Let K = ℚ[X] / (T) a number field, ℤ_K its ring of integers, T ∈ ℤ[X] is monic. Three basic number field structures can be attached to K in GP:

* nf denotes a number field, i.e. a data structure output by nfinit. This contains the basic arithmetic data attached to the number field: signature, maximal order (given by a basis nf.zk), discriminant, defining polynomial T, etc.

* bnf denotes a "Buchmann's number field", i.e. a data structure output by bnfinit. This contains nf and the deeper invariants of the field: units U(K), class group Cl(K), as well as technical data required to solve the two attached discrete logarithm problems.

* bnr denotes a "ray number field", i.e. a data structure output by bnrinit, corresponding to the ray class group structure of the field, for some modulus f. It contains a bnf, the modulus f, the ray class group Cl_f(K) and data attached to the discrete logarithm problem therein.

An algebraic number belonging to K = ℚ[X]/(T) is given as

* a t_INT, t_FRAC or t_POL (implicitly modulo T), or

* a t_POLMOD (modulo T), or

* a t_COL v of dimension N = [K:ℚ], representing the element in terms of the computed integral basis, as sum(i = 1, N, v[i] * nf.zk[i]). Note that a t_VEC will not be recognized.

An ideal is given in any of the following ways:

* an algebraic number in one of the above forms, defining a principal ideal.

* a prime ideal, i.e. a 5-component vector in the format output by idealprimedec or idealfactor.

* a t_MAT, square and in Hermite Normal Form (or at least upper triangular with nonnegative coefficients), whose columns represent a ℤ-basis of the ideal.

One may use idealhnf to convert any ideal to the last (preferred) format.

* an extended ideal is a 2-component vector [I, t], where I is an ideal as above and t is an algebraic number, representing the ideal (t)I. This is useful whenever idealred is involved, implicitly working in the ideal class group, while keeping track of principal ideals. The following multiplicative ideal operations update the principal part: idealmul, idealinv, idealsqr, idealpow and idealred; e.g. using idealmul on [I,t], [J,u], we obtain [IJ, tu]. In all other functions, the extended part is silently discarded, e.g. using idealadd with the above input produces I+J.

The "principal part" t in an extended ideal may be represented in any of the above forms, and also as a factorization matrix (in terms of number field elements, not ideals!), possibly the empty factorization matrix factor(1) representing 1; the empty matrix [;] is also accepted as a synonym for 1. When t is such a factorization matrix, elements stay in factored form, or famat for factorization matrix, which is a convenient way to avoid coefficient explosion. To recover the conventional expanded form, try nffactorback; but many functions already accept famats as input, for instance ideallog, so expanding huge elements should never be necessary.

A finite abelian group G in user-readable format is given by its Smith Normal Form as a pair [h,d] or triple [h,d,g]. Here h is the cardinality of G, (d_i) is the vector of elementary divisors, and (g_i) is a vector of generators. In short, G = ⨁ _{i ≤ n} (ℤ/d_iℤ) g_i, with d_n | ... | d₂ | d₁ and ∏_i d_i = h. This information can also be retrieved as G.no, G.cyc and G.gen.

* a character on the abelian group ⨁ (ℤ/d_jℤ) g_j is given by a row vector χ = [a₁,...,a_n] such that χ(∏_j g_j^n_j) = exp(2π i∑_j a_j n_j / d_j).

* given such a structure, a subgroup H is input as a square matrix in HNF, whose columns express generators of H on the given generators g_i. Note that the determinant of that matrix is equal to the index (G:H).

We now have a look at data structures attached to relative extensions of number fields L/K, and to projective ℤ_K-modules. When defining a relative extension L/K, the nf attached to the base field K must be defined by a variable having a lower priority (see Section se:priority) than the variable defining the extension. For example, you may use the variable name y to define the base field K, and x to define the relative extension L/K.

Basic definitions.

* rnf denotes a relative number field, i.e. a data structure output by rnfinit, attached to the extension L/K. The nf attached to be base field K is rnf.nf.

* A relative matrix is an m x n matrix whose entries are elements of K, in any form. Its m columns A_j represent elements in Kⁿ.

* An ideal list is a row vector of fractional ideals of the number field nf.

* A pseudo-matrix is a 2-component row vector (A,I) where A is a relative m x n matrix and I an ideal list of length n. If I = {𝔞₁,..., 𝔞_n} and the columns of A are (A₁,..., A_n), this data defines the torsion-free (projective) ℤ_K-module 𝔞₁ A₁⨁ 𝔞_n A_n.

* An integral pseudo-matrix is a 3-component row vector (A,I,J) where A = (a_i,j) is an m x n relative matrix and I = (𝔟₁,..., 𝔟_m), J = (𝔞₁,..., 𝔞_n) are ideal lists, such that a_i,j ∈ 𝔟_i 𝔞_j^-1 for all i,j. This data defines two abstract projective ℤ_K-modules N = 𝔞₁ω₁⨁ ...⨁ 𝔞_nω_n in Kⁿ, P = 𝔟₁η₁⨁ ...⨁ 𝔟_mη_m in K^m, and a ℤ_K-linear map f:N → P given by f(∑_j α_jω_j) = ∑_i (a_i,jα_j) η_i. This data defines the ℤ_K-module M = P/f(N).

* Any projective ℤ_K-moduleprojective module M of finite type in K^m can be given by a pseudo matrix (A,I).

* An arbitrary ℤ_K module of finite type in K^m, with nontrivial torsion, is given by an integral pseudo-matrix (A,I,J)

Algebraic numbers in relative extension.

We are given a number field K = nfinit(T), attached to K = ℚ[Y]/(T), T ∈ ℚ[Y], and a relative extension L = rnfinit(K, P), attached to L = K[X]/(P), P ∈ K[X]. In all contexts (except rnfeltabstorel and rnfeltdown, see below), an algebraic number is given as

* a t_INT, t_FRAC or t_POL in ℚ[Y] (implicitly modulo T) or a t_POL in K[X] (implicitly modulo P),

* a t_POLMOD (modulo T or P), or

* a t_COL v of dimension m = [K:ℚ], representing the element in terms of the integral basis K.zk;

* if an absolute nf structure Labs was attached to L, via Labs = nfinit(L), then we can also use a t_COL v of dimension [L:ℚ], representing the element in terms of the computed integral basis Labs.zk. Be careful that in the degenerate case L = K, then the previous interpretation (with respect to K.zk) takes precedence. This is no concern when K = ℚ or if P = X - Y (because in that case the primitive polynomial Labs.pol defining L of ℚ is nf.pol and the computation of nf.zk is deterministic); but in other cases, the integer bases attached to K and Labs may differ.

Special case: rnfeltabstorel and rnfeltdown. These two functions assume that elements are given in absolute representation (with respect to Labs.zk or modulo Labs.pol and converts them to relative representation modulo L.pol. In these two functions (only), a t_POL in X is implicitly understood modulo Labs.pol and a t_COL of length [L:ℚ] refers to the integral basis Labs.zk in all cases, including L = K.

Pseudo-bases, determinant.

* The pair (A,I) is a pseudo-basis of the module it generates if the 𝔞_j are nonzero, and the A_j are K-linearly independent. We call n the size of the pseudo-basis. If A is a relative matrix, the latter condition means it is square with nonzero determinant; we say that it is in Hermite Normal Form (HNF) if it is upper triangular and all the elements of the diagonal are equal to 1.

* For instance, the relative integer basis rnf.zk is a pseudo-basis (A,I) of ℤ_L, where A = rnf.zk[1] is a vector of elements of L, which are K-linearly independent. Most rnf routines return and handle ℤ_K-modules contained in L (e.g. ℤ_L-ideals) via a pseudo-basis (A',I'), where A' is a relative matrix representing a vector of elements of L in terms of the fixed basis rnf.zk[1]

* The determinant of a pseudo-basis (A,I) is the ideal equal to the product of the determinant of A by all the ideals of I. The determinant of a pseudo-matrix is the determinant of any pseudo-basis of the module it generates.

A modulus, in the sense of class field theory, is a divisor supported on the real and finite places of K. In PARI terms, this means either an ordinary ideal I as above (no Archimedean component), or a pair [I,a], where a is a vector with r₁ {0,1}-components, corresponding to the infinite part of the divisor. More precisely, the i-th component of a corresponds to the real embedding attached to the i-th real root of K.roots. (That ordering is not canonical, but well defined once a defining polynomial for K is chosen.) For instance, [1, [1,1]] is a modulus for a real quadratic field, allowing ramification at any of the two places at infinity, and nowhere else.

A bid or "big ideal" is a structure output by idealstar needed to compute in (ℤ_K/I)^*, where I is a modulus in the above sense. It is a finite abelian group as described above, supplemented by technical data needed to solve discrete log problems.

Finally we explain how to input ray number fields (or bnr), using class field theory. These are defined by a triple A, B, C, where the defining set [A,B,C] can have any of the following forms: [bnr], [bnr,subgroup], [bnr,character], [bnf,mod], [bnf,mod,subgroup]. The last two forms are kept for backward compatibility, but no longer serve any real purpose (see example below); no newly written function will accept them.

* bnf is as output by bnfinit, where units are mandatory unless the modulus is trivial; bnr is as output by bnrinit. This is the ground field K.

* mod is a modulus 𝔣, as described above.

* subgroup a subgroup of the ray class group modulo 𝔣 of K. As described above, this is input as a square matrix expressing generators of a subgroup of the ray class group bnr.clgp on the given generators. We also allow a t_INT n for n.Cl_f.

* character is a character χ of the ray class group modulo 𝔣, representing the subgroup Ker χ.

The corresponding bnr is the subfield of the ray class field of K modulo 𝔣, fixed by the given subgroup.

    ? K = bnfinit(y^2+1);
    ? bnr = bnrinit(K, 13)
    ? %.clgp
    %3 = [36, [12, 3]]
    ? bnrdisc(bnr); \\ discriminant of the full ray class field
    ? bnrdisc(bnr, [3,1;0,1]); \\ discriminant of cyclic cubic extension of K
    ? bnrconductor(bnr, [3,1]); \\ conductor of chi: g1->zeta_12^3, g2->zeta3

We could have written directly

    ? bnrdisc(K, 13);
    ? bnrdisc(K, 13, [3,1;0,1]);

avoiding one bnrinit, but this would actually be slower since the bnrinit is called internally anyway. And now twice!

All the functions which are specific to relative extensions, number fields, Buchmann's number fields, Buchmann's number rays, share the prefix rnf, nf, bnf, bnr respectively. They take as first argument a number field of that precise type, respectively output by rnfinit, nfinit, bnfinit, and bnrinit.

However, and even though it may not be specified in the descriptions of the functions below, it is permissible, if the function expects a nf, to use a bnf instead, which contains much more information. On the other hand, if the function requires a bnf, it will not launch bnfinit for you, which is a costly operation. Instead, it will give you a specific error message. In short, the types nf ≤ bnf ≤ bnr are ordered, each function requires a minimal type to work properly, but you may always substitute a larger type.

The data types corresponding to the structures described above are rather complicated. Thus, as we already have seen it with elliptic curves, GP provides "member functions" to retrieve data from these structures (once they have been initialized of course). The relevant types of number fields are indicated between parentheses:

 bid (bnr ) : bid ideal structure.

 bnf (bnr, bnf ) : Buchmann's number field.

 clgp (bnr, bnf ) : classgroup. This one admits the following three subclasses:

  cyc : cyclic decomposition (SNF).

  gen : generators.

  no : number of elements.

 diff (bnr, bnf, nf ) : the different ideal.

 codiff (bnr, bnf, nf ) : the codifferent (inverse of the different in the ideal group).

 disc (bnr, bnf, nf ) : discriminant.

 fu ( bnf ) : fundamental units.

 index (bnr, bnf, nf ) : index of the power order in the ring of integers.

 mod (bnr ) : modulus.

 nf (bnr, bnf, nf ) : number field.

 pol (bnr, bnf, nf ) : defining polynomial.

 r1 (bnr, bnf, nf ) : the number of real embeddings.

 r2 (bnr, bnf, nf ) : the number of pairs of complex embeddings.

 reg ( bnf ) : regulator.

 roots (bnr, bnf, nf ) : roots of the polynomial generating the field.

 sign (bnr, bnf, nf ) : signature [r1,r2].

 t2 (bnr, bnf, nf ) : the T₂ matrix (see nfinit).

 tu ( bnf ) : a generator for the torsion units.

 zk (bnr, bnf, nf ) : integral basis, i.e. a ℤ-basis of the maximal order.

 zkst (bnr ) : structure of (ℤ_K/m)^*.

The member functions .codiff, .t2 and .zk perform a computation and are relatively expensive in large degree: move them out of tight loops and store them in variables.

For instance, assume that bnf = bnfinit(pol), for some polynomial. Then bnf.clgp retrieves the class group, and bnf.clgp.no the class number. If we had set bnf = nfinit(pol), both would have output an error message. All these functions are completely recursive, thus for instance bnr.bnf.nf.zk will yield the maximal order of bnr, which you could get directly with a simple bnr.zk.

Some of the functions starting with bnf are implementations of the sub-exponential algorithms for finding class and unit groups under GRH, due to Hafner-McCurley, Buchmann and Cohen-Diaz-Olivier. The general call to the functions concerning class groups of general number fields (i.e. excluding quadclassunit) involves a polynomial P and a technical vector tech = [c₁, c₂, nrpid ], where the parameters are to be understood as follows:

P is the defining polynomial for the number field, which must be in ℤ[X], irreducible and monic. In fact, if you supply a nonmonic polynomial at this point, gp issues a warning, then transforms your polynomial so that it becomes monic. The nfinit routine will return a different result in this case: instead of res, you get a vector [res,Mod(a,Q)], where Mod(a,Q) = Mod(X,P) gives the change of variables. In all other routines, the variable change is simply lost.

The tech interface is obsolete and you should not tamper with these parameters. Indeed, from version 2.4.0 on,

* the results are always rigorous under GRH (before that version, they relied on a heuristic strengthening, hence the need for overrides).

* the influence of these parameters on execution time and stack size is marginal. They can be useful to fine-tune and experiment with the bnfinit code, but you will be better off modifying all tuning parameters in the C code (there are many more than just those three). We nevertheless describe it for completeness.

The numbers c₁ ≤ c₂ are nonnegative real numbers. By default they are chosen so that the result is correct under GRH. For i = 1,2, let B_i = c_i(log |d_K|)², and denote by S(B) the set of maximal ideals of K whose norm is less than B. We want S(B₁) to generate Cl(K) and hope that S(B₂) can be proven to generate Cl(K).

More precisely, S(B₁) is a factorbase used to compute a tentative Cl(K) by generators and relations. We then check explicitly, using essentially bnfisprincipal, that the elements of S(B₂) belong to the span of S(B₁). Under the assumption that S(B₂) generates Cl(K), we are done. User-supplied c_i are only used to compute initial guesses for the bounds B_i, and the algorithm increases them until one can prove under GRH that S(B₂) generates Cl(K). A uniform result of Grenié and Molteni says that c₂ = 4 is always suitable, but this bound is very pessimistic and a direct algorithm due to Belabas-Diaz-Friedman, improved by Grenié and Molteni, is used to check the condition, assuming GRH. The default values are c₁ = c₂ = 0. When c₁ is equal to 0 the algorithm takes it equal to c₂.

nrpid is the maximal number of small norm relations attached to each ideal in the factor base. Set it to 0 to disable the search for small norm relations. Otherwise, reasonable values are between 4 and 20. The default is 4.

Warning. Make sure you understand the above! By default, most of the bnf routines depend on the correctness of the GRH. In particular, any of the class number, class group structure, class group generators, regulator and fundamental units may be wrong, independently of each other. Any result computed from such a bnf may be wrong. The only guarantee is that the units given generate a subgroup of finite index in the full unit group. You must use bnfcertify to certify the computations unconditionally.

Remarks.

You do not need to supply the technical parameters (under the library you still need to send at least an empty vector, coded as NULL). However, should you choose to set some of them, they must be given in the requested order. For example, if you want to specify a given value of nrpid, you must give some values as well for c₁ and c₂, and provide a vector [c₁,c₂,nrpid].

Note also that you can use an nf instead of P, which avoids recomputing the integral basis and analogous quantities.

Hecke Grossencharacters are continuous characters of the id\`ele class group; they generalize classical Hecke characters on ray class groups obtained through the bnr structure.

Let K be a number field, \A^× its group of id\`eles. Every Grossencharacter

χ : \A^×/K^× → ℂ^×

can be uniquely written χ = χ₀ |.|^s for some s ∈ ℂ and some character χ₀ of the compact group \A^×/(K^×.ℝ _{> 0}), where |a |= ∏_v |a_v|_v is the id\`ele norm.

Let 𝔪 be a modulus (an integral ideal and a finite set of real places). Let U(𝔪) be the subgroup of id\`eles congruent to 1 modulo 𝔪 (units outside 𝔪, positive at real places in 𝔪). The Hecke Grossencharacters defined modulo 𝔪 are the characters of the id\`ele class group

C_K(𝔪) = \A^×/(K^×.U(𝔪)),

that is, combinations of an archimedean character χ _oo on the connected component K _oo ^{x o} and a ray class group character χ_f satisfying a compatibility condition χ _oo (a)χ_f(a) = 1 for all units a congruent to 1 modulo 𝔪.

gchar * gc denotes a structure allowing to compute with Hecke Grossencharacters.

* gcharinit(bnf,mod) initializes the structure gc. The underlying number field and modulus can be accessed using gc.bnf and gc.mod.

* gc.cyc describes the finite abelian group structure of gc, the torsion part corresponding to finite order ray class characters, the exact zeros corresponding to a lattice of infinite order Grossencharacters, and the approximate zero being a placeholder for the complex powers of the id\`ele norm.

* A Hecke character of modulus 𝔪 is described as a t_COL of coordinates corresponding to gc.cyc: all the coordinates are integers except the last one, which can be an arbitrary complex number, or omitted instead of 0.

* Hecke Grossencharacters have L-functions and can be given to all lfun functions as a 2 components vector [gc,chi], see also Section se:lfungchar.

bnf being as output by bnfinit, checks whether the result is correct, i.e. whether it is possible to remove the assumption of the Generalized Riemann Hypothesis. It is correct if and only if the answer is 1. If it is incorrect, the program may output some error message, or loop indefinitely. You can check its progress by increasing the debug level. The bnf structure must contain the fundamental units:

  ? K = bnfinit(x^3+2^2^3+1); bnfcertify(K)
    ***   at top-level: K=bnfinit(x^3+2^2^3+1);bnfcertify(K)
    ***                                        ^ —  —  —  — -
    *** bnfcertify: precision too low in makeunits [use bnfinit(,1)].
  ? K = bnfinit(x^3+2^2^3+1, 1); \\ include units
  ? bnfcertify(K)
  %3 = 1

If flag is present, only certify that the class group is a quotient of the one computed in bnf (much simpler in general); likewise, the computed units may form a subgroup of the full unit group. In this variant, the units are no longer needed:

  ? K = bnfinit(x^3+2^2^3+1); bnfcertify(K, 1)
  %4 = 1

The library syntax is long bnfcertify0(GEN bnf, long flag). Also available is GEN bnfcertify(GEN bnf) (flag = 0).

If m is a module as output in the first component of an extension given by bnrdisclist, outputs the true module.

  ? K = bnfinit(x^2+23); L = bnrdisclist(K, 10); s = L[2]
  %1 = [[[Vecsmall([8]), Vecsmall([1])], [[0, 0, 0]]],
        [[Vecsmall([9]), Vecsmall([1])], [[0, 0, 0]]]]
  ? bnfdecodemodule(K, s[1][1])
  %2 =
  [2 0]
  
  [0 1]
  ? bnfdecodemodule(K,s[2][1])
  %3 =
  [2 1]
  
  [0 1]

The library syntax is GEN decodemodule(GEN nf, GEN m).

Initializes a bnf structure. Used in programs such as bnfisprincipal, bnfisunit or bnfnarrow. By default, the results are conditional on the GRH, see se:GRHbnf. The result is a 10-component vector bnf.

This implements Buchmann's sub-exponential algorithm for computing the class group, the regulator and a system of fundamental units of the general algebraic number field K defined by the irreducible polynomial P with integer coefficients. The meaning of flag is as follows:

* flag = 0 (default). This is the historical behavior, kept for compatibility reasons and speed. It has severe drawbacks but is likely to be a little faster than the alternative, twice faster say, so only use it if speed is paramount, you obtain a useful speed gain for the fields under consideration, and you are only interested in the field invariants such as the classgroup structure or its regulator. The computations involve exact algebraic numbers which are replaced by floating point embeddings for the sake of speed. If the precision is insufficient, gp may not be able to compute fundamental units, nor to solve some discrete logarithm problems. It may be possible to increase the precision of the bnf structure using nfnewprec but this may fail, in particular when fundamental units are large. In short, the resulting bnf structure is correct and contains useful information but later function calls to bnfisprincpal or bnrclassfield may fail.

When flag = 1, we keep an exact algebraic version of all floating point data and this allows to guarantee that functions using the structure will always succeed, as well as to compute the fundamental units exactly. The units are computed in compact form, as a product of small S-units, possibly with huge exponents. This flag also allows bnfisprincipal to compute generators of principal ideals in factored form as well. Be warned that expanding such products explicitly can take a very long time, but they can easily be mapped to floating point or ℓ-adic embeddings of bounded accuracy, or to K^*/(K^*)^ℓ, and this is enough for applications. In short, this flag should be used by default, unless you have a very good reason for it, for instance building massive tables of class numbers, and you do not care about units or the effect large units would have on your computation.

tech is a technical vector (empty by default, see se:GRHbnf). Careful use of this parameter may speed up your computations, but it is mostly obsolete and you should leave it alone.

The components of a bnf are technical. In fact: never access a component directly, always use a proper member function. However, for the sake of completeness and internal documentation, their description is as follows. We use the notations explained in the book by H. Cohen, A Course in Computational Algebraic Number Theory, Graduate Texts in Maths 138, Springer-Verlag, 1993, Section 6.5, and subsection 6.5.5 in particular.

bnf[1] contains the matrix W, i.e. the matrix in Hermite normal form giving relations for the class group on prime ideal generators (𝔭_i)_{1 ≤ i ≤ r}.

bnf[2] contains the matrix B, i.e. the matrix containing the expressions of the prime ideal factorbase in terms of the 𝔭_i. It is an r x c matrix.

bnf[3] contains the complex logarithmic embeddings of the system of fundamental units which has been found. It is an (r₁+r₂) x (r₁+r₂-1) matrix.

bnf[4] contains the matrix M"_C of Archimedean components of the relations of the matrix (W|B).

bnf[5] contains the prime factor base, i.e. the list of prime ideals used in finding the relations.

bnf[6] contains a dummy 0.

bnf[7] or bnf.nf is equal to the number field data nf as would be given by nfinit.

bnf[8] is a vector containing the classgroup bnf.clgp as a finite abelian group, the regulator bnf.reg, the number of roots of unity and a generator bnf.tu, the fundamental units in expanded form bnf.fu. If the fundamental units were omitted in the bnf, bnf.fu returns the sentinel value 0. If flag = 1, this vector contain also algebraic data corresponding to the fundamental units and to the discrete logarithm problem (see bnfisprincipal). In particular, if flag = 1 we may only know the units in factored form: the first call to bnf.fu expands them, which may be very costly, then caches the result.

bnf[9] is a vector used in bnfisprincipal only and obtained as follows. Let D = U W V obtained by applying the Smith normal form algorithm to the matrix W ( = bnf[1]) and let U_r be the reduction of U modulo D. The first elements of the factorbase are given (in terms of bnf.gen) by the columns of U_r, with Archimedean component g_a; let also GD_a be the Archimedean components of the generators of the (principal) ideals defined by the bnf.gen[i]^bnf.cyc[i]. Then bnf[9] = [U_r, g_a, GD_a], followed by technical exact components which allow to recompute g_a and GD_a to higher accuracy.

bnf[10] is by default unused and set equal to 0. This field is used to store further information about the field as it becomes available, which is rarely needed, hence would be too expensive to compute during the initial bnfinit call. For instance, the generators of the principal ideals bnf.gen[i]^bnf.cyc[i] (during a call to bnrisprincipal), or those corresponding to the relations in W and B (when the bnf internal precision needs to be increased).

The library syntax is GEN bnfinit0(GEN P, long flag, GEN tech = NULL, long prec).

Also available is GEN Buchall(GEN P, long flag, long prec), corresponding to tech = NULL, where flag is either 0 (default) or nf_FORCE (include all data in algebraic form). The function GEN Buchall_param(GEN P, double c1, double c2, long nrpid, long flag, long prec) gives direct access to the technical parameters.

Computes a complete system of solutions (modulo units of positive norm) of the absolute norm equation Norm(a) = x, where a is an integer in bnf. If bnf has not been certified, the correctness of the result depends on the validity of GRH. If (optional) flag is set, allow returning solutions in factored form, which helps a lot when the fundamental units are large (equivalently, when bnf.reg is large); having an exact algebraic bnf from bnfinit(,1) is necessary in this case, else setting the flag will mostly be a no-op.

  ? bnf = bnfinit(x^4-2, 1);
  ? bnfisintnorm(bnf,7)
  %2 = [-x^2 + x - 1, x^2 + x + 1]
  ? bnfisintnorm(bnf,-7)
  %3 = [-x^3 - 1, x^3 + 2*x^2 + 2*x + 1]
  
  ? bnf = bnfinit(x^2-2305843005992468481, 1);
  ? bnfisintnorm(bnf, 2305843008139952128)
    \\ stack overflow with 100GB parisize
  ? bnf.reg \\ fundamental unit is huge
  %6 = 14054016.227457155120413774802385952043
  
  ? v = bnfisintnorm(bnf, 2305843008139952128, 1); #v
  %7 = 31   \\ succeeds instantly
  ? s = v[1]; [type(s), matsize(s)]
  %8 = ["t_MAT", [165, 2]]   \\ solution 1 is a product of 165 factors
  ? exponent(s[,2])
  %9 = 105

The exponents have 105 bits, so there is indeed little hope of writing down the solutions in expanded form.

See also bnfisnorm.

The library syntax is GEN bnfisintnorm0(GEN bnf, GEN x, long flag). The function GEN bnfisintnormabs0(GEN bnf, GEN a, long flag), where bnf is a true bnf structure, returns a complete system of solutions modulo units of the absolute norm equation |Norm(x) |= |a|. As fast as bnfisintnorm, but solves the two equations Norm(x) = ± a simultaneously. The functions GEN bnfisintnormabs(GEN bnf, GEN a), GEN bnfisintnorm(GEN bnf, GEN a) correspond to flag = 0.

Tries to tell whether the rational number x is the norm of some element y in bnf. Returns a vector [a,b] where x = Norm(a)*b. Looks for a solution which is an S-unit, with S a certain set of prime ideals containing (among others) all primes dividing x. If bnf is known to be Galois, you may set flag = 0 (in this case, x is a norm iff b = 1). If flag is nonzero the program adds to S the following prime ideals, depending on the sign of flag. If flag > 0, the ideals of norm less than flag. And if flag < 0 the ideals dividing flag.

Assuming GRH, the answer is guaranteed (i.e. x is a norm iff b = 1), if S contains all primes less than 4log(disc(Bnf))², where Bnf is the Galois closure of bnf.

See also bnfisintnorm.

The library syntax is GEN bnfisnorm(GEN bnf, GEN x, long flag).

bnf being the number field data output by bnfinit, and x being an ideal, this function tests whether the ideal is principal or not. The result is more complete than a simple true/false answer and solves a general discrete logarithm problem. Assume the class group is ⨁ (ℤ/d_iℤ)g_i (where the generators g_i and their orders d_i are respectively given by bnf.gen and bnf.cyc). The routine returns a row vector [e,t], where e is a vector of exponents 0 ≤ e_i < d_i, and t is a number field element such that x = (t) ∏_i g_i^e_i. For given g_i (i.e. for a given bnf), the e_i are unique, and t is unique modulo units.

In particular, x is principal if and only if e is the zero vector. Note that the empty vector, which is returned when the class number is 1, is considered to be a zero vector (of dimension 0).

  ? K = bnfinit(y^2+23);
  ? K.cyc
  %2 = [3]
  ? K.gen
  %3 = [[2, 0; 0, 1]]          \\ a prime ideal above 2
  ? P = idealprimedec(K,3)[1]; \\ a prime ideal above 3
  ? v = bnfisprincipal(K, P)
  %5 = [[2]~, [3/4, 1/4]~]
  ? idealmul(K, v[2], idealfactorback(K, K.gen, v[1]))
  %6 =
  [3 0]
  
  [0 1]
  ? % == idealhnf(K, P)
  %7 = 1

The binary digits of flag mean:

* 1: If set, outputs [e,t] as explained above, otherwise returns only e, which is easier to compute. The following idiom only tests whether an ideal is principal:

    is_principal(bnf, x) = !bnfisprincipal(bnf,x,0);

* 2: It may not be possible to recover t, given the initial accuracy to which the bnf structure was computed. In that case, a warning is printed and t is set equal to the empty vector []~. If this bit is set, increase the precision and recompute needed quantities until t can be computed. Warning: setting this may induce lengthy computations, and the result may be too large to be physically representable in any case. You should consider using flag = 4 instead.

