Analytic number theory

Analytic number theory studies integers, prime numbers, and arithmetic functions using tools from real analysis and complex analysis, especially Dirichlet series and their singularities.

The circle method represents a weighted additive count as an integral of products of exponential sums against a character. Near rationals with small denominator, major arcs produce a structured main term involving a singular series and an archimedean singular integral. Away from those rationals, minor arcs require cancellation bounds. The partition and approximation estimates depend on the problem; a formal integral alone is not a positivity proof.

Every sufficiently large odd integer is a sum of three primes, with repetitions allowed. The weighted count is asymptotic to $\mathfrak{S}(n)n^2/2$. Siegel–Walfisz theorem controls the major arcs, and Type I/II estimates for the prime exponential sum make the minor arcs negligible. Removing prime-power contributions gives genuine prime representations. This is the sufficiently-large theorem, not by itself a proof for every small odd integer.

A singular series assembles the local congruence densities in an additive counting asymptotic. For three prime summands it is ${\sum }_{q\ge 1}\mu (q)c_q(n)/\phi (q{)}^3$, with Ramanujan sum $c_q(n)$. Its Euler product vanishes for even $n$ and has a uniform positive lower bound for odd $n$. Different additive problems have different local factors, so this is a specific instance rather than a universal formula.

Minor arcs are the complement of the specified major arcs. They contribute an error only when a suitable pointwise or mean-value cancellation estimate is proved; their name does not imply that their measure is small.

In a circle-method problem of scale $n$, major arcs are neighborhoods of reduced rational points $a/q$ with small denominator, where the relevant exponential sum is approximated using distribution in residue classes. For the three-prime problem one may take $q\le (\log n{)}^B$ and $|\theta -a/q|\le (\log n{)}^B/n$, with suitably fixed $B$.

An L-function is an arithmetic function given initially by a Dirichlet series and often an Euler product, with analytic continuation and a functional equation in important cases. Dirichlet L-functions and L-functions of cusp forms are basic examples; p-adic L-functions interpolate suitably modified special values.

The Hasse–Weil zeta function of an algebraic variety over a number field is assembled from local point-counting factors at finite places. At a good place with residue field of size $q$, the factor is $Z(V_q,q^{-s})$, where $Z(V_q,T)=\exp ({\sum }_{n\ge 1}\#V_q({\mathbb{F}}_{q^n})T^n/n)$. Defining all bad-place factors requires a specified integral model or convention. The zeta function of an elliptic curve over a finite field is a local factor, rather than this global product.

For an elliptic curve over a number field, its Hasse-Weil L-function is the product of the local factors determined by its Tate module. At a good finite place, the factor is $(1-a_{v}N{v}^{-s}+N{v}^{1-2s}{)}^{-1}$, with $a_v=Nv+1-\#E(\kappa (v))$. At a bad place use the inertia-invariant subspace and an auxiliary prime different from the residue characteristic. For a CM elliptic curve with all endomorphisms defined over the base, this factors as the product of the two conjugate Hecke L-functions.

The L-function of a Hecke character is initially the displayed Euler product over prime ideals away from its conductor. The equivalent Dirichlet series sums $\psi (\mathfrak{a})N{\mathfrak{a}}^{-s}$ over integral ideals prime to the conductor. For a CM elliptic curve with its endomorphisms defined over the base, its Hasse-Weil L-function factors into the two conjugate Hecke L-functions.

A p-adic L-function is a $p$-adic analytic function or measure interpolating suitably adjusted special values of an L-function. The Kubota-Leopoldt p-adic L-function is the basic example attached to a Dirichlet character.

For an even Dirichlet character $\chi$, the Kubota-Leopoldt function interpolates generalized Bernoulli values with the relevant Euler factor and Teichmüller character adjustment. If $\chi \ne 1$, it can be encoded by an integral power series. For $u=1+p$, the $p$-ramified convention is $g_{\chi }(u^{1-s}-1)=L_p(\chi ,s)$. The trivial character requires separate treatment of the pole.

On $G={\mathbb{Z}}_p^{\times }/\{\pm 1\}$, this element of the total quotient ring of the Iwasawa algebra becomes an integral measure after multiplication by every difference $[a]-[1]$. Its even moments interpolate the displayed Euler-corrected zeta values. Its trivial-character component has a pole, so the pseudomeasure itself must not be treated as an element of the integral group algebra. Multiplication by the augmentation ideal removes the pole.

