytic engine drives the Sato-Tate density of Frobenius eigenvalues for modular newforms, generalising Dirichlet density from $GL_{1}$ (residue classes) to $GL_{2}$ (Frobenius eigenvalue distribution in $SU (2)$ -conjugacy classes). The forward connection: density theorems for modular $L$ -functions are the higher-rank analogues of Dirichlet's density theorem, and the analytic continuation / non-vanishing input is supplied by the modular-forms unit and its Hecke-operator successors.
Infinite series convergence 02.03.03. The analytic prerequisite for Dirichlet density is the convergence of the prime sum $\sum_{p} p^{- s}$ on $Re (s) > 1$ , which is the simplest case of a Dirichlet series convergence theorem. The sibling unit on infinite-series convergence provides the foundational tools (comparison test, Dirichlet test, Abel summation) that the density-formula proof uses repeatedly.

Historical & philosophical context Master

Dirichlet stated and proved the theorem on primes in arithmetic progressions in 1837 in the Abhandlungen der Königlich Preussischen Akademie der Wissenschaften zu Berlin ^{[Dirichlet 1837]}. The proof was the first time in mathematics that complex analysis (or its real-variable precursor in Dirichlet's hands) was applied to settle a question about whole numbers; it established analytic number theory as a discipline. Dirichlet introduced characters $χ : (Z / m)^{\times} \to C^{\times}$ (the term character and the formalism are his), constructed the $L$ -series $L (s, χ) = \sum_{n} χ (n) / n^{s}$ , proved the non-vanishing $L (1, χ) \neq = 0$ for non-principal $χ$ , and derived the infinitude of primes in any arithmetic progression $a + m Z$ with $g cd (a, m) = 1$ . The 1837 paper does not use the modern definition of Dirichlet density in the form $lim_{s \to 1^{+}} \sum_{p \in A} p^{- s} / \sum_{p} p^{- s}$ ; Dirichlet's argument proceeds via the divergence $\sum_{p \in P_{a, m}} p^{- s} \to \infty$ as $s \to 1^{+}$ , which is logically equivalent but does not extract the rate. The density-rate refinement, namely $\sum_{p \in P_{a, m}} p^{- s} \sim \frac{1}{φ ( m )} lo g (1/ (s - 1))$ , is implicit in the 1837 argument and made explicit in subsequent treatments (Dirichlet 1839/40 Crelle's Journal 19, 21 for the class-number formula refinement ^{[Dirichlet 1839]}).

The density-theoretic packaging emerged later. Chebotarev's 1922 paper in Mathematische Annalen ^{[Chebotarev 1922]}, Die Bestimmung der Dichtigkeit einer Menge von Primzahlen, welche zu einer gegebenen Substitutionsklasse gehören, generalised Dirichlet's theorem from cyclotomic extensions to arbitrary finite Galois extensions $L / K$ , with the share of primes whose Frobenius lies in a conjugacy class $C \subseteq G = Gal (L / K)$ given by $∣ C ∣/∣ G ∣$ . Chebotarev's original proof used an indirect reduction to cyclotomic extensions; the analytic proof via Hecke / Artin $L$ -functions was filled in by Artin 1924 (the analytic argument, though Artin's 1923 paper on Artin $L$ -functions ^{[source pending]} predates Chebotarev's publication). Hecke's 1918 and 1920 Mathematische Zeitschrift papers ^{[Hecke 1918]} provided the technical Hecke $L$ -function infrastructure — analytic continuation, functional equation via theta inversion, non-vanishing at $s = 1$ — on which Chebotarev's analytic proof depends. The cyclotomic special case proves the strongest possible version of Dirichlet's theorem via the route Chebotarev intended; Artin 1927 Hamburg Abh. 5 ^{[Artin 1927]} completed the picture with Artin reciprocity, identifying one-dimensional Artin $L$ -functions with Hecke $L$ on the idele class group.

