Discover millions of ebooks, audiobooks, and so much more with a free trial

Only $11.99/month after trial. Cancel anytime.

Algebraic Number Theory
Algebraic Number Theory
Algebraic Number Theory
Ebook464 pages8 hours

Algebraic Number Theory

Rating: 0 out of 5 stars

()

Read preview

About this ebook

Careful organization and clear, detailed proofs characterize this methodical, self-contained exposition of basic results of classical algebraic number theory from a relatively modem point of view. This volume presents most of the number-theoretic prerequisites for a study of either class field theory (as formulated by Artin and Tate) or the contemporary treatment of analytical questions (as found, for example, in Tate's thesis).
Although concerned exclusively with algebraic number fields, this treatment features axiomatic formulations with a considerable range of applications. Modem abstract techniques constitute the primary focus. Topics include introductory materials on elementary valuation theory, extension of valuations, local and ordinary arithmetic fields, and global, quadratic, and cyclotomic fields.
Subjects correspond to those usually covered in a one-semester, graduate level course in algebraic number theory, making this book ideal either for classroom use or as a stimulating series of exercises for mathematically minded individuals.

LanguageEnglish
Release dateJan 27, 2012
ISBN9780486154367
Algebraic Number Theory

Read more from Edwin Weiss

Related to Algebraic Number Theory

Titles in the series (100)

View More

Related ebooks

Mathematics For You

View More

Related articles

Reviews for Algebraic Number Theory

Rating: 0 out of 5 stars
0 ratings

0 ratings0 reviews

What did you think?

Tap to rate

Review must be at least 10 words

    Book preview

    Algebraic Number Theory - Edwin Weiss

    Index

    1

    Elementary Valuation Theory

    The central feature of the subject commonly known as algebraic number theory is the problem of factorization in an algebraic number field, where by an algebraic number field we mean a finite extension of the rational field Q. Of course, it will take some time before the full meaning of this statement will become apparent.

    There are several distinct approaches to our subject matter, and we have chosen to emphasize the valuation-theoretic approach. In this chapter, we begin to learn the language.

    1-1.Valuations and Prime Divisors

    Let us begin with a definition. A valuation of the field F is a function φ from F into the nonnegative reals such that

    (iii) There exists a real constant C such that

    Some obvious examples of valuations are the following: (1) Let F denote the field of real numbers R or the field of complex numbers C, and put φ(a) = |a|, C = 2. It may be remarked that the original impetus for the study of valuations of an arbitrary field arose from this example —that is, as a generalization of the notion of absolute value. (2) Let F be any field, and put φ(0) = 0, φ(a) = 1 for a F. This is known as the trivial valuation of F.

    In view of the numerous examples that will be considered later, let us be content for the moment with the simple examples above. The impatient reader may turn to Section 1-4.

    A valuation determines a homomorphism of F*, the multiplicative group of F, and φ(b/a) = φ(b)/(φ(a) for a ≠ 0. Also, since φ(1) = φ(− l)φ(− 1), it follows that φ(−l) = 1 and φ(−a) = φ(−1)φ(a) = φ(a). It is also immediate that if ζ is any root of unity belonging to F then φ(ζ) = 1 and that a finite field admits only the trivial valuation.

    1-1-1. Proposition.In the definition of a valuation, (iii) may be replaced by

    Proof. (iii) ⇒ (iii′). Suppose φ(a) = max {φ(a), φ(b)}. If φ(a) = 0 then a = b = 0 and (iii′) holds. If φ(a) ≠ 0, then

    (iii′) ⇒ (iii). If φ(a) ≤ 1, then

    In the more usual definition of valuation, condition (iii) is replaced by the so-called triangle inequality

    We shall soon see that, for all practical purposes, our valuations may be taken to satisfy the triangle inequality.

