Professional Documents
Culture Documents
Full Ebook of Real Algebra A First Course 1St Edition Manfred Knebusch Claus Scheiderer Online PDF All Chapter
Full Ebook of Real Algebra A First Course 1St Edition Manfred Knebusch Claus Scheiderer Online PDF All Chapter
Full Ebook of Real Algebra A First Course 1St Edition Manfred Knebusch Claus Scheiderer Online PDF All Chapter
https://ebookmeta.com/product/abstract-algebra-a-first-
course-2nd-edition-stephen-lovett/
https://ebookmeta.com/product/introduction-to-real-analysis-3rd-
edition-manfred-stoll/
https://ebookmeta.com/product/introduction-to-real-analysis-3rd-
edition-manfred-stoll-2/
https://ebookmeta.com/product/a-first-course-in-spectral-
theory-1st-edition-milivoje-lukic/
A First Course in Group Theory 1st Edition Bijan Davvaz
https://ebookmeta.com/product/a-first-course-in-group-theory-1st-
edition-bijan-davvaz/
https://ebookmeta.com/product/business-statistics-a-first-
course-8th-edition-david-levine/
https://ebookmeta.com/product/business-statistics-a-first-
course-8th-edition-david-levine-2/
https://ebookmeta.com/product/course-1-real-estate-essentials-
humber-real-estate-education-7th-edition-real-estate-council-of-
ontario-reco/
https://ebookmeta.com/product/a-first-course-in-functional-
analysis-1st-edition-orr-moshe-shalit/
Universitext
Manfred Knebusch
Claus Scheiderer
Real Algebra
A First Course
Translated and with Contributions by
Thomas Unger
Universitext
Series Editors
Carles Casacuberta, Universitat de Barcelona, Barcelona, Spain
John Greenlees, University of Warwick, Coventry, UK
Angus MacIntyre, Queen Mary University of London, London, UK
Claude Sabbah, École Polytechnique, CNRS, Université Paris-Saclay, Palaiseau,
France
Endre Süli, University of Oxford, Oxford, UK
Universitext is a series of textbooks that presents material from a wide variety of
mathematical disciplines at master’s level and beyond. The books, often well class-
tested by their author, may have an informal, personal even experimental approach
to their subject matter. Some of the most successful and established books in the
series have evolved through several editions, always following the evolution of
teaching curricula, into very polished texts.
Thus as research topics trickle down into graduate-level teaching, first textbooks
written for new, cutting-edge courses may make their way into Universitext.
Manfred Knebusch • Claus Scheiderer
Real Algebra
A First Course
Translated by
Thomas Unger
School of Mathematics and Statistics
University College Dublin
Dublin, Ireland
Translation from the German language edition: “Einführung in die reelle Algebra” by Manfred Knebusch
et al., © Friedr. Vieweg & Sohn Verlagsgesellschaft mbH 1989. Published by Friedr. Vieweg & Sohn,
Braunschweig/Wiesbaden. All Rights Reserved.
This Springer imprint is published by the registered company Springer Nature Switzerland AG
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
Preface
More than 30 years after its publication, we are pleased and surprised with the
ongoing interest in Einführung in die reelle Algebra. Given the vibrant development
of real algebra and geometry in the previous decades, this seems by no means self-
evident.
Real algebra has grown in many directions, partially from within itself, and
exceedingly also through nudges and stimuli from without. While some of these
developments were already discernible at the end of the 1980s, many modern ones
that currently belong to the core of the area—such as the connections with tropical
geometry, or with semidefinite optimization—would have been difficult to predict.
Hence, it is not obvious that a text that was timely and modern in 1989 still serves
current needs in a reasonable manner.
Nevertheless, we are convinced that also from today’s perspective the choice
and exposition of the material in Einführung in die reelle Algebra still constitute a
solid first course on the basic principles and important techniques of real algebra.
For this reason, we decided to keep the text essentially unchanged in the English
translation—even though it was tempting to change the wording here and there, and
to drop some sections in favour of new ones.
A small number of typos, omissions and errors that were present in the German
original have been corrected, and some notation and terminology have been brought
up to date. Several new examples (some in the form of exercises) have been added.
All definitions and propositions have been numbered in a consistent manner. The
original bibliography has been augmented with a number of newer references.
A feature of this translation is the addition of a short fourth chapter that provides
a succinct overview of the most important developments and advances in those parts
of real algebra that are directly related to material covered in Einführung in die reelle
Algebra.
We are grateful to Springer Nature, and Rémi Lodh in particular, for their
endorsement and friendly and professional assistance.
We received support and encouragement from many quarters, and are pleased to
extend special thanks to Oliver Hien and Marcus Tressl, who made valuable remarks
on the German original.
v
vi Preface
Finally, our very special and heartfelt thanks go to Thomas Unger, who translated
the original text in a conscientious and competent manner, and co-authored the final
chapter. It was always a great pleasure to work with him.
Algebra textbooks that are currently in common use present real algebra only in
later chapters and then usually rather tersely. This is so for the influential works
of van der Waerden [113], Jacobson [55] and Lang [73], although Jacobson covers
a bit more than the others. Bourbaki’s substantial multivolume text Eléments de
Mathématique shows a similar picture: the volume Algèbre contains just a short
chapter (Chapter 6, Groupes et corps ordonnés) on real algebra. In contrast, a
complete volume (currently counting nine chapters) is dedicated to commutative
algebra, even though on the whole only the elementary part is covered, and not
more than what is absolutely essential for the foundation of the theory, both by
current standards as well as Bourbaki’s own standards.
That being the case, not too many algebraists today seem to perceive real algebra
as a proper branch of algebra at all. This has not always been so. In the nineteenth
century real algebra flourished. The study of real zeroes of a real polynomial in one
variable was at the centre of algebraic interest during the whole century, and was an
indispensable part of any higher mathematical education.
Heinrich Weber’s large three-volume textbook on algebra [114] provides evi-
dence for this fact. As Weber’s research interests were primarily in number theory,
especially in complex multiplication and class field theory, he focussed his textbook
mostly on these topics. Nevertheless he devoted well over a hundred pages in the
first volume to real zeroes of real polynomials.
The twentieth century witnessed a dramatic decline in interest in real algebra.
This trend only seems to have changed since the late 1970s, which is all the more
astonishing since the essential seeds of a modern real algebra, as we understand it
today, are already present in two works by Artin and Schreier from the 1920s [3, 4].
So, what is real algebra? Instead of giving a formal definition—which would
be difficult—we rather answer the question with an analogy that puts real algebra
in parallel with commutative algebra. Commutative algebra can be seen as that
part of algebra that contains the algebraic foundations that are typically important
for algebraic geometry (in particular in its modern, abstract form); and algebraic
geometry is ultimately the study of solution sets of systems of polynomial equations
F (x1 , . . . , xn ) = 0 and non-equations F (x1 , . . . , xn ) = 0. Correspondingly,
vii
viii Preface to Einführung in die reelle Algebra
real algebra provides algebraic methods that typically function as tools for real
algebraic geometry, and in particular semialgebraic geometry, which is the study
of the solution sets of systems of polynomial inequalities F (x1 , . . . , xn ) > 0 or
F (x1 , . . . , xn ) ≥ 0, where the coefficients classically come from the field of real
numbers and more generally from any ordered field.
This analogy provides an argument for why the interest has been so much
stronger in commutative than in real algebra during our century thus far. Indeed,
algebraic geometry experienced a continuous and ultimately triumphant upswing in
the twentieth century, while real algebraic geometry was only practiced in isolation
and then mostly with transcendental methods. Thus, from the geometric point of
view, there was for many decades no engine available that could have pushed real
algebra forward.
It was not until 1987 that the first textbook [11] on real algebraic geometry
was published. Its authors’ report (in the last section of the introduction to this
commendable work [11, p. 4 ff.]) on the strange sleeping beauty slumber of real
algebraic geometry (insofar as it was practiced with algebraic methods) is worth
considering.
In the meantime this slumber has given way to a lively development. We
consider the introduction—or better: discovery—of the real spectrum Sper A of a
commutative ring A by Michel Coste and Marie-Françoise Roy around the year
1979 as the most important trigger.