There is an absolute finite exponent $L$ bounding the least prime in any reduced arithmetic progression by a constant times $q^L$. The multiplicative large sieve inequality is one of the tools in proofs of this theorem.

A sum of character values over a set. Arithmetic cancellation, rather than the termwise absolute-value bound, is the useful feature. A Gauss sum of a Dirichlet character is a weighted example, and the Pólya–Vinogradov inequality bounds interval sums.

For a primitive nonprincipal Dirichlet character modulo $q$, finite Fourier inversion theorem and the primitive Gauss magnitude give this bound uniformly in the interval. Bound the exponential sums by $C\Vert a/q{\Vert }^{-1}$ and sum the resulting harmonic series. The general Gauss identity follows from the prime-power identity by the Chinese remainder theorem.

A prime sum is a summation indexed by prime numbers, for example ${\sum }_{p\le x}1/p$. The indexing restriction distinguishes it from a sum over all positive integers.

For every $A>0$ there is $B>0$ such that, for $Q\le x^{1/2}/{\log }^{B}x$,Here $\psi (y;q,a)$ is the Chebyshev function in an arithmetic progression and $\varphi$ is the Euler totient function. Partial summation gives the corresponding result for the prime-counting function and the offset logarithmic integral function. The theorem controls the total error over many moduli; it need not give the same bound for each individual modulus.

Let $E_q={\max }_{(a,q)=1}|\pi (x;q,a)-\mathrm{Li}(x)/\varphi (q)|$. For $Q\le x^{1-\eta }$, the Brun–Titchmarsh theorem and Cauchy-Schwarz inequality giveWrite each summand as $\sqrt{E_q}{3}^{\omega (q)}\sqrt{E_q}$, then use $E_q{\ll }_{\eta }x/(\varphi (q)\log x)$.

A finite sum of complex exponentials, such as $S(\theta )={\sum }_n{a}_{n}e(n\theta )$ with $e(t)=\exp (2\pi it)$, is an exponential sum. Cancellation in an exponential sum between its terms can make it much smaller than the sum of their absolute values.

Taking a finite difference lowers the degree of a polynomial phase. For $S(\theta )={\sum }_{x=1}^{n}e(\theta P(x))$, expand $|S{|}^2$ by shifts $h$ and apply the Cauchy-Schwarz inequality to those shift sums. This gives $|S{|}^4\le 2n{\sum }_{h,l}{\sum }_{x\in I_{h,l}}e(\theta {\Delta }_h{\Delta }_{l}P(x))$, where $I_{h,l}$ enforces that $x,x+h,x+l,x+h+l$ lie in $[1,n]$. Although individual terms can be complex, the entire sum is real and nonnegative because it equals a sum of squares. For $P(x)=x^3$, the phase is $3hl(2x+h+l)$.

For a real degree-$k$ polynomial with leading coefficient $\theta$, $k\ge 2$, and coprime integers $a,q$ satisfying $|\theta -a/q|\le q^{-2}$, the displayed bound applies to its exponential sum over $1\le n\le N$. Repeated Weyl differencing reduces the phase to slope $k!\theta h_1\cdots h_{k-1}$. A geometric-sum estimate, the subpower bound for the divisor function and the reciprocal fractional-part sum near a rational complete the estimate. The arbitrary positive exponent $\eta$ absorbs divisor bounds and logarithms.

If $S(\theta )={\sum }_{x\le n}e(\theta x^3)$ and $|\theta -a/q|\le q^{-2}$ with $(a,q)=1$, the displayed estimate holds. Twice applying Weyl differencing bounds $|S{|}^4$ by $O(n^3+n{\sum }_{0<|h|,|l|<n}\min (n,\Vert 6\theta hl{\Vert }^{-1}))$. Group by $6hl$ and use the subpower bound for the divisor function. The reciprocal fractional-part sum near a rational bounds the resulting sum by $O(\log (2q)(n^3/q+n^2+n+q))$. Take fourth roots, absorbing logarithms into $n^{\eta }$ when $q\le n^3$; for larger $q$ the trivial bound suffices. Leiden lecture notes, Section 7.2 develop this cubic differencing estimate.