* 4: Return t in factored form (compact representation), as a small product of S-units for a small set of finite places S, possibly with huge exponents. This kind of result can be cheaply mapped to K^*/(K^*)^ℓ or to ℂ or ℚ_p to bounded accuracy and this is usually enough for applications. Explicitly expanding such a compact representation is possible using nffactorback but may be very costly. The algorithm is guaranteed to succeed if the bnf was computed using bnfinit(,1). If not, the algorithm may fail to compute a huge generator in this case (and replace it by []~). This is orders of magnitude faster than flag = 2 when the generators are indeed large.

The library syntax is GEN bnfisprincipal0(GEN bnf, GEN x, long flag). Instead of the above hardcoded numerical flags, one should rather use an or-ed combination of the symbolic flags nf_GEN (include generators, possibly a place holder if too difficult), nf_GENMAT (include generators in compact form) and nf_FORCE (insist on finding the generators, a no-op if nf_GENMAT is included).

This function is obsolete, use bnfisunit.

The library syntax is GEN bnfissunit(GEN bnf, GEN sfu, GEN x).

bnf being the number field data output by bnfinit and x being an algebraic number (type integer, rational or polmod), this outputs the decomposition of x on the fundamental units and the roots of unity if x is a unit, the empty vector otherwise. More precisely, if u₁,...,u_r are the fundamental units, and ζ is the generator of the group of roots of unity (bnf.tu), the output is a vector [x₁,...,x_r,x_r+1] such that x = u₁^x₁... u_r^x_r.ζ^x_r+1. The x_i are integers but the last one (i = r+1) is only defined modulo the order w of ζ and is guaranteed to be in [0,w[.

Note that bnf need not contain the fundamental units explicitly: it may contain the placeholder 0 instead:

  ? setrand(1); bnf = bnfinit(x^2-x-100000);
  ? bnf.fu
  %2 = 0
  ? u = [119836165644250789990462835950022871665178127611316131167, \
         379554884019013781006303254896369154068336082609238336]~;
  ? bnfisunit(bnf, u)
  %3 = [-1, 0]~

The given u is 1/u₁, where u₁ is the fundamental unit implicitly stored in bnf. In this case, u₁ was not computed and stored in algebraic form since the default accuracy was too low. Re-run the bnfinit command at \g1 or higher to see such diagnostics.

This function allows x to be given in factored form, but it then assumes that x is an actual unit. (Because it is general too costly to check whether this is the case.)

  ? { v = [2, 85; 5, -71; 13, -162; 17, -76; 23, -37; 29, -104; [224, 1]~, -66;
  [-86, 1]~, 86; [-241, 1]~, -20; [44, 1]~, 30; [124, 1]~, 11; [125, -1]~, -11;
  [-214, 1]~, 33; [-213, -1]~, -33; [189, 1]~, 74; [190, -1]~, 104;
  [-168, 1]~, 2; [-167, -1]~, -8]; }
  ? bnfisunit(bnf,v)
  %5 = [1, 0]~

Note that v is the fundamental unit of bnf given in compact (factored) form.

If the argument U is present, as output by bnfunits(bnf, S), then the function decomposes x on the S-units generators given in U[1].

   ? bnf = bnfinit(x^4 - x^3 + 4*x^2 + 3*x + 9, 1);
   ? bnf.sign
   %2 = [0, 2]
   ? S = idealprimedec(bnf,5); #S
   %3 = 2
   ? US = bnfunits(bnf,S);
   ? g = US[1]; #g  \\ #S = #g, four S-units generators, in factored form
   %5 = 4
   ? g[1]
   %6 = [[6, -3, -2, -2]~ 1]
   ? g[2]
   %7 =
   [[-1, 1/2, -1/2, -1/2]~ 1]
  
   [      [4, -2, -1, -1]~ 1]
   ? [nffactorback(bnf, x) | x <- g]
   %8 = [[6, -3, -2, -2]~, [-5, 5, 0, 0]~, [-1, 1, -1, 0]~,
         [1, -1, 0, 0]~]
  
   ? u = [10,-40,24,11]~;
   ? a = bnfisunit(bnf, u, US)
   %9 = [2, 0, 1, 4]~
   ? nffactorback(bnf, g, a) \\ prodi g[i]^a[i] still in factored form
   %10 =
   [[6, -3, -2, -2]~  2]
  
   [ [0, 0, -1, -1]~  1]
  
   [ [2, -1, -1, 0]~ -2]
  
   [   [1, 1, 0, 0]~  2]
  
   [  [-1, 1, 1, 1]~ -1]
  
   [  [1, -1, 0, 0]~  4]
  
   ? nffactorback(bnf,%)  \\ u = prodi g[i]^a[i]
   %11 = [10, -40, 24, 11]~

The library syntax is GEN bnfisunit0(GEN bnf, GEN x, GEN U = NULL). Also available is GEN bnfisunit(GEN bnf, GEN x) for U = NULL.

Let bnf be a bnf structure attached to the number field F and let l be a prime number (hereafter denoted ℓ for typographical reasons). Return the logarithmic ℓ-class group ~{Cl}_F of F. This is an abelian group, conjecturally finite (known to be finite if F/ℚ is abelian). The function returns if and only if the group is indeed finite (otherwise it would run into an infinite loop). Let S = { 𝔭₁,..., 𝔭_k} be the set of ℓ-adic places (maximal ideals containing ℓ). The function returns [D, G(ℓ), G'], where

* D is the vector of elementary divisors for ~{Cl}_F.

* G(ℓ) is the vector of elementary divisors for the (conjecturally finite) abelian group ~{Cl}(ℓ) = { 𝔞 = ∑_{i ≤ k} a_i 𝔭_i : deg_F 𝔞 = 0}, where the 𝔭_i are the ℓ-adic places of F; this is a subgroup of ~{Cl}.

* G' is the vector of elementary divisors for the ℓ-Sylow Cl' of the S-class group of F; the group ~{Cl} maps to Cl' with a simple co-kernel.

The library syntax is GEN bnflog(GEN bnf, GEN l).

Let nf be a nf structure attached to a number field F, and let l be a prime number (hereafter denoted ℓ). The ℓ-adified group of id\`{e}les of F quotiented by the group of logarithmic units is identified to the ℓ-group of logarithmic divisors ⨁ ℤ_ℓ [𝔭], generated by the maximal ideals of F.

The degree map deg_F is additive with values in ℤ_ℓ, defined by deg_F 𝔭 = ~{f}_𝔭 deg_ℓ p, where the integer ~{f}_𝔭 is as in bnflogef and deg_ℓ p is log_ℓ p for p ! = ℓ, log_ℓ (1 + ℓ) for p = ℓ ! = 2 and log_ℓ (1 + 2²) for p = ℓ = 2.

Let A = ∏ 𝔭^n_𝔭 be an ideal and let ~{A} = ∑ n_𝔭 [𝔭] be the attached logarithmic divisor. Return the exponential of the ℓ-adic logarithmic degree deg_F A, which is a natural number.

The library syntax is GEN bnflogdegree(GEN nf, GEN A, GEN l).

Let nf be a nf structure attached to a number field F and let pr be a prid structure attached to a maximal ideal 𝔭 / p. Return [~{e}(F_𝔭 / ℚ_p), ~{f}(F_𝔭 / ℚ_p)] the logarithmic ramification and residue degrees. Let ℚ_p^c/ℚ_p be the cyclotomic ℤ_p-extension, then ~{e} = [F_𝔭 : F_𝔭 ∩ ℚ_p^c] and ~{f} = [F_𝔭 ∩ ℚ_p^c : ℚ_p]. Note that ~{e}~{f} = e(𝔭/p) f(𝔭/p), where e(𝔭/p) and f(𝔭/p) denote the usual ramification and residue degrees.

  ? F = nfinit(y^6 - 3*y^5 + 5*y^3 - 3*y + 1);
  ? bnflogef(F, idealprimedec(F,2)[1])
  %2 = [6, 1]
  ? bnflogef(F, idealprimedec(F,5)[1])
  %3 = [1, 2]

The library syntax is GEN bnflogef(GEN nf, GEN pr).

bnf being as output by bnfinit, computes the narrow class group of bnf. The output is a 3-component row vector v analogous to the corresponding class group component bnf.clgp: the first component is the narrow class number v.no, the second component is a vector containing the SNF cyclic components v.cyc of the narrow class group, and the third is a vector giving the generators of the corresponding v.gen cyclic groups. Note that this function is a special case of bnrinit; the bnf need not contain fundamental units.

The library syntax is GEN bnfnarrow(GEN bnf).

bnf being as output by bnfinit, this computes an r₁ x (r₁+r₂-1) matrix having ±1 components, giving the signs of the real embeddings of the fundamental units. The following functions compute generators for the totally positive units:

  /* exponents of totally positive units generators on K.tu, K.fu */
  tpuexpo(K)=
  { my(M, S = bnfsignunit(K), [m,n] = matsize(S));
    \\ m = K.r1, n = r1+r2-1
    S = matrix(m,n, i,j, if (S[i,j] < 0, 1,0));
    S = concat(vectorv(m,i,1), S);   \\ add sign(-1)
    M = matkermod(S, 2);
    if (M, mathnfmodid(M, 2), 2*matid(n+1))
  }
  
  /* totally positive fundamental units of bnf K */
  tpu(K)=
  { my(ex = tpuexpo(K)[,^1]); \\ remove ex[,1], corresponds to 1 or -1
    my(v = concat(K.tu[2], K.fu));
    [ nffactorback(K, v, c) | c <- ex];
  }

The library syntax is GEN signunits(GEN bnf).

Computes the fundamental S-units of the number field bnf (output by bnfinit), where S is a list of prime ideals (output by idealprimedec). The output is a vector v with 6 components.

v[1] gives a minimal system of (integral) generators of the S-unit group modulo the unit group.

v[2] contains technical data needed by bnfissunit.

v[3] is an obsoleted component, now the empty vector.

v[4] is the S-regulator (this is the product of the regulator, the S-class number and the natural logarithms of the norms of the ideals in S).

v[5] gives the S-class group structure, in the usual abelian group format: a vector whose three components give in order the S-class number, the cyclic components and the generators.

v[6] is a copy of S.

The library syntax is GEN bnfsunit(GEN bnf, GEN S, long prec). Also available is GEN sunits_mod_units(GEN bnf, GEN S) which returns only v[1].

Return the fundamental units of the number field bnf output by bnfinit; if S is present and is a list of prime ideals, compute fundamental S-units instead. The first component of the result contains independent integral S-units generators: first nonunits, then r₁+r₂-1 fundamental units, then the torsion unit. The result may be used as an optional argument to bnfisunit. The units are given in compact form: no expensive computation is attempted if the bnf does not already contain units.

   ? bnf = bnfinit(x^4 - x^3 + 4*x^2 + 3*x + 9, 1);
   ? bnf.sign   \\ r1 + r2 - 1 = 1
   %2 = [0, 2]
   ? U = bnfunits(bnf); u = U[1];
   ? #u \\ r1 + r2 = 2 units
   %5 = 2;
   ? u[1] \\ fundamental unit as factorization matrix
   %6 =
   [[0, 0, -1, -1]~  1]
  
   [[2, -1, -1, 0]~ -2]
  
   [  [1, 1, 0, 0]~  2]
  
   [ [-1, 1, 1, 1]~ -1]
   ? u[2] \\ torsion unit as factorization matrix
   %7 =
   [[1, -1, 0, 0]~ 1]
   ? [nffactorback(bnf, z) | z <- u]  \\ same units in expanded form
   %8 = [[-1, 1, -1, 0]~, [1, -1, 0, 0]~]

Now an example involving S-units for a nontrivial S:

   ? S = idealprimedec(bnf,5); #S
   %9 = 2
   ? US = bnfunits(bnf, S); uS = US[1];
   ? g = [nffactorback(bnf, z) | z <- uS] \\ now 4 units
   %11 = [[6, -3, -2, -2]~, [-5, 5, 0, 0]~, [-1, 1, -1, 0]~, [1, -1, 0, 0]~]
   ? bnfisunit(bnf,[10,-40,24,11]~)
   %12 = []~  \\ not a unit
   ? e = bnfisunit(bnf, [10,-40,24,11]~, US)
   %13 = [2, 0, 1, 4]~  \\ ...but an S-unit
   ? nffactorback(bnf, g, e)
   %14 = [10, -40, 24, 11]~
   ? nffactorback(bnf, uS, e) \\ in factored form
   %15 =
   [[6, -3, -2, -2]~  2]
  
   [ [0, 0, -1, -1]~  1]
  
   [ [2, -1, -1, 0]~ -2]
  
   [   [1, 1, 0, 0]~  2]
  
   [  [-1, 1, 1, 1]~ -1]
  
   [  [1, -1, 0, 0]~  4]

Note that in more complicated cases, any nffactorback fully expanding an element in factored form could be very expensive. On the other hand, the final example expands a factorization whose components are themselves in factored form, hence the result is a factored form: this is a cheap operation.

The library syntax is GEN bnfunits(GEN bnf, GEN S = NULL).

Let bnr be the number field data output by bnrinit and H be a square matrix defining a congruence subgroup of the ray class group corresponding to bnr (the trivial congruence subgroup if omitted). This function returns, for each character χ of the ray class group which is trivial on H, the value at s = 1 (or s = 0) of the abelian L-function attached to χ. For the value at s = 0, the function returns in fact for each χ a vector [r_χ, c_χ] where L(s, χ) = c.s^r + O(s^{r + 1}) near 0.

The argument flag is optional, its binary digits mean 1: compute at s = 0 if unset or s = 1 if set, 2: compute the primitive L-function attached to χ if unset or the L-function with Euler factors at prime ideals dividing the modulus of bnr removed if set (that is L_S(s, χ), where S is the set of infinite places of the number field together with the finite prime ideals dividing the modulus of bnr), 3: return also the character if set.

  K = bnfinit(x^2-229);
  bnr = bnrinit(K,1);
  bnrL1(bnr)

returns the order and the first nonzero term of L(s, χ) at s = 0 where χ runs through the characters of the class group of K = ℚ(sqrt{229}). Then

  bnr2 = bnrinit(K,2);
  bnrL1(bnr2,,2)

returns the order and the first nonzero terms of L_S(s, χ) at s = 0 where χ runs through the characters of the class group of K and S is the set of infinite places of K together with the finite prime 2. Note that the ray class group modulo 2 is in fact the class group, so bnrL1(bnr2,0) returns the same answer as bnrL1(bnr,0).

This function will fail with the message

   *** bnrL1: overflow in zeta_get_N0 [need too many primes].

if the approximate functional equation requires us to sum too many terms (if the discriminant of K is too large).

The library syntax is GEN bnrL1(GEN bnr, GEN H = NULL, long flag, long prec).

Returns all characters χ on G such that χ(g_i) = e(v_i), where e(x) = exp(2iπ x). G is allowed to be a bnr struct (representing a ray class group) or a znstar (representing (ℤ/Nℤ)^*). If v is omitted, returns all characters that are trivial on the g_i. Else the vectors g and v must have the same length, the g_i must be elements of G, and each v_i is a rational number whose denominator must divide the order of g_i in G.

For convenience, the vector of the g_i can be replaced by a matrix whose columns give their discrete logarithm in G, for instance as given by bnrisprincipal if G is a bnr; in this particular case, G can be any finite abelian group given by a vector of elementary divisors.

  ? G = bnrinit(bnfinit(x), [160,[1]], 1); /* (Z/160Z)* */
  ? G.cyc
  %2 = [8, 4, 2]
  ? g = G.gen;
  ? bnrchar(G, g, [1/2,0,0])
  %4 = [[4, 0, 0]]  \\ a unique character
  ? bnrchar(G, [g[1],g[3]]) \\ all characters trivial on g[1] and g[3]
  %5 = [[0, 1, 0], [0, 2, 0], [0, 3, 0], [0, 0, 0]]
  ? bnrchar(G, [1,0,0;0,1,0;0,0,2])
  %6 = [[0, 0, 1], [0, 0, 0]]  \\ characters trivial on given subgroup
  
  ? G = znstar(75, 1);
  ? bnrchar(G, [2, 7], [11/20, 1/4])
  %8 = [[1, 1]] \\ Dirichlet char: chi(2) = e(11/20), chi(7) = e(1/4)

The library syntax is GEN bnrchar(GEN G, GEN g, GEN v = NULL).

bnr being as output by bnrinit, returns a relative equation for the class field corresponding to the congruence group defined by (bnr,subgp) (the full ray class field if subgp is omitted). The subgroup can also be a t_INT n, meaning n.Cl_f. The function also handles a vector of subgroup, e.g, from subgrouplist and returns the vector of individual results in this case.

If flag = 0, returns a vector of polynomials such that the compositum of the corresponding fields is the class field; if flag = 1 returns a single polynomial; if flag = 2 returns a single absolute polynomial.

  ? bnf = bnfinit(y^3+14*y-1); bnf.cyc
  %1 = [4, 2]
  ? pol = bnrclassfield(bnf,,1) \\ Hilbert class field
  %2 = x^8 - 2*x^7 + ... + Mod(11*y^2 - 82*y + 116, y^3 + 14*y - 1)
  ? rnfdisc(bnf,pol)[1]
  %3 = 1
  ? bnr = bnrinit(bnf,3*5*7); bnr.cyc
  %4 = [24, 12, 12, 2]
  ? bnrclassfield(bnr,2) \\ maximal 2-elementary subextension
  %5 = [x^2 + (-21*y - 105), x^2 + (-5*y - 25), x^2 + (-y - 5), x^2 + (-y - 1)]
  \\ quadratic extensions of maximal conductor
  ? bnrclassfield(bnr, subgrouplist(bnr,[2]))
  %6 = [[x^2 - 105], [x^2 + (-105*y^2 - 1260)], [x^2 + (-105*y - 525)],
        [x^2 + (-105*y - 105)]]
  ? #bnrclassfield(bnr,subgrouplist(bnr,[2],1)) \\ all quadratic extensions
  %7 = 15

When the subgroup contains n Cl_f, where n is fixed, it is advised to directly compute the bnr modulo n to avoid expensive discrete logarithms:

  ? bnf = bnfinit(y^2-5); p = 1594287814679644276013;
  ? bnr = bnrinit(bnf,p); \\ very slow
  time = 24,146 ms.
  ? bnrclassfield(bnr, 2) \\ ... even though the result is trivial
  %3 = [x^2 - 1594287814679644276013]
  ? bnr2 = bnrinit(bnf,p,,2); \\ now fast
  time = 1 ms.
  ? bnrclassfield(bnr2, 2)
  %5 = [x^2 - 1594287814679644276013]

This will save a lot of time when the modulus contains a maximal ideal whose residue field is large.

The library syntax is GEN bnrclassfield(GEN bnr, GEN subgp = NULL, long flag, long prec).

Let A, B, C define a class field L over a ground field K (of type [bnr], [bnr, subgroup], or [bnf, modulus], or [bnf, modulus,subgroup], Section se:CFT); this function returns the relative degree [L:K].

In particular if A is a bnf (with units), and B a modulus, this function returns the corresponding ray class number modulo B. One can input the attached bid (with generators if the subgroup C is non trivial) for B instead of the module itself, saving some time.

This function is faster than bnrinit and should be used if only the ray class number is desired. See bnrclassnolist if you need ray class numbers for all moduli less than some bound.

The library syntax is GEN bnrclassno0(GEN A, GEN B = NULL, GEN C = NULL). Also available is GEN bnrclassno(GEN bnf,GEN f) to compute the ray class number modulo f.

bnf being as output by bnfinit, and list being a list of moduli (with units) as output by ideallist or ideallistarch, outputs the list of the class numbers of the corresponding ray class groups. To compute a single class number, bnrclassno is more efficient.

  ? bnf = bnfinit(x^2 - 2);
  ? L = ideallist(bnf, 100, 2);
  ? H = bnrclassnolist(bnf, L);
  ? H[98]
  %4 = [1, 3, 1]
  ? l = L[1][98]; ids = vector(#l, i, l[i].mod[1])
  %5 = [[98, 88; 0, 1], [14, 0; 0, 7], [98, 10; 0, 1]]

The weird l[i].mod[1], is the first component of l[i].mod, i.e. the finite part of the conductor. (This is cosmetic: since by construction the Archimedean part is trivial, I do not want to see it). This tells us that the ray class groups modulo the ideals of norm 98 (printed as %5) have respectively order 1, 3 and 1. Indeed, we may check directly:

  ? bnrclassno(bnf, ids[2])
  %6 = 3

The library syntax is GEN bnrclassnolist(GEN bnf, GEN list).

Given two abelian extensions A = [bnr1, H1] and B = [bnr2, H2], where bnr1 and bnr2 are two bnr structures attached to the same base field, return their compositum as [bnr, H]. The modulus attached to bnr need not be the conductor of the compositum.

  ? Q = bnfinit(y);
  ? bnr1 = bnrinit(Q, [7, [1]]); bnr1.cyc
  %2 = [6]
  ? bnr2 = bnrinit(Q, [13, [1]]); bnr2.cyc
  %3 = [12]
  ? H1 = Mat(2); bnrclassfield(bnr1, H1)
  %4 = [x^2 + 7]
  ? H2 = Mat(2); bnrclassfield(bnr2, H2)
  %5 = [x^2 - 13]
  ? [bnr,H] = bnrcompositum([bnr1, H1], [bnr2,H2]);
  ? bnrclassfield(bnr,H)
  %7 = [x^2 - 13, x^2 + 7]

The library syntax is GEN bnrcompositum(GEN A, GEN B).

Conductor f of the subfield of a ray class field as defined by [A,B,C] (of type [bnr], [bnr, subgroup], [bnf, modulus] or [bnf, modulus, subgroup], Section se:CFT)

If flag = 0, returns f.

If flag = 1, returns [f, Cl_f, H], where Cl_f is the ray class group modulo f, as a finite abelian group; finally H is the subgroup of Cl_f defining the extension.

If flag = 2, returns [f, bnr(f), H], as above except Cl_f is replaced by a bnr structure, as output by bnrinit(,f), without generators unless the input contained a bnr with generators.

In place of a subgroup H, this function also accepts a character chi = (a_j), expressed as usual in terms of the generators bnr.gen: χ(g_j) = exp(2iπ a_j / d_j), where g_j has order d_j = bnr.cyc[j]. In which case, the function returns respectively

If flag = 0, the conductor f of Ker χ.

If flag = 1, [f, Cl_f, χ_f], where χ_f is χ expressed on the minimal ray class group, whose modulus is the conductor.

If flag = 2, [f, bnr(f), χ_f].

Note. Using this function with flag ! = 0 is usually a bad idea and kept for compatibility and convenience only: flag = 1 has always been useless, since it is no faster than flag = 2 and returns less information; flag = 2 is mostly OK with two subtle drawbacks:

* it returns the full bnr attached to the full ray class group, whereas in applications we only need Cl_f modulo N-th powers, where N is any multiple of the exponent of Cl_f/H. Computing directly the conductor, then calling bnrinit with optional argument N avoids this problem.

* computing the bnr needs only be done once for each conductor, which is not possible using this function.

For maximal efficiency, the recommended procedure is as follows. Starting from data (character or congruence subgroups) attached to a modulus m, we can first compute the conductors using this function with default flag = 0. Then for all data with a common conductor f | m, compute (once!) the bnr attached to f using bnrinit (modulo N-th powers for a suitable N!) and finally map original data to the new bnr using bnrmap.

The library syntax is GEN bnrconductor0(GEN A, GEN B = NULL, GEN C = NULL, long flag).

Also available are GEN bnrconductor(GEN bnr, GEN H, long flag) and GEN bnrconductormod(GEN bnr, GEN H, long flag, GEN cycmod) which returns ray class groups modulo cycmod-th powers.

This function is obsolete, use bnrconductor.

The library syntax is GEN bnrconductorofchar(GEN bnr, GEN chi).

A, B, C defining a class field L over a ground field K (of type [bnr], [bnr, subgroup], [bnr, character], [bnf, modulus] or [bnf, modulus, subgroup], Section se:CFT), outputs data [N,r₁,D] giving the discriminant and signature of L, depending on the binary digits of flag:

* 1: if this bit is unset, output absolute data related to L/ℚ: N is the absolute degree [L:ℚ], r₁ the number of real places of L, and D the discriminant of L/ℚ. Otherwise, output relative data for L/K: N is the relative degree [L:K], r₁ is the number of real places of K unramified in L (so that the number of real places of L is equal to r₁ times N), and D is the relative discriminant ideal of L/K.

* 2: if this bit is set and if the modulus is not the conductor of L, only return 0.

The library syntax is GEN bnrdisc0(GEN A, GEN B = NULL, GEN C = NULL, long flag).

bnf being as output by bnfinit (with units), computes a list of discriminants of Abelian extensions of the number field by increasing modulus norm up to bound bound. The ramified Archimedean places are given by arch; all possible values are taken if arch is omitted.

The alternative syntax bnrdisclist(bnf,list) is supported, where list is as output by ideallist or ideallistarch (with units), in which case arch is disregarded.

The output v is a vector, where v[k] is itself a vector w, whose length is the number of ideals of norm k.

* We consider first the case where arch was specified. Each component of w corresponds to an ideal m of norm k, and gives invariants attached to the ray class field L of bnf of conductor [m, arch]. Namely, each contains a vector [m,d,r,D] with the following meaning: m is the prime ideal factorization of the modulus, d = [L:ℚ] is the absolute degree of L, r is the number of real places of L, and D is the factorization of its absolute discriminant. We set d = r = D = 0 if m is not the finite part of a conductor.

* If arch was omitted, all t = 2^r₁ possible values are taken and a component of w has the form [m, [[d₁,r₁,D₁],..., [d_t,r_t,D_t]]], where m is the finite part of the conductor as above, and [d_i,r_i,D_i] are the invariants of the ray class field of conductor [m,v_i], where v_i is the i-th Archimedean component, ordered by inverse lexicographic order; so v₁ = [0,...,0], v₂ = [1,0...,0], etc. Again, we set d_i = r_i = D_i = 0 if [m,v_i] is not a conductor.

Finally, each prime ideal pr = [p,α,e,f,β] in the prime factorization m is coded as the integer p.n²+(f-1).n+(j-1), where n is the degree of the base field and j is such that

pr = idealprimedec(nf,p)[j].

m can be decoded using bnfdecodemodule.

Note that to compute such data for a single field, either bnrclassno or bnrdisc are (much) more efficient.

The library syntax is GEN bnrdisclist0(GEN bnf, GEN bound, GEN arch = NULL).

Apply the automorphism given by its matrix mat to the congruence subgroup H given as a HNF matrix. The matrix mat can be computed with bnrgaloismatrix.

The library syntax is GEN bnrgaloisapply(GEN bnr, GEN mat, GEN H).

Return the matrix of the action of the automorphism aut of the base field bnf.nf on the generators of the ray class field bnr.gen. The automorphism aut can be given as a polynomial, an algebraic number, or a vector of automorphisms and must stabilize the modulus bnr.mod. We also allow a Galois group as output by galoisinit, in which case a vector of matrices is returned corresponding to the generators aut.gen. Note: This function only makes sense when the ray class field attached to bnr is Galois, which is not checked.

The generators bnr.gen need not be explicitly computed in the input bnr, which saves time: the result is well defined in this case also.

  ? K = bnfinit(a^4-3*a^2+253009); B = bnrinit(K,9); B.cyc
  %1 = [8400, 12, 6, 3]
  ? G = nfgaloisconj(K)
  %2 = [-a, a, -1/503*a^3 + 3/503*a, 1/503*a^3 - 3/503*a]~
  ? bnrgaloismatrix(B, G[2])  \\ G[2] = Id ...
  %3 =
  [1 0 0 0]
  
  [0 1 0 0]
  
  [0 0 1 0]
  
  [0 0 0 1]
  ? bnrgaloismatrix(B, G[3]) \\ automorphism of order 2
  %4 =
  [799 0 0 2800]
  
  [  0 7 0    4]
  
  [  4 0 5    2]
  
  [  0 0 0    2]
  ? M = %^2; for (i=1, #B.cyc, M[i,] %= B.cyc[i]); M
  %5 =  \\ acts on ray class group as automorphism of order 2
  [1 0 0 0]
  
  [0 1 0 0]
  
  [0 0 1 0]
  
  [0 0 0 1]

See bnrisgalois for further examples.

The library syntax is GEN bnrgaloismatrix(GEN bnr, GEN aut). When aut is a polynomial or an algebraic number, GEN bnrautmatrix(GEN bnr, GEN aut) is available.

bnf is as output by bnfinit (including fundamental units), f is a modulus, initializes data linked to the ray class group structure corresponding to this module, a so-called bnr structure. One can input the attached bid with generators for f instead of the module itself, saving some time. (As in idealstar, the finite part of the conductor may be given by a factorization into prime ideals, as produced by idealfactor.)

If the positive integer cycmod is present, only compute the ray class group modulo cycmod, which may save a lot of time when some maximal ideals in f have a huge residue field. In applications, we are given a congruence subgroup H and study the class field attached to Cl_f/H. If that finite Abelian group has an exponent which divides cycmod, then we have changed nothing theoretically, while trivializing expensive discrete logs in residue fields (since computations can be made modulo cycmod-th powers). This is useful in bnrclassfield, for instance when computing p-elementary extensions.