    1-1-2. Proposition.A valuation φ of F determines a Hausdorff topology on F. For each a F, a fundamental system of neighborhoods of a is given by the set of all

    Proof. Let Δ denote the diagonal of F × F, and for subsets S and T of F × F let

    For each ε > 0 we put

    From the properties of φ, we have

    In fact, (1), (2), and (3) are trivial; (5) is clear since φ(a ba = b; finally, the assertion that (a, b) belongs to the left side of (4) means that there exists c F with φ(a c) < ε/C and φ(c b) < ε/C, so that φ(a b) < ε and (4) holds.

    is a filter base on F × F. This filter base defines a uniform structure on F, and according to (5) the uniform structure is separated. Therefore, the topology of F deduced from the uniform structure in the canonical way has fundamental neighborhoods as described and is a Hausdorff topology. There is no particular significance attached to the method by which we have arrived at the topology of F. It is based on facts and notation as found in N. Bourbaki’s Topologie Générale (Hermann & Cie, Paris, 1940).

    1-1-3. Corollary.Let φ be a valuation of F, and let {xn} be a sequence in F; then

    We say that two valuations φ1 and φ2 are equivalent (and ) when they determine the same topology on F—that is, if and only if 1 = 2. The equivalence classes with respect to this equivalence relation are called prime divisors of F and are denoted by P, Q, etc. The prime divisor to which the trivial valuation of F belongs is called the trivial prime divisor; all the others are nontrivial prime divisors.

    Let φ be a valuation of F with constant C, and let α > 0 be a real number; then the function φα on F given by φα(a) = [φ(a)]α is a valuation of F with constant .

    1-1-4. Theorem.Let φ1, φ2 be nontrivial valuations of F, and let a denote an arbitrary element of F; then the following statements are

    Proof. (1) ⇒ (2). For i = 1, 2 we write

    , so that

    Thus the uniform structures are equivalent, and the topologies are the same—that is, 1 = 2.

    (2) ⇒ (3). This is trivial since 1 stronger than 2 means that every open set for the topology 2 is also open for the topology 1.

    (3) ⇒ (4). Making use of (1-1-3), we have

    (4) ⇒ (5). Since φ1 is nontrivial, there exists b F* such that φ1(b) ≠ 1, and we may then take φ1(b) < 1. Consequently,

    In the same way, φ2(a−l) ≤ 1, and we conclude that φ2(a) = 1.

    (5) ⇒ (6). Since φ2 is nontrivial, there exists c F* such that 0 < φ2(c) < 1, and then

    Thus (4) holds, and as in the proof that (4) (5) we have

    Of course,

    . The validity of (6) is now clear.

    (6) ⇒ (1). Fix a F such that φ1(a) > 1. Then φ2(a) > 1, and we may put

    for this it suffices to show that if for each b F* we write

    . Let r = m/n with n > 0 denote a rational number; then

    1-1-5. Corollary.Let P be a prime divisor of F; then for any φ P

    Proof. (1-1-4) takes care of the case where P is nontrivial. Since it is easy to see that the valuation φ of F is trivial if and only if is the discrete topology, it follows that a trivial prime divisor P consists solely of the trivial valuation; thus, the assertion holds in this case also.

    In order to clarify the connection between our valuations and those satisfying the triangle inequality, it is convenient to introduce a simple definition. Given a valuation φ, we define ||φ|| (the norm of φ) by ||φ|| = inf C, where C runs over all constants that may be used in (iii) of the definition of valuation. ||φ|| is the smallest possible value for C, and (1-1-1) applies with C = ||φ||. For any real number α > 0, it follows immediately that ||φα|| = ||α||α. In view of (1-1-5) we see that, given a prime divisor P of F, there exists φ P with ||φ|| ≤ 2. We may also note that, for any φ, ||φ|| ≥ 1; in fact, 1 = φ(1) = φ(l + 0) < ||φ||.

    1-1-6. Exercise.Suppose that φ is a valuation for which condition (iii) is replaced by the triangle inequality. Give an example to show that under such circumstances φα need not be a valuation (in the sense that it need not satisfy the triangle inequality).