For $1\le j\le k$ and every $\eta >0$, the displayed estimate bounds a mean value of a polynomial exponential sum. For cubes its eighth-moment bound is $O_{\eta }(n^{5+\eta })$. It complements pointwise Weyl differencing estimates: a ninth moment over a region is bounded by its supremum there times this eighth moment. Leiden lecture notes, Section 8.1 give the cubic formulation.

Write $S(\theta )={\sum }_{x\le n}e(\theta x^3)$ and $I_j={\int }_0^1|S{|}^j$. Character orthogonality gives $I_2=n$. For $I_4$, fix the difference $D=x^3-y^3$. If $D=0$, there are $n^2$ choices for the two equal pairs. If $D\ne 0$, any solution $z^3-w^3=D$ has $z-w\mid D$; fixing this divisor gives a quadratic equation for $w$, hence at most two choices. The subpower bound for the divisor function yields $I_4{\ll }_{\eta }{n}^{2+\eta }$.

The twice-differenced inequality in Weyl differencing reads $|S{|}^4\le 2n{\sum }_v{c}_{v}e(\theta v)$, with $c_v$ counting triples $(h,l,x)$ for which $v=3hl(2x+h+l)$ and all four shifted arguments are in $[1,n]$. Thus $c_0\ll n^2$, and $c_v{\ll }_{\eta }{n}^{\eta }$ for $v\ne 0$: the product $hl$ divides $v$, and the remaining equation determines $x$. The Fourier coefficients $r(v)$ of $|S{|}^4$ are nonnegative counts, with $r(0)=I_4$ and ${\sum }_{v}r(v)=n^4$. Multiply the inequality by $|S{|}^4$ and integrate. It follows that $I_8\ll n(n^2{I}_4+n^{\eta }{n}^4){\ll }_{\eta }{n}^{5+\eta }$, after choosing the divisor-bound exponents small enough.

For $a\ne 0$ in the finite field ${\mathbb{F}}_p$, put $d=\gcd (k,p-1)$. The multiplicative group of a finite field is cyclic, and character orthogonality expresses the number of $k$th roots of $y\ne 0$ as ${\sum }_{{\chi }^d=1}\chi (y)$. The trivial character contributes $-1$, cancelling the $x=0$ term. Each of the remaining $d-1$ Gauss sums of Dirichlet characters has modulus $\sqrt{p}$, by the prime-field character Gauss sum identity. Consequently $|G_{a,p}|\le (d-1)/\sqrt{p}\le k/\sqrt{p}$.

For $S(a)={\sum }_{u}e(a{u}^k/p)$, character orthogonality counts representations of $x$ by $p^{-1}{\sum }_{a}S(a{)}^{3}e(-ax/p)$. Its zero-frequency term is $p^2$. The power Gauss sum over a prime field bounds the other terms in total by $k^3{p}^{3/2}$, strictly less than $p^2$ when $p>k^6$. Every residue class therefore has a representation.

A quadratic exponential sum has a polynomial phase of degree two. Its multiplicative derivative at lag $h$ has the linear phase $2\alpha hn$ up to a constant factor. This degree reduction lets the Van der Corput inequality for finite scalar sequences reduce its size to estimates for finite geometric series.

For coprime integers $a,q$ with $1\le q\le Q$ and $|\alpha -a/q|\le 1/(qQ)$, expand the squared quadratic exponential sum by its difference $h$. The exponential geometric sum bound gives $|S{|}^2\ll N+{\sum }_{h\le N}\min (N+1,\Vert 2\alpha h{\Vert }^{-1})$. In each block of length at most $q/4$, the rotations $2\alpha h$ are separated by at least $1/(2q)$: the reduced denominator of $2a/q$ is at least $q/2$, and the approximation error between two block points is at most $1/(2q)$. The separated reciprocal-distance sum bounds each block by $O(N+q\log (2N))$. There are $O(N/q+1)$ blocks, giving the displayed inequality. For bounded $q$ the trivial estimate supplies the same conclusion.

By orthogonality of integer Fourier modes, this integral counts pairs of $k$-tuples with the same first $r$ power sums. Diagonal pairs give $J_{k,r}(Z)\gg Z^k$. There are $O_{k,r}(Z^{r(r+1)/2})$ possible moment vectors; the Cauchy-Schwarz inequality gives $J_{k,r}(Z)\gg Z^{2k-r(r+1)/2}$. Upper bounds measure how much arithmetic coincidence remains beyond these necessary contributions, and enter the Vinogradov mean-value method for a bilinear exponential sum.