One can view Sper A as an analogue of the Zariski spectrum Spec A, intro-
duced by Grothendieck. It is well-known that the Zariski spectrum is the key
to Grothendieck’s abstract algebraic geometry. Likewise, the real spectrum is the
key to an abstract semialgebraic geometry. (There is one difference: in contrast to
Spec A, the real spectrum Sper A seems to carry (at least) two important structure
sheafs, the sheaf of abstract Nash functions, introduced by Coste and Roy [29], as
well as the sheaf of abstract semialgebraic functions, introduced by G. Brumfiel and
N. Schwartz [19, 107, 108].)
Real algebra is necessary to understand the latest developments in real algebraic
geometry and to meet the future intellectual challenges in this area. This brings us
to the objective of this book.
Our book is based on two insights: real algebra is a branch of algebra that is
largely autonomous in its foundations, with methods specific to this branch. These
foundations can be successfully taught with little more preparation than the standard
background from a typical one-semester algebra course on linear algebra, group
theory, field theory and ring theory.
We differentiate more precisely between an elementary and a higher real algebra.
The former can be developed without any special prior knowledge and used with
benefit in real algebraic geometry. The latter makes serious use of resources from
other branches of mathematics, especially real algebraic geometry, commutative
algebra, algebraic geometry, model theory and the theory of quadratic forms, but
occasionally also algebraic topology, real analysis and complex analysis. (This list
can certainly be extended.) An analogous distinction can be made in commutative
algebra. A demarcation between “elementary” and “higher” in either area is not
Preface to Einführung in die reelle Algebra ix
entirely objectively possible, but rather is subject to personal points of view and
taste. Furthermore, the higher one goes, the more fluid and arbitrary the boundary
between the two kinds of algebra and their corresponding geometries becomes.
Our book is dedicated to elementary real algebra in the above sense. Another
book on higher real algebra is planned.1 In the current book we do get by with
previous knowledge of the extent sketched above. By adding another twenty to
thirty pages, we could have reduced the requirements further and, for example,
developed everything that is needed from commutative algebra and the theory of
quadratic forms. We have not done so however, since students who are interested in
real algebra more than likely already mastered almost all of the prerequisites for the
current book, and would easily be able to find the few things they may be missing
elsewhere.
One specific feature of real algebra should be pointed out in particular: the major
role that general (Krull) valuation rings play. In most parts of commutative algebra,
valuation rings that are not discrete are only viewed as an aid that can often be
done without. In real algebra on the other hand, general valuation rings are a natural
and even central concept throughout. The reason is that every convex subring of
an ordered field is a valuation ring, but only in rare cases a discrete valuation
ring. This fact was already observed by Krull in the introduction to his pioneering
work Allgemeine Bewertungstheorie [67]. (It can already be found in embryonic
form, without the concept of valuation ring, in the work of Artin and Schreier,
cf. [4, p. 95].) Krull also identified the theory of ordered fields as an important
application area of general valuation theory, but then devoted only one section—
albeit substantial—of his great work to ordered fields [67, §12].
Our book is divided into three chapters. In addition to the Artin-Schreier theory
of ordered fields and the elementary relationships between orderings and quadratic
forms, the first chapter covers some aspects of real algebra from the nineteenth
century. Various methods for determining the number of real zeroes of a real
polynomial are treated (Sturm’s algorithm, Hermite’s method using quadratic forms,
Hurwitz’ Theorem). Another section is devoted to the relationship between the
Cauchy index and the Hankel form, as well as the Bézoutian of a rational function.
The second chapter deals with real valuation theory. Everything we need from
general valuation theory is developed from scratch. The chapter culminates in a
presentation of Artin’s solution of Hilbert’s 17th Problem.
Finally, the third chapter is devoted to the real spectrum. After a short crash
course on the Zariski spectrum, the real spectrum is examined in detail, but only
as a topological space (i.e., without the introduction of a structure sheaf). In the
geometric setting (affine algebras over real closed fields), the points as well as
certain subsets of the real spectrum are described in terms of filter sets. In the last
five sections of the chapter we come to some parts of real algebra in which the real
spectrum proves to be clarifying and helpful, such as the reduced Witt ring of a field,
Regensburg
January 1989
Contents
xi
xii Contents
References .. .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 195
Symbol Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 201
Index . . . . . . . . .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . 203
Chapter 1
Ordered Fields and Their Real Closures
Our starting point is the basic notion for the entire book, the general concept of
orderings of arbitrary fields. Conceived by Artin and Schreier in their foundational
1927 paper [4], it was successfully used by Artin in his solution of Hilbert’s 17th
Problem in the same year [3]. We introduce real closed fields and show that they
have the same algebraic properties as the field of real numbers. Moreover we prove
the existence and uniqueness of a real closure for any ordered field.
The second main topic in this chapter concerns methods for counting real roots
of real polynomials. Instead of trying to calculate or estimate the roots numerically,
we discuss purely algebraic methods that extend from the real numbers to any real
closed base field. These results were mostly found in the nineteenth century or even
earlier, and are connected to famous names such as Descartes, Sturm, Sylvester and
Hermite.
Parallel to these two subjects we give an introduction to the basic notions of
quadratic forms over fields and their algebraic theory. They will find applications in
Chaps. 2 and 3.
Let K be a field.
Definition 1.1.1 An ordering of K is a subset P of K that satisfies the properties
(O1) P + P ⊆ P , P P ⊆ P ,
(O2) P ∩ (−P ) = {0},
(O3) P ∪ (−P ) = K,
where P + P := {a + b : a, b ∈ P } and P P := {ab : a, b ∈ P }. The pair (K, P )
is called an ordered field.
Remark 1.1.2 If we assume (O1) and (O3), then (O2) is equivalent with
(O2’) −1 ∈ P .
a ≤P b :⇔ b − a ∈ P (a, b ∈ K)
defines a total order relation ≤P on the set K which satisfies the properties
(i) a ≤P b ⇒ a + c ≤P b + c,
(ii) a ≤P b, c ≥ 0 ⇒ ac ≤P bc
for all a, b, c ∈ K. Conversely, let ≤ be a total order relation on K which satisfies
(i) and (ii). Show that P := {a ∈ K : a ≥ 0} is an ordering of K.
Thus we clearly have a one-one correspondence between orderings of K and total
order relations on K that satisfy (i) and (ii). Usually the latter are called orderings
of K as well. When P is clear we simply write a ≤ b instead of a ≤P b.
Definition 1.1.4 A preordering of K is a subset T of K that satisfies the properties
(P1) T + T ⊆ T , T T ⊆ T ,
(P2) T ∩ (−T ) = {0},
(P3) a 2 ∈ T for all a ∈ K.
a ≤T b :⇔ b − a ∈ T (a, b ∈ K)
for all a ∈ K.
4 1 Ordered Fields and Their Real Closures
P = {f ∈ Q(t) : f (ϑ) ≥ 0}
and
Pa,− := {0} ∪ (a − t)r f (t) : r ∈ Z, f ∈ F (t) with f (a) = ∞ and f (a) > 0
are orderings of F (t) for every a ∈ F . Both orderings extend the ordering of F .
The reason for this notation is that a < t < b with respect to Pa,+ holds for all
b ∈ F with b > a. Thus F (t) is ordered in such a way that the transcendental t
is located “immediately to the right of a” on the “line” F . Similarly, t is located
“immediately to the left of a” with respect to Pa,− .
Notation 1.1.14 If (K, P ) is an ordered field, we denote the sign function with
respect to P by signP : K → {−1, 0, 1}. Thus, signP (0) = 0, and for a ∈ K ∗ we
have signP (a) = 1 if a ∈ P and signP (a) = −1 if a ∈ P .
As usual, | · |P : K → P , |a|P := a · signP (a) denotes the absolute value. If P
is clear, the index P may be omitted.
The meaning of the generalized intervals [a, b]P , [a, b[P , ]a, b]P , ]a, b[P for
a, b ∈ K ∪ {−∞, +∞} is also clear, namely [a, b[P = {x ∈ K : a ≤P x <P b},
]a, ∞[P = {x ∈ K : x >P a}, etc. Here too we will usually drop the index P , and
instead often write [a, b]K , etc. if several fields are considered.
