Artin L-Functions, Hecke Characters, and Modular Forms

Consider the equation \displaystyle x^3=2. We wish to understand for which rational primes p the congruence x^3\equiv 2\pmod p is soluble, how many solutions it possesses, and whether these different possibilities obey some general arithmetic law. One may begin, in the manner of Gauss, by computing the cubes of integers modulo a succession of small primes. Modulo 5 , the number 3 is a cube root of 2 , since 3^3\equiv 2\pmod5 . Modulo 7 , however, the nonzero cubes are only 1 and -1 , and the congruence has no solution. Modulo 31 we find three distinct solutions, namely 4,7,20 . These examples suggest that the primes fall into several classes, and that the arithmetic distinction between them may be more naturally expressed by studying the complete factorization of X^3-2 over the finite field \mathbb F_p , rather than merely asking whether the polynomial has a root.

There is another question which at first seems unrelated. Let \alpha=\sqrt[3]{2} , and consider the cubic number field F=\mathbb Q(\alpha) . Its Dedekind zeta function is defined by summing over the nonzero integral ideals of its ring of integers:

\displaystyle \zeta_F(s)=\sum_{\mathfrak a\neq 0}\frac{1}{(N\mathfrak a)^s}, \qquad {\mathrm{Re}}s>1.

Here the norm N\mathfrak a is the number of residue classes in the finite ring \mathcal O_F/\mathfrak a . Because every nonzero integral ideal factors uniquely into prime ideals, we have the Euler product

\displaystyle \zeta_F(s)= \prod_{\mathfrak p} \left(1-(N\mathfrak p)^{-s}\right)^{-1}.

To calculate the factor contributed by a rational prime p , we must know the prime ideals lying above p and their norms. We shall find that this information is determined by precisely the same factorization of X^3-2 which arose in our elementary investigation of cubic congruences. We shall then discover that these local Euler factors can be represented by determinants of two-dimensional matrices associated with permutations of the three roots. Finally, those determinants will be reproduced by a character of ideals in the quadratic field \mathbb Q(\sqrt{-3}) , and the character will lead us to a holomorphic modular form constructed from an explicit lattice sum. The principal identities of the investigation will be

\displaystyle \zeta_F(s)= \zeta(s)L(s,\rho) =\zeta(s)L_{\mathbb Q(\sqrt{-3})}(s,\psi),

where \rho is a two-dimensional representation of the six transformations of the roots, and \psi is a cubic character of ideals that can be calculated using finite-field exponentiation. The same L -series will have an expansion \sum_{n\geq1}a_nn^{-s} , whose coefficients occur in a normalized modular form of weight one and level 108 . Our object is not to infer these equalities from a general correspondence, but to examine the arithmetic mechanisms producing them. We shall distinguish throughout between the calculations that follow directly from finite fields and ideal multiplication, the cubic reciprocity statement needed to obtain a particularly simple congruence formula for \psi , and the further Fourier-analytic work needed to establish the modular transformation law.

Cubic Congruences

We first determine the possible numbers of solutions of x^3\equiv2\pmod p for primes p\neq2,3 . The nonzero residue classes modulo p form a cyclic multiplicative group of order p-1 . If g is a generator, we may write 2=g^m , and any nonzero solution has the form x=g^j . The original cubic congruence is consequently equivalent to the linear congruence 3j\equiv m\pmod{p-1} . Thus the entire question is controlled by the greatest common divisor of 3 and p-1 . Suppose first that p\equiv2\pmod3 . Then 3 is relatively prime to p-1 , so the map sending a nonzero residue to its cube is a permutation of the multiplicative group. In particular, every nonzero residue is a cube, and each has exactly one cube root. We can write this root explicitly. If p=3k+2 , the integer 2k+1 satisfies 3(2k+1)=2(p-1)+1 , and hence

\displaystyle x\equiv 2^{(2p-1)/3}\pmod p, \qquad p\equiv2\pmod3.

The formula is a useful illustration of how an elementary group-theoretic observation produces an explicit numerical solution. It also shows that, for all primes in this residue class, the polynomial X^3-2 has exactly one linear factor. Since its discriminant is nonzero modulo p , the remaining quadratic factor must be irreducible. Now suppose p\equiv1\pmod3 . The multiplicative group contains three distinct cube roots of unity, and the map x\mapsto x^3 has a kernel of order three. Consequently, every element in its image has exactly three cube roots, while every element outside its image has none. To decide whether 2 lies in the image, write 2=g^m . The congruence 3j\equiv m\pmod{p-1} is soluble if and only if 3\mid m . Equivalently, 2^{(p-1)/3}=g^{m(p-1)/3} is equal to 1 . We therefore obtain the criterion

\displaystyle 2\text{ is a cube modulo }p \quad\Longleftrightarrow\quad 2^{(p-1)/3}\equiv1\pmod p, \qquad p\equiv1\pmod3.

These elementary observations already give a complete classification of the possible factorizations. For p\equiv2\pmod3 , the factorization has degrees 1+2 . For p\equiv1\pmod3 , the polynomial has either three distinct linear factors, giving degrees 1+1+1 , or it is irreducible of degree three. In the latter two cases, the distinction is made by the value of 2^{(p-1)/3}\pmod p . This calculation will repeatedly reappear in different forms: as a description of prime ideals, as a permutation of three roots, as a matrix trace, and as the first coefficient of an Euler factor.

Primes and factorizations

We now pass to the cubic field F=\mathbb Q(\alpha) , with \alpha^3=2 . The polynomial X^3-2 is irreducible over \mathbb Q by Eisenstein’s criterion at 2 , and hence 1,\alpha,\alpha^2 form a basis of F over \mathbb Q . The ring \mathbb Z[\alpha] is in fact the full ring of integers of F . Before using this fact, let us calculate the discriminant of the displayed basis, since it will identify the primes at which the arithmetic requires special treatment. The three embeddings of F into \mathbb C send \alpha to \alpha,\omega\alpha,\omega^2\alpha , where \omega=(-1+\sqrt{-3})/2 . The discriminant is the square of the determinant formed from the images of 1,\alpha,\alpha^2 . Thus

\displaystyle \mathrm{disc}(1,\alpha,\alpha^2)=\left[ \det \begin{pmatrix}1&\alpha&\alpha^2\\1&\omega\alpha&\omega^2\alpha^2\\1&\omega^2\alpha&\omega\alpha^2\end{pmatrix} \right]^2.

Factoring \alpha from the second column and \alpha^2 from the third produces a factor \alpha^3=2 in the determinant. The remaining determinant is a Vandermonde determinant in 1,\omega,\omega^2 , whose square is -27 . It follows that

\displaystyle {\mathrm{disc}}(1,\alpha,\alpha^2)=-108=-2^2 3^3.

If \mathbb Z[\alpha] were a proper suborder of the full ring of integers, its index would have a square dividing 108 . Hence only the primes 2 and 3 could divide that index. At 2 , the polynomial X^3-2 is Eisenstein, and its root generates the full ring of integers of the corresponding totally ramified local extension. At 3 , one obtains an Eisenstein polynomial by changing variables. Put \beta=\alpha+1 ; then (\beta-1)^3=2 gives \beta^3-3\beta^2+3\beta-3=0. This polynomial is Eisenstein at 3 , so the local integer ring is again generated by the displayed element. Thus neither 2 nor 3 divides the index, and therefore \mathcal O_F=\mathbb Z[\alpha] . Consider now a prime p\nmid108 . Reduction modulo p gives a concrete description of the quotient ring:

\displaystyle \mathcal O_F/p\mathcal O_F \cong \mathbb F_p[X]/(X^3-2).