Expand moments of the inner sum and group equal power-sum differences. Their multiplicities are bounded by $J_{k,r}(Z)$ through the Cauchy-Schwarz inequality. Two applications of the Holder inequality yield a bound for $|U{|}^{4k^2}$ containing $Z^{8k^2-4k}{J}_{k,r}(Z{)}^2$ and a product of short geometric-sum bounds over the moment differences. Thus a sharp mean-value estimate combines with rational approximation or spacing of the coefficients to prove cancellation. When all coefficients are integers, $U=\lfloor Z{\rfloor }^2$, so mean-value estimates alone cannot force cancellation.

Translating an integer interval by $xy$ changes a sum of unit-modulus terms by at most $2xy$. Averaging gives the displayed formula. For $n\asymp N$, $Z=N^{2/5}$, expand $-t\log (1+xy/n)$ to degree $r=\lfloor 5.01\log t/\log N\rfloor$. Its remainder is at most $t{N}^{-(r+1)/5}<t^{-1/500}$. This reduces the logarithmic exponential sum to bilinear polynomial sums while keeping explicit boundary and approximation errors.

For a real $C^1$ phase with continuous monotone derivative and $|f^{\prime }|\le \delta <1$, the sum-integral discrepancy has the displayed uniform bound. Periodization and the Dirichlet-Jordan convergence theorem express it as symmetric nonzero Fourier modes. Integration by parts gives endpoint terms and reciprocal-derivative variations. Monotonicity bounds their total variation by ${\sum }_{h\ne 0}2\delta /(h^2-{\delta }^2)=O((1-\delta {)}^{-1})$. Pairing the leading $1/h$ endpoint terms reduces them to the uniformly bounded sine series. The exclusion of integer nonzero frequencies is essential.

Cancellation occurs when different complex numbers in an exponential sum have directions that reduce the absolute value of their sum. For example, the orthogonality of roots of unity makes ${\sum }_{a=0}^{p-1}e(a/p)=0$ for every integer $p>1$, despite all terms having absolute value one.

Probabilistic number theory studies arithmetic functions and integer sets through probability distributions, concentration, and limit theorems.

For a real strongly additive arithmetic function with uniformly bounded values on primes, put $A(x)={\sum }_{p\le x}f(p)/p$ and $B(x{)}^2={\sum }_{p\le x}f(p{)}^2/p$. If $B(x)\to \infty$, then $(f(n)-A(N))/B(N)$ under uniform sampling from $[N]$ converges to the standard normal distribution. For the prime omega function, the mean and variance scales are $\log \log N$.

If a strongly additive arithmetic function $g$ is supported on primes up to $y$, its $j$th uniform central moment differs from the independent Bernoulli random variables model with probabilities $1/p$ by $O(j{y}^j({\sum }_{p\le y}|g(p)|{)}^j/N)$. Expansion of divisibility indicator functions reduces the comparison to $\lfloor N/d\rfloor /N=1/d+O(1/N)$.

The Cramér model selects integers $n\ge 3$ independently with probability $1/\log n$, and conventionally sets $U_1=0$, $U_2=1$. It is a random model for the distribution of primes, not an assertion of independence for actual prime numbers.

If $P_n$ are the increasing selected integers in the Cramér model, then almost surely ${\mathrm{limsup}}_n(P_{n+1}-P_n)/(\log P_n{)}^2\le 1$. A zero block of length $(1+ϵ)(\log k{)}^2$ after $k$ has summable probability, so the Borel-Cantelli first lemma excludes these blocks eventually.

The multiplication table problem asks how many distinct integers occur as $ab$ with $1\le a,b\le N$. Counting the $N^2$ pairs greatly overcounts the distinct products.

Concentration of the prime omega function near $\log \log N$, together with $\omega (ab)=\omega (a)+\omega (b)-\omega (\gcd (a,b))$ and the bounded mean of $\omega (\gcd (a,b))$, gives $M(N)\ll N^2/\log \log N$. This elementary argument uses only distinct prime factors.

