Professional Documents
Culture Documents
Textbook Generalized Locally Toeplitz Sequences Theory and Applications Volume Ii Carlo Garoni Ebook All Chapter PDF
Textbook Generalized Locally Toeplitz Sequences Theory and Applications Volume Ii Carlo Garoni Ebook All Chapter PDF
https://textbookfull.com/product/generalized-intuitionistic-
multiplicative-fuzzy-calculus-theory-and-applications-shan-yu/
https://textbookfull.com/product/combinatorics-and-number-theory-
of-counting-sequences-discrete-mathematics-and-its-
applications-1st-edition-istvan-mezo/
https://textbookfull.com/product/structural-dynamics-
fundamentals-and-advanced-applications-volume-ii-volume-ii-1st-
edition-alvar-m-kabe/
https://textbookfull.com/product/map-projections-theory-and-
applications-frederick-pearson-ii/
Recurrent Sequences Key Results Applications and
Problems Dorin Andrica
https://textbookfull.com/product/recurrent-sequences-key-results-
applications-and-problems-dorin-andrica/
https://textbookfull.com/product/legal-theory-and-the-social-
sciences-volume-ii-maksymilian-del-mar-editor/
https://textbookfull.com/product/existence-theory-for-
generalized-newtonian-fluids-1st-edition-breit/
https://textbookfull.com/product/generalized-fractional-calculus-
new-advancements-and-applications-george-a-anastassiou/
https://textbookfull.com/product/computation-of-generalized-
matrix-inverses-and-applications-1st-edition-ivan-stanimirovic/
Carlo Garoni · Stefano Serra-Capizzano
Generalized
Locally Toeplitz
Sequences:
Theory and
Applications
Volume II
Generalized Locally Toeplitz Sequences:
Theory and Applications
Carlo Garoni Stefano Serra-Capizzano
•
123
Carlo Garoni Stefano Serra-Capizzano
Department of Science Department of Science
and High Technology and High Technology
University of Insubria University of Insubria
Como, Italy Como, Italy
This book has been realized with the financial support of the Italian INdAM (Istituto Nazionale di Alta
Matematica) and the European “Marie-Curie Actions” Programme through the Grant PCOFUND-GA-
2012-600198.
Mathematics Subject Classification (2010): 15B05, 65N06, 65N25, 65N30, 65N35, 47B06, 35P20,
15A18, 15A60, 15A69
This Springer imprint is published by the registered company Springer Nature Switzerland AG
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
Preface
Sequences of matrices with increasing size naturally arise in several contexts and
especially in the discretization of continuous problems, such as integral and dif-
ferential equations. The theory of generalized locally Toeplitz (GLT) sequences was
developed in order to compute/analyze the asymptotic spectral distribution of these
sequences of matrices, which in many cases turn out to be GLT sequences.
In the first volume [22], we presented the theory of univariate/unilevel GLT
sequences, which arise in the discretization of unidimensional integral and differ-
ential equations; this is the reason why the first volume addressed only unidi-
mensional applications. In this second volume, we present the theory of
multivariate/multilevel GLT sequences, which arise in the discretization of multi-
dimensional integral and differential equations. The focus here is accordingly on
multidimensional applications, especially partial differential equations (PDEs).
It is important to emphasize that the extension from the univariate case addressed
in [22] to the multivariate case addressed here, despite being fundamental for the
applications as it allows one to face concrete PDEs, is essentially a technical matter
whose purpose is to illustrate the appropriate generalization of ideas already pre-
sented in [22]. The fact that all the main “GLT ideas” have been covered in [22]
makes it an essential prerequisite to this book. In particular, apart from (almost)
obvious adaptations, several “multivariate proofs” are the same as their corre-
sponding “univariate versions” from [22]. We have therefore been tempted to omit
them here so as to shorten the book, but ultimately we did not opt for this solution
in order to help the reader gain familiarity with the multivariate language (espe-
cially the multi-index notation).
The book is conceptually divided into two parts. The first part (Chaps. 1–5)
covers the theory of multilevel GLT sequences, which is finally summarized in
Chap. 6. The second part (Chap. 7) is devoted to PDE applications.
