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

Only $11.99/month after trial. Cancel anytime.

Quantum Mechanics in Hilbert Space: Second Edition
Quantum Mechanics in Hilbert Space: Second Edition
Quantum Mechanics in Hilbert Space: Second Edition
Ebook1,186 pages19 hours

Quantum Mechanics in Hilbert Space: Second Edition

Rating: 4.5 out of 5 stars

4.5/5

()

Read preview

About this ebook

A critical presentation of the basic mathematics of nonrelativistic quantum mechanics, this text is suitable for courses in functional analysis at the advanced undergraduate and graduate levels. Its readable and self-contained form is accessible even to students without an extensive mathematical background. Applications of basic theorems to quantum mechanics make it of particular interest to mathematicians working in functional analysis and related areas.
This text features the rigorous proofs of all the main functional-analytic statements encountered in books on quantum mechanics. It fills the gap between strictly physics- and mathematics-oriented texts on Hilbert space theory as applied to nonrelativistic quantum mechanics. Organized in the form of definitions, theorems, and proofs of theorems, it allows readers to immediately grasp the basic concepts and results. Exercises appear throughout the text, with hints and solutions at the end.
LanguageEnglish
Release dateJul 2, 2013
ISBN9780486318059
Quantum Mechanics in Hilbert Space: Second Edition
Author

Eduard Prugovecki

EDUARD PRUGOVECKI is a scientist who acquired his Ph.D. from Princeton University. While a professor at the University of Toronto he published a host of research papers and four monographs on quantum physics. After his retirement from academic life he completed his first futuristic novel Memoirs of the Future (Cross Cultural Publications, 2001). He now lives in Chapala, Mexico, and is working on a new futuristic novel.

Read more from Eduard Prugovecki

Related to Quantum Mechanics in Hilbert Space

Titles in the series (100)

View More

Related ebooks

Physics For You

View More

Related articles

Reviews for Quantum Mechanics in Hilbert Space

Rating: 4.333333333333333 out of 5 stars
4.5/5

3 ratings0 reviews

What did you think?

Tap to rate

Review must be at least 10 words

    Book preview

    Quantum Mechanics in Hilbert Space - Eduard Prugovecki

    Introduction

    The beginning of the era of modern quantum mechanics is marked by the year 1925 and the two almost simultaneous papers of Heisenberg [1925] and Schroedinger [1926]. The first of these papers proposes essentially the formalism of matrix mechanics, while the second one proposes the formalism of wave mechanics. It was first indicated by Schroedinger that these two formulations are physically equivalent. Both can be embraced in a more general formulation of quantum mechanics, which was first proposed in a somewhat heuristic form by Dirac [1930]. A mathematically rigorous development of this general formalism of quantum mechanics can be achieved by taking advantage of Hilbert space theory, which is done in this book.¹

    Though classical and quantum mechanics are very dissimilar in many respects, they do share a lot of common features when viewed from a general structural point of view. To find these common features, one must have some insight into the mechanism of a physical theory from an abstract point of view, at which one arrives by ignoring the technical details of the theory.

    One can dissect any physical theory into the following main constituents: (1) formalism, (2) dynamical law, (3) correspondence rules.

    Viewed from an abstract point of view, the formalism consists of a set of symbols and rules of deduction. With these symbols one can build statements or propositions. The rules of deduction enable us to deduce new statements from those already given.

    Ideally, every scientific theory starts with a set of basic statements called the axioms of the theory. However, in practice, while a theory is in the growing process, it very often happens that no set of axioms is clearly stated and the axiomatic formulation of the theory is left to the future. Nonrelativistic quantum mechanics has an accepted axiomatic formulation (given in Chapter IV) which covers all the present practical needs, though proposals for new sets of axioms are still making their appearance in scientific journals. These proposals have mainly the virtue of greater generality over the conventional Hilbert space formulation adopted here (so that they contain this formulation as a special case). However, from the practical point of view they have not yet produced any new results not derivable from the conventional approach.

    The two classes of basic objects in the formalism of both classical and quantum mechanics are the states and observables. We shall illustrate these concepts by a few examples.

    In the Newtonian approach to classical physics, the state of a system is given by the family of trajectories of all particles constituting the system. For instance, the state of one particle moving in three dimensions is given by a vector-valued function r(t), r(t) ∈ ∨³ (∨n denotes an n-dimensional real Euclidean space), in the time-parameter t. In the Hamiltonian (canonical) approach to classical mechanics, the state of a single particle is also a vector-valued function

    (q1(t), (q2(t), (q3(t), (p1(t), (p2(t), (p3(t)),

    defined now in a six-dimensional real Euclidean space called the phase space of the system. In the Schroedinger formulation of quantum mechanics, the state is again a vector-valued function Ψ(t, which bears the name of Hilbert space.

    The observables of the formalism are symbols related to specific experimental procedures for measuring them. For instance, in Newtonian classical physics of a one-particle system, r(t) is the position of the particle at time t; p(t) = m = dr(t)dits kinetic energy (m denotes the mass of the particle), etc. In the Hamiltonian approach (q1(t), (q2(t), (q3(t), (p1(t), (p2(t), and (p3(t.

    Every physical theory contains a particularly important ingredient called the dynamical law (frequently referred to as the dynamics). This dynamical law is, in general, a relation which some of the basic objects of the formalism must satisfy. It is the key component of the theory because it gives the theory its predictive power. From a formal logical point of view, the dynamical law could be considered a part of the formalism. However, due to its central role in predicting the future behavior of the system from a knowledge of its present or past behavior, we prefer to single it out.