    1-1-7. Lemma.Let φ be a valuation of F; then

    Proof. By (4-9-1) we have φ(a1 + a2) ≤ 2 max {φ(a1), φ(a2)}, and by induction

    For any n > 1 there exists a unique r , and then

    1-1-8. Proposition.Let φ be a real-valued function on F which satisfies conditions (i) and (ii) of the definition of valuation; then

    is a valuation and ||φ|| ≤ 2.

    Proof. ⇒: φ(a + b) ≤ φ(a) + φ(b) ≤ 2 max {φ(a), φ(b)}; so, by (4-9-1), φ is a valuation, and ||φ|| ≤ 2.

    ⇐: For each n > 1, we have

    Taking nth roots and letting n → ∞, we arrive at the desired conclusion.

    1-1-9. Theorem.Let P be a prime divisor of F, and let φ P have ||φ|| ≤ 2. Then is a metric topology with respect to which F is a topological field and φ is uniformly continuous.

    Proof. For a, b F put d(a, b) = φ(a b, it is immediate that d is a metric on F. In the metric topology Td determined by d, a fundamental system of neighborhoods of a F is given by

    Hence, = Td. We see from

    that

    Therefore, it follows that φ is uniformly continuous on F. It is now easy to verify that the maps (a, b) → a b, (a, b) → ab, and b b−1 are continuous; that is, F is a topological field with respect to .

    Henceforth, whenever we consider a valuation φ P, it is to be understood—unless explicit mention is made to the contrary—that ||φ|| ≤ 2. Of course, there will be no harm in writing Tp in place of .

    1-2.The Approximation Theorem

    Having determined the relations between equivalent nontrivial valuations in (4-9-1), we proceed now to a description of the connections between inequivalent nontrivial valuations.

    1-2-1. Lemma.Let φ1, … , φn be inequivalent nontrivial valuations of F; then there exists an element a F with

    Proof. By induction on n. Suppose n = 2; then by (4-9-1) there exist b, c F such that

    — so that a = b/c works. Suppose now that the lemma holds for n − 1. Choose b F such that

    and c F such that

    If φn(b) ≤ 1, then surely a = bmc, for sufficiently large m, will do. It remains to consider the case where φn(b) > 1.

    In general, if φ is any valuation, then

    and also

    Therefore, for sufficiently large m, a = [bm/(l + bm)]c satisfies the required conditions.

    1-2-2. Lemma.Let φl, … , φn be inequivalent nontrivial valuations of F; then given any ε > 0, there exists b F such that

    Proof. Choose a as in (1-2-1), and put b = am/(l + am) —with m to be determined. For i > 1,

    while, for i = 1,

    Hence, we need only take m sufficiently large.

    1-2-3. Approximation Theorem.Let φ1, … , φn be inequivalent nontrivial valuations of F, and let a1, … , an be elements of F. Given any ε > 0, there exists an element a F such that

    Proof. Let M = max {φi(aj)}, where i, j = 1, … , n. According to (1-2-2) there exist b1, … , bn F such that,

    Clearly [in virtue of (is an element of the desired type.

    The approximation theorem may also be expressed in the following way: Suppose P1, … , Pn are distinct nontrivial prime divisors of F; and for each i = 1, … , n let F(i) denote F with the topology determined by the prime divisor Pi. The set F . That F is the assertion of the approximation theorem. To put it still another way, the approximation theorem implies that any finite number of inequivalent nontrivial valuations of F are completely independent; however, we shall see in Section 1-4 that this need not be valid when the number of valuations is infinite.

    It is surprising perhaps that the validity of the approximation theorem in its full generality (i.e., as we have stated it) was not discovered until 1945.

    1-2-4. Exercise.For nontrivial inequivalent valuations φ1, … , φn the topology 1 is not comparable with the topology generated by the 1 (i = 2, … , n).