1.2 Quadratic Forms, Witt Rings, Signatures 5
n
qB (x) = bij xi xj x = (x1 , . . . , xn ) ∈ K n .
i,j =1
a1 , . . . , am ⊥ b1 , . . . , bn ∼
= a1 , . . . , am , b1 , . . . , bn
6 1 Ordered Fields and Their Real Closures
and
a1 , . . . , am ⊗ b1 , . . . , bn ∼
= ⊥a b .
i,j
i j
spaces that are Witt equivalent to ϕ by [ϕ] (the Witt class of ϕ), and the set of all
Witt classes of (nondegenerate) quadratic spaces over K by W (K).
Theorem 1.2.4 The set W (K) is made into a commutative unitary ring with unit
[1] by the (well-defined) pairings
Example 1.2.7 (Milnor’s Exact Sequence) For every prime number p, we denote
the finite field with p elements by Fp . Then W (Q) can be computed explicitly as an
additive group by the split exact sequence
where i : Z → W (Q) is the unique homomorphism that sends 1 to 1, and ∂2 and
the ∂p for odd primes p are certain group homomorphisms, cf. [72, VI, §4] for the
details. Witt groups of number fields are considerably harder to understand, cf. [72,
VI, §3]. There is a similar split exact sequence for Witt groups of rational function
fields in one variable, cf. [72, IX, §3].
Next we study the relations between the Witt ring of K and orderings of K.
Definition 1.2.8 Let (K, P ) be an ordered field and let ϕ = (V , q) be a quadratic
space over K. Then ϕ (or q) is called positive definite with respect to P if q(v) >P 0
for all 0 = v ∈ V . If −q is positive definite, then q is called negative definite.
The following theorem is well-known:
Theorem 1.2.9 (Sylvester’s Inertia Theorem) Let ϕ be a possibly degenerate
quadratic space over K and P an ordering of K. Then there is an orthogonal
decomposition
ϕ∼
= ϕ+ ⊥ ϕ− ⊥ Rad(ϕ),
Lemma 1.2.11 Let (K, P ) be an ordered field. Then [ϕ] → signP ϕ defines a ring
homomorphism signP from W (K) to Z.
Proof The following equalities are easily verified:
a, b ∼
= a + b, (a + b)ab .
Proof Since a + b is represented by a, b, there exists c ∈ K ∗ such that a, b ∼
=
a + b, c. Comparing determinants gives abK ∗2 = (a + b)cK ∗2, i.e., cK ∗2 =
(a + b)abK ∗2. The statement follows.
Theorem 1.2.12 justifies:
Definition 1.2.14 A signature of a field K is a ring homomorphism W (K) → Z.
10 1 Ordered Fields and Their Real Closures
Remark 1.2.15 We see that a field admits a signature if and only if it is real.
Note that for every field K (with char K = 2) there exists a ring homomorphism
e : W (K) → Z/2Z, defined by e[ϕ] := dim(ϕ) + 2Z. Every signature σ yields a
commutative diagram
Definition 1.2.16 The map e : W (K) → Z/2Z is called the dimension index. Its
kernel is denoted I (K) and is called the fundamental ideal of W (K). (Thus I (K)
consists of the Witt classes of even dimensional quadratic spaces.)
In the following results we consider field extensions for which every ordering
extends:
Proposition 1.3.5 Let L/K be a finite field extension of odd degree. Then every
ordering of K extends to L.
By Proposition 1.3.2, this proposition follows from
Theorem 1.3.6 (T.A. Springer) If L/K is a finite field extension of odd degree and
q is an anisotropic quadratic form over K, then qL is anisotropic over L.
Proof We may assume without loss of generality that L = K(α) is a simple field
extension. Let f ∈ K[t] be the minimal polynomial of α over K. We proceed
by induction on n = [L : K] = deg f . Assume that n > 1 is odd and that the
statement is true for all smaller degrees. Let q = a1 , . . . , am be an anisotropic
form over K. Assume for the sake of contradiction that qL is isotropic. Then there
exist g1 , . . . , gm , h ∈ K[t] with deg gi < n (i = 1, . . . , m) and not all gi = 0, such
that
Remark 1.3.9 From Theorem 1.3.6 and Proposition 1.3.8 it follows in particular
that the homomorphism iL/K : W (K) → W (L) is injective in case L/K is finite of
odd degree or purely transcendental.
1.4 The Prime Ideals of the Witt Ring 13
σ
that p = Ker W (K) − → Z → Z/pZ , i.e., that p = pσ + p · W (K). It is clear that
σ and p are uniquely determined by p. This establishes the proof.
Since in every (commutative) ring the nilradical (i.e., the set of all nilpotent
elements) is equal to the intersection of all prime ideals ([55, 68, 73], see also
Proposition 3.1.11), we obtain immediately
14 1 Ordered Fields and Their Real Closures
Having determined the prime ideals of W (K), we can state a further criterion for
when orderings extend:
Proposition 1.4.4 Let L/K be a field extension. A signature of K extends to L if
and only if it vanishes on the kernel of iL/K : W (K) → W (L).
For the proof we require a simple fact from commutative algebra:
Lemma 1.4.5 If φ : A → B is an injective ring homomorphism and p a minimal
prime ideal of A, then there exists a prime ideal q of B such that p = φ −1 (q).
Proof Since S := φ(A\p) does not contain zero, we have that S −1 B = 0 and every
prime ideal q of S −1 B satisfies p = φ −1 j −1 (q ) , where j : B → S −1 B denotes
the canonical homomorphism.
Proof of Proposition 1.4.4 Let I := Ker iL/K and let σ be a signature of K. If σ
has an extension τ , then σ = τ ◦ iL/K , i.e., σ (I ) = 0.
Conversely, assume that σ (I ) = 0, i.e., I ⊆ pσ := Ker σ . Then pσ /I is a
minimal prime ideal of W (K)/I by Theorem 1.4.1. By Lemma 1.4.5 (applied to
−1
W (K)/I → W (L)) there exists a prime ideal q of W (L) such that pσ = iL/K (q).
Since Z ∼= W (K)/pσ → W (L)/q and again by Theorem 1.4.1, it follows that
W (K)/pσ → W (L)/q is an isomorphism, i.e., q determines an extension τ of σ .
In Sect. 1.3 we proved the injectivity of the map iL/K for extensions L/K that are
of odd degree or that are purely transcendental. Next, we will determine the kernel
of iL/K for quadratic extensions (and so with Proposition 1.4.4 give a new proof of
Proposition 1.3.3):
√
Proposition 1.4.6 Let a ∈ K ∗ \ K ∗2 and L = K( a). Then the kernel of iL/K is
the ideal of W (K) generated by 1, −a.
A more precise statement is:
Proposition 1.4.7 Let q be an anisotropic quadratic form over K. There exist
quadratic forms q , q over K with q ∼ = q ⊥ 1, −a ⊗ q such that qL is
anisotropic.
1.5 Real Closed Fields and Their Field Theoretic Characterization 15
Proof By induction on dim q. The case dim q = 1 is clear. Assume thus that
dim q ≥ 2 and without loss of generality that qL is isotropic. If (V , b) denotes
the bilinear space associated to q, then there exist v, w ∈ V , not both zero, such that
√ √
0 = qL (v + a w) = q(v) + a · q(w) + a · b(v, w).
It follows that q(v)+a q(w) = 0 = b(v, w). Since q is anisotropic, q(v), q(w) = 0.
Let W := Kv + Kw ⊆ V . Then dim W = 2 (if v, w were linearly dependent, then
we would have b(v, w) = 0), and q|W ∼= −ac, c = 1, −a⊗c for c := q(w). In
particular, q|W is nondegenerate. Since V = W ⊥ W ⊥ , we can apply the induction
hypothesis to q|W ⊥ and so conclude the proof.
After our excursion into the theory of quadratic forms we return to real algebra in
the narrower sense. This section is fundamental for everything that follows.
Definition 1.5.1 A field is called real closed if it is real and has no proper algebraic
extension that is real.
The field R of real numbers is a well-known example of a real closed field.