    From the mathematical point of view, the dynamical law is very frequently a differential equation which has to be satisfied by the state of the system, and in that case it is called the equation of motion. For instance, in Newtonian classical mechanics of a single particle of mass m moving under the influence of a force F, the equation of motion is given by Newton's second postulate

    In the canonical formalism of a single particle we have

    where H(qk, pk, t) is the Hamiltonian of the system. In the Schroedinger formulation of quantum mechanics, the equation of motion governs the time evolution of the state of the system, and is called the Schroedinger equation. However, in the Heisenberg formulation of quantum mechanics, the equation of motion, called the Heisenberg equation, does not involve the state (which in this case is time independent) but rather the observables of the system (see Chapter IV, §3).

    It must be realized that the dynamical law need not always be expressed by a differential equation. Very frequently, it is expressed by an integral equation, sometimes derivable from a differential equation. In modern quantum scattering theory, the dynamical law is given by analyticity conditions that certain key functions (the so-called S-matrix elements) appearing in the theory have to satisfy.

    The correspondence rules are the rules which assign empirical meaning to some of the symbols appearing in the formalism. As such, they provide the link between theory and experiment. The body of all correspondence rules of a theory is in physics better known under the name of the physical interpretation of the theory.

    In classical mechanics, which has a very direct intuitive appeal, the correspondence rules are very straightforward, and they leave no room for plausible alternatives. For instance, in Newtonian one-particle mechanics, it is well known that for a particle in the state r(t), t ¹, the components of the vectors r(t), m , etc. represent the length of the projections of the position, momentum, etc. vectors on the three axes of an inertial frame² with Cartesian coordinates.

    In the case of quantum mechanics, which does not yield itself to visualization as readily as classical mechanics, the matter of physical interpretation is much more complex and therefore open to discussion. In this case the desirability and empirical consistency of the entire body of correspondence rules is still open to scrutiny and critical appraisal, being treated in its own right in the theory of measurement of quantum mechanics. However, the main features of the now widely accepted interpretation, sometimes called the Copenhagen school interpretation (see also Chapter IV, §1), are quite clear. The most striking of these features is that this interpretation is essentially statistical; it does not provide us with definite statements about the future behavior of a quantum mechanical system, but rather with probabilistic statements about the likelihood of different patterns of behavior. It should be mentioned that this feature has given rise to much controversy as to whether quantum mechanics provides the ultimate tool in describing the behavior of atomic and subatomic systems. The ranks of physicists who have adopted a critical attitude when considering this subject include notably the names of some of the pioneers in the field, such as Einstein, Schroedinger,³ and de Broglie.

    Part of the physical interpretation of the theory consists in giving correspondence rules which assign to observables experimental procedures for measuring them. From the mathematical point of view, these observables are represented (in the conventional formalism adopted here) by self-adjoint operators in a Hilbert space. These operators will be studied from a general point of view in Chapter III. At present we shall describe some experimental procedures for measuring, in principle,⁴ the most common of observables: those of position, momentum, and spin of an object of atomic or subatomic size (we shall call such an object a microparticle, so as not to confuse it with the more special term, elementary particle).

    Measurements for determining the position of a particle can be carried out by a number of detectors (Wilson chambers, bubble chambers, Geiger–Müller counters, etc.). These detectors signal the presence of a particle within the volume enclosed by the detector by a directly noticeable (macroscopic) change of their state. Conceptually the simplest of such detectors is a photographic plate on which a landing microparticle leaves a characteristic mark.

    The essential parts of an apparatus for determining the momentum of a charged particle are depicted in Fig. 1. If particles of known charge e interact to the left of the screen⁵ S1, and if a particle is detected by the detector depicted in Fig. 1 (which can be any of the types mentioned, or some other type best suited to the needs of a particular experiment), then the determined momentum p of particle is of magnitude (in suitable units)

    FIG. 1. Experimental arrangement for determinative measurement of the momentum of a charged microparticle.

    and in the direction indicated in Fig. 1; in the above formula, H is the strength of a uniform magnetic field H, which is present to the right of the screen S2 and orthogonal to the plane of drawing.

    It must be stressed that the above-mentioned experimental arrangements are the ones which give an empirical meaning to the concepts of position and momentum of a microscopic particle. In other words, it is not meaningful to ask the question: How do we know that the macroscopic change of state taking place in a detector (the black mark on a photoplate, or the track of condensed vapor in a Wilson chamber) is in reality" due to a particle, and that the momentum of that particle is indeed given by Eq. (0.2) ?" Such a question is empirically meaningless because, in the ultimate analysis (see Heisenberg [1925, p. 174]), the only knowledge that we possess about the microscopic world is the one we can extract from macroscopic phenomena.

    The above type of measurement determines the value of observables at the instant the measurement is carried out. More precisely, from such a measurement one can compute, within bounds of accuracy inherent in the particular apparatus, what would be the value of a particular observable if there were no disturbances due to the interaction of the system (which in the above cases is the microparticle) and the apparatus. This type of measurement will be called a determinative measurement.