Suppose that the polynomial factors into distinct monic irreducible factors f_1,\ldots,f_g , of degrees d_1,\ldots,d_g . Since these factors are pairwise relatively prime, the Chinese remainder theorem gives

\displaystyle \mathbb F_p[X]/(X^3-2) \cong \prod_{j=1}^g \mathbb F_p[X]/(f_j(X)).

Each quotient on the right is a finite field with p^{d_j} elements, and its kernel corresponds to a prime ideal \mathfrak p_j of \mathcal O_F . Consequently,

\displaystyle p\mathcal O_F=\mathfrak p_1\cdots\mathfrak p_g, \qquad N\mathfrak p_j=p^{d_j}.

This is the precise arithmetic reason why the degrees of polynomial factors determine the Euler factors of a Dedekind zeta function. The irreducible factor of degree d gives a residue field of size p^d , and that cardinality is exactly the norm of the associated prime ideal.

Let us examine p=5 explicitly. A direct multiplication gives \displaystyle X^3-2\equiv (X-3)(X^2+3X+4)\pmod5. The quadratic factor is irreducible because its discriminant is 9-16=-7\equiv3\pmod5 , which is not a square. Thus

\displaystyle (5)= (5,\alpha-3) (5,\alpha^2+3\alpha+4),

and the two prime ideals have norms 5 and 25 . Every integral ideal supported above 5 has a unique expression \mathfrak p_1^a\mathfrak p_2^b , with a,b\geq0 . Its norm is 5^{a+2b} , so the local ideal-counting series is

\displaystyle \begin{aligned} Z_{F,5}(s) &=\sum_{a,b\geq0}5^{-s(a+2b)}\\ &=\left(\sum_{a\geq0}5^{-as}\right) \left(\sum_{b\geq0}5^{-2bs}\right)\\ &=\frac{1}{(1-5^{-s})(1-5^{-2s})}. \end{aligned}

At 7 , the polynomial has no root and is irreducible, so the ideal (7) remains prime and has norm 7^3 . Its Euler factor is therefore (1-7^{-3s})^{-1} . At 31 , all three roots are present modulo 31 , giving three distinct prime ideals of norm 31 and the local factor (1-31^{-s})^{-3} . In general, when the irreducible factors have degrees d_1,\ldots,d_g , the local zeta factor is

\displaystyle Z_{F,p}(s)= \prod_{j=1}^g(1-p^{-d_js})^{-1}.

Frobenius elements

There is a second way to recover these factor degrees, using the simplest possible operation in a finite field. In characteristic p , the map x\mapsto x^p preserves both addition and multiplication. If u is a root of a polynomial with coefficients in \mathbb F_p , then u^p is another root of the same polynomial, because raising the equation to the p -th power leaves its coefficients fixed. Repeating the operation gives the sequence

\displaystyle u,\quad u^p,\quad u^{p^2},\quad u^{p^3},\ldots.

If the minimal polynomial of u over \mathbb F_p has degree d , the field \mathbb F_p(u) has p^d elements, and every element of that field satisfies x^{p^d}=x . The first return of the displayed sequence occurs after exactly d steps. Indeed, an earlier return would imply that u belonged to a field of smaller degree, contradicting the degree of its minimal polynomial. Thus the roots of an irreducible polynomial of degree d form a cycle of length d under the p -power operation. It follows that the factor degrees of X^3-2 are precisely the cycle lengths of the permutation of its roots induced by raising to the p -th power. The splitting type 1+1+1 gives three fixed points, the type 1+2 gives one fixed point and one two-cycle, and the irreducible case gives a three-cycle. We have therefore obtained a description of prime decomposition entirely in terms of a permutation. The same number d is simultaneously the degree of an irreducible polynomial, the degree of a residue-field extension, and the length of a cycle of the finite-field Frobenius operation.

We now represent this permutation by a matrix. Introduce formal basis vectors e_1,e_2,e_3 , corresponding to the three roots, and let P_p be the permutation matrix describing the p -power action. A cyclic permutation P of length d has eigenvalues equal to the d -th roots of unity, and hence

\displaystyle \det(I-TP)=1-T^d.

For a permutation consisting of disjoint cycles of lengths d_1,\ldots,d_g , the matrix becomes block diagonal after rearranging the basis. Its determinant is therefore the product of the determinants of the cyclic blocks:

\displaystyle \det(I-TP_p)= \prod_{j=1}^g(1-T^{d_j}).

Comparing this formula with the factorization of the local zeta function, and substituting T=p^{-s} , gives the identity

\displaystyle Z_{F,p}(s)= \det(I-p^{-s}P_p)^{-1}, \qquad p\nmid108.

This determinant is not an artificial construction. It encodes the same cycle lengths that determine the norms of prime ideals. The advantage is that we can now examine invariant subspaces common to all the permutations, instead of treating each splitting pattern separately. Every permutation of the three basis vectors fixes their sum e_1+e_2+e_3 . It also preserves the two-dimensional plane

\displaystyle W= \left\{ x_1e_1+x_2e_2+x_3e_3: x_1+x_2+x_3=0 \right\}.

Thus the full permutation space decomposes into the one-dimensional space of constant vectors and the plane W . The action on the constant direction is always multiplication by 1 , while the action on W is represented by a two-dimensional matrix. We shall denote this latter action by \rho . The determinant consequently factors as

\displaystyle \det(I-TP_p) =(1-T)\det(I-T\rho(P_p)).

The first factor gives exactly the Euler factor (1-p^{-s})^{-1} of the Riemann zeta function. The second factor contains the information distinguishing the possible splitting types. This suggests that the Dedekind zeta function of our cubic field should be the product of the ordinary zeta function and a new two-dimensional L -function. The decomposition is particularly important because it is independent of p . We have not chosen a separate diagonalization for each Frobenius matrix. The constant direction and the plane W are preserved by every permutation of the roots, and hence the same decomposition works simultaneously for all primes. This is precisely the sort of uniformity required in an Euler product.

Galois Group

Let L=\mathbb Q(\alpha,\omega) be the splitting field of X^3-2 . Since \mathbb Q(\alpha) is a real cubic field and \omega is nonreal, adjoining \omega doubles the degree, giving [L:\mathbb Q]=6 . The transformations of L that fix rational numbers are easily described. Let r multiply \alpha by \omega while fixing \omega , and let s be complex conjugation:

\displaystyle \begin{aligned} r(\alpha)&=\omega\alpha, &r(\omega)&=\omega,\\ s(\alpha)&=\alpha, &s(\omega)&=\omega^2. \end{aligned}

Direct substitution gives r^3=s^2=1 and srs=r^{-1} . There are six transformations, namely 1,r,r^2,s,sr,sr^2 , and they form the symmetric group on the three roots. We shall not need any abstract classification of its representations: the two-dimensional representation of interest can be written down directly. Choose the basis v_1=e_1-e_2 and v_2=e_2-e_3 of the plane W . The cyclic permutation sends v_1 to v_2 and v_2 to -v_1-v_2 . Complex conjugation fixes the first root and exchanges the other two, so it sends v_1 to v_1+v_2 and v_2 to -v_2 . Their matrices are therefore

\displaystyle R= \begin{pmatrix} 0&-1\\ 1&-1 \end{pmatrix}, \qquad S= \begin{pmatrix} 1&0\\ 1&-1 \end{pmatrix}.