    1-3.Archimedean and Nonarchimedean Prime Divisors

    Let P be a prime divisor of F. We say that P is archimedean when ||φ|| > 1 for every φ P and that P is nonarchimedean when ||φ|| = 1 for every φ P. For a valuation φ of F it is only natural to call φ archimedean (nonarchimedean) when the prime divisor P to which it belongs is archimedean (nonarchimedean). Note that an archimedean valuation and a nonarchimedean one cannot be equivalent.

    It is clear that the trivial valuation is nonarchimedean and that the absolute value valuation of R or C is archimedean.

    The next result gives equivalent and more standard formulations of our definition.

    1-3-1. Proposition..Let φ be a valuation of F; then

    Proof. The first equivalence is immediate from (4-9-1) and the properties of ||φ||. As for the remaining implications, we have:

    ⇒: By induction,

    .

    for all n Z, and suppose φ(a) ≥ φ(b); then

    and letting n → ∞ gives φ(a + b) ≤ φ(a) = max {φ(a), φ(b)}.

    1-3-2. Corollary.The archimedean or nonarchimedean character of a valuation is completely determined by its action on the prime subfield. In particular, a field with characteristic p ≠ 0 can have only nonarchimedean prime divisors.

    We may now prove an elementary but extremely useful property of nonarchimedean valuations.

    1-3-3. Proposition.Let φ be a nonarchimedean valuation of F; then

    Proof. Suppose φ(a) < φ(b); then

    1-3-4. Corollary.Let φ be a nonarchimedean valuation of Fthen

    then

    is maximal for at least two of the ai.

    1-3-5. Exercise.Let φ1, … , φn be nontrivial inequivalent nonarchimedean valuations of Fthen

    i. There exists a F* such that

    ii. If εi > 0 is a value taken on by φi for i = 1, … , n, then there exists a F such that

    Suppose that P is a nonarchimedean prime divisor of F, and choose any φ P. Put

    From (, and U are independent of the choice of φ P.

    Using the properties of φ, we see that O is an integral domain with 1; it is called the valuation ring at P or the ring of integers at P. , where O−1 = {a−l|a O, a is a prime ideal in O; it is called the prime ideal at P. Clearly U is the (multiplicative) group of units of Oit is called the group of units at P. consists of all nonunits of Ois the unique maximal ideal of O, and the residue class ring Oand call it the residue class field at P. The canonical map

    will be denoted by ψ . This map is known as the residue class map at P or as the place at P. There are alternative versions of the development of tools for algebraic number theory in which the notions of valuation ring and place, rather than that of valuation, play a central role.

    For archimedean P, the objects O, U are not defined.

    In case P .

    Before going on to more significant examples, it is convenient to set up another way of looking at a nonarchimedean prime divisor P of F. We recall that for φ P . Let us put

    . All the properties of φ carry over to ν, and ν is called an exponential valuation of F because it is a function from

    F such that

    Clearly there is a 1-1 correspondence between the set of all non-archimedean valuations of F and the set of all exponential valuations of F. To carry the connection further, we say that the exponential valuations ν and ν′ are equivalent (and the corresponding valuations φ = eν and φ′ = eν′ are equivalent. Thus,

    . No confusion can arise if we write ν P when eν = φ P. Hence, if ν P, then P = {αν|α > 0}. In terms of an exponential valuation ν P we have

    Of course, O, and U are independent of the choice of ν P.

    Any ν P (the additive group of reals). The image ν(F*) is a subgroup of R+; it is called the value group of ν and is often written as G(ν). If also νPfor some α —an order isomorphism. It is well known that any subgroup of R+ is either discrete or dense in R+ and that a discrete subgroup is either trivial (consisting of {0} alone) or infinite cyclic (generated by the smallest positive element that it contains). The prime divisor P is said to be discrete or nondiscrete according is discrete or nondiscrete in R+; naturally, this does not depend on the choice of ν P.

    Suppose that P is a discrete prime divisor of F. If, for some ν P, G(ν) = ν(F*) = {0}, then φ = eν is the trivial valuation, so that P is the trivial prime divisor. On the other hand, if P is discrete and nontrivial, then there exists in P a unique exponential valuation, which we shall invariably denote by νp. νp is then known as the normalized exponential valuation belonging to P. In such a situation we have

    1-4.The Prime Divisors of Q

    In this section, we determine all the nontrivial prime divisors of the rational field Q. In particular, we shall give many concrete examples of valuations and related objects.