    A determinative measurement does not tell us, in general, what the value of the measured observable, or even what the fate of the system will be after the measurement. For example, the system might be completely destroyed or cease to be an independent entity (for instance, when the detector is a photoplate and the particle is absorbed by it) after the measurement. In general, the measurement process will disturb the system or even change its nature (particle creation and annihilation), and such a disturbance must be taken into account if the measurement is used as a preparation of a system with a certain value of the measured observable. This phenomenon is characteristic of microphysics and cannot be ignored, as is done in macrophysics; in principle it exists also in macrophysics, but from the practical point of view it is completely negligible. To realize this, one should recall that measurements in classical mechanics require seeing the system, i.e., the reflection or emission of light from the objects constituting the system. Now, light (i.e., photons) has a certain momentum, which is imparted to the objects on which it impinges or by which it is emitted, but that momentum is completely negligible when dealing with objects of macroscopic size.

    Thus, in quantum mechanics it is important to distinguish between determinative measurements of the above type and preparatory measurements.

    FIG. 2. Experimental arrangement for a preparatory measurement of the momentum of a charged microparticle.

    For instance, in the modification in Fig. 2 of the apparatus in Fig. 1, we have a device which is meant to prepare charged particles of a given momentum. If we keep the shutter open for a very short instant⁷ from t0 to t0 + Δt, then any particle that might have emerged during this period to the right of the screen S (i.e., in the interaction region, where it would interact with other particles adequately prepared) would have a momentum of magnitude |p| given by the formula (0.2), and in the direction indicated on Fig. 2.

    A striking feature of the above experimental procedure is that it does not tell us, by itself, whether there was indeed a particle which had passed through the aperture while the shutter was open. This feature is common to many preparatory measurements. It is in the determinative (later) stage of the experiment that the presence or absence of a particle is established.

    We note that the above preparatory procedure cannot prepare a sharp value p of the momentum, but rather, due to the finite size of the apertures O1, O2, and O3, it prepares a whole range Δ ⊂ ∨³ of momenta. When the size of the last aperture O3 is such that so-called diffraction effects can be neglected, this prepared range Δ consists of all the momenta that a particle of charge e would possess when traveling along all the imaginable paths that such a particle would have to follow according to classical mechanics in order to pass through the apertures O1, O2, and O3.

    The above-mentioned diffraction effects are again characteristic of microphysics, and constitute the so-called wave nature of microparticles. They correspond to the experimental observation that if a detector (e.g., a photoplate) is placed behind two parallel impenetrable screens each having one aperture, a particle might be detected not only along any straight line passing through the two apertures, but also at other points. If we imagine that a microparticle travels along a trajectory, then it would seem that the trajectory is in such cases bent after the particle has passed through the aperture. If we have a beam of particles (i.e., many independent particles), then the net effect which will be observed in a photoplate placed behind the screens is a diffraction pattern, qualitatively very similar to the diffraction pattern of a beam of light.

    We note that the experimental arrangement in Fig. 2 prepares not only a range Δ of momenta, but also a range Δ′ of the position observables, where Δ′ is the region enclosed by the aperture O3. However, while classical mechanics assumes that the linear dimensions of Δ and Δ′ could be made arbitrarily small by building sufficiently precise apparatus, it is a direct consequence of the diffraction effects that this is not the case with microparticles. This experimental finding is embodied in Heisenberg's uncertainty principle which states that no apparatus can be built which would prepare a particle to have the x coordinate (in some Cartesian frame of reference) within the interval Δx and the px momentum component within the interval Δpx, so that the inequality

    is not satisfied, where |Δx| and |Δpx| denote the lengths of the respective intervals, and h is a universal constant called the Planck constant; similar relations

    hold for the y and z Cartesian components.

    It should be mentioned that the experimental arrangements in Figs. 1 and 2 determine and prepare, respectively, also the energy of the microparticle on which the measurement is carried out. By definition, if the momentum of a free particle of mass⁹ m is p, then its energy is

    E = p²/2m.

    We turn now to measurements of spin. Assume that we have in classical mechanics a body which rotates, with respect to an inertial frame of reference, around a certain axis passing through it, which is stationary or moves at a uniform speed with respect to that frame of reference. Such a body has then an angular momentum s with respect to that axis. If the dimensions of the body are negligible with respect to other dimensions in the experiment, so that the body can be called a particle, then s is called the spin of the particle. If the particle carries a charge, then due to its spin it will have a magnetic (dipole) moment p proportional to s. When such a particle travels through an area in which there is an inhomogeneous magnetic field H, this field will act on it with a force

    F = grad(μ · H).

    In the light of the above remarks, each one of the particles of equal velocity (i.e., equal momentum) and equal spin in a beam passing through the experimental arrangement depicted in Fig. 3 would be deflected from a straight path by an amount proportional to the projection μz of the magnetic moment of that particle in the direction of the gradient H′ of the field H. If the spins of the particles were randomly oriented in all directions (as we could expect them to be if the source were a gas of such particles), then one would predict on the basis of the above considerations a continuous distribution of marks on the photoplate. However, the Stern–Gerlach experiment carried out with molecular and atomic beams reveals instead, in general, n discrete lines (in Fig. 3, n = 2)—a phenomenon which is usually described by saying that the spin is quantized.

    FIG. 3. The Stern–Gerlach experiment.

    In quantum mechanics, by definition, the spin of the above particles is taken to be

    (though, strictly speaking, for reasons to become clear when a systematic theoretical study of the spin is undertaken, the total spin is [s(s + 1)]¹/²). Thus, s , 2, ...; in .