Proposition 1.5.2 Let K be a field. The following statements are equivalent:
(i) K is real closed;
(ii) there exists an ordering P of K that cannot be extended to any proper algebraic
extension of K;
(iii) K ∗2 ∪ {0} = {a 2 : a ∈ K} is an ordering of K, and every polynomial (in one
variable) of odd degree over K has a zero in K.
Moreover, if these conditions are satisfied, then K ∗2 ∪ {0} is the only ordering of K.
Proof The final comment is clear by (iii) (every ordering P satisfies K ∗2 ∪{0} ⊆ P ).
(i) ⇒ (ii) is clear since a field has an ordering if (and only if) it is real.
(ii)√⇒ (iii): If it were true that K ∗2 ∪ √{0} = P , then there would exist d ∈ P
with d ∈ K, and P would extend to K( d) (Proposition 1.3.3), a contradiction.
Furthermore, K has no proper extensions of odd degree by Proposition 1.3.5.
(iii) ⇒ (i): Let L ⊃ K be a finite proper field extension of K. We will show that
L is not real. By (iii), [L : K] is a power of 2. Since K has an ordering by (iii), it
follows that char K = 0 and in particular that L/K is separable. Let L1 be the Galois
L over K, G = Gal(L1 /K) and H = Gal(L1 /L). Then H is a subgroup
closure of
of G, and g∈G H g = {1} by Galois theory, −1
where H g := g H g. By elementary
g
group theory it follows that |G| divides g∈G [G : H ] . By assumption (iii) this
number is a power of 2, and so G is a 2-group. From Sylow’s Theorems (elementary
16 1 Ordered Fields and Their Real Closures
group theory again) it follows that there exists a subgroup H of G of index 2 with
H ⊂ H . Let F be the fixed field of H . Then K ⊂ F ⊂ L and [F : K] = 2.
√
There exists a ∈ K with F = K( a). Since K ∗2 ∪ {0} is an √ ordering of K and
a∈ / K ∗2 ∪{0}, there exists b ∈ K ∗ with a = −b2 . Hence −1 = ( a/b)2 is a square
in F , and so F is not real. We conclude that L is not real either.
Note that ψ is additive, i.e., ψ(b + b ) = ψ(b) + ψ(b ). Consider the trace
tr = trK/L : K → L. Then tr ◦ψ = (ψ|L )◦tr. Indeed, if b ∈ K and if b1 , . . . , bp are
the L-conjugates of b in K, then ψ(b1 ), . . . , ψ(bp ) are the L-conjugates of ψ(b),
and it follows that tr (ψ(b)) = ψ(bi ) = ψ( bi ) = ψ(tr b). Hence the diagram
i i
commutes. Since ψ and tr are surjective (the second map since K/L is separable),
the map ψ|L : L → L is also surjective, a contradiction.
(3) q = 2 (and char K = 2).
L contains the q-th roots of unity since [K : L] = q and the q-th cyclotomic
polynomial is of degree q − 1. Hence, since char K = q, there exists a ∈ L with
2
K = L(a 1/q ) (see for instance [55, Vol. I, §4.7] or [73, VI, §6]). For β := a 1/q we
obtain as in (1) that (NK/L (β))q = (−1)q−1 a. Since a is not a q-th power in L, q
must be even, thus q = 2.
√
(4) K = L( −1).
√
Let i = −1 ∈ K, and let a ∈ L and β ∈ K as in (3), thus with √ β = a
4
and K = L(β ). Since NK/L (β) = −β , we have β /NK/L (β) = −1, and so
2 2 4 2
√
K = L( −1).
(5) L = K, thus K = K(i).
If K = L, then also K(i) = K. Carrying out steps (1) to (4) for K(i) instead of
K leads to a contradiction by (4).
(6) K is real (and thus real closed).
Since i ∈ K it suffices to show that the sum of two squares in K is a square
(Proposition 1.5.2). Thus, let a, b ∈ K ∗ . Since K is algebraically closed, there are
c, d ∈ K with a+bi = (c+di)2. Hence, a−bi = (c−di)2 and a 2 +b 2 = (c2 +d 2 )2 .
This proves the theorem.
Let Ks denote the separable closure of K.
Corollary 1.6.2 If char K = 0 and [Ks : K] < ∞, then K = Ks .
Proof Steps (2)–(6) of the proof can be carried out for Ks just as for K.
We finish this section with a consequence of the theorem of Artin and Schreier
for absolute Galois groups of fields. These are profinite groups (i.e., projective
limits of finite groups). For the basic aspects of infinite Galois theory, [54, Vol. III,
§IV.2], [55, Vol. II, §8.6] or [12, Ch. V, §10], for example, can be consulted, but for
understanding what follows this is not required.
1.7 Counting Real Zeroes of Polynomials (without Multiplicities) 19
Corollary 1.6.3 Let K be a field with separable algebraic closure Ks and absolute
Galois group = Gal (Ks /K). All elements of finite order > 1 in are involutions
(i.e., have order 2), and the map τ → Fix (τ ) (the fixed field of τ in Ks = K) is a
bijection from the set of involutions in to the set of real closed overfields of K in
K. In particular, contains elements of finite order > 1 if and only if K is real.
One of the oldest problems in real analysis consists of determining the number
and location of the real zeroes of a polynomial f (t) with real coefficients. The
first general answer was formulated by J. C. F. Sturm (1803–1855) in the form of
an algorithm, the so-called Sturm sequence, that determines the number of zeroes
of f in any given interval, counted without multiplicities. Another solution of
the problem, based on the examination of certain quadratic forms, was found by
Ch. Hermite (1822–1901). Both methods will be presented in this section.
It should be remarked that these results used to be part of the mathematics
curriculum at the turn of the nineteenth century (see for example H. Weber’s
Lehrbuch der Algebra [114]).
Sturm’s solution, found in 1829, initiated an historically very interesting and to
some extent stormy period in real algebra. (Sturm published the full proof only
in 1835.) Hermite’s theorem (Theorem 1.7.17) was anticipated by C.G.J. Jacobi
(1804–1851) and his student and friend C.W. Borchardt (1817–1880) (Theo-
rem 1.7.15), and was proved in 1853 by Hermite and—essentially independently—
also by J.J. Sylvester (1814–1897). In addition, Sylvester, A. Cayley (1821–1895)
and Sturm established interesting links between the theorems of Sturm and Hermite.
In addition to the original literature, the reader can find full particulars of this history
in the substantial text [65] of Krein and Naimark, as well as in [39].
We start with a number of classical results, well-known from real analysis, that
stay valid for rational functions over arbitrary real closed fields.
In the remainder of this section R denotes a real closed field.
Proposition 1.7.1 Let f (t) = t n +a1 t n−1 +· · ·+an ∈ R[t] be a monic polynomial.
Then all real zeroes of f in R) are in the interval [−M, M],
of f (i.e., all zeroes
where M := max 1, |a1 | + · · · + |an | .
Proof If 0 = a ∈ R is such that f (a) = 0, then a = −(a1 + a2a −1 + · · ·+ an a 1−n ),
and so |a| ≤ 1 or |a| ≤ |a1 | + · · · + |an | by the triangle inequality.
20 1 Ordered Fields and Their Real Closures
f (a)
r
a − aj
−1 = sign = sign .
f (b) b − aj
j =1
Proposition 1.7.4 Let 0 = f ∈ R(t) and a, b ∈ R with a < b and f (a) = f (b) =
0. If f has neither poles nor zeroes in (a, b), then the number of zeroes of f in
]a, b[, counted with multiplicities, is odd.
f (a + ε) f (b − ε)
· < 0,
f (a + ε) f (b − ε)
f0 = q1 f1 − f2 ,
f1 = q2 f2 − f3 ,
..
.
fr−2 = qr−1 fr−1 − fr ,
fr−1 = qr fr ,
For x ∈ R let W (x) := Var (g0 (x), . . . , gr (x)). We claim that g0 has exactly W (a)−
W (b) distinct zeroes in [a, b].
To prove the claim, we first note that for every c ∈ R with g0 (c) · · · gr (c) = 0
the function W (x) is constant on a neighbourhood of x = c. Let g := g0 · · · gr , and
let c ∈ [a, b] be a zero of g. We distinguish three cases:
If g0 (c) = 0, then a < c < b, and the function x → Var (g0 (x), g1 (x)) has value
1 on ]c − ε, c[ and value 0 on [c, c + ε[, for some ε > 0. This follows from (c).