Tate's 1950 Princeton PhD thesis, Fourier analysis in number fields and Hecke's zeta functions, published in Cassels-Fröhlich 1967 ^{[Tate 1950]}, recast the entire Dirichlet / Hecke $L$ -function framework as global zeta integrals on the idele class group $A_{K}^{\times} / K^{\times}$ , with the functional equation derived from Pontryagin self-duality of the adeles and Poisson summation on the discrete subgroup $K \subset A_{K}$ . In this framework, Dirichlet density is the residue at $s = 1$ of a $GL_{1}$ zeta integral, the orthogonality of characters becomes Fourier inversion on a finite quotient of the idele class group, and the Chebotarev density theorem is the Plancherel theorem on the Galois-extension idele class group. The Tate-thesis reformulation is the foundational ancestor of the Langlands programme, which extends the density framework to higher-rank automorphic $L$ -functions (Sato-Tate density for elliptic curves over $Q$ , proved by Clozel-Harris-Shepherd-Barron-Taylor 2008-2010 Publ. Math. IHES; Langlands functoriality density questions still largely open). The modern textbook accounts in Lang 1994 Algebraic Number Theory Ch. VIII ^{[Lang 1994]}, Neukirch 1999 Algebraic Number Theory Ch. VII §13 ^{[Neukirch 1999]}, and Bump 1997 Automorphic Forms and Representations Ch. 1, 3 ^{[Bump 1997]} present the Dirichlet / Chebotarev / Tate-thesis chain as one structural framework with the orthogonality of characters / Plancherel theorem as the analytic engine.

Bibliography Master

@article{Dirichlet1837,
  author  = {Dirichlet, Peter Gustav Lejeune},
  title   = {Beweis des Satzes, dass jede unbegrenzte arithmetische Progression, deren erstes Glied und Differenz ganze Zahlen ohne gemeinschaftlichen Factor sind, unendlich viele Primzahlen enth{\"a}lt},
  journal = {Abhandlungen der K{\"o}niglich Preussischen Akademie der Wissenschaften zu Berlin},
  year    = {1837},
  pages   = {45--81}
}

@article{Dirichlet1839CrelleClassNumber,
  author  = {Dirichlet, Peter Gustav Lejeune},
  title   = {Recherches sur diverses applications de l'analyse infinit{\'e}simale {\`a} la th{\'e}orie des nombres},
  journal = {Journal f{\"u}r die reine und angewandte Mathematik (Crelle's Journal)},
  volume  = {19},
  year    = {1839},
  pages   = {324--369}
}

@article{Chebotarev1922,
  author  = {Chebotarev, Nikolai G.},
  title   = {Die Bestimmung der Dichtigkeit einer Menge von Primzahlen, welche zu einer gegebenen Substitutionsklasse geh{\"o}ren},
  journal = {Mathematische Annalen},
  volume  = {95},
  year    = {1922},
  pages   = {191--228}
}

@article{Hecke1918,
  author  = {Hecke, Erich},
  title   = {Eine neue Art von Zetafunktionen und ihre Beziehungen zur Verteilung der Primzahlen, I},
  journal = {Mathematische Zeitschrift},
  volume  = {1},
  year    = {1918},
  pages   = {357--376}
}

@article{Hecke1920,
  author  = {Hecke, Erich},
  title   = {Eine neue Art von Zetafunktionen und ihre Beziehungen zur Verteilung der Primzahlen, II},
  journal = {Mathematische Zeitschrift},
  volume  = {6},
  year    = {1920},
  pages   = {11--51}
}

@article{Artin1923,
  author  = {Artin, Emil},
  title   = {{\"U}ber eine neue Art von $L$-Reihen},
  journal = {Abhandlungen aus dem Mathematischen Seminar der Universit{\"a}t Hamburg},
  volume  = {3},
  year    = {1923},
  pages   = {89--108}
}