    The experimental arrangement of Stern and Gerlach can be used as an apparatus for a determinative measurement of the spin component in the direction H′. In that case the source of particles would originate in the interaction region, where particles of known spin are interacting. If a particle of integer spin ¹⁰ s leaves a mark at O, then, by definition, it has spin zero in the H′direction; the first, second, ..., (n − 1)/2 mark above O correspond to spin components in the H′ direction equal to 1, 2, ..., (n − 1)/2, respectively, while the first, second, ..., (n − 1)/2 marks below O correspond to spin components −1, −2, ..., − (n − 1)/2, respectively. In case of a particle of half-integer spin, there will be no middle mark; the first, second, ..., (n −1)/2 marks above or below O ..., (n , ..., − (n − 1)/2, respectively. Hence we see that according to the very definition of the spin projection onto a certain axis, that projection can assume only integer values in case of integer-spin particles, and only half-integer values in case of half- integer-spin particles.

    The above experimental arrangement can be easily transformed into an apparatus for preparatory measurements of spin by replacing the photoplate with a screen which has apertures at the spots where a beam of particles from the given source had left tracks. It has to be mentioned that no simultaneous measurements of spin in two different directions can be carried out on microparticles—a feature which is in complete agreement with certain properties (noncommutativity of spin-component operators) of the formalism of quantum mechanics.

    Here we end this short survey of some of the experimental procedures for measuring some of the basic observables which occur in quantum mechanics, and which will frequently appear in the pages of this book. In Chapter I we start our systematic study of the Hilbert space formalism of quantum mechanics, and related mathematics.


    ¹ For a detailed historical survey of the origins and development of quantum theory consult any of the many standard textbooks on quantum mechanics (e.g., Messiah [1962]).

    ² We remind the reader that an inertial frame is a physical body (usually the laboratory of the experimenter) in which Newton's equation of motion is valid for macroscopic bodies moving at speeds which are low by comparison with the speed of light. It is an empirical fact that a frame of reference moving during the duration of the experiment at uniform speed with respect to the sun is a very good approximation to an inertial frame.

    ³ It is interesting to mention that one of the earlier physical interpretations of the formalism of wave mechanics (which constitutes a special case of the Hilbert space formalism of quantum mechanics) was proposed by Schroedinger, was in total disagreement with the Copenhagen school interpretation, and lacked any statistical features.

    ⁴ The experimental procedures for measuring these observables in principle may be conceptually very simple, but from the purely technical point of view they may not be the most efficient or the easiest to carry out with a high degree of accuracy and at a given cost.

    ⁵ Both screens S1 and should be impenetrable to all the particles used in the experiment. This impenetrability can be established by closing the apertures in the screens and checking that a detector placed to the right of each screen does not detect anything except the background noise, which has a definite pattern and is due to the everpresent cosmic rays.

    ⁶ It is more common to call a determinative measurement simply a measurement, and a preparatory one a preparation of state. We prefer to avoid the term preparation of state, since it suggests that such a measurement will prepare a quantum mechanical state, which is by no means always true.

    ⁷ By a very short instant is meant a time interval of duration At which is negligible in comparison with the other errors of measurement occurring in the particular experiment in which the described preparatory measurement is included.

    ⁸ We could, of course, add a detector to the above apparatus, but such a detector would significantly disturb the prepared value of the momentum.

    ⁹ The mass of a particle can be measured by the experimental arrangement in Fig. 1 if we introduce an electric field E orthogonal to the existing magnetic field H. Quantities like mass, charge, and intrinsic spin are characteristic of each kind of microparticle, and, in fact, provide the only means of differentiating types of microparticles, like electrons, positrons, neutrons, protons, hydrogen atoms, water molecules, etc.

    ¹⁰ This means that a beam of such particle with random-oriented spins would have 2s + 1 tracks on the photoplate.

    CHAPTER I

    Basic Ideas of Hilbert Space Theory

    The central object of study in this chapter is the infinite-dimensional Hilbert space. The main goal is to give a rigorous analysis of the problem of expanding a vector in a Hilbert space in terms of an orthogonal basis containing a countable infinity of vectors.

    We first review in §1 a few key theorems on vector spaces in general, and in §2 we investigate the basic properties of vector spaces on which an inner product is defined. In order to define convergence in an inner-product space, we introduce in §3 the concept of metric. In §4 we give the basic concepts and theorems on separable Hilbert spaces, concentrating especially on properties of orthonormal bases. We conclude the chapter by illustrating some of the physical applications of these mathematical results with the initial-value problem in wave mechanics.

    1.    Vector Spaces

    1.1. VECTOR SPACES OVER FIELDS OF SCALARS

    A mathematical space is in general a set endowed with some given structure. Such a structure can be given, for instance, by means of certain operations which are defined on the elements of that set. These operations are then required to obey certain general rules, which are called the postulates or the axioms of the mathematical space.

    Definition 1.1. on which the operations of vector addition and multiplication by a scalar are defined is said to be a vector space (or linear space, or linear manifold). The operation of vector addition is a mapping,¹

    , while the operation of multiplication by a scalar a from a field² F is a mapping

    ,

    . These two vector operations are required to satisfy the following axioms for any f, g, h and any scalars a, b ∈ F:

    (1) f + g = g + f(commutativity of vector addition).

    (2) (f + g) + h = f + (g + h) (associativity of vector addition).

    (3) There is a vector 0, called the zero vector, such that g satisfies the relation f + g = f if and only if g = 0.

    (4) a(f + g) = af + ag.

    (5) (a + b)f = af + bf.

    (6) (ab)f = a(bf).

    (7) 1f = f, where 1 denotes the unit element in the field.

    By following a tacit convention, we denote a mathematical space constructed from a set S by the same letter S.