The book is intended for use as a text for graduate or advanced undergraduate
courses. It should also be useful as a reference for researchers working in the fields
of linear and multilinear algebra, numerical analysis, and matrix analysis. Given its
analytic spirit, it could also be of interest to analysts, particularly those working in
the fields of measure and operator theory.
v
vi Preface
As already pointed out, the first volume [22] is an essential prerequisite to this
second volume. It also provides detailed motivations to the theory of GLT
sequences [22, pp. 1–3] which will not be repeated here for the sake of conciseness.
In addition to [22], a basic knowledge of multidimensional integro-differential
calculus (partial derivatives, multiple integrals, etc.) is required.
Assuming the reader possesses the necessary prerequisites, most of which, if not
already addressed in [22], will be tackled in Chap. 2, there exists a way of reading
this book that allows one to omit essentially all the mathematical details/
technicalities without losing the core. This is probably “the best way of reading” for
those who love practice more than theory, but it is also advisable for theorists, who
can recover the missing details afterward. It consists in reading carefully the
summary of the theory in Chap. 6 and the applications in Chap. 7.
To conclude, we wish to express our gratitude to Bruno Iannazzo, Carla Manni,
and Hendrik Speleers, who awakened the interest in the theory of GLT sequences
and ultimately inspired the writing of this book. We also wish to thank all
of our colleagues who have worked in the field of “Toeplitz matrices and spectral
distributions” and contributed to laying the foundations of the theory of GLT
sequences. We mention in particular Bernhard Beckermann, Albrecht Böttcher,
Fabio Di Benedetto, Marco Donatelli, Leonid Golinskii, Sergei Grudsky,
Arno Kuijlaars, Maya Neytcheva, Debora Sesana, Bernd Silbermann, Paolo Tilli,
Eugene Tyrtyshnikov, and Nickolai Zamarashkin. Finally, special thanks go to
those researchers who, possibly attracted by the first volume [22], decided to
enter the research field of GLT sequences. We mention in particular
Giovanni Barbarino from Scuola Normale Superiore (Pisa, Italy), Davide Bianchi
and Isabella Furci from University of Insubria (Como, Italy), Ali Dorostkar and
Sven-Erik Ekström from Uppsala University (Uppsala, Sweden), Mariarosa Mazza
and Ahmed Ratnani from the Max Planck Institute for Plasma Physics
(Munich, Germany). Several of their contributions will certainly appear in a future
edition of both volumes I and II.
vii
viii Contents
xi
Chapter 1
Notes to the Reader
The present book covers the multivariate version of the theory of Generalized Locally
Toeplitz (GLT) sequences, also known as the theory of multilevel GLT sequences. In
addition, the book presents some emblematic (multidimensional) applications of this
theory in the context of the numerical discretization of Partial Differential Equations
(PDEs).
The generalization of the theory of GLT sequences from the univariate case
addressed in [22] to the multivariate case addressed here is essentially a matter
of technicalities, which results in the technical nature of the present volume. We
therefore recommend that, before going into this book, the reader give a reading to
[22, pp. 1–3] in order to call to mind the motivations behind the theory of (unilevel
and multilevel) GLT sequences, which will not be repeated here for the sake of con-
ciseness. When reading [22, pp. 1–3] in a multidimensional perspective, the GLT
sequences and the Differential Equations (DEs) mentioned therein should be under-
stood as multilevel GLT sequences and PDEs, respectively.
After going through [22, pp. 1–3], we encourage the reader to try reading this
book according to the scheme suggested in the preface, which consists in reading
Chaps. 6 and 7 first, and then coming back to fill the gaps (if necessary or wanted).