@article{Artin1927Reciprocity,
  author  = {Artin, Emil},
  title   = {Beweis des allgemeinen Reziprozit{\"a}tsgesetzes},
  journal = {Abhandlungen aus dem Mathematischen Seminar der Universit{\"a}t Hamburg},
  volume  = {5},
  year    = {1927},
  pages   = {353--363}
}

@phdthesis{Tate1950Thesis,
  author = {Tate, John T.},
  title  = {Fourier Analysis in Number Fields and Hecke's Zeta-Functions},
  school = {Princeton University},
  year   = {1950},
  note   = {Reprinted in Cassels-Fr{\"o}hlich (eds.) \textit{Algebraic Number Theory}, Thompson Book Co., Washington, 1967, Ch. XV, 305--347}
}

@book{Serre1973CourseArithmetic,
  author    = {Serre, Jean-Pierre},
  title     = {A Course in Arithmetic},
  publisher = {Springer-Verlag},
  series    = {Graduate Texts in Mathematics},
  volume    = {7},
  year      = {1973}
}

@book{Lang1994ANT,
  author    = {Lang, Serge},
  title     = {Algebraic Number Theory},
  publisher = {Springer-Verlag},
  series    = {Graduate Texts in Mathematics},
  volume    = {110},
  edition   = {2},
  year      = {1994}
}

@book{Neukirch1999ANT,
  author    = {Neukirch, J{\"u}rgen},
  title     = {Algebraic Number Theory},
  publisher = {Springer-Verlag},
  series    = {Grundlehren der mathematischen Wissenschaften},
  volume    = {322},
  year      = {1999}
}

@book{Apostol1976IntroANT,
  author    = {Apostol, Tom M.},
  title     = {Introduction to Analytic Number Theory},
  publisher = {Springer-Verlag},
  series    = {Undergraduate Texts in Mathematics},
  year      = {1976}
}

@book{Davenport2000MNT,
  author    = {Davenport, Harold},
  title     = {Multiplicative Number Theory},
  publisher = {Springer-Verlag},
  series    = {Graduate Texts in Mathematics},
  volume    = {74},
  edition   = {3},
  year      = {2000},
  note     = {Revised by Hugh L. Montgomery}
}

@book{IwaniecKowalski2004,
  author    = {Iwaniec, Henryk and Kowalski, Emmanuel},
  title     = {Analytic Number Theory},
  publisher = {American Mathematical Society},
  series    = {Colloquium Publications},
  volume    = {53},
  year      = {2004}
}

@book{Bump1997Automorphic,
  author    = {Bump, Daniel},
  title     = {Automorphic Forms and Representations},
  publisher = {Cambridge University Press},
  series    = {Cambridge Studies in Advanced Mathematics},
  volume    = {55},
  year      = {1997}
}

@article{Linnik1944,
  author  = {Linnik, Yuri V.},
  title   = {On the least prime in an arithmetic progression, I, II},
  journal = {Recueil Math{\'e}matique (Mat. Sbornik) N.S.},
  volume  = {15(57)},
  year    = {1944},
  pages   = {139--178, 347--368}
}

@article{ClozelHarrisShepherdBarronTaylor2008SatoTate,
  author  = {Clozel, Laurent and Harris, Michael and Shepherd-Barron, Nicholas and Taylor, Richard},
  title   = {A family of Calabi-Yau varieties and potential automorphy, I, II, III},
  journal = {Publ. Math. IH{\'E}S},
  year    = {2008--2010},
  note    = {Sato-Tate conjecture for elliptic curves over $\mathbb{Q}$ without CM}
}