    When in a vector space the multiplication by a scalar is defined for scalars which are elements of the field F, we say that we are dealing with a vector space over the field F. If the field F is the field of real or complex numbers the vector space is called, respectively, a real or a complex vector space.

    As an example (see also Exercises n) of one-column real matrices and define for

    vector summation by the mapping

    and for any scalar a ¹ define multiplication of α by a as the mapping

    It is easy to check that Axioms 1–7 in Definition 1.1 are satisfied.

    nn of one-column matrices vector operations defined by the mapping (1.1) and (1.2), where now α, β n, and therefore a1, ..., an, b1, ..., bn, as well as the scalar a, are complex numbers.

    1.2.  LAINEAR INDEPENDENCE OF VECTORS

    Theorem 1.1. has only one zero vector 0, and each element f of a vector space has one and only one inverse (−f). For any f ,

    0f = 0, (−1)f = (−f).

    Proof. If there are two zero vectors 01 and 02, they both have to satisfy Axiom 3 in Definition 1.1,

    f = f + 01 = f + 02

    for all f. Hence, by taking f = 01 we get 01 = 01 + 02, and then by taking f = 02 we deduce that 02 = 02 + 01 + 02 = 01. Now

    f = 1f = (1 + 0)f = 1f + 0f = f + 0f

    and therefore 0f = 0. We have

    (−1)f + f = (−1)f + 1f = (−1 + 1)f = 0f = 0,

    which proves the existence of an inverse (−f) = (−1)f for f. This inverse (−f) is unique, because if there is another fsuch that f + f1 = 0, we have

    (−f) = (−f) + 0 = (−f) + (f + f1) = [(−f) + f] + f1

           = 0 + f1 = f1. Q.E.D.

    Definition 1.2. The vectors f1, ..., fn are said to be linearly independent if the relation

    c1f1 + ... + cnfn = 0, c1, ..., cn ∈ F,

    has c1 = ... = cn = 0 as the only solution. A subset S is called a set of linearly independent vectors if any finite number of different vectors from S are linearly independent. The dimension of a vector space is the least upper bound (which can be finite or positive infinite) of the set of all integers v for which there are v .

    1.3.  DIMENSION OF A VECTOR SPACE

    is finite and equal to nis n is said to be infinite dimensional.

    Theorem 1.2. is n dimensional (n < + ∞), then there is at least one set f1, ..., fn of linearly independent vectors, and each vector f can be expanded in the form

    where the coefficients a1, an (which are scalars) are uniquely determined by f.

    Proof. If f = 0, (1.3) is established by taking a1 = ... = an = 0. For f ≠ 0, the equation

    should have a solution with c ≠ 0 due to the assumption that f1, ..., fn are linearly independent, while f, f1, ..., fis n dimensional. From (1.4) we get

    f = (−c1/c)f1 + ... + (−cn/c)fn,

    which establishes (1.3). If we also had

    then by subtracting (1.5) from (1.3) we get

    (a1 − b1)f1 + ... + (an − bnfn = 0.

    As f1, ..., fn are linearly independent we deduce that a1 − b1 = 0, ..., an − bn = 0, thus proving that a1, ..., an are uniquely determined when f is given. Q.E.D.

    Definition 1.3. We say that the (finite or infinite) set S can be written as a linear combination

    f = a1h1 + ... + anhn, h1, ..., hn ∈ S

    of a finite number of vectors belonging to S; if S is in addition a set of linearly independent vectors, then S is called a vector basis of .

    Theorem 1.3. If the set {g1, ..., gm} is a vector basis of the n-dimensional (n < , then necessarily m n.

    Proofis n-dimensional, there must be n linearly independent vectors f1, ..., fn. If the set {g1gmwe can write

    Thus, if we try to satisfy the equation

    we get by substituting f1, ..., fn in (1.7) with the expressions in (1.6)

    Since g1, ..., gm are assumed to be linearly independent, the above equation has a solution in x1, ..., xn if and only if

    However, as f1, ..., fn are also linearly independent, (1.7) or equivalently (1.8) or (1.9) should have as the only solution the trivial one x1 = ... = xn = 0. Now, m n is n dimensional and g1, ..., gm are linearly independent (see Definition 1.2); therefore, (1.9) has only a trivial solution if and only if m = n. Q.E.D.

    Definition 1.4. is a vector subspace (linear subspaceif it is closed under the vector operations, i.e., if f + g 1 and af 1 whenever f, g 1 and for any scalar ais said to be nontrivial and from the set {0}.

    .

    1.4.  ISOMORPHISM OF VECTOR SPACES

    Definition 1.5. 2 over the same field are isomorphic 2 which has the properties that if f2 and are the images of f1 and g1, g1, respectively, then for any scalar a, af2 is the image of af1

    af1 ↔ af2,

    and f2 + g2 is the image of f1 + g1

    f1 + g1 ↔ f2 + g2.

    2 lies in the obvious fact that two such spaces have an identical vector structure. It is easy to see that the relation of isomorphism is transitive (see 3 are also isomorphic.

    Theorem 1.4. All complex (real) n-dimensional (n nn) in case of real vector spaces].

    Proof. Consider the case of an n. According to Theorem 1.2 there is a vector basis consisting of n vectors f1, ..., fn, and each vector f can be expanded in the form (1.3), where a1, ..., a¹ are uniquely determined by f. Consequently

    nonto n) because to any

    corresponds a unique f = b1f1+ ... + bnfn such that β = αf. It is also easy to see that

    f + g αf + g = αf + αg,

    af αaf = aαf.