If gi (c) = 0, 0 < i < r, then by (d) there exists ε > 0, such that gi−1 (x)gi+1 (x)
is negative on ]c − ε, c + ε[. Hence, x → Var (gi−1 (x), gi (x), gi+1 (x)) is constant
equal to 1 on ]c − ε, c + ε[ (note that for every α ∈ R, Var(−1, α, 1) =
Var(1, α, −1) = 1 !).
If gr (c) = 0, and g0 (c) = 0 or r > 1, then gr−1 (c) = 0 by (d), and x →
Var (gr−1 (x), gr (x)) is constant on {x : 0 < |x − c| < ε} for some ε > 0, since gr
is semidefinite on [a, b].
From these three cases we obtain for all c ∈ [a, b]: If g0 (c) gr (c) = 0, then W (x)
is constant on a neighbourhood of x = c; if g0 (c) = 0, then W (x) is constant on
a punctured neighbourhood of x = c; if g0 (c) = 0, there exists ε > 0 and N ∈ Z
with W (x) = N for c − ε < x < c and W (x) = N − 1 for c < x < c + ε.
The claim then follows since g0 and gr do not vanish at the boundary points a, b.
(2) Now let f be as in the statement of the theorem and let (f0 , . . . , fr ) be the
Sturm sequence of f . We let gi := fi /fr ∈ R[t] (i = 0, . . . , r) and W (x) :=
Var (g0 (x), . . . , gr (x)) (x ∈ R). Then the sequence (g0 , . . . , gr ) satisfies (a)–(d)
from (1). Indeed, since gi−1 = hi gi − gi+1 (i = 1, . . . , r − 1) and gr−1 = hr gr for
suitable hi ∈ R[t], and since gr = 1, it follows that gi−1 and gi have no zeroes in
common (i = 1, . . . , r), and so (d) follows recursively. (a) and (b) are clear, and (c)
follows from Lemma 1.7.3 since g0 g1 = ff /fr2 .
Furthermore, g0 and f0 = f have the same zeroes (if fr (c) = 0, then also
f (c) = 0, and the zero c occurs with larger multiplicity in f than in fr !). It thus
follows from (1) that f has precisely W (a) − W (b) distinct zeroes in [a, b]. The
proof then follows from the observation that V (a) = W (a) and V (b) = W (b) since
fr (a) fr (b) = 0.
We state a more general version of Sturm’s Theorem, also due to Sturm. We make
use of
Definition 1.7.9 Let f ∈ R[t] and a, b ∈ R with a < b and f (a)f (b) = 0. A
generalized Sturm sequence of f on [a, b] is a sequence (f0 , . . . , fr ) in R[t] \ {0}
with r ≥ 1 that satisfies the properties
(1) f0 = f ;
(2) fr (a)fr (b) = 0, and fr is semidefinite on [a, b];
(3) for 0 < i < r and c ∈ [a, b] with fi (c) = 0, we have fi−1 (c)fi+1 (c) < 0.
(Warning: The Sturm sequence of f is usually not a generalized Sturm sequence of
f in the sense of Definition 1.7.9!)
1.7 Counting Real Zeroes of Polynomials (without Multiplicities) 23
The statement is clear by the proof of Theorem 1.7.8! (We only have to count the
different types of sign changes of ff1 .)
Remark 1.7.11 From the proof we see that simplifications of the original Sturm
algorithm are possible. For example:
(a) The original Sturm sequence f0 = f, f1 = f , . . . may be terminated at fs
provided that fs (a) fs (b) = 0 and fs is semidefinite on [a, b].
(b) For a remainder fi+1 in the modified Euclidean algorithm (Definition 1.7.6) it
is allowed to leave out those factors that are semidefinite on [a, b] and that do
not vanish at a, b.
n
n
f = tn + ai t n−i = (t − ξj )
i=1 j =1
be a monic polynomial in K[t] that splits over K (with ai , ξj ∈ K). Prove that
ξ1 , . . . , ξn ≥ 0 ⇔ (−1)i ai ≥ 0 for i = 1, . . . , n,
sj +k−2 (f ) 1≤j,k≤n
.
n
In other words, S(f )(x1 , . . . , xn ) = sj +k−2 (f )xj xk .
j,k=1
Theorem 1.7.15 (Jacobi, Borchardt, Hermite) Let R be a real closed field and
f ∈ R[t] a monic non-constant polynomial. Then the rank of √ S(f ) equals the
number of distinct roots of f in the algebraic closure C = R( −1) of R, and
the signature of S(f ) equals the number of distinct real roots of f .
Note that therefore S(f ) always has nonnegative signature!
Proof Let β1 , . . . , βr be the pairwise distinct real roots of f , and γ1 , γ1 , . . . , γs , γs
the pairwise distinct non-real roots of f , where r, s ≥ 0 and α → α denotes
the nontrivial R-automorphism of√C. Denote the multiplicity of βj by mj and the
multiplicity of γj by nj . Let i = −1 ∈ C, and let Re α, Im α for α ∈ C have their
usual meaning. Then
S(f ) (x1 , . . . , xn ) = αjk+l−2 xk xl
1≤j,k,l≤n
n
n 2
= αjk−1 xk
j =1 k=1
r 2
s 2 2
= mj βjk−1 xk + nj γjk−1 xk + γj k−1 xk
j =1 k j =1 k k
r 2
= mj βjk−1 xk
j =1 k
s 2 2
+2 nj Re(γjk−1 )xk − Im(γjk−1 )xk ,
j =1 k k
we have
r
s
S(f ) = mj u2j + 2 nj (vj2 − wj2 ),
j =1 j =1
S(f ) ∼
= r × 1 ⊥ s × H ⊥ (n − r − 2s) × 0,
from which we immediately see that rank S(f ) = r + 2s, sign S(f ) = r.
The following modification makes it possible to also count real zeroes in
intervals:
Definition 1.7.16 Let K be a field, f ∈ K[t] a monic polynomial of degree n ≥ 1,
and λ ∈ K. The Sylvester form of f with parameter λ is the quadratic form
(x1 , . . . , xn ) → λ sj +k−2 (f ) − sj +k−1 (f ) xj xk
1≤j,k≤n
over K, denoted by Sλ (f ).
Theorem 1.7.17 (Hermite, Sylvester, 1853) Let R be real closed and f ∈ R[t] a
monic polynomial of degree n ≥ 1. For λ ∈ R we have
√
rank Sλ (f ) = # a ∈ R( −1) : f (a) = 0 and a = λ ,
sign Sλ (f ) = # a ∈ R : f (a) = 0, a < λ − # a ∈ R : f (a) = 0, a > λ .
n 2
= (λ − αj ) αjk−1 xk
j =1 k
r
= mj (λ − βj )u2j
j =1
s
+ nj (λ − γj )(vj + iwj )2 + (λ − γj )(vj − iwj )2 .
j =1
26 1 Ordered Fields and Their Real Closures
We rewrite [· · · ] as follows:
Hence,
r
s
Sλ (f ) = mj (λ − βj )u2j + 2nj ϕj (vj , wj ).
j =1 j =1
Sλ (f ) ∼
= λ − β1 , . . . , λ − βr ⊥ s × H ⊥ (n − r − 2s) × 0.
r
trA/K (a1 , . . . , ar ) = trAi /K (ai ) for all (a1 , . . . , ar ) ∈ A.
i=1
Notation 1.8.3
(a) Let A be a K-algebra and a ∈ A. Then [A, a]K denotes the quadratic space
over K defined by the quadratic form x → trA/K (ax 2 ) on A.
(b) For polynomials f, g ∈ K[t], f = 0, let
and
[A, a] ∼
= A , ϕ(a) for all a ∈ A.
28 1 Ordered Fields and Their Real Closures
[A, a]K ⊗K K ∼
= [A ⊗K K , a ⊗ 1]K (over K ).
(c) A1 × · · · × Ar , (a1 , . . . , ar ) ∼
= [A1 , a1 ] ⊥ · · · ⊥ [Ar , ar ] (ai ∈ Ai ).
(d) Nil A ⊆ Rad[A, a] for all a ∈ A.