Matrix multiplication verifies R^3=I , S^2=I , and SRS=R^{-1} . We next calculate the characteristic polynomials, since these will supply the Euler factors. For the three-cycle we have

\displaystyle \begin{aligned} \det(I-TR) &= \det \begin{pmatrix} 1&T\\ -T&1+T \end{pmatrix}\\ &=1+T+T^2. \end{aligned}

For the transposition,

\displaystyle \begin{aligned} \det(I-TS) &= \det \begin{pmatrix} 1-T&0\\ -T&1+T \end{pmatrix}\\ &=1-T^2. \end{aligned}

For the identity the matrix is I_2 , giving \det(I-TI_2)=(1-T)^2 . Thus the three possibilities for the two-dimensional local factor are

\displaystyle \frac{1}{(1-p^{-s})^2}, \qquad \frac{1}{1-p^{-2s}}, \qquad \frac{1}{1+p^{-s}+p^{-2s}}.

These correspond respectively to the identity, a transposition, and a three-cycle. The traces of the corresponding matrices are 2,0,-1 , while their determinants are 1,-1,1 . There is a particularly useful interpretation of these traces. In the original three-dimensional permutation representation, the trace of a permutation matrix is simply the number of basis vectors fixed by the permutation, which is the number of roots fixed by the Frobenius operation. Removing the universally fixed constant direction subtracts one from this trace. Consequently, for every prime p\nmid6 ,

\displaystyle \mathrm{tr}\rho(\mathrm{Frob}_p) = \# \{x\in\mathbb F_p:x^3=2\}-1.

Thus the first coefficient of the local Artin factor is obtained by counting solutions of our original congruence. The determinant has an equally simple interpretation: it is the sign of the permutation of the roots. The sign is 1 for the identity and three-cycles, and -1 for transpositions. The sign is determined by the action on \omega , or equivalently by the splitting of p in \mathbb Q(\sqrt{-3}) . Hence

\displaystyle \det\rho(\mathrm{Frob}_p)=\chi_{-3}(p),

where \chi_{-3}(p)=1 for p\equiv1\pmod3 and \chi_{-3}(p)=-1 for p\equiv2\pmod3 . Writing a_p={\mathrm{tr}}\rho(\mathrm{Frob}_p) , we obtain the uniform quadratic Euler denominator

\displaystyle 1-a_pp^{-s}+\chi_{-3}(p)p^{-2s}.

The two coefficients have been derived from elementary arithmetic. One counts roots of the cubic after subtracting one, while the other records whether the quadratic field splits or remains inert.

Exceptional Primes and Inertia

The preceding discussion applies to primes for which the reduction of the cubic polynomial has distinct roots. Its discriminant is -108 , so the only exceptional primes are 2 and 3 . At these primes the reduced roots collide, and the residue-field power operation no longer determines a unique permutation of the original roots. To understand the local Euler factor, we must distinguish between transformations visible on the residue field and transformations that become invisible upon reduction. The latter transformations form the inertia group. At 2 , the polynomial X^3-2 is Eisenstein, so adjoining \alpha produces a totally ramified extension of degree three. On the other hand, X^2+X+1 remains irreducible and separable modulo 2 , so adjoining \omega gives an unramified quadratic extension whose residue field is \mathbb F_4 . The full local splitting field therefore has ramification degree three and residue-field degree two. Its inertia group consists of the transformations 1,r,r^2 . On the two-dimensional space W , the matrix R has characteristic polynomial X^2+X+1 and no eigenvalue equal to one. Hence no nonzero vector is fixed by inertia. The local Artin Euler factor, defined using the subspace fixed by inertia, is therefore equal to one:

\displaystyle L_2(s,\rho)=1.

At 3 , both generators contribute to ramification. The shifted cubic generator \beta=\alpha+1 satisfies the Eisenstein equation \beta^3-3\beta^2+3\beta-3=0 , giving a totally ramified cubic extension. In the quadratic field, the element 1-\omega satisfies (1-\omega)^2=-3\omega , so 3 ramifies with degree two. The complete local extension has degree six and is totally ramified; its inertia group is the entire group of six transformations. Since the only vector fixed by every permutation in the original three-dimensional space is the constant vector, the plane W has no nonzero inertia-fixed vector. Consequently,

\displaystyle L_3(s,\rho)=1.

These answers agree with ideal counting in the cubic field. Both 2 and 3 are totally ramified in F , and each has a single prime ideal above it with residue-field norm equal to the rational prime. The corresponding local Dedekind zeta factor is (1-p^{-s})^{-1} , exactly the ordinary zeta factor. Dividing by \zeta(s) leaves a local factor of one in the two-dimensional Artin L -function. Thus the equality of local factors extends to the exceptional primes, provided the Artin construction retains only the directions fixed by inertia.

Conductor

The local Euler factors at 2 and 3 are both one, but these primes do not contribute equally to the conductor. The conductor records how strongly the representation is affected by ramification, including the additional contribution from wild ramification. To calculate it, one examines the successive ramification groups G_0,G_1,\ldots , where G_0 is inertia and the later groups measure increasingly delicate forms of ramification. For our two-dimensional representation, the local conductor exponent is

\displaystyle a_p(\rho)= \sum_{j\geq0} \frac{|G_j|}{|G_0|} \left(2-\dim W^{G_j}\right).

At 2 , inertia is cyclic of order three. Since its order is prime to the residue characteristic, all ramification is tame, and the positive ramification groups are trivial. Thus G_0=C_3 and G_1=1 . The inertia-fixed space is zero, so

\displaystyle a_2(\rho)=2.

At 3 , inertia is the full group S_3 , while the wild inertia subgroup is the normal cyclic subgroup C_3 . We must determine how long that subgroup persists in the ramification filtration. This can be calculated from the different of the local extension. The cubic field \mathbb Q_3(\alpha) has an integral generator satisfying f(X)=X^3-2 , and its different is generated by f'(\alpha)=3\alpha^2 . Since \alpha is a 3 -adic unit and 3 has valuation three in the totally ramified cubic field, the different exponent is three. Adjoining \omega gives a tamely ramified quadratic extension, whose different exponent is one. The tower formula therefore gives the different exponent of the complete degree-six local extension as

\displaystyle \delta_{L/\mathbb Q_3}=1+2\cdot3=7.

For a totally ramified Galois extension, the different exponent is also the sum \sum_{j\geq0}(|G_j|-1) . The zeroth group has six elements and contributes five. The remaining contribution is two. Every nontrivial positive ramification group must be the subgroup C_3 , whose contribution is exactly two, so it occurs once and only once. Thus

\displaystyle G_0=S_3,\qquad G_1=C_3,\qquad G_2=1.

Neither S_3 nor C_3 fixes a nonzero vector of W . The conductor exponent at 3 is consequently

\displaystyle \begin{aligned}a_3(\rho)&=\frac66(2-0)+\frac36(2-0)\\&=2+1\\&=3.\end{aligned}

Combining the two exceptional primes, we obtain the conductor

\displaystyle \mathfrak f(\rho)=2^2 3^3=108.

The calculation explains the exponents rather than merely identifying their product. At 2 , the entire contribution comes from tame inertia acting without invariant vectors. At 3 , there is the same initial contribution of two, followed by one additional contribution from the wild subgroup. This conductor will reappear on the Hecke-character side as the product of a quadratic discriminant and the norm of an ideal modulus.

The complete Artin Euler product

We can now define the two-dimensional Artin L -function at every prime. At an unramified prime, we use the determinant of the Frobenius matrix on the full two-dimensional space. At a ramified prime, we restrict Frobenius to the subspace fixed by inertia. Thus

\displaystyle L(s,\rho)=\prod_p\det\left(I-p^{-s}\rho(\mathrm{Frob}_p)\mid W^{I_p}\right)^{-1}.