For an additive arithmetic function $f$, $N^{-1}{\sum }_{n\le N}|f(n)-A_f(N){|}^2\ll B_f(N{)}^2$, where $A_f$ is the mean of an additive arithmetic function and $B_f(N{)}^2={\sum }_{p^k\le N}|f(p^k){|}^2/p^k$. The implied constant is absolute.

If a Dirichlet series with nonnegative coefficients has finite abscissa of convergence ${\sigma }_c$, then the function it defines has a singularity at the real point $s={\sigma }_c$. Consequently, if that function extends holomorphically across every real point, the original series converges for every real $s$.

For every nonprincipal Dirichlet character, $L(1,\chi )\ne 0$. A nonreal character is covered by Euler product positivity for L-function nonvanishing at height zero, since its squared character is nonprincipal.

For a real character, suppose $L(1,\chi )=0$. The simple zeta pole would cancel, making $F(s)=\zeta (s)L(s,\chi )$ an entire function. Its nonnegative zeta-times-real-L coefficients satisfy $a_n\ge 0$ and $a_{m^2}\ge 1$. Termwise derivatives at two give $(-1{)}^k{F}^{(k)}(2)={\sum }_n{a}_n(\log n{)}^k{n}^{-2}$. The Taylor series of this entire function at two converges at zero. Every term there is nonnegative, and Tonelli theorem identifies its sum asThe last series diverges because all square coefficients are at least one. This contradiction proves the claim. It is the positive-coefficient mechanism behind the more general Landau theorem for a Dirichlet series with nonnegative coefficients.

For $(a,q)=1$,The Orthogonality of Dirichlet characters expresses it in terms of character-twisted Von Mangoldt functions.

There is an absolute $c>0$ such that the Dirichlet L-functions modulo $q$ have no zero inapart from at most one real simple zero belonging to a real nonprincipal character.

For primitive real nonprincipal Dirichlet characters, there is an absolute $c>0$ with at most one zero in this interval, counted with multiplicity. Positivity of $-{\zeta }^{\prime }/\zeta -L^{\prime }/L$ and the local partial fractions give $0\le 1/(\sigma -1)+C\log q-\sum (\sigma -\beta {)}^{-1}$. Testing $\sigma =1+a/\log q$ with small fixed $a$ rules out two nearby real zeros. Any possible zero is consequently simple; existence is not asserted.

An exceptional zero is the possible real simple zero $\beta$ omitted from the otherwise uniform classical zero-free region for Dirichlet L-functions. It belongs to a real nonprincipal primitive Dirichlet character and satisfies $\beta <1$.

For two real primitive nonprincipal Dirichlet characters with nonprincipal product, the real zeros in this interval are counted with multiplicity. The logarithmic-derivative coefficients of $\zeta L({\chi }_1)L({\chi }_2)L({\chi }_1{\chi }_2)$ are $\Lambda (n)(1+{\chi }_1(n))(1+{\chi }_2(n))\ge 0$. The global partial-fraction expansion of a Dirichlet L-function logarithmic derivative therefore gives $0\le 1/(\sigma -1)-{\sum }_{j=1}^2(\sigma -{\beta }_j{)}^{-1}+C\log (q_1{q}_2)$ if two nearby real zeros exist. Put $\sigma =1+a/\log (q_1{q}_2)$, choose $a$ small, and test $1-{\beta }_j<a/(4\log (q_1{q}_2))$. The right side is negative, a contradiction.

// Target: analytic-number-theory.bigb

Fix the exceptional-zero threshold $\beta >1-\eta /\log q$, with $\eta$ less than one third of the constant in Landau theorem for two real Dirichlet characters. If distinct primitive conductors satisfy $q_1<q_2\le q_1^2$, then $\log (q_1{q}_2)\le 3\log q_1$; both exceptional zeros lie in the forbidden two-zero interval. Thus the conductors grow at least quadratically. The same argument with equal conductors proves at most one exceptional real primitive character for each modulus. Existence is not asserted, and a consistently sufficiently small threshold is essential.

// Target: number-theory.bigb

Uniformly for $x\ge 2$ and $(a,q)=1$,where the second term occurs only when a real character ${\chi }_1$ has an exceptional zero $\beta$.