SylK (f ; g) ⊗K K ∼
= SylK (f ; g) (over K ).
Syl(f1 · · · fr ; g) ∼
= Syl(f1 ; g) ⊥ · · · ⊥ Syl(fr ; g).
(c) Syl(f e ; g) ∼
= e ⊗ Syl(f ; g) ⊥ 0, . . . , 0 (e ∈ N).
(d) Syl(af ; g) = Syl(f ; g) and Syl(f ; ag) ∼ = a ⊗ Syl(f ; g) for any a ∈ K ∗ .
(e) Syl (t − a; g(t)) = g(a) for any a ∈ K.
Proof (a) and (b) follow from Lemma 1.8.4, and (d) and (e) are clear. We show (c):
let A = K[t]/(f e ), A = K[t]/(f ), and π : A → A the residue class map. By
Proposition 1.8.2 we have trA/K (a) = e · trA /K (πa) for all a ∈ A, from which the
proof follows.
r
∼
= ⊥
j =1
ej ⊗ Syl t − αj ; g(t) ⊥ (ej − 1) × 0
r
∼
= ⊥
j =1
ej g(αj ) ⊥ (ej − 1) × 0 ,
r
s
f (t) = (t − βj )mj qj (t)nj ,
j =1 j =1
r s
∼
= ⊥ Syl(t − β ; g) ⊥ ⊥ Syl(q ; g) ⊥ 0, . . . , 0
(c) j =1
j
j =1
j
s
∼
= g(β1 ), . . . , g(βr ) ⊥
(e)
⊥ Syl(q ; g) ⊥ 0, . . . , 0.
j =1
j
The forms√ Syl(qj ; g) are all hyperbolic (or zero in case qj | g) since R[t]/(qj ) and
C = R( −1) are isomorphic as R-algebras, and [C; α]R ∼ = 1, −1 = H for all
α ∈ C ∗ . (Choosing β ∈ C with 2αβ 2 = i, the form [C; α]R is represented by the
matrix 01 10 with respect to the basis β, −iβ.)
It follows that
Syl(f ; g) ∼
= g(β1 ), . . . , g(βr ) ⊥ t × H ⊥ (n − r − 2t) × 0
S(f ) ∼
= Syl(f ) = Syl(f ; 1) ∼
= r × 1 ⊥ s × H ⊥ (n − r − 2s) × 0
30 1 Ordered Fields and Their Real Closures
and
Syl(f ; λ − t) ∼
= λ − β1 , . . . , λ − βr ⊥ s × H ⊥ (n − r − 2s) × 0,
from which the statements of the theorems of Sylvester and Hermite can be read off
directly.
In addition to the works [65] and [39], mentioned before, the reader may want
to compare the current topic and related themes with the article [10] that contains
interesting historical remarks about Sylvester.
This section, and the next one, will not be used in the remainder of this book and can
therefore be skipped during a first reading. Having said that, its content is a classical
part of real algebra, and is essential for understanding the historical development
of mathematics in the nineteenth century. Furthermore, it has become important in
many applications such as the stability of motion or control theory.
We will follow the paper [65] of Krein and Naimark, but can only cover a small
part for lack of space. The interested reader may therefore want to consult the
original source.
R will always denote a real closed field.
Definition 1.9.1 Let ϕ(t) = g(t)/f (t) ∈ R(t) be a rational function with g and f
relatively prime polynomials over R.
(a) Let α ∈ R be a pole of ϕ (i.e., a zero of f ). The Cauchy index indα (ϕ) of ϕ at
α is defined as follows:
⎧
⎪
⎨+1 if the value of ϕ(t) jumps from −∞ to +∞ at t = α
⎪
indα (ϕ) := −1 if the value of ϕ(t) jumps from +∞ to −∞ at t = α .
⎪
⎪
⎩ 0 otherwise
(b) The Cauchy index of ϕ in an open interval ]a, b[ ⊆ R (a and b are allowed to be
±∞), denoted Iab (ϕ), is defined as the sum of the Cauchy indices of ϕ at those
poles of ϕ that are in ]a, b[. If a = −∞ and b = +∞, we speak of the global
Cauchy index of ϕ. (Note that if ϕ has a pole at ∞, this pole is ignored for the
Cauchy index, no matter what a and b are.)
1.9 Cauchy Index of a Rational Function, Bézoutian and Hankel Forms 31
r
mi
ϕ(t) = ψ(t) + ,
t − αi
i=1
for some rational function ψ without real poles. Thus ϕ = f /f has precisely
r real poles α1 , . . . , αr and its Cauchy index at each one of them is +1.
(2) More generally: let f, h ∈ R[t] with f = 0. For ϕ = hf /f we have
+∞
I−∞ (ϕ) = # α ∈ R : f (α)= 0 and h(α) > 0 − # α ∈ R : f (α)= 0 and h(α)< 0 ,
f (x)g(y) − f (y)g(x)
n−1
= cij x i y j ,
x−y
i,j =0
32 1 Ordered Fields and Their Real Closures
where
i
cij = dk,i+j +1−k with dij := an−j bn−i − an−i bn−j .
k=0
Independently from these explicit formulas (for which we will have no further use),
it is clear that cij = cj i for all i, j = 0, . . . , n − 1.
Definition 1.9.3 The symmetric n × n-matrix (cj i )0≤i,j ≤n−1 is called the Bézout
matrix of f and g, and is denoted B(f, g). The associated quadratic form over K is
called the Bézoutian of f , also denoted B(f, g). Thus,
n−1
B(f, g) (x0 , . . . , xn−1 ) = cij xi xj .
i,j =0
Since deg g ≤ deg f we can write the rational function ϕ(t) = g(t)/f (t) as a
formal power series in t −1 :
ϕ(t) = s−1 + s0 t −1 + s1 t −2 + · · · .
Definition 1.9.4 The symmetric n × n-matrix (si+j )0≤i,j ≤n−1 is called the Hankel
matrix of f and g and is denoted H (f, g). (In general, any quadratic matrix (aij )
whose entry aij only depends on i +j is called a Hankel matrix.) Again we interpret
H (f, g) as a quadratic form,
n−1
H (f, g) (x0 , . . . , xn−1 ) = si+j xi xj ,
i,j =0
and then refer to it as the Hankel form of f and g. For 1 ≤ p ≤ n the truncated
matrix (si+j )0≤i,j ≤p−1 is denoted Hp (f, g).
Remark 1.9.5 H (f, g) clearly only depends on n and the rational function ϕ =
g/f . More generally, a Hankel matrix Hn (ϕ) as in Definition 1.9.4 can be associated
to any formal power series ϕ(t) ∈ K[[t −1 ]] in t −1 and any n ≥ 0. A well-known
(elementary) theorem of Frobenius says that ϕ(t) is rational (i.e., ϕ(t) ∈ K(t)) if
and only if there exists n ≥ 0 such that det Hm (ϕ) = 0 for all m ≥ n. Furthermore,
in this case the smallest such n is the degree of the denominator of ϕ, written as
an irreducible fraction, and every matrix Hm (ϕ) with m > n has rank n. See, for
example, [40, §16.10].
Another random document with
no related content on Scribd:
None of them said a word, but all of them wore bright, knowing
smiles.
By Monday morning, however, Burbank was himself again. The
rebuff given him by Don Channing had worn off and he was
sparkling with ideas. He speared Franks with the glitter in his eye
and said: "If our beams are always on the center, why is it necessary
to use multiplex diversity?"
Franks smiled. "You're mistaken," he told Burbank. "They're not
always on the button. They vary. Therefore, we use diversity
transmission so that if one beam fails momentarily, one of the other
beams will bring the signal in. It is analogous to tying five or six
ropes onto a hoisted stone. If one breaks, you have the others."
"You have them running all the time, then?"
"Certainly. At several minutes of time-lag in transmission, to try and
establish a beam failure of a few seconds' duration is utter
foolishness."
"And you disperse the beam to a thousand miles wide to keep the
beam centered at any variation?" Burbank shot at Channing.
"Not for any variation. Make that any normal gyration and I'll buy it."
"Then why don't we disperse the beam to two or three thousand
miles and do away with diversity transmission?" asked Burbank
triumphantly.