At primes p\nmid6 , the local factor is one of the three rational functions calculated earlier. At 2 and 3 , it is one. In particular, the Euler product may be written entirely in terms of elementary congruences:

\displaystyle \begin{aligned}L(s,\rho)={}&\prod_{\substack{p\equiv2\pmod3\\p\ne2}}(1-p^{-2s})^{-1}\\&\times\prod_{\substack{p\equiv1\pmod3\\2^{(p-1)/3}\equiv1\pmod p}}(1-p^{-s})^{-2}\\&\times\prod_{\substack{p\equiv1\pmod3\\2^{(p-1)/3}\not\equiv1\pmod p}}(1+p^{-s}+p^{-2s})^{-1}.\end{aligned}

Multiplying every local factor by the ordinary factor (1-p^{-s})^{-1} gives exactly the local ideal-counting factor of F . The resulting global identity is

\displaystyle\boxed{\zeta_F(s)=\zeta(s)L(s,\rho).}

There is also a useful identity involving the full splitting field L . The six-dimensional regular permutation representation of the six-element group decomposes into the trivial representation, the one-dimensional sign representation, and two copies of the two-dimensional representation. The corresponding Dedekind zeta factorization is

\displaystyle \zeta_L(s)=\zeta(s)L(s,\chi_{-3})L(s,\rho)^2.

Since M=\mathbb Q(\omega) has \zeta_M(s)=\zeta(s)L(s,\chi_{-3}) , this may be rewritten as \zeta_L(s)=\zeta_M(s)L(s,\rho)^2 . Thus the two-dimensional Artin function measures, in a precise Euler-product sense, the additional arithmetic of the full six-degree splitting field beyond that already present in the quadratic field. The next problem is to produce the same function through arithmetic intrinsic to that quadratic field.

Eisenstein integers

Let M=\mathbb Q(\omega)=\mathbb Q(\sqrt{-3}) . Its ring of integers is \mathcal O_M=\mathbb Z[\omega] . Every element has the form a+b\omega , with a,b\in\mathbb Z , and its norm is

\displaystyle N(a+b\omega)=(a+b\omega)(a+b\omega^2)=a^2-ab+b^2.

The norm is multiplicative and is positive for every nonzero element. Geometrically, \mathbb Z[\omega] is the triangular lattice in the complex plane. Given two nonzero Eisenstein integers \gamma,\delta , we may divide \gamma by \delta in the complex plane and choose the nearest Eisenstein integer q to the quotient. The remainder r=\gamma-q\delta then has strictly smaller norm than \delta . Thus the ordinary Euclidean algorithm works for this norm, and every ideal of \mathbb Z[\omega] is principal.

The units are the elements of norm one. Solving a^2-ab+b^2=1 gives precisely six:

\displaystyle \mathcal O_M^\times=\{\pm1,\pm\omega,\pm\omega^2\}.

The behavior of rational primes in M is controlled by the polynomial X^2+X+1 . For p\neq3 , this polynomial has two distinct roots modulo p when p\equiv1\pmod3 , and is irreducible when p\equiv2\pmod3 . Consequently, in the first case (p)=\mathfrak q\bar{\mathfrak q} , with both prime ideals of norm p , while in the second case (p) remains prime and has norm p^2 . At 3 , the identity (1-\omega)^2=-3\omega shows that the prime ramifies.

The extension L=M(\alpha) is now cyclic of degree three. Indeed, its transformations fixing M are 1,r,r^2 , and they send \alpha to \alpha,\omega\alpha,\omega^2\alpha . Thus the noncommuting group of six transformations becomes a cyclic group when we work over M . This is the crucial simplification: a two-dimensional representation over \mathbb Q can be reconstructed from a one-dimensional cubic character over M .

Cubic character

Let \mathfrak q be a prime ideal of \mathbb Z[\omega] not dividing 6 , and write Q=N\mathfrak q . If the underlying rational prime splits in M , then Q=p , with p\equiv1\pmod3 . If it remains inert, then Q=p^2 , with p\equiv2\pmod3 . In both cases, Q\equiv1\pmod3 . Therefore the multiplicative group of the residue field \mathcal O_M/\mathfrak q has order divisible by three.

Consider the residue of 2^{(Q-1)/3} modulo \mathfrak q . Its cube is 2^{Q-1}\equiv1\pmod{\mathfrak q} , by the finite-field analogue of Fermat’s theorem. Hence this residue is one of the three cube roots of unity. Because \mathfrak q\nmid3 , the three residues 1,\omega,\omega^2 remain distinct, so exactly one of these complex cube roots represents the computed residue. We define

\displaystyle \psi(\mathfrak q)\equiv2^{(N\mathfrak q-1)/3}\pmod{\mathfrak q},\qquad\psi(\mathfrak q)\in\{1,\omega,\omega^2\}.

We extend this definition multiplicatively to integral ideals relatively prime to 6 . If \mathfrak a=\prod_j\mathfrak q_j^{e_j} , put \psi(\mathfrak a)=\prod_j\psi(\mathfrak q_j)^{e_j} . Unique factorization of ideals makes the definition unambiguous. The result is a cubic character of ideals, and its construction has required nothing beyond finite fields and multiplication of prime ideals.

The relation to the cubic equation becomes transparent when we examine Frobenius in the extension L/M . Choose a prime \mathfrak Q of L above \mathfrak q . In the residue field, the Frobenius operation is raising to the Q -th power. Since \alpha^3=2 , we have

\displaystyle \begin{aligned}\frac{\alpha^Q}{\alpha}&=\alpha^{Q-1}\\&=(\alpha^3)^{(Q-1)/3}\\&=2^{(Q-1)/3}.\end{aligned}

Every transformation of L/M sends \alpha to one of \alpha,\omega\alpha,\omega^2\alpha . Thus the preceding quotient identifies precisely which transformation acts as Frobenius, and gives

\displaystyle \mathrm{Frob}_{\mathfrak q}(\alpha)=\psi(\mathfrak q)\alpha.

This identity is the elementary heart of the Artin–Hecke comparison. The cubic character records the multiplier by which Frobenius moves the cubic root, while the Artin representation records the corresponding permutation of all three roots. We have obtained the character directly from finite-field arithmetic, without yet invoking cubic reciprocity or class field theory.

Comparing the Artin and Hecke Euler factors

Define the ideal-character series

\displaystyle L_M(s,\psi)=\sum_{\substack{\mathfrak a\ne0\\(\mathfrak a,6)=1}}\frac{\psi(\mathfrak a)}{(N\mathfrak a)^s}.

Unique factorization of ideals gives its Euler product

\displaystyle L_M(s,\psi)=\prod_{\mathfrak q\nmid6}\left(1-\psi(\mathfrak q)(N\mathfrak q)^{-s}\right)^{-1}.

To compare it with the Artin function, we must group the factors indexed by prime ideals of M according to the underlying rational prime. Suppose first that p\equiv1\pmod3 . Then (p)=\mathfrak q\bar{\mathfrak q} , with both ideals of norm p . Conjugation interchanges the ideals and sends the cube root \omega to \omega^2 , so \psi(\bar{\mathfrak q})=\psi(\mathfrak q)^{-1} . Writing u=\psi(\mathfrak q) and T=p^{-s} , the two ideal factors combine to give

\displaystyle \begin{aligned}L_{M,p}(s,\psi)&=\frac1{(1-uT)(1-u^{-1}T)}\\&=\frac1{1-(u+u^{-1})T+T^2}.\end{aligned}

If u=1 , this is (1-T)^{-2} , the Artin factor for the identity transformation. If u=\omega or \omega^2 , then u+u^{-1}=-1 , and the factor is (1+T+T^2)^{-1} , the Artin factor for a three-cycle. Thus the two character values on the conjugate prime ideals are exactly the two eigenvalues of the corresponding Frobenius matrix.