If some $ϵ>0$ satisfiesuniformly for sufficiently large $q$, then every exceptional zero modulo $q$ satisfiesChoose ${\chi }_1(a)=-1$ and $x=\exp (A(\log q{)}^2)$ in the exceptional-zero asymptotic, with $A$ large in terms of $ϵ$.

Pretentious number theory studies multiplicative arithmetic functions by measuring how closely they imitate structured functions such as Dirichlet characters and Archimedean characters.

A positive function $g(n)$ is a normal order of an arithmetic function $f(n)$ when, for every $ϵ>0$, all but $o(x)$ integers $n\le x$ satisfy $|f(n)-g(n)|<ϵg(n)$.

The prime omega function has normal order $\log \log n$. The elementary second-moment estimatealready proves concentration on every scale larger than $\sqrt{\log \log x}$.

If $\gcd (a,m)=1$, infinitely many primes satisfy

For fixed coprime integers $a,q$, the Prime number theorem in the arithmetic progression gives $\pi (x;q,a)\sim x/(\phi (q)\log x)$. Partial summation expresses the reciprocal-prime sum as $\pi (x;q,a)/x+{\int }_2^x\pi (t;q,a)t^{-2}dt$. The integral of the relative error is $o(\log \log x)$: fix a threshold beyond which that error is at most $\epsilon$, bound its remaining integral by $\epsilon \log \log x$, and then let $\epsilon$ decrease to zero. The modulus is fixed in this statement.

For fixed $A,C>0$,uniformly for $q\le {\log }^{C}x$ and $(a,q)=1$. The implied constant may be ineffective. This small-modulus input complements the large sieve in the proof of the Bombieri–Vinogradov theorem.

For $(a,q)=1$ and $q\le {\log }^{A}N$, apply the Siegel–Walfisz theorem to every reduced residue class with error exponent $A+B+2$. The main terms sum to $N{c}_q(a)/\phi (q)$, where the Ramanujan sum satisfies $c_q(a)=\mu (q)$ for coprime $a,q$. Nonunits with nonzero Von Mangoldt function are powers of primes dividing $q$ and contribute at most $\omega (q)\log N$, which is absorbed by the error. For arbitrary $a$, the correct main term is $N{c}_q(a)/\phi (q)$; the reduced numerator hypothesis cannot be omitted.

For real $s>1$, the logarithm defined by the Euler product of a Dirichlet L-function differs from $F_{\chi }(s)$ by a quantity of absolute value at most ${\sum }_{n\ge 2}1/(n(n-1))=1$. If $L(\chi ,1)\ne 0$, a local holomorphic logarithm differs from this continuous logarithm by one fixed multiple of $2\pi i$ on a short real interval, so $F_{\chi }(s)$ stays bounded as $s\downarrow 1$. For the principal Dirichlet character, the pole of the Riemann zeta function instead gives $F_{{\chi }_0}(s)=\log (1/(s-1))+O(1)$. Orthogonality of Dirichlet characters then makes the sum over any reduced residue class diverge, proving infinitude of its primes.

The $n$th harmonic number is $H_n={\sum }_{k=1}^{n}1/k$. Its asymptotic expansion beginswhere $\gamma$ is the Euler--Mascheroni constant.

Partial summation is the discrete analogue of integration by parts. If $A(x)={\sum }_{n\le x}{a}_n$ and $g$ is continuously differentiable, then

If $A(x)$ counts a set of positive integers and $\eta >0$, Abel summation gives ${\sum }_{n\in A,n\le X}1/n=A(X)/X+{\int }_1^{X}A(t)t^{-2}dt$. A bound $A(t)\ll t/(\log t{)}^{1+\eta }$ for $t\ge 3$ makes the integral convergent, since the substitution $u=\log t$ gives ${\int }_{\log 3}^{\infty }{u}^{-1-\eta }du<\infty$.

If $\sum a_n$ converges and $b_n$ is bounded with $\sum |b_{n+1}-b_n|<\infty$, then $\sum a_n{b}_n$ converges. For $T_j={\sum }_{k=m}^j{a}_k$, summation by parts givesThe first factor tends uniformly to zero by the Cauchy sequence criterion, and the second has a uniform bound. A bounded monotone sequence has finite total variation, so monotone multipliers are included. Eventual monotonicity also suffices, since finitely many initial terms do not affect convergence. Absolute convergence of $\sum a_n$ is unnecessary.