When reading the present book, it is advisable that the reader have at hand the first
volume [22], for at least two reasons. First, [22] is cited many times throughout the
book. Secondly, several “multivariate proofs” from Chaps. 2–5 are essentially the
same as their corresponding “univariate versions” from [22], and we recommend that
the reader compare them with each other so as to learn the way in which the multilevel
language (especially, the multi-index notation) allows one to transfer many results
from the univariate to the multivariate case. Roughly speaking, this transfer process
is carried out through a sort of “automatic procedure” consisting in turning some
letters (n, i, j, x, θ, etc.) in boldface (n, i, j , x, θ , etc.). Finally, we remark that, as
highlighted in the preface, the first volume [22] is an essential prerequisite to this
second volume. In addition to [22], the other necessary prerequisite for reading this
book is a basic knowledge of multidimensional integro-differential calculus (partial
derivatives, multiple integrals, etc.).
This chapter collects the necessary preliminaries to develop the multivariate version
of the theory of GLT sequences. The reader is supposed to be familiar with the univari-
ate version of the theory [22] and to possess a basic knowledge of multidimensional
integro-differential calculus (partial derivatives, multiple integrals, etc.).
For the reader’s convenience, we report in this section some of the most common
notations and terminologies that will be used throughout the book. Special attention
is devoted to the multi-index notation and the notion of multilevel matrix-sequences.
Together with the index at the end, this section can be used as a reference whenever
an unknown notation/terminology is encountered.
The 2-norm | · |2 is also known as the spectral (or Euclidean) norm and it will be
preferably denoted by · . For more on p-norms, see [22, Sect. 2.4.1].
• Given X ∈ Cm×m and 1 ≤ p ≤ ∞, X p denotes the Schatten p-norm of X ,
which is defined as the p-norm of the vector (σ1 (X ), . . . , σm (X )) formed by the
singular values of X . The Schatten 1-norm is also known under the names of trace-
norm and nuclear norm. For more on Schatten p-norms, see [22, Sect. 2.4.3].
• (X ) and (X ) are, respectively, the real and the imaginary part of the square
matrix X :
X + X∗ X − X∗
(X ) = , (X ) = ,
2 2i
where i is the imaginary unit (i2 = −1). Note that (X ), (X ) are Hermitian and
X = (X ) + i (X ) for all square matrices X .
• If X ∈ Cm×m is diagonalizable and f : Λ(X ) → C, we denote by f (X ) the matrix
obtained by applying the function f to X . For more on matrix functions, see [22,
Sect. 2.4.6].
• We use the abbreviations FDs, FEs, IgA for “Finite Differences”, “Finite Ele-
ments”, “Isogeometric Analysis”.
• Given two sequences {ζn }n and {ξn }n , with ζn ≥ 0 and ξn > 0 for all n, the notation
ζn = O(ξn ) means that there exists a constant C, independent of n, such that
ζn ≤ Cξn for all n; and the notation ζn = o(ξn ) means that ζn /ξn → 0 as n → ∞.
2.1 Notation and Terminology 5
M D = { f : D → C : f is measurable},
L (D) = f ∈ M D :
p
|f| < ∞ ,
p
1 ≤ p < ∞,
D
L ∞ (D) = f ∈ M D : ess sup D | f | < ∞ .
6 2 Mathematical Background
If D is the special domain [0, 1]d × [−π, π ]d , we preferably use the notation Md
instead of M D :
If f ∈ L p (D) and the domain D is clear from the context, we write f Lp instead
of f L p (D) to indicate the L p -norm of f , which is defined as
( D | f | p )1/ p , if 1 ≤ p < ∞,
f Lp =
ess sup D | f |, if p = ∞.
• We use a notation borrowed from probability theory to indicate sets. For example,
if f, g : D ⊆ Rk → C, then
{ f = 1} = {x ∈ D : f (x) = 1},
{0 ≤ f ≤ 1, g > 2} = {x ∈ D : 0 ≤ f (x) ≤ 1, g(x) > 2},
μk { f > 0, g < 0} is the measure of the set {x ∈ D : f (x) > 0, g(x) < 0},
χ{ f =0} is the characteristic function of the set where f vanishes,
...
• A functional φ is any function defined on some vector space (such as, for example,
Cc (C) or Cc (R)) and taking values in C.