The following member functions are available on the result: .bnf is the underlying bnf, .mod the modulus, .bid the bid structure attached to the modulus; finally, .clgp, .no, .cyc, .gen refer to the ray class group (as a finite abelian group), its cardinality, its elementary divisors, its generators (only computed if flag = 1).

The last group of functions are different from the members of the underlying bnf, which refer to the class group; use bnr.bnf.xxx to access these, e.g. bnr.bnf.cyc to get the cyclic decomposition of the class group.

They are also different from the members of the underlying bid, which refer to (ℤ_K/f)^*; use bnr.bid.xxx to access these, e.g. bnr.bid.no to get φ(f).

If flag = 0 (default), the generators of the ray class group are not explicitly computed, which saves time. Hence bnr.gen would produce an error. Note that implicit generators are still fixed and stored in the bnr (and guaranteed to be the same for fixed bnf and bid inputs), in terms of bnr.bnf.gen and bnr.bid.gen. The computation which is not performed is the expansion of such products in the ray class group so as to fix eplicit ideal representatives.

If flag = 1, as the default, except that generators are computed.

The library syntax is GEN bnrinitmod(GEN bnf, GEN f, long flag, GEN cycmod = NULL). Instead of the above hardcoded numerical flags, one should rather use GEN Buchraymod(GEN bnf, GEN module, long flag, GEN cycmod) where an omitted cycmod is coded as NULL and flag is an or-ed combination of nf_GEN (include generators) and nf_INIT (if omitted, return just the cardinality of the ray class group and its structure), possibly 0. Or simply GEN Buchray(GEN bnf, GEN module, long flag) when cycmod is NULL.

Fast variant of bnrconductor(A,B,C); A, B, C represent an extension of the base field, given by class field theory (see Section se:CFT). Outputs 1 if this modulus is the conductor, and 0 otherwise. This is slightly faster than bnrconductor when the character or subgroup is not primitive.

The library syntax is long bnrisconductor0(GEN A, GEN B = NULL, GEN C = NULL).

Check whether the class field attached to the subgroup H is Galois over the subfield of bnr.nf fixed by the group gal, which can be given as output by galoisinit, or as a matrix or a vector of matrices as output by bnrgaloismatrix, the second option being preferable, since it saves the recomputation of the matrices. Note: The function assumes that the ray class field attached to bnr is Galois, which is not checked.

In the following example, we lists the congruence subgroups of subextension of degree at most 3 of the ray class field of conductor 9 which are Galois over the rationals.

  ? K = bnfinit(a^4-3*a^2+253009); B = bnrinit(K,9); G = galoisinit(K);
  ? [H | H<-subgrouplist(B,3), bnrisgalois(B,G,H)];
  time = 160 ms.
  ? M = bnrgaloismatrix(B,G);
  ? [H | H<-subgrouplist(B,3), bnrisgalois(B,M,H)]
  time = 1 ms.

The second computation is much faster since bnrgaloismatrix(B,G) is computed only once.

The library syntax is long bnrisgalois(GEN bnr, GEN gal, GEN H).

Let bnr be the ray class group data output by bnrinit(,,1) and let x be an ideal in any form, coprime to the modulus f = bnr.mod. Solves the discrete logarithm problem in the ray class group, with respect to the generators bnr.gen, in a way similar to bnfisprincipal. If x is not coprime to the modulus of bnr the result is undefined. Note that bnr need not contain the ray class group generators, i.e. it may be created with bnrinit(,,0); in that case, although bnr.gen is undefined, we can still fix natural generators for the ray class group (in terms of the generators in bnr.bnf.gen and bnr.bid.gen) and compute with respect to them.

The binary digits of flag (default flag = 1) mean:

* 1: If set returns a 2-component vector [e,α] where e is the vector of components of x on the ray class group generators, α is an element congruent to 1 mod^* f such that x = α ∏_i g_i^e_i. If unset, returns only e.

* 4: If set, returns [e,α] where α is given in factored form (compact representation). This is orders of magnitude faster.

  ? K = bnfinit(x^2 - 30); bnr = bnrinit(K, [4, [1,1]]);
  ? bnr.clgp \\ ray class group is isomorphic to Z/4 x Z/2 x Z/2
  %2 = [16, [4, 2, 2]]
  ? P = idealprimedec(K, 3)[1]; \\ the ramified prime ideal above 3
  ? bnrisprincipal(bnr,P) \\ bnr.gen undefined !
  %5 = [[3, 0, 0]~, 9]
  ? bnrisprincipal(bnr,P, 0) \\ omit principal part
  %5 = [3, 0, 0]~
  ? bnr = bnrinit(bnr, bnr.bid, 1); \\ include explicit generators
  ? bnrisprincipal(bnr,P) \\ ... alpha is different !
  %7 = [[3, 0, 0]~, 1/128625]

It may be surprising that the generator α is different although the underlying bnf and bid are the same. This defines unique generators for the ray class group as ideal classes, whether we use bnrinit(,0) or bnrinit(,1). But the actual ideal representatives (implicit if flag = 0, computed and stored in the bnr if flag = 1) are in general different and this is what happens here. Indeed, the implicit generators are naturally expressed in terms of bnr.bnf.gen and bnr.bid.gen and then expanded and simplified (in the same ideal class) so that we obtain ideal representatives for bnr.gen which are as simple as possible. And indeed the quotient of the two α found is 1 modulo the conductor (and positive at the infinite places it contains), and this is the only guaranteed property.

Beware that, when bnr is generated using bnrinit(, cycmod), the results are given in Cl_f modulo cycmod-th powers:

  ? bnr2 = bnrinit(K, bnr.mod,, 2);  \\ modulo squares
  ? bnr2.clgp
  %9 = [8, [2, 2, 2]]  \\ bnr.clgp tensored by Z/2Z
  ? bnrisprincipal(bnr2,P, 0)
  %10 = [1, 0, 0]~

The library syntax is GEN bnrisprincipal(GEN bnr, GEN x, long flag). Instead of hardcoded numerical flags, one should rather use GEN isprincipalray(GEN bnr, GEN x) for flag = 0, and if you want generators:

    bnrisprincipal(bnr, x, nf_GEN)

Also available is GEN bnrisprincipalmod(GEN bnr, GEN x, GEN mod, long flag) that returns the discrete logarithm of x modulo the t_INT mod; the value mod = NULL is treated as 0 (full discrete logarithm), and flag = 1 is not allowed if mod is set.

This function has two different uses:

* if A and B are bnr structures for the same bnf attached to moduli m_A and m_B with m_B | m_A, return the canonical surjection from A to B, i.e. from the ray class group moodulo m_A to the ray class group modulo m_B. The map is coded by a triple [M,cyc_A,cyc_B]: M gives the image of the fixed ray class group generators of A in terms of the ones in B, cyc_A and cyc_B are the cyclic structures A.cyc and B.cyc respectively. Note that this function does not need A or B to contain explicit generators for the ray class groups: they may be created using bnrinit(,0).

If B is only known modulo N-th powers (from bnrinit(,N)), the result is correct provided N is a multiple of the exponent of A.

* if A is a projection map as above and B is either a congruence subgroup H, or a ray class character χ, or a discrete logarithm (from bnrisprincipal) modulo m_A whose conductor divides m_B, return the image of the subgroup (resp. the character, the discrete logarighm) as defined modulo m_B. The main use of this variant is to compute the primitive subgroup or character attached to a bnr modulo their conductor. This is more efficient than bnrconductor in two respects: the bnr attached to the conductor need only be computed once and, most importantly, the ray class group can be computed modulo N-th powers, where N is a multiple of the exponent of Cl_{m_{A}} / H (resp. of the order of χ). Whereas bnrconductor is specified to return a bnr attached to the full ray class group, which may lead to untractable discrete logarithms in the full ray class group instead of a tiny quotient.

The library syntax is GEN bnrmap(GEN A, GEN B).

If χ = chi is a character over bnr, not necessarily primitive, let L(s,χ) = ∑_id χ(id) N(id)^-s be the attached Artin L-function. Returns the so-called Artin root number, i.e. the complex number W(χ) of modulus 1 such that

Λ(1-s,χ) = W(χ) Λ(s,χ)

where Λ(s,χ) = A(χ)^s/2γ_χ(s) L(s,χ) is the enlarged L-function attached to L.

You can set flag = 1 if the character is known to be primitive. Example:

  bnf = bnfinit(x^2 - x - 57);
  bnr = bnrinit(bnf, [7,[1,1]]);
  bnrrootnumber(bnr, [2,1])

returns the root number of the character χ of Cl_{7 oo _{1} oo ₂}(ℚ(sqrt{229})) defined by χ(g₁^ag₂^b) = ζ₁^2aζ₂^b. Here g₁, g₂ are the generators of the ray-class group given by bnr.gen and ζ₁ = e^2iπ/N₁, ζ₂ = e^2iπ/N₂ where N₁, N₂ are the orders of g₁ and g₂ respectively (N₁ = 6 and N₂ = 3 as bnr.cyc readily tells us).

The library syntax is GEN bnrrootnumber(GEN bnr, GEN chi, long flag, long prec).

bnr being as output by bnrinit, finds a relative equation for the class field corresponding to the modulus in bnr and the given congruence subgroup (as usual, omit subgroup if you want the whole ray class group).

The main variable of bnr must not be x, and the ground field and the class field must be totally real. When the base field is ℚ, the vastly simpler galoissubcyclo is used instead. Here is an example:

  bnf = bnfinit(y^2 - 3);
  bnr = bnrinit(bnf, 5);
  bnrstark(bnr)

returns the ray class field of ℚ(sqrt{3}) modulo 5. Usually, one wants to apply to the result one of

  rnfpolredbest(bnf, pol)    \\  compute a reduced relative polynomial
  rnfpolredbest(bnf, pol, 2) \\  compute a reduced absolute polynomial

The routine uses Stark units and needs to find a suitable auxiliary conductor, which may not exist when the class field is not cyclic over the base. In this case bnrstark is allowed to return a vector of polynomials defining independent relative extensions, whose compositum is the requested class field. We decided that it was useful to keep the extra information thus made available, hence the user has to take the compositum herself, see nfcompositum.

Even if it exists, the auxiliary conductor may be so large that later computations become unfeasible. (And of course, Stark's conjecture may simply be wrong.) In case of difficulties, try bnrclassfield:

  ? bnr = bnrinit(bnfinit(y^8-12*y^6+36*y^4-36*y^2+9,1), 2);
  ? bnrstark(bnr)
    ***   at top-level: bnrstark(bnr)
    ***                 ^ —  —  —  — -
    *** bnrstark: need 3919350809720744 coefficients in initzeta.
    *** Computation impossible.
  ? bnrclassfield(bnr)
  time = 20 ms.
  %2 = [x^2 + (-2/3*y^6 + 7*y^4 - 14*y^2 + 3)]

The library syntax is GEN bnrstark(GEN bnr, GEN subgroup = NULL, long prec).

bnr being as output by bnrinit, returns the characteristic polynomial of the (conjectural) Stark unit corresponding to the modulus in bnr and the given congruence subgroup (as usual, omit subgroup if you want the whole ray class group).

The ground field attached to bnr must be totally real and all but one infinite place must become complex in the class field, which must be a quadratic extension of its totally real subfield. Finally, the output is given as a polynomial in x, so the main variable of bnr must not be x. Here is an example:

  ? bnf = bnfinit(y^2 - 2);
  ? bnr = bnrinit(bnf, [15, [1,0]]);
  ? lift(bnrstarkunit(bnr))
  %3 = x^8 + (-9000*y - 12728)*x^7 + (57877380*y + 81850978)*x^6 + ... + 1

The library syntax is GEN bnrstarkunit(GEN bnr, GEN subgroup = NULL).

Gives as a vector the first b coefficients of the Dedekind zeta function of the number field nf considered as a Dirichlet series.

The library syntax is GEN dirzetak(GEN nf, GEN b).

This function is obsolete, use nffactor.

factorization of the univariate polynomial x over the number field defined by the (univariate) polynomial t. x may have coefficients in ℚ or in the number field. The algorithm reduces to factorization over ℚ (Trager's trick). The direct approach of nffactor, which uses van Hoeij's method in a relative setting, is in general faster.

The main variable of t must be of lower priority than that of x (see Section se:priority). However if nonrational number field elements occur (as polmods or polynomials) as coefficients of x, the variable of these polmods must be the same as the main variable of t. For example

  ? factornf(x^2 + Mod(y, y^2+1), y^2+1);
  ? factornf(x^2 + y, y^2+1); \\  these two are OK
  ? factornf(x^2 + Mod(z,z^2+1), y^2+1)
    ***   at top-level: factornf(x^2+Mod(z,z
    ***                 ^ —  —  —  —  —  — --
    *** factornf: inconsistent data in rnf function.
  ? factornf(x^2 + z, y^2+1)
    ***   at top-level: factornf(x^2+z,y^2+1
    ***                 ^ —  —  —  —  —  — --
    *** factornf: incorrect variable in rnf function.

The library syntax is GEN polfnf(GEN x, GEN t).

Let G be the group attached to the galoisinit structure gal, and let χ be the character of some representation ρ of the group G, where a polynomial variable is to be interpreted as an o-th root of 1. For instance, if [T,o] = galoischartable(gal) the characters χ are input as the columns of T.

Return the degree-1 character detρ as the list of det ρ(g), where g runs through representatives of the conjugacy classes in galoisconjclasses(gal), with the same ordering.

  ? P = x^5 - x^4 - 5*x^3 + 4*x^2 + 3*x - 1;
  ? polgalois(P)
  %2 = [10, 1, 1, "D(5) = 5:2"]
  ? K = nfsplitting(P);
  ? gal = galoisinit(K);  \\ dihedral of order 10
  ? [T,o] = galoischartable(gal);
  ? chi = T[,1]; \\ trivial character
  ? galoischardet(gal, chi, o)
  %7 = [1, 1, 1, 1]~
  ? [galoischardet(gal, T[,i], o) | i <- [1..#T]] \\ all characters
  %8 = [[1, 1, 1, 1]~, [1, 1, -1, 1]~, [1, 1, -1, 1]~, [1, 1, -1, 1]~]

The library syntax is GEN galoischardet(GEN gal, GEN chi, long o).

Let G be the group attached to the galoisinit structure gal, and let χ be the character of some representation ρ of the group G, where a polynomial variable is to be interpreted as an o-th root of 1, e.g., if [T,o] = galoischartable(gal) and χ is a column of T. Return the list of characteristic polynomials det(1 - ρ(g)T), where g runs through representatives of the conjugacy classes in galoisconjclasses(gal), with the same ordering.

  ? T = x^5 - x^4 - 5*x^3 + 4*x^2 + 3*x - 1;
  ? polgalois(T)
  %2 = [10, 1, 1, "D(5) = 5:2"]
  ? K = nfsplitting(T);
  ? gal = galoisinit(K);  \\ dihedral of order 10
  ? [T,o] = galoischartable(gal);
  ? o
  %5 = 5
  ? galoischarpoly(gal, T[,1], o)  \\ T[,1] is the trivial character
  %6 = [-x + 1, -x + 1, -x + 1, -x + 1]~
  ? galoischarpoly(gal, T[,3], o)
  %7 = [x^2 - 2*x + 1,
        x^2 + (y^3 + y^2 + 1)*x + 1,
        -x^2 + 1,
        x^2 + (-y^3 - y^2)*x + 1]~

The library syntax is GEN galoischarpoly(GEN gal, GEN chi, long o).

Compute the character table of G, where G is the underlying group of the galoisinit structure gal. The input gal is also allowed to be a t_VEC of permutations that is closed under products. Let N be the number of conjugacy classes of G. Return a t_VEC [M,e] where e ≥ 1 is an integer and M is a square t_MAT of size N giving the character table of G.

* Each column corresponds to an irreducible character; the characters are ordered by increasing dimension and the first column is the trivial character (hence contains only 1's).

* Each row corresponds to a conjugacy class; the conjugacy classes are ordered as specified by galoisconjclasses(gal), in particular the first row corresponds to the identity and gives the dimension χ(1) of the irreducible representation attached to the successive characters χ.

The value M[i,j] of the character j at the conjugacy class i is represented by a polynomial in y whose variable should be interpreted as an e-th root of unity, i.e. as the lift of

    Mod(y, polcyclo(e,'y))

(Note that M is the transpose of the usual orientation for character tables.)

The integer e divides the exponent of the group G and is chosen as small as posible; for instance e = 1 when the characters are all defined over ℚ, as is the case for S_n. Examples:

  ? K = nfsplitting(x^4+x+1);
  ? gal = galoisinit(K);
  ? [M,e] = galoischartable(gal);
  ? M~  \\ take the transpose to get the usual orientation
  %4 =
  [1  1  1  1  1]
  
  [1 -1 -1  1  1]
  
  [2  0  0 -1  2]
  
  [3 -1  1  0 -1]
  
  [3  1 -1  0 -1]
  ? e
  %5 = 1
  ? {G = [Vecsmall([1, 2, 3, 4, 5]), Vecsmall([1, 5, 4, 3, 2]),
          Vecsmall([2, 1, 5, 4, 3]), Vecsmall([2, 3, 4, 5, 1]),
          Vecsmall([3, 2, 1, 5, 4]), Vecsmall([3, 4, 5, 1, 2]),
          Vecsmall([4, 3, 2, 1, 5]), Vecsmall([4, 5, 1, 2, 3]),
          Vecsmall([5, 1, 2, 3, 4]), Vecsmall([5, 4, 3, 2, 1])];}
    \\G = D10
  ? [M,e] = galoischartable(G);
  ? M~
  %8 =
  [1  1              1              1]
  
  [1 -1              1              1]
  
  [2  0 -y^3 - y^2 - 1      y^3 + y^2]
  
  [2  0      y^3 + y^2 -y^3 - y^2 - 1]
  ? e
  %9 = 5

The library syntax is GEN galoischartable(GEN gal).

gal being output by galoisinit, return the list of conjugacy classes of the underlying group. The ordering of the classes is consistent with galoischartable and the trivial class comes first.

  ? G = galoisinit(x^6+108);
  ? galoisidentify(G)
  %2 = [6, 1]  \\ S3
  ? S = galoisconjclasses(G)
  %3 = [[Vecsmall([1,2,3,4,5,6])],
        [Vecsmall([3,1,2,6,4,5]),Vecsmall([2,3,1,5,6,4])],
        [Vecsmall([6,5,4,3,2,1]),Vecsmall([5,4,6,2,1,3]),
                                 Vecsmall([4,6,5,1,3,2])]]
  ? [[permorder(c[1]),#c] | c <- S ]
  %4 = [[1,1], [3,2], [2,3]]

This command also accepts subgroups returned by galoissubgroups:

  ? subs = galoissubgroups(G); H = subs[5];
  ? galoisidentify(H)
  %2 = [2, 1]  \\ Z/2
  ? S = galoisconjclasses(subgroups_ofG[5]);
  ? [[permorder(c[1]),#c] | c <- S ]
  %4 = [[1,1], [2,1]]

The library syntax is GEN galoisconjclasses(GEN gal).

gal being be a Galois group as output by galoisinit, export the underlying permutation group as a string suitable for (no flags or flag = 0) GAP or (flag = 1) Magma. The following example compute the index of the underlying abstract group in the GAP library:

  ? G = galoisinit(x^6+108);
  ? s = galoisexport(G)
  %2 = "Group((1, 2, 3)(4, 5, 6), (1, 4)(2, 6)(3, 5))"
  ? extern("echo \"IdGroup("s");\" | gap -q")
  %3 = [6, 1]
  ? galoisidentify(G)
  %4 = [6, 1]

This command also accepts subgroups returned by galoissubgroups.

To import a GAP permutation into gp (for galoissubfields for instance), the following GAP function may be useful:

  PermToGP := function(p, n)
    return Permuted([1..n],p);
  end;
  
  gap> p:= (1,26)(2,5)(3,17)(4,32)(6,9)(7,11)(8,24)(10,13)(12,15)(14,27)
    (16,22)(18,28)(19,20)(21,29)(23,31)(25,30)
  gap> PermToGP(p,32);
  [ 26, 5, 17, 32, 2, 9, 11, 24, 6, 13, 7, 15, 10, 27, 12, 22, 3, 28, 20, 19,
    29, 16, 31, 8, 30, 1, 14, 18, 21, 25, 23, 4 ]

The library syntax is GEN galoisexport(GEN gal, long flag).

gal being be a Galois group as output by galoisinit and perm an element of gal.group, a vector of such elements or a subgroup of gal as returned by galoissubgroups, computes the fixed field of gal by the automorphism defined by the permutations perm of the roots gal.roots. P is guaranteed to be squarefree modulo gal.p.

If no flags or flag = 0, output format is the same as for nfsubfield, returning [P,x] such that P is a polynomial defining the fixed field, and x is a root of P expressed as a polmod in gal.pol.

If flag = 1 return only the polynomial P.

If flag = 2 return [P,x,F] where P and x are as above and F is the factorization of gal.pol over the field defined by P, where variable v (y by default) stands for a root of P. The priority of v must be less than the priority of the variable of gal.pol (see Section se:priority). In this case, P is also expressed in the variable v for compatibility with F. Example:

  ? G = galoisinit(x^4+1);
  ? galoisfixedfield(G,G.group[2],2)
  %2 = [y^2 - 2, Mod(- x^3 + x, x^4 + 1), [x^2 - y*x + 1, x^2 + y*x + 1]]

computes the factorization x⁴+1 = (x²-sqrt{2}x+1)(x²+sqrt{2}x+1)

The library syntax is GEN galoisfixedfield(GEN gal, GEN perm, long flag, long v = -1) where v is a variable number.

Query the galpol package for a group of order a with index b in the GAP4 Small Group library, by Hans Ulrich Besche, Bettina Eick and Eamonn O'Brien.

The current version of galpol supports groups of order a ≤ 143. If b is omitted, return the number of isomorphism classes of groups of order a.

The library syntax is GEN galoisgetgroup(long a, long b). Also available is GEN galoisnbpol(long a) when b is omitted.

Query the galpol package for a string describing the group of order a with index b in the GAP4 Small Group library, by Hans Ulrich Besche, Bettina Eick and Eamonn O'Brien. The strings were generated using the GAP4 function StructureDescription. The command below outputs the names of all abstract groups of order 12:

  ? o = 12; N = galoisgetgroup(o); \\ # of abstract groups of order 12
  ? for(i=1, N, print(i, ". ", galoisgetname(o,i)))
  1. C3 : C4
  2. C12
  3. A4
  4. D12
  5. C6 x C2

The current version of galpol supports groups of order a ≤ 143. For a ≥ 16, it is possible for different groups to have the same name:

  ? o = 20; N = galoisgetgroup(o);
  ? for(i=1, N, print(i, ". ", galoisgetname(o,i)))
  1. C5 : C4
  2. C20
  3. C5 : C4
  4. D20
  5. C10 x C2

The library syntax is GEN galoisgetname(long a, long b).

Query the galpol package for a polynomial with Galois group isomorphic to GAP4(a,b), totally real if s = 1 (default) and totally complex if s = 2. The current version of galpol supports groups of order a ≤ 143. The output is a vector [pol, den] where

* pol is the polynomial of degree a

* den is the denominator of nfgaloisconj(pol). Pass it as an optional argument to galoisinit or nfgaloisconj to speed them up:

  ? [pol,den] = galoisgetpol(64,4,1);
  ? G = galoisinit(pol);
  time = 352ms
  ? galoisinit(pol, den);  \\ passing 'den' speeds up the computation
  time = 264ms
  ? % == %`
  %4 = 1  \\ same answer

If b and s are omitted, return the number of isomorphism classes of groups of order a.

The library syntax is GEN galoisgetpol(long a, long b, long s). Also available is GEN galoisnbpol(long a) when b and s are omitted.

gal being be a Galois group as output by galoisinit, output the isomorphism class of the underlying abstract group as a two-components vector [o,i], where o is the group order, and i is the group index in the GAP4 Small Group library, by Hans Ulrich Besche, Bettina Eick and Eamonn O'Brien.

This command also accepts subgroups returned by galoissubgroups.

The current implementation is limited to degree less or equal to 127. Some larger "easy" orders are also supported.

The output is similar to the output of the function IdGroup in GAP4. Note that GAP4 IdGroup handles all groups of order less than 2000 except 1024, so you can use galoisexport and GAP4 to identify large Galois groups.

The library syntax is GEN galoisidentify(GEN gal).

Computes the Galois group and all necessary information for computing the fixed fields of the Galois extension K/ℚ where K is the number field defined by pol (monic irreducible polynomial in ℤ[X] or a number field as output by nfinit). The extension K/ℚ must be Galois with Galois group "weakly" super-solvable, see below; returns 0 otherwise. Hence this permits to quickly check whether a polynomial of order strictly less than 48 is Galois or not.

The algorithm used is an improved version of the paper "An efficient algorithm for the computation of Galois automorphisms", Bill Allombert, Math. Comp, vol. 73, 245, 2001, pp. 359–375.

A group G is said to be "weakly" super-solvable if there exists a normal series

{1} = H₀ ◃ H₁ ◃ ... ◃ H_n-1 ◃ H_n

such that each H_i is normal in G and for i < n, each quotient group H_i+1/H_i is cyclic, and either H_n = G (then G is super-solvable) or G/H_n is isomorphic to either A₄, S₄ or the group (3 x 3):4 (GAP4(36,9)).

In practice, almost all small groups are WKSS, the exceptions having order 48(2), 56(1), 60(1), 72(3), 75(1), 80(1), 96(10), 112(1), 120(3) and ≥ 144.

This function is a prerequisite for most of the galoisxxx routines. For instance:

  P = x^6 + 108;
  G = galoisinit(P);
  L = galoissubgroups(G);
  vector(#L, i, galoisisabelian(L[i],1))
  vector(#L, i, galoisidentify(L[i]))

The output is an 8-component vector gal.

gal[1] contains the polynomial pol (gal.pol).

gal[2] is a three-components vector [p,e,q] where p is a prime number (gal.p) such that pol totally split modulo p , e is an integer and q = p^e (gal.mod) is the modulus of the roots in gal.roots.

gal[3] is a vector L containing the p-adic roots of pol as integers implicitly modulo gal.mod. (gal.roots).

gal[4] is the inverse of the Vandermonde matrix of the p-adic roots of pol, multiplied by gal[5].

gal[5] is a multiple of the least common denominator of the automorphisms expressed as polynomial in a root of pol.

gal[6] is the Galois group G expressed as a vector of permutations of L (gal.group).

gal[7] is a generating subset S = [s₁,...,s_g] of G expressed as a vector of permutations of L (gal.gen).

gal[8] contains the relative orders [o₁,...,o_g] of the generators of S (gal.orders).

Let H_n be as above, we have the following properties:

  * if G/H_n ~ A₄ then [o₁,...,o_g] ends by [2,2,3].

  * if G/H_n ~ S₄ then [o₁,...,o_g] ends by [2,2,3,2].

  * if G/H_n ~ (3 x 3):4 (GAP4(36,9)) then [o₁,...,o_g] ends by [3,3,4].

  * for 1 ≤ i ≤ g the subgroup of G generated by [s₁,...,s_i] is normal, with the exception of i = g-2 in the A₄ and (3 x 3):4 cases and of i = g-3 in the S₄ case.

  * the relative order o_i of s_i is its order in the quotient group G/<s₁,...,s_i-1>, with the same exceptions.

  * for any x ∈ G there exists a unique family [e₁,...,e_g] such that (no exceptions):

-- for 1 ≤ i ≤ g we have 0 ≤ e_i < o_i

-- x = g₁^e₁g₂^e₂...g_n^e_n

If present den must be a suitable value for gal[5].

The library syntax is GEN galoisinit(GEN pol, GEN den = NULL).

gal being as output by galoisinit, return 0 if gal is not an abelian group, and the HNF matrix of gal over gal.gen if flag = 0, 1 if flag = 1, and the SNF matrix of gal if flag = 2.

This command also accepts subgroups returned by galoissubgroups.

The library syntax is GEN galoisisabelian(GEN gal, long flag).

gal being as output by galoisinit, and subgrp a subgroup of gal as output by galoissubgroups,return 1 if subgrp is a normal subgroup of gal, else return 0.

This command also accepts subgroups returned by galoissubgroups.

The library syntax is long galoisisnormal(GEN gal, GEN subgrp).

gal being a Galois group as output by galoisinit and perm a element of gal.group, return the polynomial defining the Galois automorphism, as output by nfgaloisconj, attached to the permutation perm of the roots gal.roots. perm can also be a vector or matrix, in this case, galoispermtopol is applied to all components recursively.

Note that

  G = galoisinit(pol);
  galoispermtopol(G, G[6])~

is equivalent to nfgaloisconj(pol), if degree of pol is greater or equal to 2.

The library syntax is GEN galoispermtopol(GEN gal, GEN perm).

Compute the Galois group over Q of the splitting field of P, that is the smallest field over which P is totally split. P is assumed to be integral, monic and irreducible; it can also be given by a nf structure. If d is given, it must be a multiple of the splitting field degree. The output is compatible with functions expecting a galoisinit structure.