    First of all, the ordinary absolute value function, which we write as φ∞ [that is, φ∞(a) = |a| for a Q], is a valuation. The prime divisor which it determines is archimedean and will be denoted by oo. It will usually be called the infinite prime of Q, and φ∞ will be known as the normed infinite valuation.

    To construct nonarchimedean valuations of Q, we make use of a procedure which applies to any field F which is the quotient field of a unique factorization domain. Let p Z be any prime number. Since Z is a unique factorization domain, every element of its quotient field Q can be expressed uniquely as a product of primes from Z—where, of course, negative exponents may appear. We define a function

    by putting for a Q*

    Thus

    and p runs over all primes of Z. Since all the primes p Z are taken to be > 0, ε is +1 or −1 according as a is > 0 or < 0. If we also put νpis an exponential valuation. The prime divisor to which νp belongs will be denoted simply by p. Clearly p is nonarchimedean—in fact it is a discrete nontrivial prime divisor of Q—and νp is the normalized exponential valuation belonging to p. Note that an element of Q is near 0 (for p) when it is divisible by a high power of p.

    From is a nonarchimedean valuation belonging to p; however, for reasons which will appear soon, we define a function φp by

    In other words, φp = φα, where α = loge p is a valuation belonging to p—it is called the normed, or normalized, p-adic valuation. If p and q are distinct primes of Z, then the prime divisors p and q are distinct; this follows from (4-9-1) and the fact that φP = 1/p ≤ 1, while φq(p) = 1.

    For each prime p, we have O, and U defined. From their description (as given in Section 1-3) one sees easily that

    .

    1-4-1. Exercise.Show that the residue class field of Q at the prime p is a field with p elements which may be written as

    1-4-2. Theorem.denote the set of all nontrivial prime divisors of Q; then

    For each prime divisor p , let φp be the normed valuation belonging to p; then for each a Q we have

    Proof. Before proceeding to the details of the proof, let us remark that the assertion of (i) and (ii) is usually stated in the form: The product formula holds in Q. It is the product formula which accounts for the way the normed valuations were selected. The fact that the product formula can be carried over to arbitrary algebraic number fields may be used as the central theme in the proof of some of the most important theorems of algebraic number theory; this will be done in Chapter 5.

    Coming back to the proof, let P be any nontrivial prime divisor of Q, and choose φ P with ||φ|| ≤ 2. Since Q is the quotient field of Z, it suffices to determine φ on Zfor some α on Q.

    If P is nonarchimedean, then ||φ|| = 1 and φ(n) ≤ 1 for all n Z. Consider

    We have I ≠ (0) because φ is nontrivial, and I Z because φ(l) = 1. Moreover, I is a prime ideal of Z and is therefore generated by a prime p. For any n Z with (n′, p, we have φ(nwhere 0 < (φ(p) < 1. Therefore, φ(p) = (l/p)α for some α > 0, and then φ(n) = φp(n)α for all n Z.

    Suppose then that P is archimedean. Fix any integers m and n both > 1. For any integer t > 0 we can write

    . Since a, ≥ 1, we have n* ≤ mt and s ≤ t(log m/log n, it follows that

    Taking ith roots and letting t → ∞ yields

    We assert that φ(n) > 1 for every n > 1; for if there exists n0 > 1 with φ(no) ≤ 1, then from the above relation φ(m) ≤ 1 for all m Z, which contradicts the archimedean character of φ, and because the assumptions are symmetric in m and n, it follows that

    for some α > 0 and all m, n > 1. Consequently, φ(n) = for n > 1, and then, for any n , and thus complete the proof of the first part of the theorem.

    for almost all p. Finally, we note that it suffices to verify the

    Enjoying the preview?
    Page 1 of 1