• If K is either R or C and g : D ⊂ Rk → K is a measurable function defined on a
set D with 0 < μk (D) < ∞, we denote by φg the functional
1
φg : Cc (K) → C, φg (F) = F(g(x))dx. (2.2)
μk (D) D
Throughout this book, we will systematically use the multi-index notation. When
discretizing a linear PDE over a d-dimensional domain ⊂ Rd by means of a
linear numerical method, the actual computation of the numerical solution reduces to
solving a linear system whose coefficient matrix usually possesses a d-level structure
(see Example 2.5 below). As we shall see later on, especially in Chap. 7, the multi-
index notation is a powerful tool that allows one to give a compact expression of this
matrix by treating the dimensionality parameter d as any other parameter involved
in the considered discretization. In this way, the dependence of the matrix structure
on d is highlighted and a compact presentation is made possible.
A multi-index i of size d, also called a d-index, is simply a (row) vector in Zd ;
its components are denoted by i 1 , . . . , i d .
• 0, 1, 2, . . . are the vectors of all zeros, all ones, all twos, . . . (their size will be
clear from the context).
• For any d-index m, we set N (m) = dj=1 m j and we write m → ∞ to indicate
that min(m) → ∞. The notation N (α) = dj=1 α j will be actually used for any
vector α with d components and not only for d-indices.
• Let {an }n∈Nd be a family of numbers parameterized by a d-index n. The limit of
an as n → ∞ is defined, as in the case of a traditional sequence {an }n∈N , in the
following way: lim n→∞ an = a if and only if for every neighborhood U of a there
exists N such that an ∈ U for n ≥ N. Moreover, we define
(h 1 , h 2 ), (h 1 , h 2 + 1), . . . , (h 1 , k2 ),
(h 1 + 1, h 2 ), (h 1 + 1, h 2 + 1), . . . , (h 1 + 1, k2 ),
. . . . . . . . . , (k1 , h 2 ), (k1 , h 2 + 1), . . . , (k1 , k2 ).
X = [x i j ]m
i, j =1 , (2.4)
Note that i ∧ j is the minimum among i and j with respect to the lexicographic
ordering. In the case where i and j are 1-indices (i.e., normal scalar indices), it is
clear that i ∧ j = min(i, j).
• Operations involving d-indices that have no meaning in the vector space Zd
must always be interpreted in the componentwise sense. For instance, n p =
(n 1 p1 , . . . , n d pd ), αi/ j = (αi 1 /j1 , . . . , αi d /jd ) for all α ∈ C, i 2 = (i 12 , . . . , i d2 ),
max(i, j ) = (max(i 1 , j1 ), . . . , max(i d , jd )), i mod m = (i 1 mod m 1 , . . . , i d mod
m d ), etc.
• When a multi-index appears as subscript or superscript, we sometimes suppress
the brackets to simplify the notation. For instance, the component of the vector
x = [x i ]m i=1 corresponding to the d-index i is denoted by x i or x i 1 ,...,i d , and we
often avoid the heavy notation x(i1 ,...,id ) .
We provide below a few examples to help the reader become familiar with the multi-
index notation.
Example 2.1 Let h = 1 = (1, 1) and k = (4, 2). The multi-index range h, . . . , k
consists of N (k − h + 1) = N (k) = 8 elements which are sorted according to the
lexicographic ordering (2.3) as follows:
(1, 1), (1, 2), (2, 1), (2, 2), (3, 1), (3, 2), (4, 1), (4, 2).
k
Note that j =h j 2 = (60, 20).
Example 2.2 Let a : [0, 1]2 → C and n ∈ N2 . Set
i n
x= a .