The library syntax is GEN galoissplittinginit(GEN P, GEN d = NULL).

Computes the subextension L of ℚ(ζ_n) fixed by the subgroup H ⊂ (ℤ/nℤ)^*. By the Kronecker-Weber theorem, all abelian number fields can be generated in this way (uniquely if n is taken to be minimal). This function output is somewhat canonical, as it returns the minimal polynomial of a Gaussian period Tr_{ℚ(ζ_{n})/L}(ζ_n).

The pair (n, H) is deduced from the parameters (N, H) as follows

* N an integer: then n = N; H is a generator, i.e. an integer or an integer modulo n; or a vector of generators.

* N the output of znstar(n) or znstar(n,1). H as in the first case above, or a matrix, taken to be a HNF left divisor of the SNF for (ℤ/nℤ)^* (N.cyc), giving the generators of H in terms of N.gen.

* N the output of bnrinit(bnfinit(y), m) where m is a module. H as in the first case, or a matrix taken to be a HNF left divisor of the SNF for the ray class group modulo m (of type N.cyc), giving the generators of H in terms of N.bid.gen ( = N.gen if N includes generators).

In this last case, beware that H is understood relatively to N; in particular, if the infinite place does not divide the module, e.g if m is an integer, then it is not a subgroup of (ℤ/nℤ)^*, but of its quotient by {± 1}.

If flag = 0, computes a polynomial (in the variable v) defining the subfield of ℚ(ζ_n) fixed by the subgroup H of (ℤ/nℤ)^*.

If flag = 1, computes only the conductor of the abelian extension, as a module.

If flag = 2, outputs [pol, N], where pol is the polynomial as output when flag = 0 and N the conductor as output when flag = 1.

If flag = 3; outputs galoisinit(pol).

The following function can be used to compute all subfields of ℚ(ζ_n) (of exact degree d, if d is set):

  subcyclo(n, d = -1)=
  { my(bnr,L,IndexBound);
    IndexBound = if (d < 0, n, [d]);
    bnr = bnrinit(bnfinit(y), [n,[1]]);
    L = subgrouplist(bnr, IndexBound, 1);
    vector(#L,i, galoissubcyclo(bnr,L[i]));
  }

Setting L = subgrouplist(bnr, IndexBound) would produce subfields of exact conductor n oo .

The library syntax is GEN galoissubcyclo(GEN N, GEN H = NULL, long flag, long v = -1) where v is a variable number.

Outputs all the subfields of the Galois group G, as a vector. This works by applying galoisfixedfield to all subgroups. The meaning of flag is the same as for galoisfixedfield.

The library syntax is GEN galoissubfields(GEN G, long flag, long v = -1) where v is a variable number.

Outputs all the subgroups of the Galois group gal. A subgroup is a vector [gen, orders], with the same meaning as for gal.gen and gal.orders. Hence gen is a vector of permutations generating the subgroup, and orders is the relatives orders of the generators. The cardinality of a subgroup is the product of the relative orders. Such subgroup can be used instead of a Galois group in the following command: galoisisabelian, galoissubgroups, galoisexport and galoisidentify.

To get the subfield fixed by a subgroup sub of gal, use

  galoisfixedfield(gal,sub[1])

The library syntax is GEN galoissubgroups(GEN G).

gc being the structure returned by gcharinit, returns a t_MAT whose columns form a basis of the subgroup of algebraic Grossencharacters in gc (Weil type A0). The last component is interpreted as a power of the norm.

If type is a t_VEC of length gc.r1+gc.r2, containing a pair of integers [p_τ,q_τ] for each complex embedding τ, returns a t_VEC containing a character whose infinity type at τ is z ⟼ z^-p_τz^-q_τ if such a character exists, or empty otherwise. The full set of characters of that infinity type is obtained by multiplying by the group of finite order characters.

  ? bnf = bnfinit(x^4-2*x^3+23*x^2-22*x+6,1);
  ? gc = gcharinit(bnf,1);
  ? gc.cyc
  % = [6, 0, 0, 0, 0.E-57]
  ? gcharalgebraic(gc)
  % =
  [1 0    0  0]
  [0 1    0  0]
  [0 0    1  0]
  [0 0    0  0]
  [0 0 -1/2 -1]
  ? gcharalgebraic(gc,[[1,1],[0,1]])
  % = [] \\   pτ+qτ must be constant for an algebraic character to exist
  ? chi = gcharalgebraic(gc,[[1,1],[0,2]])[1]
  % = [0, 1, 2, 0, -1]~
  ? for(i=0,5,print(lfuneuler([gc,chi+[i,0,0,0,0]~],3)));
  \\  all characters with this infinity type: multiply by finite order characters

When the torsion subgroup is not cyclic, we can enumerate the characters of a given type with forvec.

  ? bnf = bnfinit(x^4+15*x^2+45,1);
  ? gc = gcharinit(bnf,1);
  ? gc.cyc
  % = [2, 2, 0, 0, 0, 0.E-57]
  ? [chi] = gcharalgebraic(gc,[[2,0],[2,0]]);
  ? {forvec(v=vectorv(2,i,[0,gc.cyc[i]-1]),
       print(round(lfunan([gc,chi+concat(v,[0,0,0,0]~)],20)));
     )};
    [1, 0, 0, 4, -5, 0, 0, 0, -9, 0, 16, 0, 0, 0, 0, 16, 0, 0, 16, -20]
    [1, 0, 0, -4, 5, 0, 0, 0, 9, 0, 16, 0, 0, 0, 0, 16, 0, 0, -16, -20]
    [1, 0, 0, 4, 5, 0, 0, 0, 9, 0, -16, 0, 0, 0, 0, 16, 0, 0, 16, 20]
    [1, 0, 0, -4, -5, 0, 0, 0, -9, 0, -16, 0, 0, 0, 0, 16, 0, 0, -16, 20]

Some algebraic Hecke characters are related to CM Abelian varieties. We first show an example with an elliptic curve.

  ? E = ellinit([0, 0, 1, -270, -1708]); \\  elliptic curve with potential CM by ℚ(sqrt{-3})
  ? bnf = bnfinit(x^2+3,1);
  ? p3 = idealprimedec(bnf,3)[1];
  ? gc = gcharinit(bnf,Mat([p3,2]));
  ? gc.cyc
  % = [0, 0.E-57]
  ? [chi] = gcharalgebraic(gc,[[1,0]])
  % = [[-1, -1/2]~]
  ? LE = lfuncreate(E);
  ? lfunan(LE,20)
  % = [1, 0, 0, -2, 0, 0, -1, 0, 0, 0, 0, 0, 5, 0, 0, 4, 0, 0, -7, 0]
  ? Lchi = lfuncreate([gc,chi]);
  ? round(lfunan(Lchi,20))
  % = [1, 0, 0, -2, 0, 0, -1, 0, 0, 0, 0, 0, 5, 0, 0, 4, 0, 0, -7, 0]

Here is an example with a CM Abelian surface.

  ? L = lfungenus2([-2*x^4 - 2*x^3 + 2*x^2 + 3*x - 2, x^3]);
  ? bnf = bnfinit(a^4 - a^3 + 2*a^2 + 4*a + 3, 1);
  ? pr = idealprimedec(bnf,13)[1];
  ? gc = gcharinit(bnf,pr);
  ? gc.cyc
  % = [3, 0, 0, 0, 0.E-57]
  ? chitors = [1,0,0,0,0]~;
  ? typ = [[1,0],[1,0]];
  ? [chi0] = gcharalgebraic(gc,typ);
  ? igood = oo; nbgood = 0;
  ? {for(i=0,gc.cyc[1]-1,
       chi = chi0 + i*chitors;
       Lchi = lfuncreate([gc,chi]);
       if(lfunparams(L) == lfunparams(Lchi)
         && exponent(lfunan(L,10) - lfunan(Lchi,10)) < -50,
         igood=i; nbgood++
       );
    )};
  ? nbgood
  % = 1
  ? chi = chi0 + igood*chitors;
  ? Lchi = lfuncreate([gc,chi]);
  ? lfunan(L,30)
  % = [1, 0, -3, 0, 0, 0, 0, 0, 4, 0, 0, 0, 0, 0, 0, -4, 0, 0,
    0, 0, 0, 0, 0, 0, 0, 0, -6, 0, -3, 0]
  ? round(lfunan(Lchi,30))
  % = [1, 0, -3, 0, 0, 0, 0, 0, 4, 0, 0, 0, 0, 0, 0, -4, 0, 0,
    0, 0, 0, 0, 0, 0, 0, 0, -6, 0, -3, 0]

The library syntax is GEN gcharalgebraic(GEN gc, GEN type = NULL).

Returns the conductor of chi, as a modulus over gc.bnf. This is the minimum modulus 𝔪 such that U(𝔪) ⊂ ker(chi) indicating the exact ramification of chi.

* for a real place v, v | 𝔪 iff χ_v(-1) = -1.

* for a finite place 𝔭, the prime power 𝔭^e divides exactly 𝔪 if e ≥ 0 is the smallest integer such that χ_𝔭 (U_e) = 1 where U₀ = ℤ_𝔭^× and U_i = 1+𝔭ⁱℤ_𝔭 for i > 0.

  ? bnf = bnfinit(x^2-5,1);
  ? gc = gcharinit(bnf,[(13*19)^2,[1,1]]);
  ? gc.cyc
  % = [8892, 6, 2, 0, 0.E-57]
  ? chi = [0,0,1,1]~;
  ? gcharconductor(gc,chi)
  % = [[61009, 7267; 0, 169], [1, 0]]
  ? gcharconductor(gc,13*chi)
  % = [[4693, 559; 0, 13], [1, 0]]
  ? gcharconductor(gc,13*19*chi)
  % = [[247, 65; 0, 13], [1, 0]]
  ? gcharconductor(gc,13*19*168*chi)
  % = [[19, 5; 0, 1], [0, 0]]

The library syntax is GEN gchar_conductor(GEN gc, GEN chi).

Returns internal logarithm vector of character chi as a t_VEC in ℝⁿ, so that for all x, gchareval(gc,chi,x,0) is equal to gcharduallog(gc,chi) * gcharlog(gc,x) in ℝ/ ℤ.

The components are organized as follows:

* the first ns components are in ℝ and describe the character on the class group generators: θ encodes 𝔭 ⟼ exp(2iπθ),

* the next nc components are in ℝ and describe the idealstar group character via its image on generators: θ encodes the image exp(2iπθ),

* the next r₁+r₂ components are in ℝ and correspond to characters of ℝ for each infinite place: ϕ encodes x ⟼ |x|^iϕ in the real case and z ⟼ |z|^2iϕ in the complex case,

* the last r₂ components are in ℤ and correspond to characters of ℝ/ℤ for each complex place: k encodes z ⟼ (z/|z|)^k.

* the last component s is in ℂ and corresponds to a power |.|^s of the adélic norm.

See also gcharlog.

  ? bnf = bnfinit(x^3+4*x-1,1);
  ? gc = gcharinit(bnf,[1,[1]]);
  ? gc.cyc
  % = [2, 0, 0, 0.E-57]
  ? chi = [0,1,0]~;
  ? f = gcharduallog(gc,chi)
  % = [0.153497221319231, 1/2, 0.776369647248353, -0.388184823624176, 1, 0]
  ? pr = idealprimedec(bnf,2)[1];
  ? v = gcharlog(gc,pr);
  ? exp(2*I*Pi*f*v)
  % = -0.569867696226731232993110144 - 0.821736459454756074068598760*I
  ? gchareval(gc,chi,pr)
  % = -0.569867696226731232993110144 - 0.821736459454756074068598760*I

The library syntax is GEN gcharduallog(GEN gc, GEN chi).

gc being the structure returned by gcharinit, chi a character in gc, and x an ideal of the base field, returns the value χ(x). If flag = 1 (default), returns a value in ℂ^×; if flag = 0, returns a value in ℂ/ℤ, normalized so that the real part is between -1/2 and 1/2.

  ? bnf = bnfinit(x^2-5);
  ? gc = gcharinit(bnf,1);
  ? chi = [1]~;
  ? pr = idealprimedec(bnf,11)[1];
  ? a = gchareval(gc,chi,pr)
  % = -0.3804107379142448929315340886 - 0.9248176417432464199580504588*I
  ? b = gchareval(gc,chi,pr,0)
  % = -0.3121086861831031476247589216
  ? a == exp(2*Pi*I*b)
  %7 = 1

The library syntax is GEN gchareval(GEN gc, GEN chi, GEN x, long flag).

gc being a Grossencharacter group as output by gcharinit, Lv being t_VEC of places v encoded by a t_INT (infinite place) or a prime ideal structure representing a prime not dividing the modulus of gc (finite place), and Lchiv being a t_VEC of local characters χ_v encoded by [k,ϕ] with k a t_INT and ϕ a t_REAL or t_COMPLEX representing x ⟼ sign(x)^k|x|^iϕ (real place) or z ⟼ (z/|z|)^k|z|^2iϕ(complex place) or by a t_REAL or t_COMPLEX θ representing 𝔭 ⟼ exp(2iπ θ) (finite place), returns a Grossencharacter ψ belonging to g such that ψ_v ~ χ_v for all v. At finite places, in place of a scalar one can provide a t_VEC whose last component is θ, as output by gcharlocal. To ensure proper identification, it is recommended to provide all infinite places together with a set of primes that generate the ray class group of modulus gc.mod.

  ? bnf = bnfinit(x^2-5,1);
  ? gc = gcharinit(bnf,1);
  ? chi = gcharidentify(gc,[2],[[0,13.]]);
  ? gcharlocal(gc,chi,2)
  % = [0, 13.057005210545987626926134713745179631]
  ? pr = idealprimedec(bnf,11)[1];
  ? chi = gcharidentify(gc,[pr],[0.3]);
  ? gchareval(gc,chi,pr,0)
  % = 0.30000006229129706787363344444425752636

If you know only few digits, it may be a good idea to reduce the current precision to obtain a meaningful result.

  ? bnf = bnfinit(x^2-5,1);
  ? gc = gcharinit(bnf,1);
  ? pr = idealprimedec(bnf,11)[1];
  ? chi = gcharidentify(gc,[pr],[0.184760])
  % = [-420226]~ \\   unlikely to be meaningful
  ? gchareval(gc,chi,pr,0)
  % = 0.18475998070331376194260927294721168954
  ? \p 10
    realprecision = 19 significant digits (10 digits displayed)
  ? chi = gcharidentify(gc,[pr],[0.184760])
  % = [-7]~ \\   probably what we were looking for
  ? gchareval(gc,chi,pr,0)
  % = 0.1847608033
  ? \p 38
    realprecision = 38 significant digits
  ? gchareval(gc,chi,pr,0)
  % = 0.18476080328172203337331245154966763237

The output may be a quasi-character.

  ? bnf = bnfinit(x^2-2,1);
  ? gc = gcharinit(bnf,1); gc.cyc
  % = [0, 0.E-57]
  ? gcharidentify(gc,[1,2],[[0,3.5+1/3*I],[0,-3.5+1/3*I]])
  % = [-1, 1/3]~

The library syntax is GEN gchar_identify(GEN gc, GEN Lv, GEN Lchiv, long prec).

bnf being a number field output by bnfinit (including fundamental units), f a modulus, initializes a structure (gc) describing the group of Hecke Grossencharacters of modulus f. (As in idealstar, the finite part of the conductor may be given by a factorization into prime ideals, as produced by idealfactor.)

The following member functions are available on the result: .bnf is the underlying bnf, .mod the modulus, .cyc its elementary divisors.

The internal representation uses a logarithm map on ideals ℒ: I → ℝⁿ, so that a Hecke Grossencharacter χ can be described by a n components vector v via χ: a ∈ I ⟼ exp(2iπ v.{ℒ(a)}).

See gcharlog for more details on the map ℒ.

  ? bnf = bnfinit(polcyclo(5),1); \\   initializes number field ℚ(ζ5)
  ? pr = idealprimedec(bnf,5)[1]; \\   prime 𝔭 = (1-ζ5) above 5
  ? gc = gcharinit(bnf,idealpow(bnf,pr,2)); \\   characters of modulus dividing 𝔭2
  ? gc.cyc \\   structure as an abelian group
  % = [0,0,0,0.E-57]
  ? chi = [1,1,-1,0]~; \\   a character
  ? gcharconductor(gc,chi)[1]
  % =
  [5 4 1 4]
  [0 1 0 0]
  [0 0 1 0]
  [0 0 0 1]

Currently, gc is a row vector with 11 components:

gc[1] is a matrix whose rows describe a system of generators of the characters as vectors of ℝⁿ, under the above description.

gc[2] contains the underlying number field bnf (gc.bnf).

gc[3] contains the underlying number field nf (gc.nf), possibly stored at higher precision than bnf.

gc[4] contains data for computing in (ℤ_K/f)^×.

gc[5] is a vector S of prime ideals which generate the class group.

gc[6] contains data to compute discrete logarithms with respect to S in the class group.

gc[7] is a vector [Sunits,m], where Sunits describes the S-units of bnf and m is a relation matrix for internal usage.

gc[8] is [Vecsmall([evalprec,prec,nfprec]), Vecsmall([ntors,nfree,nalg])] caching precisions and various dimensions.

gc[9] is a vector describing gc as a ℤ-module via its SNF invariants (gc.cyc), the last component representing the norm character.

gc[10] is a vector [R,U,Ui] allowing to convert characters from SNF basis to internal combination of generators.

Specifically, a character chi in SNF basis has coordinates chi*Ui in internal basis (the rows of gc[1]).

gc[11] = m is the matrix of ℒ(v) for all S-units v.

gc[12] = u is an integral base change matrix such that gc[1] corresponds to (mu)^-1.

The library syntax is GEN gcharinit(GEN bnf, GEN f, long prec).

gc being the structure returned by gcharinit and chi a character on gc, returns 1 if and only if chi is an algebraic (Weil type A0) character, so that its infinity type at every complex embedding τ can be written z ⟼ z^-p_τz^-q_τ for some pair of integers (p_τ,q_τ).

If type is given, it is set to the t_VEC of exponents [p_τ,q_τ].

  ? bnf = bnfinit(x^4+1,1);
  ? gc = gcharinit(bnf,1);
  ? gc.cyc
  % = [0, 0, 0, 0.E-57]
  ? chi1 = [0,0,1]~;
  ? gcharisalgebraic(gc,chi1)
  % = 0
  ? gcharlocal(gc,chi1,1)
  % = [-3, -0.89110698909568455588720672648627467040]
  ? chi2 = [1,0,0,-3]~;
  ? gcharisalgebraic(gc,chi2,&typ)
  % = 1
  ? typ
  % = [[6, 0], [2, 4]]
  ? gcharlocal(gc,chi2,1)
  % = [-6, 3*I]

The library syntax is GEN gcharisalgebraic(GEN gc, GEN chi, GEN *type = NULL).

gc being a gchar structure initialised by gcharinit, returns the local component χ_v, where v is either an integer between 1 and r₁+r₂ encoding an infinite place, or a prime ideal structure encoding a finite place.

* if v is a real place, χ_v(x) = {\rm sign}(x)^k |x|^iϕ is encoded as [k,ϕ];

* if v is a complex place, χ_v(z) = (z/|z|)^k |z|^2iϕ is encoded as [k,ϕ];

* if v = 𝔭 is a finite place not dividing gc.mod, χ_v(π_v) = exp(2iπ θ) is encoded as [θ];

* if v = 𝔭 is a finite place dividing gc.mod, we can define a bid structure attached to the multiplicative group G = (ℤ_K/𝔭^k)^*, where 𝔭^k divides exactly gc.mod (see idealstar). Then χ_v is encoded as [c₁,...,c_n,θ] where [c₁,...,c_n] defines a character on G (see gchareval) and χ_v(π_v) = exp(2iπθ). This bid structure only depends on gc and v (and not on the character χ); it can be recovered through the optional argument BID.

  ? bnf = bnfinit(x^3-x-1);
  ? gc = gcharinit(bnf,1);
  ? gc.cyc
  % = [0, 0, 0.E-57]
  ? chi = [0,1,1/3]~;
  ? pr = idealprimedec(bnf,5)[1];
  ? gcharlocal(gc,chi,1)
  % = [0, -4.8839310048284836274074581373242545693 - 1/3*I]
  ? gcharlocal(gc,chi,2)
  % = [6, 2.4419655024142418137037290686621272847 - 1/3*I]
  ? gcharlocal(gc,chi,pr)
  % = [0.115465135184293124024408915 + 0.0853833331211293579127218326*I]
  ? bnf = bnfinit(x^2+1,1);
  ? pr3 = idealprimedec(bnf,3)[1];
  ? pr5 = idealprimedec(bnf,5)[1];
  ? gc = gcharinit(bnf,[pr3,2;pr5,3]);
  ? gc.cyc
  % = [600, 3, 0, 0.E-57]
  ? chi = [1,1,1]~;
  ? gcharlocal(gc,chi,pr3,&bid)
  % = [1, 1, -21/50]
  ? bid.cyc
  % = [24, 3]
  ? gcharlocal(gc,chi,pr5,&bid)
  % = [98, -0.30120819117478336291229946188762973702]
  ? bid.cyc
  % = [100]

The library syntax is GEN gcharlocal(GEN gc, GEN chi, GEN v, long prec, GEN *BID = NULL).

Returns the internal (logarithmic) representation of the ideal x suitable for computations in gc, as a t_COL in ℝⁿ.

Its n = ns+nc+(r₁+r₂)+r₂+1 components correspond to a logarithm map on the group of fractional ideals ℒ: I → ℝⁿ, see gcharinit.

More precisely, let x = (α) ∏ 𝔭_i^a_i a principalization of x on a set S of primes generating the class group (see bnfisprincipal), then the logarithm of x is the t_COL

ℒ(x) = [ (a_i), log_f(α), (log|x/α|_τ)/(2π), (arg(x/α)_τ)/(2π), (log N(x))/(2π).i ]

where

* the exponent vector (a_i) has ns components, where ns = #S is the number of prime ideals used to generate the class group,

* log_f(α) is a discrete logarithm of α in the idealstar group (ℤ_K/f)^×, with nc components,

* log|x/α|_τ has r₁+r₂ components, one for each real embedding and pair of complex embeddings τ: K → ℂ (and |z|_τ = |z|² for complex τ).

* arg{(x/α)_τ} has r₂ components, one for each pair of complex embeddings τ: K → ℂ.

* N(x) is the norm of the ideal x.

  ? bnf = bnfinit(x^3-x^2+5*x+1,1);
  ? gc = gcharinit(bnf,3);
  ? gc.cyc
  % = [3, 0, 0, 0.E-57]
  ? chi = [1,1,0,-1]~;
  ? f = gcharduallog(gc,chi);
  ? pr = idealprimedec(bnf,5)[1];
  ? v = gcharlog(gc,pr)
  % = [2, -5, -1, 0.0188115475004995312411, -0.0188115475004995312411,
       -0.840176314833856764413, 0.256149999363388073738*I]~
  ? exp(2*I*Pi*f*v)
  % = -4.5285995080704456583673312 + 2.1193835177957097598574507*I
  ? gchareval(gc,chi,pr)
  % = -4.5285995080704456583673312 + 2.1193835177957097598574507*I

The library syntax is GEN gcharlog(GEN gc, GEN x, long prec).

gc being a Grossencharacter group output by gcharinit, recomputes its archimedean components ensuring accurate computations to current precision.

It is advisable to increase the precision before computing several values at large ideals.

The library syntax is GEN gcharnewprec(GEN gc, long prec).

Sum of the two ideals x and y in the number field nf. The result is given in HNF.

   ? K = nfinit(x^2 + 1);
   ? a = idealadd(K, 2, x + 1)  \\ ideal generated by 2 and 1+I
   %2 =
   [2 1]
  
   [0 1]
   ? pr = idealprimedec(K, 5)[1];  \\ a prime ideal above 5
   ? idealadd(K, a, pr)     \\ coprime, as expected
   %4 =
   [1 0]
  
   [0 1]

This function cannot be used to add arbitrary ℤ-modules, since it assumes that its arguments are ideals:

    ? b = Mat([1,0]~);
    ? idealadd(K, b, b)     \\ only square t_MATs represent ideals
    *** idealadd: nonsquare t_MAT in idealtyp.
    ? c = [2, 0; 2, 0]; idealadd(K, c, c)   \\ nonsense
    %6 =
    [2 0]
  
    [0 2]
    ? d = [1, 0; 0, 2]; idealadd(K, d, d)   \\ nonsense
    %7 =
    [1 0]
  
    [0 1]
  

In the last two examples, we get wrong results since the matrices c and d do not correspond to an ideal: the ℤ-span of their columns (as usual interpreted as coordinates with respect to the integer basis K.zk) is not an ℤ_K-module. To add arbitrary ℤ-modules generated by the columns of matrices A and B, use mathnf(concat(A,B)).

The library syntax is GEN idealadd(GEN nf, GEN x, GEN y).

x and y being two co-prime integral ideals (given in any form), this gives a two-component row vector [a,b] such that a ∈ x, b ∈ y and a+b = 1.

The alternative syntax idealaddtoone(nf,v), is supported, where v is a k-component vector of ideals (given in any form) which sum to ℤ_K. This outputs a k-component vector e such that e[i] ∈ x[i] for 1 ≤ i ≤ k and ∑_{1 ≤ i ≤ k}e[i] = 1.

The library syntax is GEN idealaddtoone0(GEN nf, GEN x, GEN y = NULL).

If x is a fractional ideal (given in any form), gives an element α in nf such that for all prime ideals 𝔭 such that the valuation of x at 𝔭 is nonzero, we have v_𝔭(α) = v_𝔭(x), and v_𝔭(α) ≥ 0 for all other 𝔭.

The argument x may also be given as a prime ideal factorization, as output by idealfactor, but allowing zero exponents. This yields an element α such that for all prime ideals 𝔭 occurring in x, v_𝔭(α) = v_𝔭(x); for all other prime ideals, v_𝔭(α) ≥ 0.

flag is deprecated (ignored), kept for backward compatibility.

The library syntax is GEN idealappr0(GEN nf, GEN x, long flag). Use directly GEN idealappr(GEN nf, GEN x) since flag is ignored.

x being a prime ideal factorization (i.e. a 2-columns matrix whose first column contains prime ideals and the second column contains integral exponents), y a vector of elements in nf indexed by the ideals in x, computes an element b such that

v_𝔭(b - y_𝔭) ≥ v_𝔭(x) for all prime ideals in x and v_𝔭(b) ≥ 0 for all other 𝔭.

  ? K = nfinit(t^2-2);
  ? x = idealfactor(K, 2^2*3)
  %2 =
  [[2, [0, 1]~, 2, 1, [0, 2; 1, 0]] 4]
  
  [           [3, [3, 0]~, 1, 2, 1] 1]
  ? y = [t,1];
  ? idealchinese(K, x, y)
  %4 = [4, -3]~

The argument x may also be of the form [x, s] where the first component is as above and s is a vector of signs, with r₁ components s_i in {-1,0,1}: if σ_i denotes the i-th real embedding of the number field, the element b returned satisfies further sign(σ_i(b)) = s_i for all i such that s_i = ±1. In other words, the sign is fixed to s_i at the i-th embedding whenever s_i is nonzero.

  ? idealchinese(K, [x, [1,1]], y)
  %5 = [16, -3]~
  ? idealchinese(K, [x, [-1,-1]], y)
  %6 = [-20, -3]~
  ? idealchinese(K, [x, [1,-1]], y)
  %7 = [4, -3]~

If y is omitted, return a data structure which can be used in place of x in later calls and allows to solve many chinese remainder problems for a given x more efficiently. In this case, the right hand side y is not allowed to have denominators, unless they are coprime to x.

  ? C = idealchinese(K, [x, [1,1]]);
  ? idealchinese(K, C, y) \\ as above
  %9 = [16, -3]~
  ? for(i=1,10^4, idealchinese(K,C,y))  \\ ... but faster !
  time = 80 ms.
  ? for(i=1,10^4, idealchinese(K,[x,[1,1]],y))
  time = 224 ms.

Finally, this structure is itself allowed in place of x, the new s overriding the one already present in the structure. This allows to initialize for different sign conditions more efficiently when the underlying ideal factorization remains the same.

  ? D = idealchinese(K, [C, [1,-1]]);   \\ replaces [1,1]
  ? idealchinese(K, D, y)
  %13 = [4, -3]~
  ? for(i=1,10^4,idealchinese(K,[C,[1,-1]]))
  time = 40 ms.   \\ faster than starting from scratch
  ? for(i=1,10^4,idealchinese(K,[x,[1,-1]]))
  time = 128 ms.

The library syntax is GEN idealchinese(GEN nf, GEN x, GEN y = NULL). Also available is GEN idealchineseinit(GEN nf, GEN x) when y = NULL.

Given two integral ideals x and y in the number field nf, returns a β in the field, such that β.x is an integral ideal coprime to y. In fact, β is also guaranteed to be integral outside primes dividing y.

The library syntax is GEN idealcoprime(GEN nf, GEN x, GEN y).

Quotient x.y^-1 of the two ideals x and y in the number field nf. The result is given in HNF.