Now suppose p\equiv2\pmod3 , with p\neq2 . The rational prime remains prime in M , and its ideal norm is p^2 . We claim that its cubic character value is one. Indeed,

\displaystyle \frac{p^2-1}{3}=(p-1)\frac{p+1}{3},

and the second factor is an integer because p\equiv2\pmod3 . Fermat’s theorem therefore gives 2^{(p^2-1)/3}\equiv1\pmod p , so \psi((p))=1 . The local factor is

\displaystyle L_{M,p}(s,\psi)=\frac1{1-p^{-2s}}.

This is exactly the Artin factor of a transposition. The appearance of p^{-2s} has a simple explanation: the quadratic field contains no prime ideal of norm p in this case, but it contains one of norm p^2 . Thus the factorization of primes in the quadratic field automatically reproduces the quadratic denominator of the two-dimensional representation.

At 2 and 3 , the cubic character is ramified and the primitive Hecke factors are one. The corresponding Artin factors were calculated independently to be one. We have therefore matched every Euler factor, including those at ramified primes, and conclude that

\displaystyle\boxed{L(s,\rho)=L_M(s,\psi).}

Combining this with the factorization of the cubic Dedekind zeta function gives the principal identity

\displaystyle \zeta_{\mathbb Q(\sqrt[3]{2})}(s)=\zeta(s)L_{\mathbb Q(\sqrt{-3})}(s,\psi).

The equality has a concrete local explanation. Split primes in the quadratic field contribute two character values, which become the two eigenvalues of the Artin matrix. Inert primes contribute a single prime ideal of norm p^2 , producing the transposition factor. Ramified primes are handled by the invariant-space calculation. No general nonabelian reciprocity theorem is required for this Euler-factor comparison.

Explicit calculations at the primes (5), (7), and (31)

It is useful to see the character calculation in several examples, since these illustrate all the possible local factors. At p=5 , the prime remains inert in \mathbb Q(\omega) , and the ideal (5) has norm 25 . Its cubic character is determined by 2^{(25-1)/3}=2^8\equiv1\pmod5 , giving \psi((5))=1 . Hence the Hecke Euler factor is (1-5^{-2s})^{-1} , corresponding to a transposition. There is no coefficient of 5^{-s} in the expansion because there is no ideal of norm 5 in the quadratic field.

At p=7 , we have the factorization

\displaystyle7=(3+\omega)(3+\omega^2).

Both factors have norm 7 , since N(3+\omega)=9-3+1=7 . Modulo the ideal (3+\omega) , we have \omega\equiv-3\equiv4\pmod7 . On the other hand, 2^{(7-1)/3}=2^2=4 . Therefore \psi((3+\omega))=\omega , and the conjugate ideal has character value \omega^2 . The combined factor is

\displaystyle \begin{aligned}L_{M,7}(s,\psi)&=\frac1{(1-\omega7^{-s})(1-\omega^27^{-s})}\\&=\frac1{1+7^{-s}+7^{-2s}}.\end{aligned}

This is the Artin factor for a three-cycle, and its first prime coefficient is \omega+\omega^2=-1 . This agrees with the fact that X^3-2 has no root modulo 7 : the number of roots minus one is -1 .

At p=31 , the prime also splits in the quadratic field:

\displaystyle31=(6+\omega)(6+\omega^2).

Since 2^{10}=1024\equiv1\pmod{31} , the cubic character is trivial at both prime ideals. Thus the Euler factor is (1-31^{-s})^{-2} , and the first coefficient is a_{31}=2 . The original cubic has three distinct roots modulo 31 , so its Frobenius permutation is the identity, whose two-dimensional trace is indeed two. These examples show that the character values, matrix traces, and root counts are three descriptions of one arithmetic phenomenon.

Cubic reciprocity

The finite-field definition of \psi is direct, but it requires exponentiation modulo a variable prime ideal. We shall now construct another cubic character, this time using only the residue of a normalized generator modulo a fixed ideal. The comparison of the two constructions is a special case of cubic reciprocity, and it is important to distinguish this deeper statement from the elementary Frobenius calculation already completed.

Every ideal of \mathbb Z[\omega] is principal, but a generator is determined only up to multiplication by one of the six units. We need a normalization that selects one generator from the six possibilities. Reduction modulo 3 provides exactly such a rule. The quotient ring \mathbb Z[\omega]/(3) has nine elements, of which six are units, and the six global units give precisely these six invertible residues. Consequently, for every \beta relatively prime to 3 , there is a unique unit u such that u\beta\equiv1\pmod3 . We denote this uniquely normalized generator by \beta^\circ=u\beta .

Now reduce \beta^\circ modulo 2 . Since X^2+X+1 is irreducible over \mathbb F_2 , the quotient ring is the field \mathbb F_4 , whose nonzero elements are represented by 1,\omega,\omega^2 . We therefore obtain a character on principal ideals coprime to 6 by defining

\displaystyle \psi_{\mathrm{cong}}((\beta))=\text{the cube root represented by }\beta^\circ\pmod2.

This character is multiplicative because the product of two generators congruent to one modulo 3 is again normalized modulo 3 . The construction is also independent of the original choice of generator, since the normalization removes the ambiguity arising from units.

The calculation for \beta=3+\omega is instructive. Multiplication by \omega^2 gives

\displaystyle \beta^\circ=\omega^2(3+\omega)=1+3\omega^2.

This element is congruent to one modulo 3 . Modulo 2 , however, it becomes 1+\omega^2=\omega , and hence \psi_{\mathrm{cong}}((3+\omega))=\omega . This agrees with the finite-field power calculation at the prime above 7 .

More generally, let \pi be a generator normalized by \pi\equiv1\pmod3 . Cubic reciprocity asserts, in the present special case, that the cube root represented by 2^{(N\pi-1)/3} modulo \pi is the same as the cube root represented by \pi modulo 2 . In our notation, the statement is

\displaystyle \psi((\pi))=\psi_{\mathrm{cong}}((\pi)).

The significance of the identity lies in the exchange of the two moduli. The original character asks for the cubic residue of the fixed integer 2 modulo a variable prime \pi . The congruence character instead reduces the normalized variable prime \pi modulo the fixed integer 2 . That reversal is the substance of reciprocity. It is not implied by the fact that both constructions are multiplicative, nor by checking a few small primes. A proof can be developed from cubic Gauss sums and their transformation properties, but that additional argument is distinct from the Artin–Hecke Euler-factor comparison. We shall use the reciprocity identity to obtain an explicit periodic description of the character, while keeping clear that its proof is a separate classical number-theoretic task.

Conductor of cubic ideal character

The congruence formula shows that \psi depends only on the residue of a generator modulo 6 . Indeed, normalization requires its residue modulo 3 , while the resulting cube root is determined by reduction modulo 2 . If \beta\equiv1\pmod6 , then \beta is already normalized modulo 3 and reduces to one modulo 2 , giving \psi((\beta))=1 . Thus the ideal (6) is a modulus for the character.