n i=1
Then, x is the vector of size N (n) given by
2.1 Notation and Terminology 9
⎡ ⎤
a( n11 , n12 )
⎢ 1 2 ⎥
⎢ a( n 1 , n 2 ) ⎥
⎢ ⎥
⎢ .. ⎥
⎢ . ⎥
⎢ ⎥
⎢ ⎥
⎢ a( n 1 , 1) ⎥
1
⎢ ⎥
⎢ 2 1 ⎥
⎢ a( n , n ) ⎥
⎢ 1 2 ⎥
⎢ 2 2 ⎥
⎢ a( , ) ⎥ ⎡ ⎤
⎢ n1 n2 ⎥ ⎡ ⎤ i
a( n11 , n12 )
⎢ .. ⎥
⎢ . ⎥ x1 ⎢ i ⎥
⎢ ⎥ ⎢x ⎥ ⎢ a( 1 , 2 ) ⎥
⎢ ⎥ ⎢ 2⎥ ⎢ n1 n2 ⎥
x=⎢
⎢ a( n 1 , 1) ⎥ = ⎢
⎥ xi 1 = ⎢ ⎥,
2
. ⎥, ⎢ .. ⎥ i 1 = 1, . . . , n 1 .
⎢ . ⎥ ⎣ .. ⎦ ⎢ . ⎥
⎢ .. ⎥ ⎣ ⎦
⎢ ⎥ xn 1
⎢ .. ⎥ i
a( n11 , 1)
⎢ ⎥
⎢ . ⎥
⎢ .. ⎥
⎢ ⎥
⎢ . ⎥
⎢ ⎥
⎢ a(1, 1 ) ⎥
⎢ n2 ⎥
⎢ ⎥
⎢ ⎥
⎢ a(1, n22 ) ⎥
⎢ ⎥
⎢ .. ⎥
⎣ . ⎦
a(1, 1)
n n
Moreover, x 2
= i=1 |x i |2 = i=1 |a(i/n)|2 .
Instead of indexing the entries of A in the standard way, i.e., by means of two
traditional scalar indices i, j = 1, . . . , 4, we can decide to index the entries of A by
means of two 2-indices (or bi-indices) i, j = 1, . . . , 2 = (1, 1), (1, 2), (2, 1), (2, 2).
This is possible because both the ranges 1, . . . , 4 and 1, . . . , 2 have 4 elements, and
4 is the size of A. The two writings A = [ai j ]i,4 j=1 and A = [a i j ]2i, j =1 correspond to
the two different indicizations. We have
The indicization of the entries of A with the bi-indices i, j reflects the fact that we
are thinking of A as a block matrix partitioned into 4 blocks as indicated in (2.6): for
all i, j = 1, . . . , 2, the entry ai j is the (i 2 , j2 ) entry of the (i 1 , j1 ) block of A. For
example,
a(2,1),(2,2) = entry (1, 2) in the (2, 2) block of A = 1.
Note that a i j depends only on the first components of the bi-indices i and j .
Example 2.4 (tensor products) Let X ∈ Cm 1 ×m 2 and Y ∈ C1 ×2 , and define the
block matrix
⎡ ⎤
x11 Y x12 Y · · · x1m 2 Y
⎢ x21 Y x22 Y · · · x2m 2 Y ⎥
⎢ ⎥ m ×m
Z = [xi j Y ] i=1,...,m 1 = ⎢ . .. .. ⎥ ∈ C 1 1 2 2 . (2.7)
j=1,...,m 2 ⎣ .. . . ⎦
xm 1 1 Y xm 1 2 Y · · · xm 1 m 2 Y
which are satisfied for all integers r ≥ 0 and s ≥ 1, for every i = 1, . . . , m 1 1 and
j = 1, . . . , m 2 2 we can write
It is clear that this expression is rather complicated. Now, suppose we decide to index
the entries of Z by two bi-indices i, j such that i = 1, . . . , n and j = 1, . . . , k, where
n = (m 1 , 1 ) and k = (m 2 , 2 ). This indicization, which is possible because
reflects the fact that we are thinking of Z as an m 1 × m 2 block matrix in which each
of the m 1 m 2 blocks is of size 1 × 2 . Actually, this is the natural way of thinking in
view of the block structure of Z ; see (2.7). With such indicization, for all i = 1, . . . , n
and j = 1, . . . , k we have
We then see that z i j has a much simpler expression than z i j . To conclude, we remark
that the matrix Z is the so-called tensor (Kronecker) product of X and Y , and it is
usually denoted by X ⊗ Y ; we shall come back to tensor products in Sect. 2.5.