If flag is nonzero, the quotient x.y^-1 is assumed to be an integral ideal. This can be much faster when the norm of the quotient is small even though the norms of x and y are large. More precisely, the algorithm cheaply removes all maximal ideals above rational primes such that v_p(Nx) = v_p(Ny).

The library syntax is GEN idealdiv0(GEN nf, GEN x, GEN y, long flag). Also available are GEN idealdiv(GEN nf, GEN x, GEN y) (flag = 0) and GEN idealdivexact(GEN nf, GEN x, GEN y) (flag = 1).

Let nf be a number field as output by nfinit, and x a fractional ideal. This function returns the nonnegative rational generator of x ∩ ℚ. If x is an extended ideal, the extended part is ignored.

  ? nf = nfinit(y^2+1);
  ? idealdown(nf, -1/2)
  %2 = 1/2
  ? idealdown(nf, (y+1)/3)
  %3 = 2/3
  ? idealdown(nf, [2, 11]~)
  %4 = 125
  ? x = idealprimedec(nf, 2)[1]; idealdown(nf, x)
  %5 = 2
  ? idealdown(nf, [130, 94; 0, 2])
  %6 = 130

The library syntax is GEN idealdown(GEN nf, GEN x).

Factors into prime ideal powers the ideal x in the number field nf. The output format is similar to the factor function, and the prime ideals are represented in the form output by the idealprimedec function. If lim is set, return partial factorization, including only prime ideals above rational primes < lim.

  ? nf = nfinit(x^3-2);
  ? idealfactor(nf, x) \\ a prime ideal above 2
  %2 =
  [[2, [0, 1, 0]~, 3, 1, ...] 1]
  
  ? A = idealhnf(nf, 6*x, 4+2*x+x^2)
  %3 =
  [6 0 4]
  
  [0 6 2]
  
  [0 0 1]
  
  ? idealfactor(nf, A)
  %4 =
   [[2, [0, 1, 0]~, 3, 1, ...] 2]
  
   [[3, [1, 1, 0]~, 3, 1, ...] 2]
  
  ? idealfactor(nf, A, 3) \\ restrict to primes above p < 3
  %5 =
  [[2, [0, 1, 0]~, 3, 1, ...] 2]

The library syntax is GEN gpidealfactor(GEN nf, GEN x, GEN lim = NULL). This function should only be used by the gp interface. Use directly GEN idealfactor(GEN nf, GEN x) or GEN idealfactor_limit(GEN nf, GEN x, ulong lim).

Gives back the ideal corresponding to a factorization. The integer 1 corresponds to the empty factorization. If e is present, e and f must be vectors of the same length (e being integral), and the corresponding factorization is the product of the f[i]^e[i].

If not, and f is vector, it is understood as in the preceding case with e a vector of 1s: we return the product of the f[i]. Finally, f can be a regular factorization, as produced by idealfactor.

  ? nf = nfinit(y^2+1); idealfactor(nf, 4 + 2*y)
  %1 =
  [[2, [1, 1]~, 2, 1, [1, 1]~] 2]
  
  [[5, [2, 1]~, 1, 1, [-2, 1]~] 1]
  
  ? idealfactorback(nf, %)
  %2 =
  [10 4]
  
  [0  2]
  
  ? f = %1[,1]; e = %1[,2]; idealfactorback(nf, f, e)
  %3 =
  [10 4]
  
  [0  2]
  
  ? % == idealhnf(nf, 4 + 2*y)
  %4 = 1

If flag is nonzero, perform ideal reductions (idealred) along the way. This is most useful if the ideals involved are all extended ideals (for instance with trivial principal part), so that the principal parts extracted by idealred are not lost. Here is an example:

  ? f = vector(#f, i, [f[i], [;]]);  \\ transform to extended ideals
  ? idealfactorback(nf, f, e, 1)
  %6 = [[1, 0; 0, 1], [2, 1; [2, 1]~, 1]]
  ? nffactorback(nf, %[2])
  %7 = [4, 2]~

The extended ideal returned in %6 is the trivial ideal 1, extended with a principal generator given in factored form. We use nffactorback to recover it in standard form.

The library syntax is GEN idealfactorback(GEN nf, GEN f, GEN e = NULL, long flag).

Let K be the number field defined by nf and assume K/ℚ be a Galois extension with Galois group given gal = galoisinit(nf), and that pr is an unramified prime ideal 𝔭 in prid format. This function returns a permutation of gal.group which defines the Frobenius element Frob_𝔭 attached to 𝔭. If p is the unique prime number in 𝔭, then Frob(x) = x^p mod 𝔭 for all x ∈ ℤ_K.

  ? nf = nfinit(polcyclo(31));
  ? gal = galoisinit(nf);
  ? pr = idealprimedec(nf,101)[1];
  ? g = idealfrobenius(nf,gal,pr);
  ? galoispermtopol(gal,g)
  %5 = x^8

This is correct since 101 = 8 mod 31.

The library syntax is GEN idealfrobenius(GEN nf, GEN gal, GEN pr).

Gives the Hermite normal form of the ideal uℤ_K+vℤ_K, where u and v are elements of the number field K defined by nf.

  ? nf = nfinit(y^3 - 2);
  ? idealhnf(nf, 2, y+1)
  %2 =
  [1 0 0]
  
  [0 1 0]
  
  [0 0 1]
  ? idealhnf(nf, y/2, [0,0,1/3]~)
  %3 =
  [1/3 0 0]
  
  [0 1/6 0]
  
  [0 0 1/6]

If v is omitted, returns the HNF of the ideal defined by u: u may be an algebraic number (defining a principal ideal), a maximal ideal (as given by idealprimedec or idealfactor), or a matrix whose columns give generators for the ideal. This last format is a little complicated, but useful to reduce general modules to the canonical form once in a while:

* if strictly less than N = [K:ℚ] generators are given, u is the ℤ_K-module they generate,

* if N or more are given, it is assumed that they form a ℤ-basis of the ideal, in particular that the matrix has maximal rank N. This acts as mathnf since the ℤ_K-module structure is (taken for granted hence) not taken into account in this case.

  ? idealhnf(nf, idealprimedec(nf,2)[1])
  %4 =
  [2 0 0]
  
  [0 1 0]
  
  [0 0 1]
  ? idealhnf(nf, [1,2;2,3;3,4])
  %5 =
  [1 0 0]
  
  [0 1 0]
  
  [0 0 1]

Finally, when K is quadratic with discriminant D_K, we allow u = Qfb(a,b,c), provided b² - 4ac = D_K. As usual, this represents the ideal a ℤ + (1/2)(-b + sqrt{D_K}) ℤ.

  ? K = nfinit(x^2 - 60); K.disc
  %1 = 60
  ? idealhnf(K, qfbprimeform(60,2))
  %2 =
  [2 1]
  
  [0 1]
  ? idealhnf(K, Qfb(1,2,3))
    ***   at top-level: idealhnf(K,Qfb(1,2,3
    ***                 ^ —  —  —  —  —  — --
    *** idealhnf: Qfb(1, 2, 3) has discriminant != 60 in idealhnf.

The library syntax is GEN idealhnf0(GEN nf, GEN u, GEN v = NULL). Also available is GEN idealhnf(GEN nf, GEN a), where nf is a true nf structure.

Intersection of the two ideals A and B in the number field nf. The result is given in HNF.

  ? nf = nfinit(x^2+1);
  ? idealintersect(nf, 2, x+1)
  %2 =
  [2 0]
  
  [0 2]

This function does not apply to general ℤ-modules, e.g. orders, since its arguments are replaced by the ideals they generate. The following script intersects ℤ-modules A and B given by matrices of compatible dimensions with integer coefficients:

  ZM_intersect(A,B) =
  { my(Ker = matkerint(concat(A,B)));
    mathnf( A * Ker[1..#A,] )
  }

The library syntax is GEN idealintersect(GEN nf, GEN A, GEN B).

Inverse of the ideal x in the number field nf, given in HNF. If x is an extended ideal, its principal part is suitably updated: i.e. inverting [I,t], yields [I^-1, 1/t].

The library syntax is GEN idealinv(GEN nf, GEN x).

Given nf a number field as output by nfinit and an ideal x, return 0 if x is not a maximal ideal. Otherwise return a prid structure nf attached to the ideal. This function uses ispseudoprime and may return a wrong result in case the underlying rational pseudoprime is not an actual prime number: apply isprime(pr.p) to guarantee correctness. If x is an extended ideal, the extended part is ignored.

  ? K = nfinit(y^2 + 1);
  ? idealismaximal(K, 3) \\ 3 is inert
  %2 = [3, [3, 0]~, 1, 2, 1]
  ? idealismaximal(K, 5) \\ 5 is not
  %3 = 0
  ? pr = idealprimedec(K,5)[1] \\ already a prid
  %4 = [5, [-2, 1]~, 1, 1, [2, -1; 1, 2]]
  ? idealismaximal(K, pr) \\ trivial check
  %5 = [5, [-2, 1]~, 1, 1, [2, -1; 1, 2]]
  ? x = idealhnf(K, pr)
  %6 =
  [5 3]
  
  [0 1]
  ? idealismaximal(K, x) \\ converts from matrix form to prid
  %7 = [5, [-2, 1]~, 1, 1, [2, -1; 1, 2]]

This function is noticeably faster than idealfactor since it never involves an actually factorization, in particular when x ∩ ℤ is not a prime number.

The library syntax is GEN idealismaximal(GEN nf, GEN x).

Let nf be a number field and n > 0 be a positive integer. Return 1 if the fractional ideal A = Bⁿ is an n-th power and 0 otherwise. If the argument B is present, set it to the n-th root of A, in HNF.

  ? K = nfinit(x^3 - 2);
  ? A = [46875, 30966, 9573; 0, 3, 0; 0, 0, 3];
  ? idealispower(K, A, 3, &B)
  %3 = 1
  ? B
  %4 =
  [75 22 41]
  
  [ 0  1  0]
  
  [ 0  0  1]
  
  ? A = [9375, 2841, 198; 0, 3, 0; 0, 0, 3];
  ? idealispower(K, A, 3)
  %5 = 0

The library syntax is long idealispower(GEN nf, GEN A, long n, GEN *B = NULL).

Computes the list of all ideals of norm less or equal to bound in the number field nf. The result is a row vector with exactly bound components. Each component is itself a row vector containing the information about ideals of a given norm, in no specific order. The information is inferred from local data and Chinese remainders and less expensive than computing than a direct global computation.

The binary digits of flag mean:

* 1: if the ideals are given by a bid, include generators; otherwise don't.

* 2: if this bit is set, nf must be a bnf with units. Each component is of the form [bid,U], where bid is attached to an ideal f and U is a vector of discrete logarithms of the units in (ℤ_K/f)^*. More precisely, U gives the ideallogs with respect to bid of (ζ,u₁,...,u_r) where ζ is the torsion unit generator bnf.tu[2] and (u_i) are the fundamental units in bnf.fu. This structure is technical, meant to be used in conjunction with bnrclassnolist or bnrdisclist.

* 4: give only the ideal (in HNF), else a bid.

* 8: omit ideals which cannot be conductors, i.e. divisible exactly my a prime ideal of norm 2.

  ? nf = nfinit(x^2+1);
  ? L = ideallist(nf, 100);
  ? L[1]
  %3 = [[1, 0; 0, 1]]  \\  A single ideal of norm 1
  ? #L[65]
  %4 = 4               \\  There are 4 ideals of norm 65 in ℤ[i]

If one wants more information:

  ? L = ideallist(nf, 100, 0);
  ? l = L[25]; vector(#l, i, l[i].clgp)
  %6 = [[20, [20]], [16, [4, 4]], [20, [20]]]
  ? l[1].mod
  %7 = [[25, 18; 0, 1], []]
  ? l[2].mod
  %8 = [[5, 0; 0, 5], []]
  ? l[3].mod
  %9 = [[25, 7; 0, 1], []]

where we ask for the structures of the (ℤ[i]/f)^* for all three ideals of norm 25. In fact, for all moduli with finite part of norm 25 and trivial Archimedean part, as the last 3 commands show. See ideallistarch to treat general moduli.

Finally, one can input a negative bound. The function then returns the ideals of norm |bound|, given by their factorization matrix. The only valid value of flag is then the default. If needed, one can obtain their HNF using idealfactorback, and the corresponding bid structures using idealstar (which accepts ideals in factored form).

The library syntax is GEN gideallist(GEN nf, GEN bound, long flag). Also available is GEN ideallist0(GEN nf,long bound, long flag) for a non-negative bound.

list is a vector of vectors of bid's, as output by ideallist with flag 0 to 3. Return a vector of vectors with the same number of components as the original list. The leaves give information about moduli whose finite part is as in original list, in the same order, and Archimedean part is now arch (it was originally trivial). The information contained is of the same kind as was present in the input; see ideallist, in particular the meaning of flag.

  ? bnf = bnfinit(x^2-2);
  ? bnf.sign
  %2 = [2, 0]                         \\  two places at infinity
  ? L = ideallist(bnf, 100, 0);
  ? l = L[98]; vector(#l, i, l[i].clgp)
  %4 = [[42, [42]], [36, [6, 6]], [42, [42]]]
  ? La = ideallistarch(bnf, L, [1,1]); \\  add them to the modulus
  ? l = La[98]; vector(#l, i, l[i].clgp)
  %6 = [[168, [42, 2, 2]], [144, [6, 6, 2, 2]], [168, [42, 2, 2]]]

Of course, the results above are obvious: adding t places at infinity will add t copies of ℤ/2ℤ to (ℤ_K/f)^*. The following application is more typical:

  ? L = ideallist(bnf, 100, 2);        \\  units are required now
  ? La = ideallistarch(bnf, L, [1,1]);
  ? H = bnrclassnolist(bnf, La);
  ? H[98];
  %4 = [2, 12, 2]

The library syntax is GEN ideallistarch(GEN nf, GEN list, GEN arch).

nf is a number field, bid is as output by idealstar(nf, D,...) and x an element of nf which must have valuation equal to 0 at all prime ideals in the support of D and need not be integral. This function computes the discrete logarithm of x on the generators given in bid.gen. In other words, if g_i are these generators, of orders d_i respectively, the result is a column vector of integers (x_i) such that 0 ≤ x_i < d_i and x = ∏_i g_i^x_i (mod ^*D) . Note that when the support of D contains places at infinity, this congruence implies also sign conditions on the attached real embeddings. See znlog for the limitations of the underlying discrete log algorithms.

When nf is omitted, take it to be the rational number field. In that case, x must be a t_INT and bid must have been initialized by znstar(N,1).

The library syntax is GEN ideallog(GEN nf = NULL, GEN x, GEN bid). Also available are GEN Zideallog(GEN bid, GEN x) when nf is NULL, and GEN ideallogmod(GEN nf, GEN x, GEN bid, GEN mod) that returns the discrete logarithm of x modulo the t_INT mod; the value mod = NULL is treated as 0 (full discrete logarithm), but nf = NULL is not implemented with nonzero mod.

This function is useless and kept for backward compatibility only, use idealred. Computes a pseudo-minimum of the ideal x in the direction vdir in the number field nf.

The library syntax is GEN idealmin(GEN nf, GEN ix, GEN vdir = NULL).

Ideal multiplication of the ideals x and y in the number field nf; the result is the ideal product in HNF. If either x or y are extended ideals, their principal part is suitably updated: i.e. multiplying [I,t], [J,u] yields [IJ, tu]; multiplying I and [J, u] yields [IJ, u].

  ? nf = nfinit(x^2 + 1);
  ? idealmul(nf, 2, x+1)
  %2 =
  [4 2]
  
  [0 2]
  ? idealmul(nf, [2, x], x+1)        \\ extended ideal * ideal
  %3 = [[4, 2; 0, 2], x]
  ? idealmul(nf, [2, x], [x+1, x])   \\ two extended ideals
  %4 = [[4, 2; 0, 2], [-1, 0]~]

If flag is nonzero, reduce the result using idealred.

The library syntax is GEN idealmul0(GEN nf, GEN x, GEN y, long flag).

See also GEN idealmul(GEN nf, GEN x, GEN y) (flag = 0) and GEN idealmulred(GEN nf, GEN x, GEN y) (flag ! = 0).

Computes the norm of the ideal x in the number field nf.

The library syntax is GEN idealnorm(GEN nf, GEN x).

Returns [A,B], where A,B are coprime integer ideals such that x = A/B, in the number field nf.

  ? nf = nfinit(x^2+1);
  ? idealnumden(nf, (x+1)/2)
  %2 = [[1, 0; 0, 1], [2, 1; 0, 1]]

The library syntax is GEN idealnumden(GEN nf, GEN x).

Computes the k-th power of the ideal x in the number field nf; k ∈ ℤ. If x is an extended ideal, its principal part is suitably updated: i.e. raising [I,t] to the k-th power, yields [I^k, t^k].

If flag is nonzero, reduce the result using idealred, throughout the (binary) powering process; in particular, this is not the same as idealpow(nf,x,k) followed by reduction.

The library syntax is GEN idealpow0(GEN nf, GEN x, GEN k, long flag).

See also GEN idealpow(GEN nf, GEN x, GEN k) and GEN idealpows(GEN nf, GEN x, long k) (flag = 0). Corresponding to flag = 1 is GEN idealpowred(GEN nf, GEN vp, GEN k).

Computes the prime ideal decomposition of the (positive) prime number p in the number field K represented by nf. If a nonprime p is given the result is undefined. If f is present and nonzero, restrict the result to primes of residue degree ≤ f.

The result is a vector of prid structures, each representing one of the prime ideals above p in the number field nf. The representation pr = [p,a,e,f,mb] of a prime ideal means the following: a is an algebraic integer in the maximal order ℤ_K and the prime ideal is equal to 𝔭 = pℤ_K + aℤ_K; e is the ramification index; f is the residual index; finally, mb is the multiplication table attached to an algebraic integer b such that 𝔭^-1 = ℤ_K+ b/ pℤ_K, which is used internally to compute valuations. In other words if p is inert, then mb is the integer 1, and otherwise it is a square t_MAT whose j-th column is b.nf.zk[j].

The algebraic number a is guaranteed to have a valuation equal to 1 at the prime ideal (this is automatic if e > 1).

The components of pr should be accessed by member functions: pr.p, pr.e, pr.f, and pr.gen (returns the vector [p,a]):

  ? K = nfinit(x^3-2);
  ? P = idealprimedec(K, 5);
  ? #P       \\ 2 primes above 5 in Q(2^(1/3))
  %3 = 2
  ? [p1,p2] = P;
  ? [p1.e, p1.f]    \\ the first is unramified of degree 1
  %5 = [1, 1]
  ? [p2.e, p2.f]    \\ the second is unramified of degree 2
  %6 = [1, 2]
  ? p1.gen
  %7 = [5, [2, 1, 0]~]
  ? nfbasistoalg(K, %[2])  \\ a uniformizer for p1
  %8 = Mod(x + 2, x^3 - 2)
  ? #idealprimedec(K, 5, 1) \\ restrict to f = 1
  %9 = 1            \\ now only p1

The library syntax is GEN idealprimedec_limitf(GEN nf, GEN p, long f).

Given a prime ideal in idealprimedec format, returns the multiplicative group (1 + pr) / (1 + pr^k) as an abelian group. This function is much faster than idealstar when the norm of pr is large, since it avoids (useless) work in the multiplicative group of the residue field.

  ? K = nfinit(y^2+1);
  ? P = idealprimedec(K,2)[1];
  ? G = idealprincipalunits(K, P, 20);
  ? G.cyc
  %4 = [512, 256, 4]   \\ Z/512 x Z/256 x Z/4
  ? G.gen
  %5 = [[-1, -2]~, 1021, [0, -1]~] \\ minimal generators of given order

The library syntax is GEN idealprincipalunits(GEN nf, GEN pr, long k).

Let K be the number field defined by nf and assume that K/ℚ is Galois with Galois group G given by gal = galoisinit(nf). Let pr be the prime ideal 𝔓 in prid format. This function returns a vector g of subgroups of gal as follows:

* g[1] is the decomposition group of 𝔓,

* g[2] is G₀(𝔓), the inertia group of 𝔓,

and for i ≥ 2,

* g[i] is G_i-2(𝔓), the i-2-th ramification group of 𝔓.

The length of g is the number of nontrivial groups in the sequence, thus is 0 if e = 1 and f = 1, and 1 if f > 1 and e = 1. The following function computes the cardinality of a subgroup of G, as given by the components of g:

  card(H) =my(o=H[2]); prod(i=1,#o,o[i]);

  ? nf=nfinit(x^6+3); gal=galoisinit(nf); pr=idealprimedec(nf,3)[1];
  ? g = idealramgroups(nf, gal, pr);
  ? apply(card,g)
  %3 = [6, 6, 3, 3, 3] \\ cardinalities of the Gi

  ? nf=nfinit(x^6+108); gal=galoisinit(nf); pr=idealprimedec(nf,2)[1];
  ? iso=idealramgroups(nf,gal,pr)[2]
  %5 = [[Vecsmall([2, 3, 1, 5, 6, 4])], Vecsmall([3])]
  ? nfdisc(galoisfixedfield(gal,iso,1))
  %6 = -3

The field fixed by the inertia group of 2 is not ramified at 2.

The library syntax is GEN idealramgroups(GEN nf, GEN gal, GEN pr).

LLL reduction of the ideal I in the number field K attached to nf, along the direction v. The v parameter is best left omitted, but if it is present, it must be an nf.r1 + nf.r2-component vector of nonnegative integers. (What counts is the relative magnitude of the entries: if all entries are equal, the effect is the same as if the vector had been omitted.)

This function finds an a ∈ K^* such that J = (a)I is "small" and integral (see the end for technical details). The result is the Hermite normal form of the "reduced" ideal J.

  ? K = nfinit(y^2+1);
  ? P = idealprimedec(K,5)[1];
  ? idealred(K, P)
  %3 =
  [1 0]
  
  [0 1]

More often than not, a principal ideal yields the unit ideal as above. This is a quick and dirty way to check if ideals are principal, but it is not a necessary condition: a nontrivial result does not prove that the ideal is nonprincipal. For guaranteed results, see bnfisprincipal, which requires the computation of a full bnf structure.

If the input is an extended ideal [I,s], the output is [J, sa]; in this way, one keeps track of the principal ideal part:

  ? idealred(K, [P, 1])
  %5 = [[1, 0; 0, 1], [2, -1]~]

meaning that P is generated by [2, -1] . The number field element in the extended part is an algebraic number in any form or a factorization matrix (in terms of number field elements, not ideals!). In the latter case, elements stay in factored form, which is a convenient way to avoid coefficient explosion; see also idealpow.

Technical note. The routine computes an LLL-reduced basis for the lattice I^-1 equipped with the quadratic form || x ||_v² = ∑_{i = 1}^r₁+r₂ 2^v_iϵ_i|σ_i(x)|², where as usual the σ_i are the (real and) complex embeddings and ϵ_i = 1, resp. 2, for a real, resp. complex place. The element a is simply the first vector in the LLL basis. The only reason you may want to try to change some directions and set some v_i ! = 0 is to randomize the elements found for a fixed ideal, which is heuristically useful in index calculus algorithms like bnfinit and bnfisprincipal.

Even more technical note. In fact, the above is a white lie. We do not use ||.||_v exactly but a rescaled rounded variant which gets us faster and simpler LLLs. There's no harm since we are not using any theoretical property of a after all, except that it belongs to I^-1 and that a I is "expected to be small".

The library syntax is GEN idealred0(GEN nf, GEN I, GEN v = NULL).

Let nf be a number field, x an ideal in nf and n > 0 be a positive integer. Return a number field element b such that x bⁿ = v is small. If x is integral, then v is also integral.

More precisely, idealnumden reduces the problem to x integral. Then, factoring out the prime ideals dividing a rational prime p ≤ B, we rewrite x = I Jⁿ where the ideals I and J are both integral and I is B-smooth. Then we return a small element b in J^-1.

The bound B avoids a costly complete factorization of x; as soon as the n-core of x is B-smooth (i.e., as soon as I is n-power free), then J is as large as possible and so is the expected reduction.

  ? T = x^6+108; nf = nfinit(T); a = Mod(x,T);
  ? setrand(1); u = (2*a^2+a+3)*random(2^1000*x^6)^6;
  ? sizebyte(u)
  %3 = 4864
  ? b = idealredmodpower(nf,u,2);
  ? v2 = nfeltmul(nf,u, nfeltpow(nf,b,2))
  %5 = [34, 47, 15, 35, 9, 3]~
  ? b = idealredmodpower(nf,u,6);
  ? v6 = nfeltmul(nf,u, nfeltpow(nf,b,6))
  %7 = [3, 0, 2, 6, -7, 1]~

The last element v6, obtained by reducing modulo 6-th powers instead of squares, looks smaller than v2 but its norm is actually a little larger:

  ? idealnorm(nf,v2)
  %8 = 81309
  ? idealnorm(nf,v6)
  %9 = 731781

The library syntax is GEN idealredmodpower(GEN nf, GEN x, ulong n, ulong B).

Outputs a bid structure, necessary for computing in the finite abelian group G = (ℤ_K/N)^*. Here, nf is a number field and N is a modulus: either an ideal in any form, or a row vector whose first component is an ideal and whose second component is a row vector of r₁ 0 or 1. Ideals can also be given by a factorization into prime ideals, as produced by idealfactor.

If the positive integer cycmod is present, only compute the group modulo cycmod-th powers, which may save a lot of time when some maximal ideals in the modulus have a huge residue field. Whereas you might only be interested in quadratic or cubic residuosity; see also bnrinit for applications in class field theory.

This bid is used in ideallog to compute discrete logarithms. It also contains useful information which can be conveniently retrieved as bid.mod (the modulus), bid.clgp (G as a finite abelian group), bid.no (the cardinality of G), bid.cyc (elementary divisors) and bid.gen (generators).

If flag = 1 (default), the result is a bid structure without generators: they are well defined but not explicitly computed, which saves time.

If flag = 2, as flag = 1, but including generators.

If flag = 0, only outputs (ℤ_K/N)^* as an abelian group, i.e as a 3-component vector [h,d,g]: h is the order, d is the vector of SNF cyclic components and g the corresponding generators.

If nf is omitted, we take it to be the rational number fields, N must be an integer and we return the structure of (ℤ/Nℤ)^*. In other words idealstar(, N, flag) is short for

    idealstar(nfinit(x), N, flag)

but faster. The alternative syntax znstar(N, flag) is also available for an analogous effect but, due to an unfortunate historical oversight, the default value of flag is different in the two functions (znstar does not initialize by default, you probably want znstar(N,1)).

The library syntax is GEN idealstarmod(GEN nf = NULL, GEN N, long flag, GEN cycmod = NULL). Instead the above hardcoded numerical flags, one should rather use GEN Idealstarmod(GEN nf, GEN ideal, long flag, GEN cycmod) or GEN Idealstar(GEN nf, GEN ideal, long flag) (cycmod is NULL), where flag is an or-ed combination of nf_GEN (include generators) and nf_INIT (return a full bid, not a group), possibly 0. This offers one more combination: gen, but no init. The nf argument must be a true nf structure.

Computes a two-element representation of the ideal x in the number field nf, combining a random search and an approximation theorem; x is an ideal in any form (possibly an extended ideal, whose principal part is ignored)

* When called as idealtwoelt(nf,x), the result is a row vector [a,α] with two components such that x = aℤ_K+αℤ_K and a is chosen to be the positive generator of x∩ℤ, unless x was given as a principal ideal in which case we may choose a = 0. The algorithm uses a fast lazy factorization of x∩ ℤ and runs in randomized polynomial time.

  ? K = nfinit(t^5-23);
  ? x = idealhnf(K, t^2*(t+1), t^3*(t+1))
  %2 =  \\ some random ideal of norm 552*23
  [552 23 23 529 23]
  
  [  0 23  0   0  0]
  
  [  0  0  1   0  0]
  
  [  0  0  0   1  0]
  
  [  0  0  0   0  1]
  
  ? [a,alpha] = idealtwoelt(K, x)
  %3 = [552, [23, 0, 1, 0, 0]~]
  ? nfbasistoalg(K, alpha)
  %4 = Mod(t^2 + 23, t^5 - 23)

* When called as idealtwoelt(nf,x,a) with an explicit nonzero a supplied as third argument, the function assumes that a ∈ x and returns α ∈ x such that x = aℤ_K + αℤ_K. Note that we must factor a in this case, and the algorithm is generally slower than the default variant and gives larger generators:

  ? alpha2 = idealtwoelt(K, x, 552)
  %5 = [-161, -161, -183, -207, 0]~
  ? idealhnf(K, 552, alpha2) == x
  %6 = 1

Note that, in both cases, the return value is not recognized as an ideal by GP functions; one must use idealhnf as above to recover a valid ideal structure from the two-element representation.

The library syntax is GEN idealtwoelt0(GEN nf, GEN x, GEN a = NULL). Also available are GEN idealtwoelt(GEN nf, GEN x) and GEN idealtwoelt2(GEN nf, GEN x, GEN a).

Gives the valuation of the ideal x at the prime ideal pr in the number field nf, where pr is in idealprimedec format. The valuation of the 0 ideal is +oo.

The library syntax is GEN gpidealval(GEN nf, GEN x, GEN pr). Also available is long idealval(GEN nf, GEN x, GEN pr), which returns LONG_MAX if x = 0 and the valuation as a long integer.