To understand why this modulus is minimal, consider the invertible residue classes. The quotient modulo 2 is \mathbb F_4 , whose unit group has three elements. The quotient modulo 3 has six invertible elements. By the Chinese remainder theorem, the unit group modulo 6 has 3\cdot6=18 elements. The six global units act on these classes, and after quotienting by them there remain exactly three classes, which our cubic character distinguishes by the values 1,\omega,\omega^2 . Examination of these classes shows that neither the factor modulo 2 nor the normalization factor modulo 3 can be omitted. Hence the exact conductor ideal is

\displaystyle\boxed{\mathfrak f_\psi=(6),\qquad N\mathfrak f_\psi=36.}

The discriminant of the quadratic field M=\mathbb Q(\sqrt{-3}) is D_M=-3 . We therefore have

\displaystyle |D_M|N\mathfrak f_\psi=3\cdot36=108.

This is the same integer obtained earlier by the ramification-group calculation for the two-dimensional Artin representation. On the Galois side, the factor 2^2 3^3 arose from inertia and wild ramification. On the Hecke side, the quadratic discriminant contributes a factor 3 , while the character conductor contributes 36 . The equality of these quantities is an additional arithmetic consistency check on the correspondence, and it will determine the level of the modular form.

Ideals and quadratic lattice sum

We now turn from algebraic arithmetic to analytic sums. Since every ideal of \mathbb Z[\omega] is principal, the Hecke series can be rewritten as a sum over nonzero elements a+b\omega . Each nonzero principal ideal has six generators, differing by the units \pm1,\pm\omega,\pm\omega^2 . The norm of its generator is a^2-ab+b^2 . Therefore

\displaystyle L_M(s,\psi)=\frac16\sum_{\substack{(a,b)\in\mathbb Z^2\\(a+b\omega,6)=1}}\frac{\psi((a+b\omega))}{(a^2-ab+b^2)^s}.

The factor 1/6 corrects the sixfold counting of each ideal. We may extend the character by zero to elements not relatively prime to 6 , and write w(a,b)=\psi((a+b\omega)) on admissible pairs and w(a,b)=0 otherwise. By the congruence description, w is periodic in both variables modulo 6 . The Hecke series becomes

\displaystyle L_M(s,\psi)=\frac16\sum_{\substack{(a,b)\in\mathbb Z^2\\(a,b)\ne(0,0)}}\frac{w(a,b)}{(a^2-ab+b^2)^s}.

We have thus converted the two-dimensional Artin L -function into a weighted sum over an ordinary lattice in the plane. Its denominator is a positive definite quadratic form, while its numerator is a bounded periodic function. These are precisely the features that make Fourier analysis useful.

Let Q(a,b)=a^2-ab+b^2 . Its associated symmetric matrix is

\displaystyle A=\begin{pmatrix}2&-1\\-1&2\end{pmatrix},\qquad Q(a,b)=\frac12\begin{pmatrix}a&b\end{pmatrix}A\begin{pmatrix}a\\b\end{pmatrix}.

The matrix has determinant three and positive eigenvalues, so the quadratic form is positive definite. Its inverse is

\displaystyle A^{-1}=\frac13\begin{pmatrix}2&1\\1&2\end{pmatrix}.

This inverse matrix will appear naturally when we apply Poisson summation. The importance of the explicit quadratic form is that the analytic transformation can now be calculated directly rather than described only through abstract ideal theory.

Theta series and the Fourier coefficients

Replace the denominator in the lattice sum by an exponential weight. For z in the upper half-plane, define

\displaystyle f(z)=\frac16\sum_{a,b\in\mathbb Z}w(a,b)e^{2\pi i(a^2-ab+b^2)z}.

The term corresponding to a=b=0 vanishes because the extended character is zero there. Since Q(a,b) is positive definite and w(a,b) is bounded, the series converges normally on compact subsets of the upper half-plane and hence defines a holomorphic function. Writing q=e^{2\pi iz} and collecting lattice points of equal norm gives the Fourier expansion

\displaystyle\boxed{f(z)=\sum_{n\geq1}a_nq^n,}

with coefficients

\displaystyle a_n=\frac16\sum_{\substack{a,b\in\mathbb Z\\a^2-ab+b^2=n}}w(a,b).

Equivalently, in ideal-theoretic language,

\displaystyle a_n=\sum_{\substack{N\mathfrak a=n\\(\mathfrak a,6)=1}}\psi(\mathfrak a).

This formula allows us to calculate the coefficients by finite arithmetic. For n=1 , the solutions of a^2-ab+b^2=1 are the six units of \mathbb Z[\omega] . Each generates the unit ideal and has character value one, so a_1=1 . At n=7 , the two ideals of norm seven have character values \omega,\omega^2 , giving a_7=-1 . At n=31 , the two ideals of norm thirty-one both have character value one, giving a_{31}=2 . At n=5 , there is no ideal of norm five, and therefore a_5=0 . These are exactly the traces of the corresponding Frobenius matrices calculated earlier.

The Mellin transform of the theta series returns the original Dirichlet series. Indeed, putting z=iy , with y>0 , gives f(iy)=\sum_{n\geq1}a_ne^{-2\pi ny} . For \mathrm{Re}s>1 , absolute convergence allows us to interchange summation and integration, and the gamma integral gives

\displaystyle \begin{aligned}\int_0^\infty f(iy)y^{s-1}\,dy&=\sum_{n\ge1}a_n\int_0^\infty e^{-2\pi ny}y^{s-1}\,dy\\&=(2\pi)^{-s}\Gamma(s)\sum_{n\ge1}\frac{a_n}{n^s}.\end{aligned}

Consequently,

\displaystyle \int_0^\infty f(iy)y^{s-1}\,dy=(2\pi)^{-s}\Gamma(s)L(s,\rho).

The gamma factor has appeared from a simple integral, and the Dirichlet coefficients have become Fourier coefficients of a holomorphic function. The advantage is that the theta series can be transformed by Fourier analysis, and these transformations can then be translated into analytic properties of the L -function.

Hecke Operators

Before examining modular transformations, we can derive the principal arithmetic relations among the Fourier coefficients directly from the Euler product. For a prime p\nmid6 , the local factor is

\displaystyle\frac{1}{1-a_pp^{-s}+\chi_{-3}(p)p^{-2s}}.

Put T=p^{-s} and expand the factor as a formal power series:

\displaystyle \frac1{1-a_pT+\chi_{-3}(p)T^2}=\sum_{r\ge0}a_{p^r}T^r.

Multiplying through by the denominator and comparing coefficients gives a_{p^2}=a_p^2-\chi_{-3}(p) , followed by the general recurrence

\displaystyle a_{p^{r+1}}=a_pa_{p^r}-\chi_{-3}(p)a_{p^{r-1}}.

For example, at p=7 we have a_7=-1 and \chi_{-3}(7)=1 , so a_{49}=(-1)^2-1=0 . Direct ideal counting gives the same result: the three ideals of norm 49 are \mathfrak q^2,\mathfrak q\bar{\mathfrak q},\bar{\mathfrak q}^2 , with character values \omega^2,1,\omega , whose sum is zero. At p=5 , we have a_5=0 and \chi_{-3}(5)=-1 , so a_{25}=1 , agreeing with the single ideal (5) of norm twenty-five.

Unique factorization of ideals also gives a_{mn}=a_ma_n whenever (m,n)=1 . Combining this with the prime-power recurrence yields, for integers m,n relatively prime to 6 ,

\displaystyle a_ma_n=\sum_{d\mid(m,n)}\chi_{-3}(d)a_{mn/d^2}.