    Since isomorphism of vector spaces is a transitive relation (see Exercise 1.6) we can conclude that all nn). Q.E.D.

    EXERCISES

    1.1. Check that the set of all m × n complex matrices constitutes an m · n dimensional complex vector space if vector addition is defined as being addition of matrices, and multiplication by a scalar is multiplication of a matrix by a complex number.

    1.2. ¹ of all complex numbers becomes a two-dimensional real vector space if vector addition is identical to addition of complex numbers, and multiplication by a scalar is multiplication of a complex number (the vector) by a real number (the scalar).

    ¹) of all complex-valued continuous functions defined on the real line is an infinite-dimensional vector space if the vector sumf + g off(x), g(x¹) is the function (f + g)(x) = f(x) + g(x), and the product af of f(x) ¹) with a¹ is the function (af)(x) = af(x). The zero vector is taken to be the function

    1.4. is a family of linear subspaces L L .

    1.5. Show that if S s (called the vector subspace spanned by S).

    1.6. Verify that the relation of isomorphism of vector spaces is:

    is isomorphic to itself;

    1;

    3.

    1.7. ¹) (see ¹):

    ∞ of all polynomials with complex coefficients;

    n of all polynomials of at most degree n.

    n.

    2.    Euclidean (Pre-Hilbert) Spaces

    2.1.  INNER PRODUCTS ON VECTOR SPACES

    A Euclidean (or pre-Hilbert or inner productor unitaryis a vect or space on which an inner product is defined. The Euclidean space is called real or complex if the vector space on which the inner product is defined is, respectively, real or complex.

    Definition 2.1. An inner(or scalar¹ of complex numbers

    which satisfies the following requirements:

    (1)  〈f | f〉 > 0, for all f ≠ 0,

    (2)  〈f | g〉 = 〈g | f〉*,

    (3)  〈f | ag〉 = af | g〉, a ¹,

    (4)  〈f | g + h〉 = 〈f | g〉 + 〈f | h〉.

    Note that by inserting f = g = h = 0 in Point 4 we get 〈0 | 0) = 0.

    Following a notation first introduced by Dirac [1930] and widely adopted by physicists, we denote the inner product of / and g by 〈f | g). Mathematicians often prefer the notation (f, g) and replace Point 3 in Definition 2.1 by

    (af, g) = a(f, g).

    The above definition can be easily specialized to real vector spaces, in which case the inner product 〈f | g〉 is a real number, and Point 2 of Definition 2.1 becomes 〈f | g〉 = 〈g | f〉. As in quantum physics we deal almost exclusively with complex Euclidean spaces, we limit ourselves from now on to the complex case. Consequently, if not otherwise stated, whenever we talk about a Euclidean space, we shall mean a complex Euclidean space.

    Theorem 2.1. , the inner product 〈f | g〉 satisfies the relations

    (a)  〈af | g〉 = a*(f | g〉,

    (b)  〈f + g | h) = 〈f | h〉 + 〈g | h〉.

    The proof is obtained by a straightforward application of Points 1–4 in Definition 2.1:

    n) defined in the preceding section, in which we introduce as the inner product of the vectors α and β th components ak and bk,

    α | β〉 = a1 * b1 a2 * b2 +... + an * bn.

    ¹ satisfies the four requirements of Definition 2.1. We shall denote the above Euclidean space with the symbol l²(n).

    ¹)] of continuous complex-valued functions f(x) on the real line which satisfy

    in which the inner product (see Exercise 2.1) is

    Theorem 2.2. Any two elements f, g satisfy the Schwarz-Cauchy inequality

    |〈f | g〉|² ≤ 〈f | f〉 〈g | g〉.

    Proof. For any given f, g, and any complex number a we have, from property 1 in Definition 2.1 and the comment following it,

    f + ag | f + ag〉 ≥ 0.

    In particular, if we take in the above inequality

    we easily show that the inequality

    g>(λ) = λ²〈g | g〉 + f | g〉| + 〈f | f〉 ≥ 0

    is true for all real values of λ. A necessary and sufficient condition that g(λ) ≥ 0 is that the discriminant of the quadratic polynomial g(λ) is not positive

    f | g〉|² – 〈f | f〉〈g | g〉 ≤ 0,

    from which the Schwarz-Cauchy inequality follows immediately. Q.E.D.

    2.2.  THE CONCEPT OF NORM

    The family of all Euclidean spaces is obviously contained in the family of vector spaces. There is another family of vector spaces with special properties which is of great importance in mathematics: the family of normed spaces.

    Definition 2.2. A mapping

    into the set of real numbers is called a normif it satisfies the following conditions:

    (1)  || f || > 0 for f ¬ 0,

    (2)  || 0 || = 0,

    (3)  || af || = | a | || f || for all a¹,

    (4)  || f + g || ≤ || f || + || g || (the triangle inequality).

    We denote the above norm by || · ||.

    For a real vector space, we require in Point 3 that a¹.

    The last requirement in Definition 2.2 is known as the triangle inequality because it represents in a two- or three-dimensional real vector space a relation satisfied by the sides of a triangle formed by three vectors f, g and f + g.

    A real (complex) vector space on which a particular norm is given is called a real (complex) normed vector space. A Euclidean space is a special case of a normed space; this can be seen from the following theorem.

    Theorem 2.3. with the inner product 〈f | g〉 the real-valued function

    is a norm.