This function is deprecated, use apply.

nf being a number field in nfinit format, and x a (row or column) vector or matrix, apply nfalgtobasis to each entry of x.

The library syntax is GEN matalgtobasis(GEN nf, GEN x).

This function is deprecated, use apply.

nf being a number field in nfinit format, and x a (row or column) vector or matrix, apply nfbasistoalg to each entry of x.

The library syntax is GEN matbasistoalg(GEN nf, GEN x).

Let z = Mod(A, T) be a polmod, and Q be its minimal polynomial, which must satisfy deg(Q) = deg(T). Returns a "reverse polmod" Mod(B, Q), which is a root of T.

This is quite useful when one changes the generating element in algebraic extensions:

  ? u = Mod(x, x^3 - x -1); v = u^5;
  ? w = modreverse(v)
  %2 = Mod(x^2 - 4*x + 1, x^3 - 5*x^2 + 4*x - 1)

which means that x³ - 5x² + 4x -1 is another defining polynomial for the cubic field ℚ(u) = ℚ[x]/(x³ - x - 1) = ℚ[x]/(x³ - 5x² + 4x - 1) = ℚ(v), and that u → v² - 4v + 1 gives an explicit isomorphism. From this, it is easy to convert elements between the A(u) ∈ ℚ(u) and B(v) ∈ ℚ(v) representations:

  ? A = u^2 + 2*u + 3; subst(lift(A), 'x, w)
  %3 = Mod(x^2 - 3*x + 3, x^3 - 5*x^2 + 4*x - 1)
  ? B = v^2 + v + 1;   subst(lift(B), 'x, v)
  %4 = Mod(26*x^2 + 31*x + 26, x^3 - x - 1)

If the minimal polynomial of z has lower degree than expected, the routine fails

  ? u = Mod(-x^3 + 9*x, x^4 - 10*x^2 + 1)
  ? modreverse(u)
   *** modreverse: domain error in modreverse: deg(minpoly(z)) < 4
   ***   Break loop: type 'break' to go back to GP prompt
  break> Vec( dbg_err() ) \\ ask for more info
  ["e_DOMAIN", "modreverse", "deg(minpoly(z))", "<", 4,
    Mod(-x^3 + 9*x, x^4 - 10*x^2 + 1)]
  break> minpoly(u)
  x^2 - 8

The library syntax is GEN modreverse(GEN z).

Gives the vector of the slopes of the Newton polygon of the polynomial x with respect to the prime number p. The n components of the vector are in decreasing order, where n is equal to the degree of x. Vertical slopes occur iff the constant coefficient of x is zero and are denoted by +oo.

The library syntax is GEN newtonpoly(GEN x, GEN p).

Given an algebraic number x in the number field nf, transforms it to a column vector on the integral basis nf.zk.

  ? nf = nfinit(y^2 + 4);
  ? nf.zk
  %2 = [1, 1/2*y]
  ? nfalgtobasis(nf, [1,1]~)
  %3 = [1, 1]~
  ? nfalgtobasis(nf, y)
  %4 = [0, 2]~
  ? nfalgtobasis(nf, Mod(y, y^2+4))
  %5 = [0, 2]~

This is the inverse function of nfbasistoalg.

The library syntax is GEN algtobasis(GEN nf, GEN x).

Let T(X) be an irreducible polynomial with integral coefficients. This function returns an integral basis of the number field defined by T, that is a ℤ-basis of its maximal order. If present, dK is set to the discriminant of the returned order. The basis elements are given as elements in K = ℚ[X]/(T), in Hermite normal form with respect to the ℚ-basis (1,X,...,X^{deg T-1}) of K, lifted to ℚ[X]. In particular its first element is always 1 and its i-th element is a polynomial of degree i-1 whose leading coefficient is the inverse of an integer: the product of those integers is the index of ℤ[X]/(T) in the maximal order ℤ_K:

  ? nfbasis(x^2 + 4) \\ Z[X]/(T) has index 2 in ZK
  %1 = [1, x/2]
  ? nfbasis(x^2 + 4, &D)
  %2 = [1, x/2]
  ? D
  %3 = -4

This function uses a modified version of the round 4 algorithm, due to David Ford, Sebastian Pauli and Xavier Roblot.

Local basis, orders maximal at certain primes.

Obtaining the maximal order is hard: it requires factoring the discriminant D of T. Obtaining an order which is maximal at a finite explicit set of primes is easy, but it may then be a strict suborder of the maximal order. To specify that we are interested in a given set of places only, we can replace the argument T by an argument [T,listP], where listP encodes the primes we are interested in: it must be a factorization matrix, a vector of integers or a single integer.

* Vector: we assume that it contains distinct prime numbers.

* Matrix: we assume that it is a two-column matrix of a (partial) factorization of D; namely the first column contains distinct primes and the second one the valuation of D at each of these primes.

* Integer B: this is replaced by the vector of primes up to B. Note that the function will use at least O(B) time: a small value, about 10⁵, should be enough for most applications. Values larger than 2³² are not supported.

In all these cases, the primes may or may not divide the discriminant D of T. The function then returns a ℤ-basis of an order whose index is not divisible by any of these prime numbers. The result may actually be a global integral basis, in particular if all the prime divisors of the field discriminant are included, but this is not guaranteed! Note that nfinit has built-in support for such a check:

  ? K = nfinit([T, listP]);
  ? nfcertify(K)   \\ we computed an actual maximal order
  %2 = [];

The first line initializes a number field structure incorporating nfbasis([T, listP] in place of a proven integral basis. The second line certifies that the resulting structure is correct. This allows to create an nf structure attached to the number field K = ℚ[X]/(T), when the discriminant of T cannot be factored completely, whereas the prime divisors of disc K are known. If present, the argument dK is set to the discriminant of the returned order, and is equal to the field discriminant if and only if the order is maximal.

Of course, if listP contains a single prime number p, the function returns a local integral basis for ℤ_p[X]/(T):

  ? nfbasis(x^2+x-1001)
  %1 = [1, 1/3*x - 1/3]
  ? nfbasis( [x^2+x-1001, [2]] )
  %2 = [1, x]

The following function computes the index i_T of ℤ[X]/(T) in the order generated by the ℤ-basis B:

  nfbasisindex(T, B) = vecprod([denominator(pollead(Q)) | Q <- B]);

In particular, B is a basis of the maximal order if and only if poldisc(T) / i_T² is equal to the field discriminant. More generally, this formula gives the square of index of the order given by B in ℤ_K. For instance, assume that P is a vector of prime numbers containing (at least) all prime divisors of the field discriminant, then the following construct allows to provably compute the field discriminant and to check whether the returned basis is actually a basis of the maximal order

  ? B = nfbasis([T, P], &D);
  ? dK = sign(D) * vecprod([p^valuation(D,p) | p<-P]);
  ? dK * nfbasisindex(T, B)^2 == poldisc(T)

The variable dK contains the field discriminant and the last command returns 1 if and only if B is a ℤ-basis of the maximal order. Of course, the nfinit / nfcertify approach is simpler, but it is also more costly.

The Buchmann-Lenstra algorithm.

We now complicate the picture: it is in fact allowed to include composite numbers instead of primes in listP (Vector or Matrix case), provided they are pairwise coprime. The result may still be a correct integral basis if the field discriminant factors completely over the actual primes in the list; again, this is not guaranteed. Adding a composite C such that C² divides D may help because when we consider C as a prime and run the algorithm, two good things can happen: either we succeed in proving that no prime dividing C can divide the index (without actually needing to find those primes), or the computation exhibits a nontrivial zero divisor, thereby factoring C and we go on with the refined factorization. (Note that including a C such that C² does not divide D is useless.) If neither happen, then the computed basis need not generate the maximal order. Here is an example:

  ? B = 10^5;
  ? listP = factor(poldisc(T), B); \\ primes <= B dividing D + cofactor
  ? basis = nfbasis([T, listP], &D)

If the computed discriminant D factors completely over the primes less than B (together with the primes contained in the addprimes table), then everything is certified: D is the field discriminant and basis generates the maximal order. This can be tested as follows:

    F = factor(D, B); P = F[,1]; E = F[,2];
    for (i = 1, #P,
      if (P[i] > B && !isprime(P[i]), warning("nf may be incorrect")));

This is a sufficient but not a necessary condition, hence the warning, instead of an error.

The function nfcertify speeds up and automates the above process:

  ? B = 10^5;
  ? nf = nfinit([T, B]);
  ? nfcertify(nf)
  %3 = []      \\ nf is unconditionally correct
  ? [basis, disc] = [nf.zk, nf.disc];

The library syntax is GEN nfbasis(GEN T, GEN *dK = NULL).

Given an algebraic number x in the number field nf, transforms it into t_POLMOD form.

  ? nf = nfinit(y^2 + 4);
  ? nf.zk
  %2 = [1, 1/2*y]
  ? nfbasistoalg(nf, [1,1]~)
  %3 = Mod(1/2*y + 1, y^2 + 4)
  ? nfbasistoalg(nf, y)
  %4 = Mod(y, y^2 + 4)
  ? nfbasistoalg(nf, Mod(y, y^2+4))
  %5 = Mod(y, y^2 + 4)

This is the inverse function of nfalgtobasis.

The library syntax is GEN basistoalg(GEN nf, GEN x).

nf being as output by nfinit, checks whether the integer basis is known unconditionally. This is in particular useful when the argument to nfinit was of the form [T, listP], specifying a finite list of primes when p-maximality had to be proven, or a list of coprime integers to which Buchmann-Lenstra algorithm was to be applied.

The function returns a vector of coprime composite integers. If this vector is empty, then nf.zk and nf.disc are correct. Otherwise, the result is dubious. In order to obtain a certified result, one must completely factor each of the given integers, then addprime each of their prime factors, then check whether nfdisc(nf.pol) is equal to nf.disc.

The library syntax is GEN nfcertify(GEN nf).

Let nf be a number field structure attached to the field K and let P and Q be squarefree polynomials in K[X] in the same variable. Outputs the simple factors of the étale K-algebra A = K[X, Y] / (P(X), Q(Y)). The factors are given by a list of polynomials R in K[X], attached to the number field K[X]/ (R), and sorted by increasing degree (with respect to lexicographic ordering for factors of equal degrees). Returns an error if one of the polynomials is not squarefree.

Note that it is more efficient to reduce to the case where P and Q are irreducible first. The routine will not perform this for you, since it may be expensive, and the inputs are irreducible in most applications anyway. In this case, there will be a single factor R if and only if the number fields defined by P and Q are linearly disjoint (their intersection is K).

The binary digits of flag mean

1: outputs a vector of 4-component vectors [R,a,b,k], where R ranges through the list of all possible compositums as above, and a (resp. b) expresses the root of P (resp. Q) as an element of K[X]/(R). Finally, k is a small integer such that b + ka = X modulo R.

2: assume that P and Q define number fields that are linearly disjoint: both polynomials are irreducible and the corresponding number fields have no common subfield besides K. This allows to save a costly factorization over K. In this case return the single simple factor instead of a vector with one element.

A compositum is often defined by a complicated polynomial, which it is advisable to reduce before further work. Here is an example involving the field K(ζ₅, 5^1/10), K = ℚ(sqrt{5}):

  ? K = nfinit(y^2-5);
  ? L = nfcompositum(K, x^5 - y, polcyclo(5), 1); \\  list of [R,a,b,k]
  ? [R, a] = L[1];  \\  pick the single factor, extract R,a (ignore b,k)
  ? lift(R)         \\  defines the compositum
  %4 = x^10 + (-5/2*y + 5/2)*x^9 + (-5*y + 20)*x^8 + (-20*y + 30)*x^7 + \
  (-45/2*y + 145/2)*x^6 + (-71/2*y + 121/2)*x^5 + (-20*y + 60)*x^4 +    \
  (-25*y + 5)*x^3 + 45*x^2 + (-5*y + 15)*x + (-2*y + 6)
  ? a^5 - y         \\  a fifth root of y
  %5 = 0
  ? [T, X] = rnfpolredbest(K, R, 1);
  ? lift(T)     \\  simpler defining polynomial for K[x]/(R)
  %7 = x^10 + (-11/2*y + 25/2)
  ? liftall(X)  \\   root of R in K[x]/(T(x))
  %8 = (3/4*y + 7/4)*x^7 + (-1/2*y - 1)*x^5 + 1/2*x^2 + (1/4*y - 1/4)
  ? a = subst(a.pol, 'x, X);  \\  a in the new coordinates
  ? liftall(a)
  %10 = (-3/4*y - 7/4)*x^7 - 1/2*x^2
  ? a^5 - y
  %11 = 0

The main variables of P and Q must be the same and have higher priority than that of nf (see varhigher and varlower).

The library syntax is GEN nfcompositum(GEN nf, GEN P, GEN Q, long flag).

Given a pseudo-matrix x, computes a nonzero ideal contained in (i.e. multiple of) the determinant of x. This is particularly useful in conjunction with nfhnfmod.

The library syntax is GEN nfdetint(GEN nf, GEN x).

field discriminant of the number field defined by the integral, preferably monic, irreducible polynomial T(X). Returns the discriminant of the number field ℚ[X]/(T), using the Round 4 algorithm.

Local discriminants, valuations at certain primes.

As in nfbasis, the argument T can be replaced by [T,listP], where listP is as in nfbasis: a vector of pairwise coprime integers (usually distinct primes), a factorization matrix, or a single integer. In that case, the function returns the discriminant of an order whose basis is given by nfbasis(T,listP), which need not be the maximal order, and whose valuation at a prime entry in listP is the same as the valuation of the field discriminant.

In particular, if listP is [p] for a prime p, we can return the p-adic discriminant of the maximal order of ℤ_p[X]/(T), as a power of p, as follows:

  ? padicdisc(T,p) = p^valuation(nfdisc([T,[p]]), p);
  ? nfdisc(x^2 + 6)
  %2 = -24
  ? padicdisc(x^2 + 6, 2)
  %3 = 8
  ? padicdisc(x^2 + 6, 3)
  %4 = 3

The following function computes the discriminant of the maximal order under the assumption that P is a vector of prime numbers containing (at least) all prime divisors of the field discriminant:

  globaldisc(T, P) =
  { my (D = nfdisc([T, P]));
    sign(D) * vecprod([p^valuation(D,p) | p <-P]);
  }
  ? globaldisc(x^2 + 6, [2, 3, 5])
  %1 = -24

The library syntax is nfdisc(GEN T). Also available is GEN nfbasis(GEN T, GEN *d), which returns the order basis, and where *d receives the order discriminant.

Given a polynomial T with integer coefficients, return [D, faD] where D is nfdisc(T) and faD is the factorization of |D|. All the variants [T,listP] are allowed (see ??nfdisc), in which case faD is the factorization of the discriminant underlying order (which need not be maximal at the primes not specified by listP) and the factorization may contain large composites.

  ? T = x^3 - 6021021*x^2 + 12072210077769*x - 8092423140177664432;
  ? [D,faD] = nfdiscfactors(T); print(faD); D
  [3, 3; 500009, 2]
  %2 = -6750243002187]
  
  ? T = x^3 + 9*x^2 + 27*x - 125014250689643346789780229390526092263790263725;
  ? [D,faD] = nfdiscfactors(T); print(faD); D
  [3, 3; 1000003, 2]
  %4 = -27000162000243
  
  ? [D,faD] = nfdiscfactors([T, 10^3]); print(faD)
  [3, 3; 125007125141751093502187, 2]

In the final example, we only get a partial factorization, which is only guaranteed correct at primes ≤ 10³.

The function also accept number field structures, for instance as output by nfinit, and returns the field discriminant and its factorization:

  ? T = x^3 + 9*x^2 + 27*x - 125014250689643346789780229390526092263790263725;
  ? nf = nfinit(T); [D,faD] = nfdiscfactors(T); print(faD); D
  %2 = -27000162000243
  ? nf.disc
  %3 = -27000162000243

The library syntax is GEN nfdiscfactors(GEN T).

Given two elements x and y in nf, computes their sum x+y in the number field nf.

  ? nf = nfinit(1+x^2);
  ? nfeltadd(nf, 1, x) \\ 1 + I
  %2 = [1, 1]~

The library syntax is GEN nfadd(GEN nf, GEN x, GEN y).

Given two elements x and y in nf, computes their quotient x/y in the number field nf.

The library syntax is GEN nfdiv(GEN nf, GEN x, GEN y).

Given two elements x and y in nf, computes an algebraic integer q in the number field nf such that the components of x-qy are reasonably small. In fact, this is functionally identical to round(nfdiv(nf,x,y)).

The library syntax is GEN nfdiveuc(GEN nf, GEN x, GEN y).

This function is obsolete, use nfmodpr.

Given two elements x and y in nf and pr a prime ideal in modpr format (see nfmodprinit), computes their quotient x / y modulo the prime ideal pr.

The library syntax is GEN nfdivmodpr(GEN nf, GEN x, GEN y, GEN pr). This function is normally useless in library mode. Project your inputs to the residue field using nf_to_Fq, then work there.

Given two elements x and y in nf, gives a two-element row vector [q,r] such that x = qy+r, q is an algebraic integer in nf, and the components of r are reasonably small.

The library syntax is GEN nfdivrem(GEN nf, GEN x, GEN y).

Given an element x in the number field nf, return the (real or) complex embeddings of x specified by optional argument pl, at the current realprecision:

* pl omitted: return the vector of embeddings at all r₁+r₂ places;

* pl an integer between 1 and r₁+r₂: return the i-th embedding of x, attached to the i-th root of nf.pol, i.e. nf.roots[i];

* pl a vector or t_VECSMALL: return the vector of embeddings; the i-th entry gives the embedding at the place attached to the pl[i]-th real root of nf.pol.

  ? nf = nfinit('y^3 - 2);
  ? nf.sign
  %2 = [1, 1]
  ? nfeltembed(nf, 'y)
  %3 = [1.25992[...], -0.62996[...] + 1.09112[...]*I]]
  ? nfeltembed(nf, 'y, 1)
  %4 = 1.25992[...]
  ? nfeltembed(nf, 'y, 3) \\ there are only 2 arch. places
   ***   at top-level: nfeltembed(nf,'y,3)
   ***                 ^ —  —  —  —  — --
   *** nfeltembed: domain error in nfeltembed: index > 2

The library syntax is GEN nfeltembed(GEN nf, GEN x, GEN pl = NULL, long prec).

Returns 1 if x is an n-th power in the number field nf (and sets y to an n-th root if the argument is present), else returns 0.

  ? nf = nfinit(1+x^2);
  ? nfeltispower(nf, -4, 4, &y)
  %2 = 1
  ? y
  %3 = [-1, -1]~

The library syntax is long nfispower(GEN nf, GEN x, long n, GEN *y = NULL).

Returns 1 if x is a square in nf (and sets y to a square root if the argument is present), else returns 0.

  ? nf = nfinit(1+x^2);
  ? nfeltissquare(nf, -1, &y)
  %2 = 1
  ? y
  %3 = [0, -1]~

The library syntax is long nfissquare(GEN nf, GEN x, GEN *y = NULL).

Given two elements x and y in nf, computes an element r of nf of the form r = x-qy with q and algebraic integer, and such that r is small. This is functionally identical to x - nfmul(nf,round(nfdiv(nf,x,y)),y).

The library syntax is GEN nfmod(GEN nf, GEN x, GEN y).

Given two elements x and y in nf, computes their product x*y in the number field nf.

The library syntax is GEN nfmul(GEN nf, GEN x, GEN y).

This function is obsolete, use nfmodpr.

Given two elements x and y in nf and pr a prime ideal in modpr format (see nfmodprinit), computes their product x*y modulo the prime ideal pr.

The library syntax is GEN nfmulmodpr(GEN nf, GEN x, GEN y, GEN pr). This function is normally useless in library mode. Project your inputs to the residue field using nf_to_Fq, then work there.

Returns the absolute norm of x.

The library syntax is GEN nfnorm(GEN nf, GEN x).

Given an element x in nf, and a positive or negative integer k, computes x^k in the number field nf.

The library syntax is GEN nfpow(GEN nf, GEN x, GEN k). GEN nfinv(GEN nf, GEN x) correspond to k = -1, and GEN nfsqr(GEN nf,GEN x) to k = 2.

This function is obsolete, use nfmodpr.

Given an element x in nf, an integer k and a prime ideal pr in modpr format (see nfmodprinit), computes x^k modulo the prime ideal pr.

The library syntax is GEN nfpowmodpr(GEN nf, GEN x, GEN k, GEN pr). This function is normally useless in library mode. Project your inputs to the residue field using nf_to_Fq, then work there.

Given an ideal id in Hermite normal form and an element a of the number field nf, finds an element r in nf such that a-r belongs to the ideal and r is small.

The library syntax is GEN nfreduce(GEN nf, GEN a, GEN id).

This function is obsolete, use nfmodpr.

Given an element x of the number field nf and a prime ideal pr in modpr format compute a canonical representative for the class of x modulo pr.

The library syntax is GEN nfreducemodpr(GEN nf, GEN x, GEN pr). This function is normally useless in library mode. Project your inputs to the residue field using nf_to_Fq, then work there.

Given an element x in the number field nf, returns the signs of the real embeddings of x specified by optional argument pl:

* pl omitted: return the vector of signs at all r₁ real places;

* pl an integer between 1 and r₁: return the sign of the i-th embedding of x, attached to the i-th real root of nf.pol, i.e. nf.roots[i];

* pl a vector or t_VECSMALL: return the vector of signs; the i-th entry gives the sign at the real place attached to the pl[i]-th real root of nf.pol.

  ? nf = nfinit(polsubcyclo(11,5,'y)); \\ Q(cos(2 pi/11))
  ? nf.sign
  %2 = [5, 0]
  ? x = Mod('y, nf.pol);
  ? nfeltsign(nf, x)
  %4 = [-1, -1, -1, 1, 1]
  ? nfeltsign(nf, x, 1)
  %5 = -1
  ? nfeltsign(nf, x, [1..4])
  %6 = [-1, -1, -1, 1]
  ? nfeltsign(nf, x, 6) \\ there are only 5 real embeddings
   ***   at top-level: nfeltsign(nf,x,6)
   ***                 ^ —  —  —  —  — --
   *** nfeltsign: domain error in nfeltsign: index > 5

The library syntax is GEN nfeltsign(GEN nf, GEN x, GEN pl = NULL).

Returns the absolute trace of x.

The library syntax is GEN nftrace(GEN nf, GEN x).

Given an element x in nf and a prime ideal pr in the format output by idealprimedec, computes the valuation v at pr of the element x. The valuation of 0 is +oo.

  ? nf = nfinit(x^2 + 1);
  ? P = idealprimedec(nf, 2)[1];
  ? nfeltval(nf, x+1, P)
  %3 = 1

This particular valuation can also be obtained using idealval(nf,x,pr), since x is then converted to a principal ideal.

If the y argument is present, sets y = x τ^v, where τ is a fixed "anti-uniformizer" for pr: its valuation at pr is -1; its valuation is 0 at other prime ideals dividing pr.p and nonnegative at all other primes. In other words y is the part of x coprime to pr. If x is an algebraic integer, so is y.

  ? nfeltval(nf, x+1, P, &y); y
  %4 = [0, 1]~

For instance if x = ∏_i x_i^e_i is known to be coprime to pr, where the x_i are algebraic integers and e_i ∈ ℤ then, if v_i = nfeltval(nf, x_i, pr, &y_i), we still have x = ∏_i y_i^e_i, where the y_i are still algebraic integers but now all of them are coprime to pr. They can then be mapped to the residue field of pr more efficiently than if the product had been expanded beforehand: we can reduce mod pr after each ring operation.

The library syntax is GEN gpnfvalrem(GEN nf, GEN x, GEN pr, GEN *y = NULL). Also available are long nfvalrem(GEN nf, GEN x, GEN pr, GEN *y = NULL), which returns LONG_MAX if x = 0 and the valuation as a long integer, and long nfval(GEN nf, GEN x, GEN pr), which only returns the valuation (y = NULL).

Factorization of the univariate polynomial (or rational function) T over the number field nf given by nfinit; T has coefficients in nf (i.e. either scalar, polmod, polynomial or column vector). The factors are sorted by increasing degree.

The main variable of nf must be of lower priority than that of T, see Section se:priority. However if the polynomial defining the number field occurs explicitly in the coefficients of T as modulus of a t_POLMOD or as a t_POL coefficient, its main variable must be the same as the main variable of T. For example,

  ? nf = nfinit(y^2 + 1);
  ? nffactor(nf, x^2 + y); \\  OK
  ? nffactor(nf, x^2 + Mod(y, y^2+1)); \\   OK
  ? nffactor(nf, x^2 + Mod(z, z^2+1)); \\   WRONG

It is possible to input a defining polynomial for nf instead, but this is in general less efficient since parts of an nf structure will then be computed internally. This is useful in two situations: when you do not need the nf elsewhere, or when you cannot initialize an nf due to integer factorization difficulties when attempting to compute the field discriminant and maximal order. In all cases, the function runs in polynomial time using Belabas's variant of van Hoeij's algorithm, which copes with hundreds of modular factors.

Caveat. nfinit([T, listP]) allows to compute in polynomial time a conditional nf structure, which sets nf.zk to an order which is not guaranteed to be maximal at all primes. Always either use nfcertify first (which may not run in polynomial time) or make sure to input nf.pol instead of the conditional nf: nffactor is able to recover in polynomial time in this case, instead of potentially missing a factor.

The library syntax is GEN nffactor(GEN nf, GEN T).

Gives back the nf element corresponding to a factorization. The integer 1 corresponds to the empty factorization.

If e is present, e and f must be vectors of the same length (e being integral), and the corresponding factorization is the product of the f[i]^e[i].

If not, and f is vector, it is understood as in the preceding case with e a vector of 1s: we return the product of the f[i]. Finally, f can be a regular factorization matrix.

  ? nf = nfinit(y^2+1);
  ? nffactorback(nf, [3, y+1, [1,2]~], [1, 2, 3])
  %2 = [12, -66]~
  ? 3 * (I+1)^2 * (1+2*I)^3
  %3 = 12 - 66*I

The library syntax is GEN nffactorback(GEN nf, GEN f, GEN e = NULL).

This routine is obsolete, use nfmodpr and factormod.

Factors the univariate polynomial Q modulo the prime ideal pr in the number field nf. The coefficients of Q belong to the number field (scalar, polmod, polynomial, even column vector) and the main variable of nf must be of lower priority than that of Q (see Section se:priority). The prime ideal pr is either in idealprimedec or (preferred) modprinit format. The coefficients of the polynomial factors are lifted to elements of nf:

  ? K = nfinit(y^2+1);
  ? P = idealprimedec(K, 3)[1];
  ? nffactormod(K, x^2 + y*x + 18*y+1, P)
  %3 =
  [x + (2*y + 1) 1]
  
  [x + (2*y + 2) 1]
  ? P = nfmodprinit(K, P);  \\ convert to nfmodprinit format
  ? nffactormod(K, x^2 + y*x + 18*y+1)
  %5 =
  [x + (2*y + 1) 1]
  
  [x + (2*y + 2) 1]

Same result, of course, here about 10% faster due to the precomputation.

The library syntax is GEN nffactormod(GEN nf, GEN Q, GEN pr).

Let nf be a number field as output by nfinit, and let aut be a Galois automorphism of nf expressed by its image on the field generator (such automorphisms can be found using nfgaloisconj). The function computes the action of the automorphism aut on the object x in the number field; x can be a number field element, or an ideal (possibly extended). Because of possible confusion with elements and ideals, other vector or matrix arguments are forbidden.

   ? nf = nfinit(x^2+1);
   ? L = nfgaloisconj(nf)
   %2 = [-x, x]~
   ? aut = L[1]; /* the nontrivial automorphism */
   ? nfgaloisapply(nf, aut, x)
   %4 = Mod(-x, x^2 + 1)
   ? P = idealprimedec(nf,5); /* prime ideals above 5 */
   ? nfgaloisapply(nf, aut, P[2]) == P[1]
   %6 = 0 \\ !!!!
   ? idealval(nf, nfgaloisapply(nf, aut, P[2]), P[1])
   %7 = 1

The surprising failure of the equality test (%7) is due to the fact that although the corresponding prime ideals are equal, their representations are not. (A prime ideal is specified by a uniformizer, and there is no guarantee that applying automorphisms yields the same elements as a direct idealprimedec call.)

The automorphism can also be given as a column vector, representing the image of Mod(x, nf.pol) as an algebraic number. This last representation is more efficient and should be preferred if a given automorphism must be used in many such calls.

   ? nf = nfinit(x^3 - 37*x^2 + 74*x - 37);
   ? aut = nfgaloisconj(nf)[2]; \\   an automorphism in basistoalg form
   %2 = -31/11*x^2 + 1109/11*x - 925/11
   ? AUT = nfalgtobasis(nf, aut); \\   same in algtobasis form
   %3 = [16, -6, 5]~
   ? v = [1, 2, 3]~; nfgaloisapply(nf, aut, v) == nfgaloisapply(nf, AUT, v)
   %4 = 1 \\   same result...
   ? for (i=1,10^5, nfgaloisapply(nf, aut, v))
   time = 463 ms.
   ? for (i=1,10^5, nfgaloisapply(nf, AUT, v))
   time = 343 ms.  \\   but the latter is faster

The library syntax is GEN galoisapply(GEN nf, GEN aut, GEN x).

nf being a number field as output by nfinit, computes the conjugates of a root r of the nonconstant polynomial x = nf[1] expressed as polynomials in r. This also makes sense when the number field is not Galois since some conjugates may lie in the field. nf can simply be a polynomial.