In particular,

\displaystyle a_{pn}+\chi_{-3}(p)a_{n/p}=a_pa_n,

where a_{n/p}=0 if p\nmid n . This is exactly the coefficient identity expected of a normalized eigenform for the Hecke operator T_p .

To see the analytic meaning, write f(z)=\sum_{n\geq1}a_nq^n . For a modular form of weight one and quadratic character \chi_{-3} , the Hecke operator at a prime p away from the level acts on Fourier expansions by

\displaystyle T_pf(z)=\sum_{n\ge1}\left(a_{pn}+\chi_{-3}(p)a_{n/p}\right)q^n.

Our ideal-multiplication calculation therefore gives T_pf=a_pf as an identity of Fourier series. The calculation was obtained before proving that f is modular. The remaining analytic work is to establish that the theta series transforms under the appropriate fractional linear substitutions, thereby placing it in the modular-form space on which these Hecke operators act.

Poisson Summation and Modularity

The basic analytic mechanism behind the modular transformation law is Poisson summation. Consider the Gaussian on \mathbb R^2 associated with the quadratic form Q :

\displaystyle g_y(x)=e^{-\pi yx^TAx},\qquad A=\begin{pmatrix}2&-1\\-1&2\end{pmatrix},\qquad y>0.

Use the Fourier transform \widehat g_y(\xi)=\int_{\mathbb R^2}g_y(x)e^{-2\pi ix\cdot\xi},dx . The Gaussian integral gives

\displaystyle \widehat g_y(\xi)=\frac1{y\sqrt3}e^{-\pi\xi^TA^{-1}\xi/y}.

The structure of this formula is worth understanding. The factor y^{-1} comes from integrating over two real variables: a scaling by y^{1/2} in each variable contributes y^{-1/2} , and the two contributions multiply. The factor 1/\sqrt3 comes from the determinant of the quadratic-form matrix. Finally, the inverse matrix A^{-1} appears because Fourier transformation exchanges a Gaussian with another Gaussian whose quadratic form is dual to the original one.

Applying the two-dimensional Poisson summation formula to the integer lattice gives

\displaystyle \sum_{m\in\mathbb Z^2}g_y(m)=\sum_{k\in\mathbb Z^2}\widehat g_y(k),

and therefore

\displaystyle \sum_{m\in\mathbb Z^2}e^{-\pi ym^TAm}=\frac1{y\sqrt3}\sum_{k\in\mathbb Z^2}e^{-\pi k^TA^{-1}k/y}.

This is the elementary transformation formula for the unweighted theta series of the Eisenstein lattice. It already explains why a rank-two lattice naturally produces a factor of weight one under inversion of the variable. A lattice of rank r produces a scaling factor y^{-r/2} ; for r=2 , this is y^{-1} .

Our actual theta series contains the periodic weight w(a,b) , so we must go further. The correct method is to separate the sum into residue classes modulo 6 , apply Poisson summation to each class, and then recombine the results using the finite Fourier transform of the character.

For a residue vector r\in(\mathbb Z/6\mathbb Z)^2 , define

\displaystyle \Theta_r(y)=\sum_{\substack{m\in\mathbb Z^2\\m\equiv r\pmod6}}e^{-\pi ym^TAm}.

Writing m=r+6n , this is a Gaussian sum over a translate of the lattice 6\mathbb Z^2 . The covolume of that lattice is 36 , and Poisson summation gives

\displaystyle \Theta_r(y)=\frac1{36y\sqrt3}\sum_{k\in\mathbb Z^2}e^{2\pi ik\cdot r/6}e^{-\pi k^TA^{-1}k/(36y)}.

Our weighted theta series is obtained by multiplying \Theta_r(y) by w(r) , summing over the residue classes, and dividing by six. Define the finite Fourier transform of the periodic weight by

\displaystyle \widehat w(k)=\sum_{r\in(\mathbb Z/6\mathbb Z)^2}w(r)e^{2\pi ik\cdot r/6}.

The resulting Poisson formula is

\displaystyle f(iy)=\frac1{216y\sqrt3}\sum_{k\in\mathbb Z^2}\widehat w(k)e^{-\pi k^TA^{-1}k/(36y)}.

This is the concrete analytic calculation underlying the theta transformation. The transformed expression contains the inverse quadratic form, a scaling by the lattice modulus, and a finite Fourier transform of the cubic character. To obtain the full modular transformation law, one must evaluate these finite character sums and understand how the resulting collection of residue-class theta functions transforms under the appropriate matrices.

One immediate cancellation is worth recording. Since the nonzero character values occur equally often among the admissible residue classes, their sum is zero:

\displaystyle \widehat w(0)=6(1+\omega+\omega^2)=0.

Thus the zero-frequency term in the Poisson expansion disappears. This cancellation is an analytic reflection of the nontriviality of the character, and it is related to the cuspidal character of the resulting modular form. Nevertheless, vanishing of this single Fourier coefficient is not sufficient by itself to prove cuspidality at every cusp. The complete argument requires the finite Fourier transformations associated with all the relevant cusps.

Weight one and level 108

The modular transformation law we seek is

\displaystyle f\left(\frac{az+b}{cz+d}\right)=\chi_{-3}(d)(cz+d)f(z).

for every matrix \begin{pmatrix}a&b\\c&d\end{pmatrix} in the congruence subgroup \Gamma_0(108) , where ad-bc=1 and 108\mid c . The factor cz+d indicates weight one, while \chi_{-3}(d) is the quadratic character associated with \mathbb Q(\sqrt{-3}) .

The weight can be understood from the Gaussian calculation. Our theta series is built from a positive definite quadratic form in two real variables, and the Fourier transform of its Gaussian contributes the factor y^{-1} . For a quadratic theta series of rank r , the corresponding factor is y^{-r/2} , so the rank-two case naturally produces weight one. This analytic observation explains the weight without treating it as an arbitrary parameter attached to the L -function.

The level has a different origin. The lattice is associated with a quadratic field of discriminant -3 , and the character has conductor ideal (6) of norm 36 . The finite Fourier-transform calculation combines these arithmetic quantities into the integer

\displaystyle |D_M|N\mathfrak f_\psi=3\cdot36=108.

This is the same conductor obtained from the ramification groups of the two-dimensional Artin representation. The equality is not accidental: the arithmetic conductor and the analytic level measure the same restrictions on the local transformations, expressed in two different languages.

The classical theorem concerning theta series of finite-order characters of imaginary quadratic fields now gives the modular transformation law. Since our character has trivial restriction to rational principal ideals, the resulting nebentypus is exactly the quadratic character \chi_{-3} . The theta series is therefore a weight-one modular form of level 108 and character \chi_{-3} :

\displaystyle\boxed{f\in S_1(\Gamma_0(108),\chi_{-3}).}

The fact that it is cuspidal follows from the nontriviality and genuinely cubic nature of the ideal character. More precisely, the character is not obtained merely by composing a character of \mathbb Q with the norm map from M , so the associated weight-one theta series is cuspidal.

It is important to identify the precise point at which a general theorem enters. We have explicitly calculated the Gaussian Fourier transform, the Poisson formula on residue classes, the character modulus, and the resulting conductor. What has not been carried out in full here is the evaluation of every finite Fourier transform needed to establish the transformation law at every cusp. The classical theta-series theorem supplies that remaining calculation. Thus the modularity assertion rests on a definite Fourier-analytic result, not merely on the existence of an Euler product or on the formal resemblance of the coefficient recurrence to the Hecke relations.