"Ever heard of fading?" asked Channing with a grin. "Your signal
comes and goes. Not gyration, it just gets weaker. It fails for want of
something to eat, I guess, and takes off after a wandering cosmic
ray. At any rate, there are many times per minute that one beam will
be right on the nose and yet so weak that our strippers cannot clean
it enough to make it usable. Then the diversity system comes in
handy. Our coupling detectors automatically select the proper signal
channel. It takes the one that is the strongest and subdues the rest
within itself."
"Complicated?"
"It was done in the heyday of radio—1935 or so. Your two channels
come in to a common detector. Automatic volume control voltage
comes from the single detector and is applied to all channels. This
voltage is proper for the strongest channel, but is too high for the
ones receiving the weaker signal; blocking them by rendering them
insensitive. When the strong channel fades and the weak channel
rises, the detector follows down until the two signal channels are
equal and then it rises with the stronger channel."
"I see," said Burbank, "Has anything been done about fading?"
"It is like the weather, according to Mark Twain," smiled Channing.
"'Everybody talks about it, but nobody does anything about it.' About
all we've learned is that we can cuss it out and it doesn't cuss back."
"I think it should be tried," said Burbank.
"If you'll pardon me, it has been tried. The first installation at Venus
Equilateral was made that way. It didn't work, though we used more
power than all of our diversity transmitters together. Sorry."
"Have you anything to report?" Burbank asked Channing.
"Nothing. I've been more than busy investigating the trouble we've
had in keeping the beams centered."
Burbank said nothing. He was stopped. He hoped that the secret of
his failure was not generally known, but he knew at the same time
that when three hundred men are aware of something interesting,
some of them will see to it that all the others involved will surely
know. He looked at the faces of the men around the table and saw
suppressed mirth in every one of them. Burbank writhed in inward
anger. He was a good poker player. He didn't show it at all.
He then went on to other problems. He ironed some out, others he
shelved for the time being. Burbank was a good business man. But
like so many other businessmen, Burbank had the firm conviction
that if he had the time to spare and at the same time was free of the
worries and paper work of his position, he could step into the
laboratory and show the engineers how to make things hum. He was
infuriated every time he saw one of the engineering staff sitting with
hands behind head, lost in a gazy, unreal land of deep thought.
Though he knew better, he was often tempted to raise hell because
the man was obviously loafing.
But give him credit. He could handle business angles to perfection.
In spite of his tangle over the beam control, he had rebounded
excellently and had ironed out all of the complaints that had poured
in. Ironed it out to the satisfaction of the injured party as well as the
Interplanetary Communications Commission, who were interested in
anything that cost money.
He dismissed the conference and went to thinking. And he assumed
the same pose that infuriated him in other men under him; hands
behind head, feet upon desk.
The moving picture theater was dark. The hero reached longing
arms to the heroine, and there was a sort of magnetic attraction.
They approached one another. But the spark misfired. It was blacked
out with a nice slice of utter blackness that came from the screen
and spread its lightlessness all over the theater. In the ensuing
darkness, there were several osculations that were more personal
and more satisfying than the censored clinch. The lights flashed on
and several male heads moved back hastily. Female lips smiled
happily. Some of them parted in speech.
One of them said: "Why, Mr. Channing!"
"Shut up, Arden," snapped the man. "People will think that I've been
kissing you."
"If someone else was taking advantage of the situation," she said,
"you got gypped. I thought I was kissing you and I cooked with gas!"
"Did you ever try that before?" asked Channing interestedly.
"Why?" she asked.
"I liked it. I merely wondered, if you'd worked it on other men, what
there was about you that kept you single."
"They all died after the first application," she said. "They couldn't
take it."
"Let me outta here! I get the implication. I am the first bird that hasn't
died, hey?" He yawned luxuriously.
"Company or the hour?" asked Arden.
"Can't be either," he said. "Come on, let's break a bottle of beer
open. I'm dry!"
"I've got a slight headache," she told him, "From what, I can't
imagine."
"I haven't a headache, but I'm sort of logy."
"What have you been doing?" asked Arden. "Haven't seen you for a
couple of days."
"Nothing worth mentioning. Had an idea a couple of days ago and
went to work on it."
"Haven't been working overtime or missing breakfast?"
"Nope."
"Then I don't see why you should be ill. I can explain my headache
away by attributing it to eyestrain. Since Billyboy came here, and
censored the movies to the bone, the darned things flicker like
anything. But eyestrain doesn't create an autointoxication. So, my
fine fellow, what have you been drinking?"
"Nothing that I haven't been drinking since I first took to my second
bottlehood some years ago."
"You wouldn't be suffering from a hangover from that hangover you
had a couple of weeks ago?"
"Nope. I swore off. Never again will I try to drink a whole quart of Two
Moons in one evening. It got me."
"It had you for a couple of days," laughed Arden. "All to itself."
Don Channing said nothing. He recalled, all too vividly, the rolling of
the tummy that ensued after that session with the only fighter that
hadn't yet been beaten: Old John Barleycorn.
"How are you coming on with Burbank?" asked Arden. "I haven't
heard a rave for—well, ever since Monday morning's conference.
Three days without a nasty dig at Our Boss. That's a record."
"Give the devil his due. He's been more than busy placating irate
citizens. That last debacle with the beam control gave him a real
Moscow winter. His reforms came to a stop whilst he entrenched.
But he's been doing an excellent job of squirming out from under. Of
course, it has been helped by the fact that even though the service
was rotten for a few hours, the customers couldn't rush out to some
other agency to get communications with the other planets."
"Sort of: 'Take us, as lousy as we are?'"
"That's it."
Channing opened the door to his apartment and Arden went in.
Channing followed, and then stopped cold.
"Great Jeepers!" he said in an awed tone. "If I didn't know—"
"Why, Don! What's so startling?"
"Have you noticed?" he asked. "It smells like the inside of a chicken
coop in here!"
Arden sniffed. "It does sort of remind me of something that died and
couldn't get out of its skin." Arden smiled. "I'll hold my breath. Any
sacrifice for a drink."
"That isn't the point. This is purified air. It should be as sweet as a
baby's breath."
"Some baby," whistled Arden. "What's baby been drinking?"
"It wasn't cow-juice. What I've been trying to put over is that the air
doesn't seem to have been changed in here for nine weeks."
Channing went to the ventilator and lit a match. The flame bent over,
flickered, and went out.
"Air intake is O.K.," he said. "Maybe it is I. Bring on that bottle,
Channing; don't keep the lady waiting."
He yawned again, deeply and jaw-stretchingly. Arden yawned, too,
and the thought of both of them stretching their jaws to the breaking-
off point made both of them laugh foolishly.
"Arden, I'm going to break one bottle of beer with you, after which I'm
going to take you home, kiss you good night, and toss you into your
own apartment. Then I'm coming back here and I'm going to hit the
hay!"
Arden took a long, deep breath. "I'll buy that," she said. "And tonight,
it wouldn't take much persuasion to induce me to snooze right here
in this chair!"
"Oh, fine," cheered Don. "That would fix me up swell with the
neighbors. I'm not going to get shotgunned into anything like that!"
"Don't be silly," said Arden.
"From the look in your eye," said Channing, "I'd say that you were
just about to do that very thing. I was merely trying to dissolve any
ideas that you might have."
"Don't bother," she said pettishly. "I haven't any ideas. I'm as free as
you are, and I intend to stay that way!"
Channing stood up. "The next thing we know, we'll be fighting," he
observed. "Stand up, Arden. Shake."
Arden stood up, shook herself, and then looked at Channing with a
strange light in her eyes. "I feel sort of dizzy," she admitted. "And
everything irritates me."
She passed a hand over her eyes wearily. Then, with a visible effort,
she straightened. She seemed to throw off her momentary ill feeling
instantly. She smiled at Channing and was her normal self in less
than a minute.
"What is it?" she asked. "Do you feel funny, too?"
"I do!" he said. "I don't want that beer. I want to snooze."
"When Channing would prefer snoozing to boozing he is sick," she
said. "Come on, fellow, take me home."
Slowly they walked down the long hallway. They said nothing. Arm in
arm they went, and when they reached Arden's door, their good-
night kiss lacked enthusiasm. "See you in the morning," said Don.