If no flags or flag = 0, use a combination of flag 4 and 1 and the result is always complete. There is no point whatsoever in using the other flags.

If flag = 1, use nfroots: a little slow, but guaranteed to work in polynomial time.

If flag = 4, use galoisinit: very fast, but only applies to (most) Galois fields. If the field is Galois with weakly super-solvable Galois group (see galoisinit), return the complete list of automorphisms, else only the identity element. If present, d is assumed to be a multiple of the least common denominator of the conjugates expressed as polynomial in a root of pol.

This routine can only compute ℚ-automorphisms, but it may be used to get K-automorphism for any base field K as follows:

  rnfgaloisconj(nfK, R) = \\ K-automorphisms of L = K[X] / (R)
  {
    my(polabs, N,al,S, ala,k, vR);
    R *= Mod(1, nfK.pol); \\ convert coeffs to polmod elts of K
    vR = variable(R);
    al = Mod(variable(nfK.pol),nfK.pol);
    [polabs,ala,k] = rnfequation(nfK, R, 1);
    Rt = if(k==0,R,subst(R,vR,vR-al*k));
    N = nfgaloisconj(polabs) % Rt; \\ Q-automorphisms of L
    S = select(s->subst(Rt, vR, Mod(s,Rt)) == 0, N);
    if (k==0, S, apply(s->subst(s,vR,vR+k*al)-k*al,S));
  }
  K  = nfinit(y^2 + 7);
  rnfgaloisconj(K, x^4 - y*x^3 - 3*x^2 + y*x + 1)  \\ K-automorphisms of L

The library syntax is GEN galoisconj0(GEN nf, long flag, GEN d = NULL, long prec). Use directly GEN galoisconj(GEN nf, GEN d), corresponding to flag = 0, the others only have historical interest.

Given nf a number field in nf or bnf format, a t_VEC Lpr of primes of nf and a t_VEC Ld of positive integers of the same length, a t_VECSMALL pl of length r₁ the number of real places of nf, computes a polynomial with coefficients in nf defining a cyclic extension of nf of minimal degree satisfying certain local conditions:

* at the prime Lpr[i], the extension has local degree a multiple of Ld[i];

* at the i-th real place of nf, it is complex if pl[i] = -1 (no condition if pl[i] = 0).

The extension has degree the LCM of the local degrees. Currently, the degree is restricted to be a prime power for the search, and to be prime for the construction because of the rnfkummer restrictions.

When nf is ℚ, prime integers are accepted instead of prid structures. However, their primality is not checked and the behavior is undefined if you provide a composite number.

Warning. If the number field nf does not contain the n-th roots of unity where n is the degree of the extension to be computed, the function triggers the computation of the bnf of nf(ζ_n), which may be costly.

  ? nf = nfinit(y^2-5);
  ? pr = idealprimedec(nf,13)[1];
  ? pol = nfgrunwaldwang(nf, [pr], [2], [0,-1], 'x)
  %3 = x^2 + Mod(3/2*y + 13/2, y^2 - 5)

The library syntax is GEN nfgrunwaldwang(GEN nf, GEN Lpr, GEN Ld, GEN pl, long v = -1) where v is a variable number.

If pr is omitted, compute the global quadratic Hilbert symbol (a,b) in nf, that is 1 if x² - a y² - b z² has a non trivial solution (x,y,z) in nf, and -1 otherwise. Otherwise compute the local symbol modulo the prime ideal pr, as output by idealprimedec.

The library syntax is long nfhilbert0(GEN nf, GEN a, GEN b, GEN pr = NULL).

Also available is long nfhilbert(GEN nf,GEN a,GEN b) (global quadratic Hilbert symbol), where nf is a true nf structure.

Given a pseudo-matrix (A,I), finds a pseudo-basis (B,J) in Hermite normal form of the module it generates. If flag is nonzero, also return the transformation matrix U such that AU = [0|B].

The library syntax is GEN nfhnf0(GEN nf, GEN x, long flag). Also available:

GEN nfhnf(GEN nf, GEN x) (flag = 0).

GEN rnfsimplifybasis(GEN bnf, GEN x) simplifies the pseudo-basis x = (A,I), returning a pseudo-basis (B,J). The ideals in the list J are integral, primitive and either trivial (equal to the full ring of integer) or nonprincipal.

Given a pseudo-matrix (A,I) and an ideal detx which is contained in (read integral multiple of) the determinant of (A,I), finds a pseudo-basis in Hermite normal form of the module generated by (A,I). This avoids coefficient explosion. detx can be computed using the function nfdetint.

The library syntax is GEN nfhnfmod(GEN nf, GEN x, GEN detx).

pol being a nonconstant irreducible polynomial in ℚ[X], preferably monic and integral, initializes a number field (or nf) structure attached to the field K defined by pol. As such, it's a technical object passed as the first argument to most nfxxx functions, but it contains some information which may be directly useful. Access to this information via member functions is preferred since the specific data organization given below may change in the future. Currently, nf is a row vector with 9 components:

nf[1] contains the polynomial pol (nf.pol).

nf[2] contains [r1,r2] (nf.sign, nf.r1, nf.r2), the number of real and complex places of K.

nf[3] contains the discriminant d(K) (nf.disc) of K.

nf[4] contains the index of nf[1] (nf.index), i.e. [ℤ_K : ℤ[θ]], where θ is any root of nf[1].

nf[5] is a vector containing 7 matrices M, G, roundG, T, MD, TI, MDI and a vector vP defined as follows:

  * M is the (r1+r2) x n matrix whose columns represent the numerical values of the conjugates of the elements of the integral basis.

  * G is an n x n matrix such that T2 = ^t G G, where T2 is the quadratic form T₂(x) = ∑ |σ(x)|², σ running over the embeddings of K into ℂ.

  * roundG is a rescaled copy of G, rounded to nearest integers.

  * T is the n x n matrix whose coefficients are Tr(ω_iω_j) where the ω_i are the elements of the integral basis. Note also that det(T) is equal to the discriminant of the field K. Also, when understood as an ideal, the matrix T^-1 generates the codifferent ideal.

  * The columns of MD (nf.diff) express a ℤ-basis of the different of K on the integral basis.

  * TI is equal to the primitive part of T^-1, which has integral coefficients.

  * MDI is a two-element representation (for faster ideal product) of d(K) times the codifferent ideal (nf.disc*nf.codiff, which is an integral ideal). This is used in idealinv.

  * vP is the list of prime divisors of the field discriminant, i.e, the ramified primes (nf.p); nfdiscfactors(nf) is the preferred way to access that information.

nf[6] is the vector containing the r1+r2 roots (nf.roots) of nf[1] corresponding to the r1+r2 embeddings of the number field into ℂ (the first r1 components are real, the next r2 have positive imaginary part).

nf[7] is a ℤ-basis for dℤ_K, where d = [ℤ_K:ℤ(θ)], expressed on the powers of θ. The multiplication by d ensures that all polynomials have integral coefficients and nf[7] / d (nf.zk) is an integral basis for ℤ_K. Its first element is guaranteed to be 1. This basis is LLL-reduced with respect to T₂ (strictly speaking, it is a permutation of such a basis, due to the condition that the first element be 1).

nf[8] is the n x n integral matrix expressing the power basis in terms of the integral basis, and finally

nf[9] is the n x n² matrix giving the multiplication table of the integral basis.

If a non monic or non integral polynomial is input, nfinit will transform it, and return a structure attached to the new (monic integral) polynomial together with the attached change of variables, see flag = 3. It is allowed, though not very useful given the existence of nfnewprec, to input a nf or a bnf instead of a polynomial. It is also allowed to input a rnf, in which case an nf structure attached to the absolute defining polynomial polabs is returned (flag is then ignored).

  ? nf = nfinit(x^3 - 12); \\ initialize number field Q[X] / (X^3 - 12)
  ? nf.pol   \\ defining polynomial
  %2 = x^3 - 12
  ? nf.disc  \\ field discriminant
  %3 = -972
  ? nf.index \\ index of power basis order in maximal order
  %4 = 2
  ? nf.zk    \\ integer basis, lifted to Q[X]
  %5 = [1, x, 1/2*x^2]
  ? nf.sign  \\ signature
  %6 = [1, 1]
  ? factor(abs(nf.disc ))  \\ determines ramified primes
  %7 =
  [2 2]
  
  [3 5]
  ? idealfactor(nf, 2)
  %8 =
  [[2, [0, 0, -1]~, 3, 1, [0, 1, 0]~] 3]  \\   𝔭23

Huge discriminants, helping nfdisc.

In case pol has a huge discriminant which is difficult to factor, it is hard to compute from scratch the maximal order. The following special input formats are also accepted:

* [pol, B] where pol is a monic integral polynomial and B is the lift of an integer basis, as would be computed by nfbasis: a vector of polynomials with first element 1 (implicitly modulo pol). This is useful if the maximal order is known in advance.

* [pol, B, P] where pol and B are as above (a monic integral polynomial and the lift of an integer basis), and P is the list of ramified primes in the extension.

* [pol, listP] where pol is a rational polynomial and listP specifies a list of primes as in nfbasis. Instead of the maximal order, nfinit then computes an order which is maximal at these particular primes as well as the primes contained in the private prime table, see addprimes. The result has a good chance of being correct when the discriminant nf.disc factors completely over this set of primes but this is not guaranteed. The function nfcertify automates this:

  ? pol = polcompositum(x^5 - 101, polcyclo(7))[1];
  ? nf = nfinit( [pol, 10^3] );
  ? nfcertify(nf)
  %3 = []

A priori, nf.zk defines an order which is only known to be maximal at all primes ≤ 10³ (no prime ≤ 10³ divides nf.index). The certification step proves the correctness of the computation. Had it failed, that particular nf structure could not have been trusted and may have caused routines using it to fail randomly. One particular function that remains trustworthy in all cases is idealprimedec when applied to a prime included in the above list of primes or, more generally, a prime not dividing any entry in nfcertify output.

In order to explain the meaning of flag, let P = polredbest(pol), a polynomial defining the same number field obtained using the LLL algorithm on the lattice (ℤ_K, T₂), which may be equal to pol but is usually different and simpler. Binary digits of flag mean:

* 1: return [nf,Mod(a,P)], where nf is nfinit(P) and Mod(a,P) = Mod(x,pol) gives the change of variables. If only this bit is set, the behaviour is useless since we have P = pol.

* 2: return nfinit(P).

Both flags are set automatically when pol is not monic or not integral: first a linear change of variables is performed, to get a monic integral polynomial, then polredbest.

* 4: do not LLL-reduce nf.zk, which saves time in large degrees, you may expect to gain a factor 2 or so in degree n ≥ 100 or more, at the expense of possibly slowing down later uses of the nf structure. Use this flag if you only need basic arithmetic (the nfelt*, nfmodpr* and ideal* functions); or if you expect the natural basis of the maximal order to contain small elements, this will be the case for cyclotomic fields for instance. On the other hand, functions involving LLL reduction of rank n lattices should be avoided since each call will be about as costly as the initial LLL reduction that the flag prevents and may become more costly because of this missing initial reduction. In particular it is silly to use this flag in addition to the first two, although GP will not protest.

  ? T = polcyclo(307);
  ? K = nfinit(T);
  time = 19,390 ms.
  ? a = idealhnf(K,1-x);
  time = 477ms
  ? idealfactor(K, a)
  time = 294ms
  
  ? Kno = nfinit(T, 4);
  time = 11,256 ms.
  ? ano = idealhnf(Kno,1-x); \\ no slowdown, even sligthly faster
  time = 460ms
  ? idealfactor(Kno, ano)
  time = 264ms
  
  ? nfinit(T, 2); \\ polredbest is very slow in high degree
  time = 4min, 34,870 ms.
  ? norml2(%.pol) == norml2(T) \\ and gains nothing here
  %9 = 1

The library syntax is GEN nfinit0(GEN pol, long flag, long prec). Also available are GEN nfinit(GEN x, long prec) (flag = 0), GEN nfinitred(GEN x, long prec) (flag = 2), GEN nfinitred2(GEN x, long prec) (flag = 3). Instead of the above hardcoded numerical flags in nfinit0, one should rather use an or-ed combination of

* nf_RED: find a simpler defining polynomial,

* nf_ORIG: also return the change of variable,

* nf_NOLLL: do not LLL-reduce the maximal order ℤ-basis.

Returns 1 if x is an ideal in the number field nf, 0 otherwise.

The library syntax is long isideal(GEN nf, GEN x).

Let f and g define number fields, where f and g are irreducible polynomials in ℚ[X] and nf structures as output by nfinit. If either f or g is not irreducible, the result is undefined. Tests whether the number field f is conjugate to a subfield of the field g. If not, the output is the integer 0; if it is, the output depends on the value of flag:

* flag = 0 (default): return a vector of polynomials [a₁,...,a_n] with rational coefficients, representing all distinct embeddings: we have g | f o a_i for all i.

* flag = 1: return a single polynomial a representing a single embedding; this can be n times faster than the default when the embeddings have huge coefficients.

* flag = 2: return a vector of rational functions [r₁,...,r_n] whose denominators are coprime to g and such that r_i % g is the polynomial a_i from flag = 0. This variant is always faster than flag = 0 but produces results which are harder to use. If the denominators are hard to invert in ℚ[X]/(g), this may be even faster than flag = 1.

  ? T = x^6 + 3*x^4 - 6*x^3 + 3*x^2 + 18*x + 10;
  ? U = x^3 + 3*x^2 + 3*x - 2
  ? nfisincl(U, T)
  %3 = [24/179*x^5-27/179*x^4+80/179*x^3-234/179*x^2+380/179*x+94/179]
  ? a = nfisincl(U, T, 1)
  %4 = 24/179*x^5-27/179*x^4+80/179*x^3-234/179*x^2+380/179*x+94/179
  ? subst(U, x, Mod(a,T))
  %5 = Mod(0, x^6 + 3*x^4 - 6*x^3 + 3*x^2 + 18*x + 10)
  ? nfisincl(U, T, 2) \\ a as a t_RFRAC
  %6 = [(2*x^3 - 3*x^2 + 2*x + 4)/(3*x^2 - 1)]
  ? (a - %[1]) % T
  %7 = 0
  ? #nfisincl(x^2+1, T) \\ two embeddings
  %8 = 2
  
  \\ same result with nf structures
  ? L = nfinit(T); K = nfinit(U); v = [a];
  ? nfisincl(U, L) == v
  %10 = 1
  ? nfisincl(K, T) == v
  %11 = 1
  ? nfisincl(K, L) == v
  %12 = 1
  
  \\ comparative bench: an nf is a little faster, esp. for the subfield
  ? B = 2000;
  ? for (i=1, B, nfisincl(U,T))
  time = 1,364 ms.
  ? for (i=1, B, nfisincl(K,T))
  time = 988 ms.
  ? for (i=1, B, nfisincl(U,L))
  time = 1,341 ms.
  ? for (i=1, B, nfisincl(K,L))
  time = 880 ms.

Using an nf structure for the tentative subfield is faster if the structure is already available. On the other hand, the gain in nfisincl is usually not sufficient to make it worthwhile to initialize only for that purpose.

  ? for (i=1, B, nfinit(U))
  time = 590 ms.

A final more complicated example

  ? f = x^8 - 72*x^6 + 1944*x^4 - 30228*x^2 - 62100*x - 34749;
  ? g = nfsplitting(f); poldegree(g)
  %2 = 96
  ? #nfisincl(f, g)
  time = 559 ms.
  %3 = 8
  ? nfisincl(f,g,1);
  time = 172 ms.
  ? v = nfisincl(f,g,2);
  time = 199 ms.
  ? apply(x->poldegree(denominator(x)), v)
  %6 = [81, 81, 81, 81, 81, 81, 80, 81]
  ? v % g;
  time = 407 ms.

This final example shows that mapping rational functions to ℚ[X]/(g) can be more costly than that the rest of the algorithm. Note that nfsplitting also admits a flag yielding an embedding.

The library syntax is GEN nfisincl0(GEN f, GEN g, long flag). Also available is GEN nfisisom(GEN a, GEN b) (flag = 0).

As nfisincl, but tests for isomorphism. More efficient if f or g is a number field structure.

  ? f = x^6 + 30*x^5 + 495*x^4 + 1870*x^3 + 16317*x^2 - 22560*x + 59648;
  ? g = x^6 + 42*x^5 + 999*x^4 + 8966*x^3 + 36117*x^2 + 21768*x + 159332;
  ? h = x^6 + 30*x^5 + 351*x^4 + 2240*x^3 + 10311*x^2 + 35466*x + 58321;
  
  ? #nfisisom(f,g)  \\ two isomorphisms
  %3 = 2
  ? nfisisom(f,h) \\ not isomorphic
  %4 = 0
  \\ comparative bench
  ? K = nfinit(f); L = nfinit(g); B = 10^3;
  ? for (i=1, B, nfisisom(f,g))
  time = 6,124 ms.
  ? for (i=1, B, nfisisom(K,g))
  time = 3,356 ms.
  ? for (i=1, B, nfisisom(f,L))
  time = 3,204 ms.
  ? for (i=1, B, nfisisom(K,L))
  time = 3,173 ms.

The function is usually very fast when the fields are nonisomorphic, whenever the fields can be distinguished via a simple invariant such as degree, signature or discriminant. It may be slower when the fields share all invariants, but still faster than computing actual isomorphisms:

  \\ usually very fast when the answer is 'no':
  ? for (i=1, B, nfisisom(f,h))
  time = 32 ms.
  
  \\ but not always
  ? u = x^6 + 12*x^5 + 6*x^4 - 377*x^3 - 714*x^2 + 5304*x + 15379
  ? v = x^6 + 12*x^5 + 60*x^4 + 166*x^3 + 708*x^2 + 6600*x + 23353
  ? nfisisom(u,v)
  %13 = 0
  ? polsturm(u) == polsturm(v)
  %14 = 1
  ? nfdisc(u) == nfdisc(v)
  %15 = 1
  ? for(i=1,B, nfisisom(u,v))
  time = 1,821 ms.
  ? K = nfinit(u); L = nfinit(v);
  ? for(i=1,B, nfisisom(K,v))
  time = 232 ms.

The library syntax is GEN nfisisom(GEN f, GEN g).

Let nf be a nf structure attached to a number field K, let a ∈ K and let pr be a prid structure attached to a maximal ideal v. Return 1 if a is an n-th power in the completed local field K_v, and 0 otherwise.

  ? K = nfinit(y^2+1);
  ? P = idealprimedec(K,2)[1]; \\ the ramified prime above 2
  ? nfislocalpower(K,P,-1, 2) \\ -1 is a square
  %3 = 1
  ? nfislocalpower(K,P,-1, 4) \\ ... but not a 4-th power
  %4 = 0
  ? nfislocalpower(K,P,2, 2)  \\ 2 is not a square
  %5 = 0
  
  ? Q = idealprimedec(K,5)[1]; \\ a prime above 5
  ? nfislocalpower(K,Q, [0, 32]~, 30)  \\ 32*I is locally a 30-th power
  %7 = 1

The library syntax is long nfislocalpower(GEN nf, GEN pr, GEN a, GEN n).

This function is obsolete, use nfmodpr.

Kernel of the matrix a in ℤ_K/pr, where pr is in modpr format (see nfmodprinit).

The library syntax is GEN nfkermodpr(GEN nf, GEN x, GEN pr). This function is normally useless in library mode. Project your inputs to the residue field using nfM_to_FqM, then work there.

Finds number fields (up to isomorphism) with Galois group of Galois closure isomorphic to G with s complex places. The number fields are given by polynomials. This function supports the following groups:

* degree 2: C₂ = 2T1;

* degree 3: C₃ = 3T1 and S₃ = 3T2;

* degree 4: C₄ = 4T1, V₄ = 4T2, D₄ = 4T3, A₄ = 4T4 and S₄ = 4T5;

* degree 5: C₅ = 5T1, D₅ = 5T2, F₅ = M₂₀ = 5T3 and A₅ = 5T4;

* degree 6: C₆ = 6T1, S₃(6) = D₆(6) = 6T2, D₆(12) = 6T3, A₄(6) = 6T4, S₃ x C₃ = 6T5, A₄(6) x C₂ = 6T6, S₄(6)⁺ = 6T7, S₄(6)⁻ = 6T8, S₃² = 6T9, C₃²:C₄ = 6T10, S₄(6) x C₂ = 6T11, A₅(6) = PSL₂(5) = 6T12 and C₃²:D₄ = 6T13;

* degree 7: C₇ = 7T1, D₇ = 7T2, M₂₁ = 7T3 and M₄₂ = 7T4;

* degree 9: C₉ = 9T1, C₃ x C₃ = 9T2 and D₉ = 9T3;

* degree ℓ with ℓ prime: C_ℓ = ℓ T1 and D_ℓ = ℓ T2.

The groups A₅ and A₅(6) require the optional package nflistdata.

In addition, if N is a polynomial, all transitive subgroups of S_n with n ≤ 15, as well as alternating groups A_n and the full symmetric group S_n for all n (see below for details and explanations).

The groups are coded as [n,k] using the nTk format where n is the degree and k is the T-number, the index in the classification of transitive subgroups of S_n.

Alternatively, the groups C_n, D_n, A_n, S_n, V₄, F₅ = M₂₀, M₂₁ and M₄₂ can be input as character strings exactly as written, lifting subscripts; for instance "S4" or "M21". If the group is not recognized or is unsupported the function raises an exception.

The number fields are computed on the fly (and not from a preexisting table) using a variety of algorithms, with the exception of A₅ and A₅(6) which are obtained by table lookup. The algorithms are recursive and use the following ingredients: build distinguished subfields (or resolvent fields in Galois closures) of smaller degrees, use class field theory to build abelian extensions over a known base, select subfields using Galois theory. Because of our use of class field theory, and ultimately bnfinit, all results depend on the GRH in degree n > 3.

To avoid wasting time, the output polynomials defining the number fields are usually not the simplest possible, use polredbest or polredabs to reduce them.

The non-negative integer s specifies the number of complex places, between 0 and n/2. Additional supported values are:

* s = -1 (default) all signatures;

* s = -2 all signatures, given by increasing number of complex places; in degree n, this means a vector with 1 + floor(n/2) components: the i-th entry corresponds to s = i - 1.

If the irreducible monic polynomial F ∈ ℤ[X] is specified, gives only number fields having ℚ[X]/(F) as a subfield, or in the case of S₃, D_ℓ, A₄, S₄, F₅, M₂₁ and M₄₂, as a resolvent field (see also the function nfresolvent for these cases).

The parameter N can be the following:

* a positive integer: finds all fields with absolute discriminant N (recall that the discriminant over ℚ is (-1)^s N).

* a pair of non-negative real numbers [a,b] specifying a real interval: finds all fields with absolute value of discriminant between a and b. For most Galois groups, this is faster than iterating on individual N.

* omitted (default): a few fields of small discriminant (not always those with smallest absolute discriminant) are output with given G and s; usually about 10, less if too difficult to find. The parameter F is ignored.

* a polynomial with main variable, say t, of priority lower than x. The program outputs a regular polynomial in ℚ(t)[x] (in fact in ℤ[x,t]) with the given Galois group. By Hilbert irreducibility, almost all specializations of t will give suitable polynomials. The parameters s and F are ignored. This is implemented for all transitive subgroups of S_n with n ≤ 15 as well as for the alternating and symmetric groups A_n and S_n for all n. Polynomials for A_n were inspired by J.-F. Mestre, a few polynomials in degree ≤ 8 come from G. W. Smith, "Some polynomials over ℚ(t) and their Galois groups", Math. Comp., 69 (230), 1999, pp. 775–796 most others in degree ≤ 11 were provided by J. Klüners and G. Malle (see G. Malle and B. H. Matzat, Inverse Galois Theory, Springer, 1999) and T. Dokchitser completed the list up to degree 15. But for A_n and S_n, subgroups of S_n for n > 7 require the optional nflistdata package.

Complexity. : For a positive integer N, the complexity is subexponential in log N (and involves factoring N). For an interval [a,b], the complexity is roughly as follows, ignoring terms which are subexponential in log b. It is usually linear in the output size.

* C_n: O(b^1/φ(n)) for n = 2, 4, 6, 9 or any odd prime;

* D_n: O(b^2/φ(n)) for n = 4 or any odd prime;

* V₄, A₄: O(b^1/2), S₄: O(b); N.B. The subexponential terms are expensive for A₄ and S₄.

* M₂₀: O(b).

* S₄(6)⁻, S₄(6)⁺ A₄(6) x C₂, S₃ x S₃, S₄(6) x C₂ : O(b), D₆(12), A₄(6), S₃(6), S₃ x C₃, C₃²:C₄: O(b^1/2).

* M₂₁, M₄₂: O(b).

* C₃ x C₃: O(b^1/3), D₉: O(b^5/12).

  ? #nflist("S3", [1, 10^5]) \\ S3 cubic fields
  %1 = 21794
  ? #nflist("S3", [1, 10^5], 0) \\ real S3 cubic fields (0 complex place)
  %2 = 4753
  ? #nflist("S3", [1, 10^5], 1) \\ complex cubic fields (1 complex place)
  %3 = 17041
  ? v = nflist("S3", [1, 10^5], -2); apply(length,v)
  %4 = [4753, 17041]
  ? nflist("S4") \\ a few S4 fields
  %5 = [x^4 + 12*x^2 - 8*x + 16, x^4 - 2*x^2 - 8*x + 25, ...]
  ? nflist("S4",,0) \\ a few real S4 fields
  %6 = [x^4 - 52*x^2 - 56*x + 48, x^4 - 26*x^2 - 8*x + 1, ...]
  ? nflist("S4",,-2) \\ a few real S4 fields, by signature
  %7 = [[x^4 - 52*x^2 - 56*x + 48, ...],
        [x^4 - 8*x - 16, ... ],
        [x^4 + 138*x^2 - 8*x + 4541, ...]]
  ? nflist("S3",,,x^2+23) \\ a few cubic fields with resolvent Q(sqrt(-23))
  %8 = [x^3 + x + 1, x^3 + 2*x + 1, ...]
  ? nflist("C3", 3969) \\ C3 fields of given discriminant
  %9 = [x^3 - 21*x + 28, x^3 - 21*x - 35]
  ? nflist([3,1], 3969) \\ C3 fields, using nTt label
  %10 = [x^3 - 21*x + 28, x^3 - 21*x - 35]
  ? P = nflist([8,12],t) \\ geometric 8T12 polynomial
  %11 = x^8 + (-t^2 - 803)*x^6 + (264*t^2 + 165528)*x^4
        + (-2064*t^2 - 1724976)*x^2 + 4096*t^2
  ? polgalois(subst(P, t, 11))
  %12 = [24, 1, 12, "2A4(8)=[2]A(4)=SL(2,3)"]
  ? nflist("S11")
   ***   at top-level: nflist("S11")
   ***                 ^ —  —  —  — -
   *** nflist: unsupported group (S11). Use one of
   "C1"=[1,1];
   "C2"=[2,1];
   "C3"=[3,1], "S3"=[3,2];
   "C4"=[4,1], "V4"=[4,2], "D4"=[4,3], "A4"=[4,4], "S4"=[4,5];
   "C5"=[5,1], "D5"=[5,2], "F5"="M20"=[5,3], "A5"=[5,4];
   "C6"=[6,1], "D6"=[6,2], [6,3], ..., [6,13];
   "C7"=[7,1], "D7"=[7,2], "M21"=[7,3], "M42"=[7,4];
   "C9"=[9,1], [9,2], "D9"=[9,3]."
   Also supported are "Cp"=[p,1] and "Dp"=[p,2] for any odd prime p.
  
  ? nflist("S25", 't)
  %13 = x^25 + x*t + 1

The library syntax is GEN nflist(GEN G, GEN N = NULL, long s, GEN F = NULL).

Map x to a t_FFELT in the residue field modulo pr. The argument pr is either a maximal ideal in idealprimedec format or, preferably, a modpr structure from nfmodprinit. The function nfmodprlift allows to lift back to ℤ_K.

Note that the function applies to number field elements and not to vector / matrices / polynomials of such. Use apply to convert recursive structures.

  ? K = nfinit(y^3-250);
  ? P = idealprimedec(K, 5)[2];
  ? modP = nfmodprinit(K, P, 't);
  ? K.zk
  %4 = [1, 1/5*y, 1/25*y^2]
  ? apply(t->nfmodpr(K,t,modP), K.zk)
  %5 = [1, t, 2*t + 1]
  ? %[1].mod
  %6 = t^2 + 3*t + 4
  ? K.index
  %7 = 125

For clarity, we represent elements in the residue field 𝔽₅[t]/(T) as polynomials in the variable t. Whenever the underlying rational prime does not divide K.index, it is actually the case that t is the reduction of y in ℚ[y]/(K.pol) modulo an irreducible factor of K.pol over 𝔽_p. In the above example, 5 divides the index and t is actually the reduction of y/5.