Completed L-function and its functional equation

The completed L -function provides another comparison between the arithmetic and analytic constructions. The cubic field F=\mathbb Q(\sqrt[3]{2}) has one real embedding and one pair of complex embeddings. Its completed Dedekind zeta function is

\displaystyle \Lambda_F(s)=|D_F|^{s/2}\Gamma_{\mathbb R}(s)\Gamma_{\mathbb C}(s)\zeta_F(s),

where

\displaystyle \Gamma_{\mathbb R}(s)=\pi^{-s/2}\Gamma(s/2),\qquad\Gamma_{\mathbb C}(s)=2(2\pi)^{-s}\Gamma(s).

The completed Riemann zeta function is \Lambda_{\mathbb Q}(s)=\Gamma_{\mathbb R}(s)\zeta(s) . Since \zeta_F(s)=\zeta(s)L(s,\rho) , taking the quotient gives

\displaystyle \frac{\Lambda_F(s)}{\Lambda_{\mathbb Q}(s)}=108^{s/2}\Gamma_{\mathbb C}(s)L(s,\rho).

We therefore define the completed Artin function by

\displaystyle \Lambda(s,\rho)=108^{s/2}\,2(2\pi)^{-s}\Gamma(s)L(s,\rho).

Both completed Dedekind zeta functions satisfy functional equations under the substitution s\mapsto1-s . Their quotient consequently satisfies

\displaystyle\boxed{\Lambda(s,\rho)=\Lambda(1-s,\rho).}

The root number is therefore +1 . This calculation also identifies the archimedean gamma factor. Complex conjugation acts on our two-dimensional representation with eigenvalues 1 and -1 , so the two corresponding real gamma factors are \Gamma_{\mathbb R}(s) and \Gamma_{\mathbb R}(s+1) . Their product is \Gamma_{\mathbb C}(s) , by the duplication formula for the gamma function. Thus the archimedean factor obtained from the Galois representation agrees with the one obtained from the quotient of Dedekind zeta functions and with the factor expected for a holomorphic modular form of weight one.

We have now matched not only the finite Euler factors and the conductor, but also the gamma factor and the functional equation. These are essential parts of the analytic correspondence. An equality of formal prime coefficients would not by itself provide this global analytic structure.

The complete correspondence

We can now collect the constructions and see precisely what has been established. The factorization of X^3-2 modulo a prime determines the decomposition of that prime into ideals of the cubic field. The degrees of the factors are the cycle lengths of the finite-field Frobenius operation on the roots. Representing this permutation on the three-dimensional root space, and removing the universally fixed constant direction, produces a two-dimensional representation \rho . Its trace is the number of roots of the cubic modulo p minus one, while its determinant is the quadratic character \chi_{-3}(p) . Its Euler factors are therefore determined explicitly by elementary congruences.

Over the quadratic field M=\mathbb Q(\omega) , the degree-six splitting field becomes a cyclic cubic extension. The multiplier by which Frobenius acts on \alpha is the cubic character value \psi(\mathfrak q) , calculated by finite exponentiation. At rational primes splitting in M , the two conjugate character values produce the two eigenvalues of the Artin matrix. At inert primes, the single prime ideal has norm p^2 and produces the transposition factor. This proves the Artin–Hecke identity at every Euler factor.

Cubic reciprocity then identifies the finite-field power-residue character with a congruence character obtained from normalized generators modulo 6 . This makes the character periodic on the Eisenstein lattice, so the Hecke L -series can be rewritten as a weighted sum involving the quadratic form a^2-ab+b^2 . Replacing the denominator by an exponential produces a holomorphic theta series, whose Fourier coefficients are the weighted ideal counts. The multiplicative relations of these coefficients come directly from the Euler product and coincide with the Hecke eigenvalue relations. Finally, the Fourier transformation of the quadratic Gaussian, together with the finite Fourier analysis of the character, gives the modular transformation law and the level 108 .

The resulting identities are

\displaystyle \begin{aligned}\zeta_{\mathbb Q(\sqrt[3]{2})}(s)&=\zeta(s)L(s,\rho),\\L(s,\rho)&=L_{\mathbb Q(\omega)}(s,\psi),\\L(s,\rho)&=\sum_{n\ge1}\frac{a_n}{n^s},\\f(z)&=\sum_{n\ge1}a_ne^{2\pi inz},\\f&\in S_1(\Gamma_0(108),\chi_{-3}).\end{aligned}

The value of this example lies in the fact that the correspondence is visible through finite calculations. The Artin side begins with permutations of roots and produces determinants. The Hecke side begins with power residues and produces character values on prime ideals. The modular side begins with a weighted quadratic lattice and produces Fourier coefficients. These apparently different objects agree because they preserve the same local arithmetic: Frobenius eigenvalues, residue-field degrees, ramification behavior, conductor exponents, and multiplicative relations.

The method also shows the special feature that makes this example accessible. The six-element group of the splitting field contains a cyclic subgroup of index two. Restricting the two-dimensional representation to that subgroup separates it into two one-dimensional characters, exchanged by the remaining involution. Thus a nonabelian problem over \mathbb Q can be reconstructed from abelian arithmetic over the quadratic field \mathbb Q(\omega) . For a general Galois representation, such a reduction need not be possible, and one cannot expect to obtain the corresponding modular form from a character of an imaginary quadratic field.

The broader reciprocity problem asks whether other analytic constructions can reproduce the same local characteristic polynomials when no such reduction to one-dimensional characters exists. Our calculations indicate what those constructions must preserve. They must recover the entire Euler factor, not merely its first coefficient; they must give the correct local behavior at ramified primes and the correct conductor; they must produce the appropriate gamma factors and functional equation; and their Fourier coefficients must obey multiplication laws compatible with the prime decomposition of the original arithmetic problem. In the present cubic case, these requirements are met through finite fields, the arithmetic of Eisenstein integers, and Fourier analysis on a two-dimensional lattice. That is the concrete classical model from which one can begin to understand the more general relation between Galois representations and automorphic forms.

Although the Artin Euler factors, the conductor, the Hecke ideal series, the theta coefficients, and the basic Poisson transformation have been calculated explicitly, two deeper parts of the argument deserve independent treatment if one wishes to make the entire article self-contained without invoking named theorems. The first is the cubic reciprocity identity relating 2^{(N\pi-1)/3}\pmod\pi to the normalized generator \pi\pmod2 . A classical proof would introduce cubic Gauss sums, examine their behavior under multiplication and conjugation, and derive the congruence relation that exchanges the fixed integer 2 with the varying Eisenstein prime \pi . This is where reciprocity enters as a genuinely new arithmetic assertion, beyond the formal properties of finite-field Frobenius. The second is the complete theta transformation calculation. The ordinary Gaussian Fourier transform gives the factor y^{-1} and the dual quadratic form, but the character-weighted theta series requires the evaluation of finite Fourier sums modulo 6 and the verification of the transformations at the cusps of \Gamma_0(108) . Carrying out these sums explicitly would reveal how the finite arithmetic of the character determines the modular transformation constants. It would also provide a direct proof of cuspidality, rather than appealing to the general theorem for theta series attached to nontrivial imaginary-quadratic Hecke characters. These two calculations are particularly worthwhile because they mark the points at which new mathematical information enters. The Frobenius identity follows directly from \alpha^3=2 ; cubic reciprocity goes further by expressing a variable-prime power residue through a fixed modulus. The Mellin transform follows from the gamma integral; modularity goes further by asserting that the weighted lattice sum possesses a rigid transformation law under a large collection of fractional linear substitutions.

Leave a Reply