Prerequisites

21.03.01
21.03.02
02.03.03

Tier anchors

beginner: Serre 1973 *A Course in Arithmetic* (GTM 7) Ch. VI §1 informal opening — counting primes by residue class via a Dirichlet-series weight; Apostol *Introduction to Analytic Number Theory* Ch. 7 §7.4 informal motivation
intermediate: Serre 1973 *A Course in Arithmetic* (GTM 7) Ch. VI §§3-4 (originator anchor — Dirichlet density, the logarithmic singularity of $\log \zeta(s)$ at $s = 1^+$, the density-$1/\varphi(m)$ theorem); Apostol 1976 *Introduction to Analytic Number Theory* (Springer UTM) Ch. 7 §§7.4-7.5 (Dirichlet density inside the proof of Dirichlet's theorem); Davenport 2000 *Multiplicative Number Theory* (Springer GTM 74) Ch. 1, 7
master: Serre 1973 *A Course in Arithmetic* Ch. VI §§3-4 (originator anchor for the Dirichlet-density viewpoint, with the cyclotomic / Galois-theoretic packaging carried forward in Ch. VI §3 Theorem 2); Dirichlet 1837 *Abhandl. Königl. Preuss. Akad.* 45-81 (originator of primes-in-AP in density-counting form); Chebotarev 1922 *Math. Ann.* 95, 191-228 (Chebotarev density theorem; Galois-theoretic generalisation); Hecke 1918 *Math. Z.* 1, 357-376 + 1920 *Math. Z.* 6, 11-51 (Hecke $L$-functions for number fields, the analytic-continuation engine driving Chebotarev); Tate 1950 Princeton PhD thesis (published in Cassels-Fröhlich 1967 *Algebraic Number Theory* Ch. XV, 305-347; adelic / Fourier-analytic reformulation, idele-class-group zeta integrals); Lang 1994 *Algebraic Number Theory* (Springer GTM 110, 2nd ed.) Ch. VIII (modern proof of the Chebotarev density theorem from Hecke / Artin $L$); Neukirch 1999 *Algebraic Number Theory* (Springer Grundlehren 322) Ch. VII §13 (Chebotarev + density formula in modern form); Iwaniec-Kowalski 2004 *Analytic Number Theory* (AMS Colloquium 53) Ch. 5 §5.6 (quantitative versions, the PNT in arithmetic progressions); Davenport 2000 *Multiplicative Number Theory* (Springer GTM 74) Ch. 7; Bump 1997 *Automorphic Forms and Representations* (Cambridge Studies in Advanced Mathematics 55) Ch. 1, 3 (Tate-thesis density viewpoint)