The library syntax is GEN nfmodpr(GEN nf, GEN x, GEN pr).

Transforms the prime ideal pr into modpr format necessary for all operations modulo pr in the number field nf. The functions nfmodpr and nfmodprlift allow to project to and lift from the residue field. The variable v is used to display finite field elements (see ffgen).

  ? K = nfinit(y^3-250);
  ? P = idealprimedec(K, 5)[2];
  ? modP = nfmodprinit(K, P, 't);
  ? K.zk
  %4 = [1, 1/5*y, 1/25*y^2]
  ? apply(t->nfmodpr(K,t,modP), K.zk)
  %5 = [1, t, 2*t + 1]
  ? %[1].mod
  %6 = t^2 + 3*t + 4
  ? K.index
  %7 = 125

For clarity, we represent elements in the residue field 𝔽₅[t]/(T) as polynomials in the variable t. Whenever the underlying rational prime does not divide K.index, it is actually the case that t is the reduction of y in ℚ[y]/(K.pol) modulo an irreducible factor of K.pol over 𝔽_p. In the above example, 5 divides the index and t is actually the reduction of y/5.

The library syntax is GEN nfmodprinit0(GEN nf, GEN pr, long v) = -1) where v) is a variable number.

Lift the t_FFELT x (from nfmodpr) in the residue field modulo pr to the ring of integers. Vectors and matrices are also supported. For polynomials, use apply and the present function.

The argument pr is either a maximal ideal in idealprimedec format or, preferably, a modpr structure from nfmodprinit. There are no compatibility checks to try and decide whether x is attached the same residue field as defined by pr: the result is undefined if not.

The function nfmodpr allows to reduce to the residue field.

  ? K = nfinit(y^3-250);
  ? P = idealprimedec(K, 5)[2];
  ? modP = nfmodprinit(K,P);
  ? K.zk
  %4 = [1, 1/5*y, 1/25*y^2]
  ? apply(t->nfmodpr(K,t,modP), K.zk)
  %5 = [1, y, 2*y + 1]
  ? nfmodprlift(K, %, modP)
  %6 = [1, 1/5*y, 2/5*y + 1]
  ? nfeltval(K, %[3] - K.zk[3], P)
  %7 = 1

The library syntax is GEN nfmodprlift(GEN nf, GEN x, GEN pr).

Transforms the number field nf into the corresponding data using current (usually larger) precision. This function works as expected if nf is in fact a bnf, a bnr or a rnf (update structure to current precision). If the original bnf structure was not computed by bnfinit(,1), then this may be quite slow and even fail: many generators of principal ideals have to be computed and the algorithm may fail because the accuracy is not sufficient to bootstrap the required generators and fundamental units.

The library syntax is GEN nfnewprec(GEN nf, long prec). See also GEN bnfnewprec(GEN bnf, long prec) and GEN bnrnewprec(GEN bnr, long prec).

Given a polynomial T with coefficients in the number field nf, returns the number of real roots of the s(T) where s runs through the real embeddings of the field specified by optional argument pl:

* pl omitted: all r₁ real places;

* pl an integer between 1 and r₁: the embedding attached to the i-th real root of nf.pol, i.e. nf.roots[i];

* pl a vector or t_VECSMALL: the embeddings attached to the pl[i]-th real roots of nf.pol.

  ? nf = nfinit('y^2 - 2);
  ? nf.sign
  %2 = [2, 0]
  ? nf.roots
  %3 = [-1.414..., 1.414...]
  ? T = x^2 + 'y;
  ? nfpolsturm(nf, T, 1) \\ subst(T,y,sqrt(2)) has two real roots
  %5 = 2
  ? nfpolsturm(nf, T, 2) \\ subst(T,y,-sqrt(2)) has no real root
  %6 = 0
  ? nfpolsturm(nf, T) \\ all embeddings together
  %7 = [2, 0]
  ? nfpolsturm(nf, T, [2,1]) \\ second then first embedding
  %8 = [0, 2]
  ? nfpolsturm(nf, x^3)  \\ number of distinct roots !
  %9 = [1, 1]
  ? nfpolsturm(nf, x, 6) \\ there are only 2 real embeddings !
   ***   at top-level: nfpolsturm(nf,x,6)
   ***                 ^ —  —  —  —  — --
   *** nfpolsturm: domain error in nfpolsturm: index > 2

The library syntax is GEN nfpolsturm(GEN nf, GEN T, GEN pl = NULL).

Let pol be an irreducible integral polynomial defining a number field K with Galois closure ~{K}. This function is limited to the Galois groups supported by nflist; in the following ℓ denotes an odd prime. If Gal(~{K}/ℚ) is D_ℓ, A₄, S₄, F₅ (M₂₀), A₅, M₂₁ or M₄₂, returns a polynomial R defining the corresponding resolvent field (quadratic for D_ℓ, cyclic cubic for A₄ and M₂₁, noncyclic cubic for S₄, cyclic quartic for F₅, A₅(6) sextic for A₅, and cyclic sextic for M₄₂). In the A₅(6) case, returns the A₅ field of which it is the resolvent. Otherwise, gives a "canonical" subfield, or 0 if the Galois group is not supported.

The binary digits of flag correspond to 1: returns a pair [R,f] where f is a "conductor" whose definition is specific to each group and given below; 2: returns all "canonical" subfields.

Let D be the discriminant of the resolvent field nfdisc(R):

* In cases C_ℓ, D_ℓ, A₄, or S₄, disc(K) = (Df²)^m with m = (ℓ-1)/2 in the first two cases, and 1 in the last two.

* In cases where K is abelian over the resolvent subfield, the conductor of the relative extension.

* In case F₅, disc(K) = Df⁴ if f > 0 or 5²Df⁴ if f < 0.

* In cases M₂₁ or M₄₂, disc(K) = D^mf⁶ if f > 0 or 7³D^mf⁶ if f < 0, where m = 2 for M₂₁ and m = 1 for M₄₂.

* In cases A₅ and A₅(6), flag is currently ignored.

   ? pol = x^6-3*x^5+7*x^4-9*x^3+7*x^2-3*x+1; \\ Galois closure D6
   ? nfresolvent(pol)
   %2 = x^3 + x - 1
   ? nfresolvent(pol,1)
   %3 = [x^3 + x - 1, [[31, 21, 3; 0, 1, 0; 0, 0, 1], [1]]]

The library syntax is GEN nfresolvent(GEN pol, long flag).

Roots of the polynomial x in the number field nf given by nfinit without multiplicity (in ℚ if nf is omitted). x has coefficients in the number field (scalar, polmod, polynomial, column vector). The main variable of nf must be of lower priority than that of x (see Section se:priority). However if the coefficients of the number field occur explicitly (as polmods) as coefficients of x, the variable of these polmods must be the same as the main variable of t (see nffactor).

It is possible to input a defining polynomial for nf instead, but this is in general less efficient since parts of an nf structure will then be computed internally. This is useful in two situations: when you do not need the nf elsewhere, or when you cannot initialize an nf due to integer factorization difficulties when attempting to compute the field discriminant and maximal order.

Caveat. nfinit([T, listP]) allows to compute in polynomial time a conditional nf structure, which sets nf.zk to an order which is not guaranteed to be maximal at all primes. Always either use nfcertify first (which may not run in polynomial time) or make sure to input nf.pol instead of the conditional nf: nfroots is able to recover in polynomial time in this case, instead of potentially missing a factor.

The library syntax is GEN nfroots(GEN nf = NULL, GEN x). See also GEN nfrootsQ(GEN x), corresponding to nf = NULL.

Returns a two-component vector [w,z] where w is the number of roots of unity in the number field nf, and z is a primitive w-th root of unity. It is possible to input a defining polynomial for nf instead.

  ? K = nfinit(polcyclo(11));
  ? nfrootsof1(K)
  %2 = [22, [0, 0, 0, 0, 0, -1, 0, 0, 0, 0]~]
  ? z = nfbasistoalg(K, %[2])   \\ in algebraic form
  %3 = Mod(-x^5, x^10 + x^9 + x^8 + x^7 + x^6 + x^5 + x^4 + x^3 + x^2 + x + 1)
  ? [lift(z^11), lift(z^2)]     \\ proves that the order of z is 22
  %4 = [-1, -x^9 - x^8 - x^7 - x^6 - x^5 - x^4 - x^3 - x^2 - x - 1]

This function guesses the number w as the gcd of the #k(v)^* for unramified v above odd primes, then computes the roots in nf of the w-th cyclotomic polynomial. The algorithm is polynomial time with respect to the field degree and the bitsize of the multiplication table in nf (both of them polynomially bounded in terms of the size of the discriminant). Fields of degree up to 100 or so should require less than one minute.

The library syntax is GEN nfrootsof1(GEN nf).

Given a torsion ℤ_K-module x attached to the square integral invertible pseudo-matrix (A,I,J), returns an ideal list D = [d₁,...,d_n] which is the Smith normal form of x. In other words, x is isomorphic to ℤ_K/d₁⨁ ...⨁ ℤ_K/d_n and d_i divides d_i-1 for i ≥ 2. If flag is nonzero return [D,U,V], where UAV is the identity.

See Section se:ZKmodules for the definition of integral pseudo-matrix; briefly, it is input as a 3-component row vector [A,I,J] where I = [b₁,...,b_n] and J = [a₁,...,a_n] are two ideal lists, and A is a square n x n matrix with columns (A₁,...,A_n), seen as elements in Kⁿ (with canonical basis (e₁,...,e_n)). This data defines the ℤ_K module x given by (b₁e₁⨁ ...⨁ b_ne_n) / (a₁A₁⨁ ...⨁ a_nA_n) , The integrality condition is a_i,j ∈ b_i a_j^-1 for all i,j. If it is not satisfied, then the d_i will not be integral. Note that every finitely generated torsion module is isomorphic to a module of this form and even with b_i = Z_K for all i.

The library syntax is GEN nfsnf0(GEN nf, GEN x, long flag). Also available:

GEN nfsnf(GEN nf, GEN x) (flag = 0).

This function is obsolete, use nfmodpr.

Let P be a prime ideal in modpr format (see nfmodprinit), let a be a matrix, invertible over the residue field, and let b be a column vector or matrix. This function returns a solution of a.x = b; the coefficients of x are lifted to nf elements.

  ? K = nfinit(y^2+1);
  ? P = idealprimedec(K, 3)[1];
  ? P = nfmodprinit(K, P);
  ? a = [y+1, y; y, 0]; b = [1, y]~
  ? nfsolvemodpr(K, a,b, P)
  %5 = [1, 2]~

The library syntax is GEN nfsolvemodpr(GEN nf, GEN a, GEN b, GEN P). This function is normally useless in library mode. Project your inputs to the residue field using nfM_to_FqM, then work there.

Defining polynomial S over ℚ for the splitting field of P ∈ ℚ[x], that is the smallest field over which P is totally split. If irreducible, the polynomial P can also be given by a nf structure, which is more efficient. If d is given, it must be a multiple of the splitting field degree. Note that if P is reducible the splitting field degree can be smaller than the degree of P.

If flag is non-zero, we assume P to be monic, integral and irreducible and the return value depends on flag:

* flag = 1: return [S,C] where S is as before and C is an embedding of ℚ[x]/(P) in its splitting field given by a polynomial (implicitly modulo S, as in nfisincl).

* flag = 2: return [S,C] where C is vector of rational functions whose image in ℚ[x]/(S) yields the embedding; this avoids inverting the denominator, which is costly. when the degree of the splitting field is huge.

* flag = 3: return [S, v, p] a data structure allowing to quickly compute the Galois group of the splitting field, which is used by galoissplittinginit; more precisely, p is a prime splitting completely in the splitting field and v is a vector with deg S elements describing the automorphisms of S acting on the roots of S modulo p.

  ? K = nfinit(x^3 - 2);
  ? nfsplitting(K)
  %2 = x^6 + 108
  ? nfsplitting(x^8 - 2)
  %3 = x^16 + 272*x^8 + 64
  ? S = nfsplitting(x^6 - 8) \\ reducible
  %4 = x^4 + 2*x^2 + 4
  ? lift(nfroots(subst(S,x,a),x^6-8))
  %5 = [-a, a, -1/2*a^3 - a, -1/2*a^3, 1/2*a^3, 1/2*a^3 + a]
  
  ? P = x^8-2;
  ? [S,C] = nfsplitting(P,,1)
  %7 = [x^16 + 272*x^8 + 64, -7/768*x^13 - 239/96*x^5 + 1/2*x]
  ? subst(P, x, Mod(C,S))
  %8 = Mod(0, x^16 + 272*x^8 + 64)

Specifying the degree d of the splitting field can make the computation faster; if d is not a multiple of the true degree, it will be ignored with a warning.

  ? nfsplitting(x^17-123);
  time = 3,607 ms.
  ? poldegree(%)
  %2 = 272
  ? nfsplitting(x^17-123,272);
  time = 150 ms.
  ? nfsplitting(x^17-123,273);
   *** nfsplitting: Warning: ignoring incorrect degree bound 273
  time = 3,611 ms.

The complexity of the algorithm is polynomial in the degree d of the splitting field and the bitsize of T; if d is large the result will likely be unusable, e.g. nfinit will not be an option:

  ? nfsplitting(x^6-x-1)
  [... degree 720 polynomial deleted ...]
  time = 11,020 ms.

Variant: Also available is GEN nfsplitting(GEN T, GEN D) for flag = 0.

The library syntax is GEN nfsplitting0(GEN P, GEN d = NULL, long fl).

Finds all subfields of degree d of the number field defined by the (monic, integral) polynomial pol (all subfields if d is null or omitted). The result is a vector of subfields, each being given by [g,h] (default) or simply g (flag = 1), where g is an absolute equation and h expresses one of the roots of g in terms of the root x of the polynomial defining nf. This routine uses

* Allombert's galoissubfields when nf is Galois (with weakly supersolvable Galois group).

* Klüners's or van Hoeij-Klüners-Novocin algorithm in the general case. The latter runs in polynomial time and is generally superior unless there exists a small unramified prime p such that pol has few irreducible factors modulo p.

An input of the form [nf, fa] is also allowed, where fa is the factorisation of nf.pol over nf, expressed as a famat of polynomials with coefficients in the variable of nf, in which case the van Hoeij-Klüners-Novocin algorithm is used.

  ? pol = x^4 - x^3 - x^2 + x + 1;
  ? nfsubfields(pol)
  %2 = [[x, 0], [x^2 - x + 1, x^3 - x^2 + 1], [x^4 - x^3 - x^2 + x + 1, x]]
  ? nfsubfields(pol,,1)
  %2 = [x, x^2 - x + 1, x^4 - x^3 - x^2 + x + 1]
  ? y=varhigher("y"); fa = nffactor(pol,subst(pol,x,y));
  ? #nfsubfields([pol,fa])
  %5 = 3

The library syntax is GEN nfsubfields0(GEN pol, long d, long flag). Also available is GEN nfsubfields(GEN nf, long d), corresponding to flag = 0.

Computes the maximal CM subfield of nf. Returns 0 if nf does not have a CM subfield, otherwise returns [g,h] (default) or g (flag = 1) where g is an absolute equation and h expresses a root of g in terms of the generator of nf. Moreover, the CM involution is given by X mod g(X) ⟼ -X mod g(X), i.e. X mod g(X) is a totally imaginary element.

An input of the form [nf, fa] is also allowed, where fa is the factorisation of nf.pol over nf, and nf is also allowed to be a monic defining polynomial for the number field.

  ? nf = nfinit(x^8 + 20*x^6 + 10*x^4 - 4*x^2 + 9);
  ? nfsubfieldscm(nf)
  %2 = [x^4 + 4480*x^2 + 3612672, 3*x^5 + 58*x^3 + 5*x]
  ? pol = y^16-8*y^14+29*y^12-60*y^10+74*y^8-48*y^6+8*y^4+4*y^2+1;
  ? fa = nffactor(pol, subst(pol,y,x));
  ? nfsubfieldscm([pol,fa])
  %5 = [y^8 + ... , ...]

The library syntax is GEN nfsubfieldscm(GEN nf, long flag).

Computes the list of maximal subfields of nf. The result is a vector as in nfsubfields.

An input of the form [nf, fa] is also allowed, where fa is the factorisation of nf.pol over nf, and nf is also allowed to be a monic defining polynomial for the number field.

The library syntax is GEN nfsubfieldsmax(GEN nf, long flag).

Let nf be attached to a number field K, let v be a vector of elements of K, not all of them 0, seen as element of the projective space of dimension #v - 1. Return the absolute logarithmic Weil height of that element, which does not depend on the number field used to compute it.

When the entries of v are rational, the height is log(normlp(v / content(v), oo)).

  ? v = [1, 2, -3, 101]; Q = nfinit(x); Qi = nfinit(x^2 + 1);
  ? exponent(nfweilheight(Q, v) - log(101))
  %2 = -125
  ? exponent(nfweilheight(Qi, v) - log(101))
  %3 = -125

The library syntax is GEN nfweilheight(GEN nf, GEN v, long prec).

P and Q being squarefree polynomials in ℤ[X] in the same variable, outputs the simple factors of the étale ℚ-algebra A = ℚ(X, Y) / (P(X), Q(Y)). The factors are given by a list of polynomials R in ℤ[X], attached to the number field ℚ(X)/ (R), and sorted by increasing degree (with respect to lexicographic ordering for factors of equal degrees). Returns an error if one of the polynomials is not squarefree.

Note that it is more efficient to reduce to the case where P and Q are irreducible first. The routine will not perform this for you, since it may be expensive, and the inputs are irreducible in most applications anyway. In this case, there will be a single factor R if and only if the number fields defined by P and Q are linearly disjoint (their intersection is ℚ).

Assuming P is irreducible (of smaller degree than Q for efficiency), it is in general much faster to proceed as follows

  nf = nfinit(P); L = nffactor(nf, Q)[,1];
  vector(#L, i, rnfequation(nf, L[i]))

to obtain the same result. If you are only interested in the degrees of the simple factors, the rnfequation instruction can be replaced by a trivial poldegree(P) * poldegree(L[i]).

The binary digits of flag mean

1: outputs a vector of 4-component vectors [R,a,b,k], where R ranges through the list of all possible compositums as above, and a (resp. b) expresses the root of P (resp. Q) as an element of ℚ(X)/(R). Finally, k is a small integer such that b + ka = X modulo R.

2: assume that P and Q define number fields which are linearly disjoint: both polynomials are irreducible and the corresponding number fields have no common subfield besides ℚ. This allows to save a costly factorization over ℚ. In this case return the single simple factor instead of a vector with one element.

A compositum is often defined by a complicated polynomial, which it is advisable to reduce before further work. Here is an example involving the field ℚ(ζ₅, 5^1/5):

  ? L = polcompositum(x^5 - 5, polcyclo(5), 1); \\  list of [R,a,b,k]
  ? [R, a] = L[1];  \\  pick the single factor, extract R,a (ignore b,k)
  ? R               \\  defines the compositum
  %3 = x^20 + 5*x^19 + 15*x^18 + 35*x^17 + 70*x^16 + 141*x^15 + 260*x^14\
  + 355*x^13 + 95*x^12 - 1460*x^11 - 3279*x^10 - 3660*x^9 - 2005*x^8    \
  + 705*x^7 + 9210*x^6 + 13506*x^5 + 7145*x^4 - 2740*x^3 + 1040*x^2     \
  - 320*x + 256
  ? a^5 - 5         \\  a fifth root of 5
  %4 = 0
  ? [T, X] = polredbest(R, 1);
  ? T     \\  simpler defining polynomial for ℚ[x]/(R)
  %6 = x^20 + 25*x^10 + 5
  ? X     \\   root of R in ℚ[y]/(T(y))
  %7 = Mod(-1/11*x^15 - 1/11*x^14 + 1/22*x^10 - 47/22*x^5 - 29/11*x^4 + 7/22,\
  x^20 + 25*x^10 + 5)
  ? a = subst(a.pol, 'x, X)  \\  a in the new coordinates
  %8 = Mod(1/11*x^14 + 29/11*x^4, x^20 + 25*x^10 + 5)
  ? a^5 - 5
  %9 = 0

In the above example, x⁵-5 and the 5-th cyclotomic polynomial are irreducible over ℚ; they have coprime degrees so define linearly disjoint extensions and we could have started by

  ? [R,a] = polcompositum(x^5 - 5, polcyclo(5), 3); \\  [R,a,b,k]

The library syntax is GEN polcompositum0(GEN P, GEN Q, long flag). Also available are GEN compositum(GEN P, GEN Q) (flag = 0) and GEN compositum2(GEN P, GEN Q) (flag = 1).

Galois group of the nonconstant polynomial T ∈ ℚ[X]. In the present version 2.17.0, T must be irreducible and the degree d of T must be less than or equal to 7. If the galdata package has been installed, degrees 8, 9, 10 and 11 are also implemented. By definition, if K = ℚ[x]/(T), this computes the action of the Galois group of the Galois closure of K on the d distinct roots of T, up to conjugacy (corresponding to different root orderings).

The output is a 4-component vector [n,s,k,name] with the following meaning: n is the cardinality of the group, s is its signature (s = 1 if the group is a subgroup of the alternating group A_d, s = -1 otherwise) and name is a character string containing name of the transitive group according to the GAP 4 transitive groups library by Alexander Hulpke.

k is more arbitrary and the choice made up to version 2.2.3 of PARI is rather unfortunate: for d > 7, k is the numbering of the group among all transitive subgroups of S_d, as given in "The transitive groups of degree up to eleven", G. Butler and J. McKay, Communications in Algebra, vol. 11, 1983, pp. 863–911 (group k is denoted T_k there). And for d ≤ 7, it was ad hoc, so as to ensure that a given triple would denote a unique group. Specifically, for polynomials of degree d ≤ 7, the groups are coded as follows, using standard notations

In degree 1: S₁ = [1,1,1].

In degree 2: S₂ = [2,-1,1].

In degree 3: A₃ = C₃ = [3,1,1], S₃ = [6,-1,1].

In degree 4: C₄ = [4,-1,1], V₄ = [4,1,1], D₄ = [8,-1,1], A₄ = [12,1,1], S₄ = [24,-1,1].

In degree 5: C₅ = [5,1,1], D₅ = [10,1,1], M₂₀ = [20,-1,1], A₅ = [60,1,1], S₅ = [120,-1,1].

In degree 6: C₆ = [6,-1,1], S₃ = [6,-1,2], D₆ = [12,-1,1], A₄ = [12,1,1], G₁₈ = [18,-1,1], S₄⁻ = [24,-1,1], A₄ x C₂ = [24,-1,2], S₄⁺ = [24,1,1], G₃₆⁻ = [36,-1,1], G₃₆⁺ = [36,1,1], S₄ x C₂ = [48,-1,1], A₅ = PSL₂(5) = [60,1,1], G₇₂ = [72,-1,1], S₅ = PGL₂(5) = [120,-1,1], A₆ = [360,1,1], S₆ = [720,-1,1].

In degree 7: C₇ = [7,1,1], D₇ = [14,-1,1], M₂₁ = [21,1,1], M₄₂ = [42,-1,1], PSL₂(7) = PSL₃(2) = [168,1,1], A₇ = [2520,1,1], S₇ = [5040,-1,1].

This is deprecated and obsolete, but for reasons of backward compatibility, we cannot change this behavior yet. So you can use the default new_galois_format to switch to a consistent naming scheme, namely k is always the standard numbering of the group among all transitive subgroups of S_n. If this default is in effect, the above groups will be coded as:

In degree 1: S₁ = [1,1,1].

In degree 2: S₂ = [2,-1,1].

In degree 3: A₃ = C₃ = [3,1,1], S₃ = [6,-1,2].

In degree 4: C₄ = [4,-1,1], V₄ = [4,1,2], D₄ = [8,-1,3], A₄ = [12,1,4], S₄ = [24,-1,5].

In degree 5: C₅ = [5,1,1], D₅ = [10,1,2], M₂₀ = [20,-1,3], A₅ = [60,1,4], S₅ = [120,-1,5].

In degree 6: C₆ = [6,-1,1], S₃ = [6,-1,2], D₆ = [12,-1,3], A₄ = [12,1,4], G₁₈ = [18,-1,5], A₄ x C₂ = [24,-1,6], S₄⁺ = [24,1,7], S₄⁻ = [24,-1,8], G₃₆⁻ = [36,-1,9], G₃₆⁺ = [36,1,10], S₄ x C₂ = [48,-1,11], A₅ = PSL₂(5) = [60,1,12], G₇₂ = [72,-1,13], S₅ = PGL₂(5) = [120,-1,14], A₆ = [360,1,15], S₆ = [720,-1,16].

In degree 7: C₇ = [7,1,1], D₇ = [14,-1,2], M₂₁ = [21,1,3], M₄₂ = [42,-1,4], PSL₂(7) = PSL₃(2) = [168,1,5], A₇ = [2520,1,6], S₇ = [5040,-1,7].

Warning. The method used is that of resolvent polynomials and is sensitive to the current precision. The precision is updated internally but, in very rare cases, a wrong result may be returned if the initial precision was not sufficient.

The library syntax is GEN polgalois(GEN T, long prec). To enable the new format in library mode, set the global variable new_galois_format to 1.

This function is deprecated, use polredbest instead. Finds polynomials with reasonably small coefficients defining subfields of the number field defined by T. One of the polynomials always defines ℚ (hence has degree 1), and another always defines the same number field as T if T is irreducible.

All T accepted by nfinit are also allowed here; in particular, the format [T, listP] is recommended, e.g. with listP = 10⁵ or a vector containing all ramified primes. Otherwise, the maximal order of ℚ[x]/(T) must be computed.

The following binary digits of flag are significant:

1: Possibly use a suborder of the maximal order. The primes dividing the index of the order chosen are larger than primelimit or divide integers stored in the addprimes table. This flag is deprecated, the [T, listP] format is more flexible.

2: gives also elements. The result is a two-column matrix, the first column giving primitive elements defining these subfields, the second giving the corresponding minimal polynomials.

  ? M = polred(x^4 + 8, 2)
  %1 =
  [           1         x - 1]
  
  [ 1/2*x^2 + 1 x^2 - 2*x + 3]
  
  [-1/2*x^2 + 1 x^2 - 2*x + 3]
  
  [     1/2*x^2       x^2 + 2]
  
  [     1/4*x^3       x^4 + 2]
  ? minpoly(Mod(M[4,1], x^4+8))
  %2 = x^2 + 2

The library syntax is polred(GEN T) (flag = 0). Also available is GEN polred2(GEN T) (flag = 2). The function polred0 is deprecated, provided for backward compatibility.

Returns a canonical defining polynomial P for the number field ℚ[X]/(T) defined by T, such that the sum of the squares of the modulus of the roots (i.e. the T₂-norm) is minimal. Different T defining isomorphic number fields will yield the same P. All T accepted by nfinit are also allowed here, e.g. nonmonic polynomials, or pairs [T, listP] specifying that a nonmaximal order may be used. For convenience, any number field structure (nf, bnf,...) can also be used instead of T.

  ? polredabs(x^2 + 16)
  %1 = x^2 + 1
  ? K = bnfinit(x^2 + 16); polredabs(K)
  %2 = x^2 + 1

Warning 1. Using a t_POL T requires computing and fully factoring the discriminant d_K of the maximal order which may be very hard. You can use the format [T, listP], where listP encodes a list of known coprime divisors of disc(T) (see ??nfbasis), to help the routine, thereby replacing this part of the algorithm by a polynomial time computation But this may only compute a suborder of the maximal order, when the divisors are not squarefree or do not include all primes dividing d_K. The routine attempts to certify the result independently of this order computation as per nfcertify: we try to prove that the computed order is maximal. If the certification fails, the routine then fully factors the integers returned by nfcertify. You can also use polredbest to avoid this factorization step; in this case, the result is small but no longer canonical.

Warning 2. Apart from the factorization of the discriminant of T, this routine runs in polynomial time for a fixed degree. But the complexity is exponential in the degree: this routine may be exceedingly slow when the number field has many subfields, hence a lot of elements of small T₂-norm. If you do not need a canonical polynomial, the function polredbest is in general much faster (it runs in polynomial time), and tends to return polynomials with smaller discriminants.

The binary digits of flag mean

1: outputs a two-component row vector [P,a], where P is the default output and Mod(a, P) is a root of the original T.

4: gives all polynomials of minimal T₂ norm; of the two polynomials P(x) and ± P(-x), only one is given.

16: (OBSOLETE) Possibly use a suborder of the maximal order, without attempting to certify the result as in Warning 1. This makes polredabs behave like polredbest. Just use the latter.

  ? T = x^16 - 136*x^14 + 6476*x^12 - 141912*x^10 + 1513334*x^8 \
        - 7453176*x^6 + 13950764*x^4 - 5596840*x^2 + 46225
  ? T1 = polredabs(T); T2 = polredbest(T);
  ? [ norml2(polroots(T1)), norml2(polroots(T2)) ]
  %3 = [88.0000000, 120.000000]
  ? [ sizedigit(poldisc(T1)), sizedigit(poldisc(T2)) ]
  %4 = [75, 67]

The precise definition of the output of polredabs is as follows.