    Proof. The only one of the four properties of a norm which is not satisfied by (2.3) in an evident way is the triangle inequality. We easily get

    From the Schwarz-Cauchy inequality we have

    | Re〈f | g〉| ≤ | <,〈f | g〉 ≤ || f || || g ||,

    which when inserted in (2.4) yields

    || f + f || ² ≤ || f || ² + 2|| f || || g || + || g || ² = (|| f || + || g ||)².

    The above relation leads immediately to the triangle inequality. Q.E.D.

    2.3.  ORTHOGONAL VECTORS AND ORTHONORMAL BASES

    Some elementary geometrical concepts valid for real two- or three- dimensional Euclidean spaces can be generalized in a straightforward manner to any Euclidean space.

    Definition 2.3. two vectors f and g are called orthogonal, symbolically f g, if 〈f | g〉 = 0. Two subsets R and S are said to be orthogonal (symbolically, R S) if each vector in R is orthogonal to each vector in S. A set of vectors in which any two vectors are orthogonal is called an orthogonal system of vectors. A vector f is said to be normalized if || f || = 1. An orthogonal system of vectors is called an orthonormal system if each vector in the system is normalized.

    Theorem 2.4. If S and (S) is the vector subspace of spanned by S > then there is an orthonormal system T of vectors which spans (S), i.e., for which (T) = (S); T is a finite set when S is a finite set.

    Proof. As the set S is at most countable we can write it in the form

    S = {f1, f2,...}

    by assigning each vector in S to a natural number. In general some of the vectors in S might be linearly dependent. We can build from S another set S0 of linearly independent vectors spanning the same subspace (S) i.e., such that (S0) = (S), by the following procedure (which should be applied consecutively on n = 1, 2,...): if fn is the zero vector or is linearly dependent on f1,..., fn − 1, then discard it; otherwise include it in SQ. Thus we get a set SQ of linearly independent vectors

    S0 = {g1, g2,...}, (S0) = (S).

    We can obtain from S0 an orthonormal set T such that (T) = (S0) by the so-called Schmidt (or Gram-Schmidt) orthonormalization procedure.

    Since g1 ≠ 0, we can introduce the vector

    which is normalized. Proceeding by induction, assume that we have obtained the orthonormal system of vectors e1,..., en − 1. Then en is given by

    The above vector is certainly well defined, since the denominator of the above expression is different from zero; namely, if it were zero, then we would have

    gn − 〈en − 1 | gn) en − 1 −... − 〈e1 | gne1 = 0,

    i.e. gn would depend on e1,..., en − 1. However, by solving the equations for e1,..., en − 1, it is easy to see that we have

    and therefore if gn depended on e1,..., en − 1, then it would also depend on g1,..., gn − 1, contrary to the fact that S0 consists only of linearly independent vectors.

    The vectors of T are obviously normalized. In order to prove that T is an orthonormal system, assume that we have proved that 〈ei | ej〉 = δij for i, j = 1,..., n − 1. Then we have for m < n

    which proves that 〈ei | ej〉 = δij for i, j = 1,..., n. Thus, by induction T is orthonormal.

    As we have for any n that e1,..., en can be expressed in terms of g1,..., gn, and vice versa, we can conclude that (T) = (S0). Q.E.D.

    2.4.  ISOMORPHISM OF EUCLIDEAN SPACES

    We introduce now a concept of isomorphism of Euclidean spaces, which makes two isomorphic Euclidean spaces identical from the point of view of their vector structure as well as from the point of view of the structure induced by the inner product.

    Definition 2.4. 2 with inner products 〈 · | · 〉1 and 〈 · | · 〉2 respectively, are isomorphic (or unitarily equivalent1 onto 2

    such that if for any f1, g1the vector f2 is the image of f1 and the vector g2 is the image of g1, then

    A mapping having the above properties is called a unitary 2.

    Theorem 2.5. All complex Euclidean n-dimensional spaces are isomorphic to l²(n), and consequently (see Exercise 2.8) mutually isomorphic.

    Proofis an n-dimensional Euclidean space, there is according to Theorem 1.2 a set of n vectors f1,...,fn spanning According to Theorem 2.4, we can find an orthonormal system of n vectors e1en . It is easy to check (see Exercise 2.7) that the mapping

    and l²(n). Q.E.D.

    Obviously, a similar theorem can be proved for real Euclidean spaces.

    Theorem 2.6. A unitary transformation

    1.

    Proof. We note that since

    || f1 − g1 || 1 = || f2 − g2 || 2,

    the images f2 and g2 of f1 and g1respectively, are distinct whenever f1 ≠ g2, we conclude that the inverse of the mapping (2.6) exists.

    We leave to the reader the details of the remainder of the Proof.

    EXERCISES

    2.1. Show that for a finite interval I

    °(I).

    ¹)

    .

    2.4. Show that

    if and only if either f is a multiple of g, i.e., if = ag, a ¹, or g = 0, and if in addition a ≥ 0 in case of the second relation.

    2.5. Show that if T is an orthonormal system of vectors, then all the vectors in T are necessarily linearly independent.

    2.6. Prove that a subspace of a Euclidean space is also a Euclidean space.

    2.7. Show that the mapping (onto l²(n), and that it satisfies the requirements of isomorphism given in Definition 2.4.

    2.8. Show that the relation of isomorphism of inner-product spaces is an equivalence relation, i.e., it is (see Exercise 1.6) reflexive, symmetric, and transitive.