References

Serre — A Course in Arithmetic · Springer Graduate Texts in Mathematics 7, 1973; Ch. VI §§3-4 (Dirichlet density, the logarithmic singularity $\sum_p p^{-s} = \log(1/(s-1)) + O(1)$ as $s \to 1^+$, Theorem 2 — the density-$1/\varphi(m)$ formula for primes in arithmetic progressions, derived from the non-vanishing $L(1, \chi) \neq 0$ for non-principal $\chi$ established in Ch. V).
Dirichlet 1837 — Beweis des Satzes, dass jede unbegrenzte arithmetische Progression… · *Abhandlungen der Königlich Preussischen Akademie der Wissenschaften zu Berlin*, 1837, 45-81; originator paper for the theorem on primes in arithmetic progressions; the analytic statement, in modern Dirichlet-density language, is $d(P_{a, m}) = 1/\varphi(m)$ for every residue $a$ with $\gcd(a, m) = 1$.
Chebotarev 1922 — Die Bestimmung der Dichtigkeit einer Menge von Primzahlen, welche zu einer gegebenen Substitutionsklasse gehören · *Mathematische Annalen* 95 (1922), 191-228; originator paper for the Chebotarev density theorem: for a finite Galois extension $L/K$ with $G = \mathrm{Gal}(L/K)$ and any conjugacy class $C \subseteq G$, the density of unramified primes $\mathfrak{p}$ with $\mathrm{Frob}_\mathfrak{p} \in C$ is $|C|/|G|$. Dirichlet's theorem on primes in arithmetic progressions is the $K = \mathbb{Q}$, $L = \mathbb{Q}(\zeta_m)$ case.
Hecke 1918 / 1920 — Eine neue Art von Zetafunktionen und ihre Beziehungen zur Verteilung der Primzahlen · *Mathematische Zeitschrift* 1 (1918), 357-376 and 6 (1920), 11-51; Hecke characters of the idele class group of a number field, Hecke $L$-functions, analytic continuation and functional equation. The number-field generalisation of Dirichlet $L$ that powers Chebotarev density at the level of Hecke / Artin $L$.
Tate 1950 — Fourier analysis in number fields and Hecke's zeta functions · Princeton PhD thesis, 1950; published in Cassels-Fröhlich (eds.) *Algebraic Number Theory* (Thompson Book Co., Washington, 1967), Ch. XV, 305-347. Adelic / Fourier-analytic reformulation of the analytic continuation and functional equation of Hecke $L$-functions; the Dirichlet-density formula becomes the residue at $s = 1$ of a $\mathrm{GL}_1$ zeta integral over $\mathbb{A}_K^\times$ against a Schwartz-Bruhat function.
Lang — Algebraic Number Theory · Springer Graduate Texts in Mathematics 110, 2nd edition, 1994; Ch. VIII (Cyclotomic fields, Dirichlet's theorem, Chebotarev density theorem with proof from the analytic-continuation properties of Hecke $L$-functions).
Neukirch — Algebraic Number Theory · Springer Grundlehren 322, 1999; Ch. VII §13 (Density theorems: Dirichlet density, the Chebotarev density theorem with full proof via Hecke / Artin $L$-functions, and the consequences for the distribution of Frobenius elements).
Apostol — Introduction to Analytic Number Theory · Springer Undergraduate Texts in Mathematics, 1976; Ch. 7 §§7.4-7.5 (Dirichlet's theorem in density form, with the orthogonality-average derivation and the dependence on $L(1, \chi) \neq 0$).
Davenport — Multiplicative Number Theory · Springer Graduate Texts in Mathematics 74, 3rd ed. (revised Montgomery), 2000; Ch. 1, 7 (Dirichlet's theorem on primes in arithmetic progressions, density-form proofs, PNT in arithmetic progressions).
Iwaniec-Kowalski — Analytic Number Theory · *AMS Colloquium Publications* 53, 2004; Ch. 5 §5.6 (quantitative density: the prime-number theorem in arithmetic progressions, $\pi(x; m, a) \sim x/(\varphi(m) \log x)$, with the error term controlled by the zero-free region of $L(s, \chi)$).
Bump — Automorphic Forms and Representations · *Cambridge Studies in Advanced Mathematics* 55, Cambridge University Press, 1997; Ch. 1 §1.6, Ch. 3 (Tate-thesis density viewpoint: idele-class-group Hecke characters, $\mathrm{GL}_1$ zeta integrals, and the Mellin-transform recasting of the Dirichlet-density formula).
Dirichlet 1839/40 — Recherches sur diverses applications de l'analyse infinitésimale à la théorie des nombres · *Journal für die reine und angewandte Mathematik* (Crelle's Journal) 19 (1839), 324-369 and 21 (1840), 1-12, 134-155; the analytic class-number formula $\mathrm{res}_{s = 1} \zeta_K(s) = 2^{r_1} (2\pi)^{r_2} h_K R_K / (w_K \sqrt{|d_K|})$, the deeper $s = 1$ residue analysis whose first-order statement is the density-$1/\varphi(m)$ formula.
Artin 1927 — Beweis des allgemeinen Reziprozitätsgesetzes · *Abhandlungen aus dem Mathematischen Seminar der Universität Hamburg* 5 (1927), 353-363; Artin reciprocity, the isomorphism $C_K/N_{L/K}C_L \cong \mathrm{Gal}(L/K)^{\mathrm{ab}}$ on the idele class group, identifying one-dimensional Artin $L$ with Hecke $L$ and supplying the analytic input for Chebotarev density at one dimension before Brauer induction.

Estimated time

beginner: 18m
intermediate: 40m
master: 80m