Arden looked at him. "That was a little flat. We'll try it again—
tomorrow or next week."
Don Channing's sleep was broken by dreams. He was warm. His
dreams depicted him in a humid, airless chamber, and he was forced
to breathe that same stale air again and again. He awoke in a hot
sweat, weak and feeling—lousy!
He dressed carelessly. He shaved hit-or-miss. His morning coffee
tasted flat and sour. He left the apartment in a bad mood, and
bumped into Arden at the corner of the hall.
"Hello," she said. "I feel rotten. But you have improved. Or is that
passionate breathing just a lack of fresh air?"
"Hell! That's it!" he said. He snapped up his wrist watch, which was
equipped with a stop-watch hand. He looked about, and finding a
man sitting on a bench, apparently taking it easy while waiting for
someone, Channing clicked the sweep hand into gear. He started to
count the man's respiration.
"What gives?" asked Arden, "What's 'It'? Why are you so excited?
Did I say something?"
"You did," said Channing after fifteen seconds. "That bird's
respiration is better than fifty! This whole place is filled to the gills
with carbon dioxide. Come on, Arden, let's get going!"
Channing led the girl by several yards by the time that they were
within sight of the elevator. He waited for her, and then sent the car
upward at a full throttle. Minutes passed, and they could feel that
stomach-rising sensation that comes when gravity is lessened.
Arden clasped her hands over her middle and hugged. She
squirmed and giggled.
"You've been up to the axis before," said Channing. "Take long, deep
breaths."
The car came to a stop with a slowing effect. A normal braking stop
would have catapulted them against the ceiling. "Come on," he
grinned at her, "here's where we make time!"
Channing looked up at the little flight of stairs that led to the
innermost level. He winked at Arden and jumped. He passed up
through the opening easily. "Jump," he commanded. "Don't use the
stairs."
Arden jumped. She sailed upward, and as she passed through the
opening, Channing caught her by one arm and stopped her flight. "At
that speed you'd go right on across," he said.
She looked up, and there about two hundred feet overhead she
could see the opposite wall.
Channing snapped on the lights. They were in a room two hundred
feet in diameter and three hundred feet long. "We're at the center of
the station," Channing informed her. "Beyond that bulkhead is the air
lock. On the other side of the other bulkhead, we have the air plant,
the storage spaces, and several rooms of machinery."
"Come on," he said. He took her by the hand and with a kick he
propelled himself along on a long, curving course to the opposite
side of the inner cylinder. He gained the opposite bulkhead as well.
"Now, that's what I call traveling," said Arden. "But my tummy goes
whoosh, whoosh every time we cross the center."
Channing operated a heavy door. They went in through rooms full of
machinery and into rooms stacked to the center with boxes; stacked
from the wall to the center and then packed with springs. Near the
axis of the cylinder, things weighed so little that packing was
necessary to keep them from floating around.
"I feel giddy," said Arden.
"High in oxygen," said he. "The CO2, drops to the bottom, being
heavier. Then, too, the air is thinner up here because centrifugal
force swings the whole out to the rim. Out there we are so used to
'down' that here, a half mile above—or to the center, rather—we
have trouble in saying, technically, what we mean. Watch!"
He left Arden standing and walked rapidly around the inside of the
cylinder. Soon he was standing on the steel plates directly over her
head. She looked up, and shook her head.
"I know why," she called, "but it still makes me dizzy. Come down
from up there or I'll be sick."
Channing made a neat dive from his position above her head. He did
it merely by jumping upward from his place toward her place,
apparently hanging head down from the ceiling. He turned a neat
flip-flop in the air and landed easily beside her. Immediately, for both
of them, things became right-side-up again.
Channing opened the door to the room marked: "Air Plant." He
stepped in, snapped on the lights, and gasped in amazement.
"Hell!" he groaned. The place was empty. Completely empty.
Absolutely, and irrevocably vacant. Oh, there was some dirt on the
floor and some trash in the corners, and a trail of scratches on the
floor to show that the life giving air plant had been removed, hunk by
hunk, out through another door at the far end of the room.
"Whoa, Tillie!" screamed Don. "We've been stabbed! Arden, get on
the type and have ... no, wait a minute until we find out a few more
things about this!"
They made record time back to the office level. They found Burbank
in his office, leaning back, and talking to someone on the phone.
Channing tried to interrupt, but Burbank removed his nose from the
telephone long enough to snarl, "Can't you see I'm busy? Have you
no manners or respect?"
Channing, fuming inside, swore inwardly. He sat down with a show
of being calm and folded his hands over his abdomen like the famed
statue of Buddha. Arden looked at him, and for all the trouble they
were in, she couldn't help giggling. Channing, tall, lanky, and strong,
looked as little as possible like the popular, pudgy figure of the Sitting
Buddha.
A minute passed.
Burbank hung up the phone.
"Where does Venus Equilateral get its air from?" snapped Burbank.
"That's what I want—"
"Answer me, please. I'm worried."
"So am I. Something—"
"Tell me first, from what source does Venus Equilateral get its fresh
air?"
"From the air plant. And that is—"
"There must be more than one," said Burbank thoughtfully.
"There's only one."
"There must be more than one. We couldn't live if there weren't,"
said the Director.
"Wishing won't make it so. There is only one."
"I tell you, there must be another. Why, I went into the one up at the
axis day before yesterday and found that instead of a bunch of
machinery, running smoothly, purifying air, and sending it out to the
various parts of the station, all there was was a veritable jungle of
weeds. Those weeds, Mr. Channing, looked as though they must
have been put in there years ago. Now, where did the air-purifying
machinery go?"
Channing listened to the latter half of Burbank's speech with his chin
at half-mast. He looked as though a feather would knock him clear
across the office.
"I had some workmen clear the weeds out. I intend to replace the air
machinery as soon as I can get some new material sent from Terra."
Channing managed to blink. It was an effort. "You had workmen toss
the weeds out—" he repeated dully. "The weeds—"
There was silence for a minute, Burbank studied the man in the chair
as though Channing were a piece of statuary. Channing was just as
motionless. "Channing, man, what ails you—" began Burbank. The
sound of Burbank's voice aroused Channing from his shocked
condition.
Channing leaped to his feet. He landed on his heels, spun, and
snapped at Arden: "Get on the type. Have 'em slap as many oxy-
drums on the fastest ship as they've got! Get 'em here at full throttle.
Tell 'em to load up the pilot and crew with gravanol and not to spare
the horsepower! Scram!"
Arden gasped. She fled from the office.
"Burbank, what did you think an air plant was?" snapped Channing.
"Why, isn't it some sort of purifying machinery?" asked the wondering
Director.
"What better purifying machine is there than a plot of grass?"
shouted Channing. "Weeds, grass, flowers, trees, alfalfa, wheat, or
anything that grows and uses chlorophyll. We breathe oxygen,
exhale CO2. Plants inhale CO2, and exude oxygen. An air plant
means just that. It is a specialized type of Martian sawgrass that is
more efficient than anything else in the system for inhaling dead air
and revitalizing it. And you've tossed the weeds out!" Channing
snorted in anger. "We've spent years getting that plant so that it will
grow just right. It got so good that the CO2 detectors weren't even
needed. The balance was so adjusted that they haven't even been
turned on for three or four years. They were just another source of
unnecessary expense. Why, save for a monthly inspection, that room
isn't even opened, so efficient is the Martian sawgrass. We, Burbank,
are losing oxygen!"
The Director grew white. "I didn't know," he said.
"Well, you know now. Get on your horse and do something. At least,
Burbank, stay out of my way while I do something."
"You have a free hand," said Burbank. His voice sounded beaten.
Channing left the office of the Director and headed for the chem lab.
"How much potassium chlorate, nitrate, sulphate, and other oxygen-
bearing compounds have you?" he asked. "That includes mercuric
oxide, spare water, or anything else that will give us oxygen if broken
down."
There was a ten-minute wait until the members of the chem lab took
a hurried inventory.
"Good," said Channing. "Start breaking it down. Collect all the
oxygen you can in containers. This is the business! It has priority!
Anything, no matter how valuable, must be scrapped if it can
facilitate the gathering of oxygen. God knows, there isn't by half
enough—not even a tenth. But try, anyway."