    3.    Metric Spaces

    3.1.  CONVERGENCE IN METRIC SPACES

    In an nwe can always find, due to Theorems 1.2 and 2.4, a basis of n vectors e1,..., en which constitute an orthonormal system. We can then expand any vector f in that basis

    We easily see that ak = 〈ek | f〉.

    In an infinite-dimensional Euclidean space not every vector can be expanded in general in terms of a finite number of vectors. We can hope, however, to replace (3.1) with the formula

    but then we meet with the problem of giving a precise meaning to the convergence of the above series. This problem is solved in its most general form in topology, but for our purposes it will be sufficient to solve it within the context of metric spaces.

    Definition 3.1. If S is a given set, a function d(ξ, η) on S × S is a metric (or distance function) if it fulfills the following requirements for any ξ, η,ζ ∈ S:

    (1)  d(ξ, η) > 0 if ξ η,

    (2)  d(ξ, ξ) = 0,

    (3)  d(ξ, η) = d(η, ξ),

    (4)  d(ξ, ζ) ≤ d(ξ, η) + d(η, ζ)    (triangle inequality).

    A set S on which a metric is defined is called a metric space.

    A metric space does not have to be a linear space. For instance, a bounded open domain in the plane becomes a metric space if the metric is taken to be the distance between each pair of points belonging to that domain; such a domain obviously is not closed under the operations of adding vectors in the plane, but it provides a metric space.

    Generalizing from the case of one-, two-, or three-dimensional real Euclidean spaces, we introduce the following notions.

    Definition 3.2. An infinite sequence ξ1, ξis said to converge to the point ξ ∈ > 0 there is a positive number N) such that d(ξ, ξnfor all n > N). An infinite sequence, ξ1, ξ2,... is called a Cauchy sequence (or a fundamental sequence> 0 there is a positive number M) such that d(ξm, ξnfor all m,n > M).

    Theorem 3.1. If a sequence ξ1, ξξ ∈ , then its limit ξ is unique, and the sequence is a Cauchy sequence.

    Proof. If ξ1, ξ2,... converges to ξ ∈ and to η ∈ N) and N) such that d(ξ, ξnfor n > N) and d(η, ξnfor n > N). Consequently, for n > max(N), N)) we get by applying the triangle inequality of Definition 3.1, Point 4,

    d(ξ, η) ≤ d(ξ, ξn) + d(ξn, η.

    > 0 can be chosen arbitrarily small, we get d(ξ, η) = 0, which, according to Definition 3.1, can be true only if ξ = η.

    Similarly we get

    d(ξm, ξn) ≤ d(ξm, ξ) + d(ξ, ξn

    if m, n > N/2); i.e., the sequence ξ1, ξ2,... is also a Cauchy sequence.

    Q.E.D.

    3.2  COMPLETE METRIC SPACES

    ¹ of all real numbers is complete. We state this generally in Definition 3.3.

    Definition 3.3. is complete .

    of all rational numbers with the metric d(m1/n1, m2/n2) = | m1/n1 − m2/n¹; we state this generally as follows:

    Definition 3.4. A subset S is (everywhere) dense > 0 and any ξ there is an element η belonging to S for which d(ξ, η.

    We can reexpress the above definition after introducing a few topological concepts, generalized from the case of sets in one, two, or three real dimensions.

    Definition 3.5. If ξ , then the set of all points η satisfying the inequality d(ξ, ηneighborhood of ξ. If S , a point ζ ∈ is called an accumulation (or cluster or limit) point of S if every e neighborhood of ζ contains a point of Sconsisting of all the cluster points of S is called the closure of S. Obviously always S ⊂ if S then S is called a closed set.

    We say that the subset S is (everywhere) dense

    The procedure of completing the set of rational numbers by embedding it in the set of all real numbers can be generalized.

    Definition 3.6. is said to be densely embedded .

    A one-to-one mapping ξ ξ is called isometric if it preserves distances, i.e., if d(ξ, η) = ξ for ξ, η ∈ and ξ, ∈. whenever ξ ξ and η η.

    3.3.  COMPLETION OF A METRIC SPACE

    *Theorem 3.2. , called the completion .

    The Proof of this theorem can be given by generalizing Cantor’s construction, by which one builds the set of real numbers from the rational numbers.

    are two such sequences, we say that they are equivalent if and only if

    (see Exercise 3.1) if we recall (see Exercises 1.6 and 2.8) the general definition of an equivalence relation.

    Definition 3.7. A relation f ξ η holding between any two ordered elements of a set S is called an equivalence relation if it is

    (1)  reflexive: ξ ξ for all ξ S;

    (2)  symmetric: ξ η implies that η ξ;

    (3)  transitive: ξ η and η ζ implies that ξ ζ.

    A subset X of S having the property that all the elements of X are equivalent and that if η ξ and ξ X then η X is called an equivalence class (with respect to the equivalence relation ∼).

    [with respect to the equivalence relation given by (, are equivalent, i.e., satisfy (3.2).

    we employ the relation (see Exercise 3.2)

    to show that, d(ξ1, η1), d(ξ2, η2),... is a Cauchy sequence of numbers, and therefore has a limit; namely as ξ1, ξ2,... and η1, η2,... are Cauchy sequences, we can make d(ξm, ξif m, n > N),d(ηm, ηif m, n > N), which, used in conjunction with (3.4), proves the statement.

    from the inequality (see Exercise 3.3)

    → 0 as n belong to the same

    Enjoying the preview?
    Page 1 of 1