The level orders n 1 , n 2 , . . . , n d and the order N are also referred to as partial orders
and total order, respectively. Indexing the entries of a d-level matrix A by two tradi-
tional scalar indices i, j = 1, . . . , N is a nightmare. On the contrary, A admits a natu-
ral indicization by means of two d-indices i, j = 1, . . . , n, where n = (n 1 , . . . , n d ).
Indeed, A can be written in the form (2.4) as follows:
A = [A i j ]ni, j =1 ,
We recall from Sect. 2.1.1 that a sequence of matrices is a sequence of the form
{An }n , where n varies in some infinite subset of N and An is a square matrix of size
dn → ∞. A d-level matrix-sequence is a special sequence of matrices of the form
{A n }n , where:
• n varies in some infinite subset of N;
• n = n(n) is a d-index with positive components which depends on n and satisfies
n → ∞ as n → ∞;
• A n is a square matrix of size N (n).
Recall from Sect. 2.1.2 that n → ∞ means min(n) → ∞.
The name “d-level matrix-sequence” is due to the fact that, in practical applica-
tions, especially in the context of PDE discretizations, each matrix A n of a d-level
matrix-sequence {A n }n is normally a d-level matrix with level orders (n 1 , . . . , n d ) =
n; see Example 2.5. Throughout this book, we often use the abbreviation “matrix-
sequence” for both “d-level matrix-sequence” and “multilevel matrix-sequence”.
N
f (θ) = f k eik·θ , f −N , . . . , f N ∈ C, N ∈ Nd . (2.8)
k=−N
2.2 Multivariate Trigonometric Polynomials 13
the numbers f −N , . . . , f N appearing in (2.8) are the (only possible nonzero) Fourier
coefficients of f according to the definition (2.1). Note also that a 1-variate (or
univariate) trigonometric polynomial is just a trigonometric polynomial in the clas-
sical sense. In what follows, we say that a d-variate trigonometric polynomial is
d-separable (or simply separable) if it is a d-separable function from Rd to C accord-
ing to the definition in Sect. 2.1.1.
Proof Since f 2 , . . . , f d are not identically 0, there exists (ϑ2 , . . . , ϑd ) such that
f 2 (ϑ2 ) · · · f d (ϑd ) = 0. The definition of d-variate trigonometric polynomials implies
that θ1 → f (θ1 , ϑ2 , . . . , ϑd ) = f 1 (θ1 ) f 2 (ϑ2 ) · · · f d (ϑd ) is a (univariate) trigonomet-
ric polynomial, and this means that f 1 is a trigonometric polynomial. With the same
argument, one can show that f 2 , . . . , f d are trigonometric polynomials as well.
The next lemma shows that the set of zeros of every non-trivial d-variate trigono-
metric polynomial has zero measure.
Proof The proof proceeds by induction on d. For d = 1, the result has already been
proved in [22, solution of Exercise 6.2, pp. 286–287]. Supposed > 1 and assume
N
that the lemma holds for dimensions up to d − 1. Let f (θ) = k=−N f k eik·θ and
set Z = { f = 0}. By Fubini’s theorem,
μd (Z ) = χ Z (θ1 , . . . , θd )dθ1 . . . dθd = dθ2 . . . dθd χ Z (θ1 , . . . , θd )dθ1
R Rd−1 R
d
= dθ2 . . . dθd dθ1 , (2.10)
Rd−1 Z θ2 ,...,θd
where, for each fixed (θ2 , . . . , θd ) ∈ Rd−1 , the set Z θ2 ,...,θd is defined by
Write
N1
f (θ ) = pk1 (θ2 , . . . , θd ) eik1 θ1 , (2.11)
k1 =−N1
(N2
,...,Nd )
pk1 (θ2 , . . . , θd ) = f k ei(k2 ,...,kd )·(θ2 ,...,θd ) , k1 = −N1 , . . . , N1 .
(k2 ,...,kd )=−(N2 ,...,Nd )
Let
Since not all the Fourier coefficients f k are equal to 0, at least one of the polyno-
mials pk1 has at least a nonzero Fourier coefficient. Thus, by induction hypothesis,
μd−1 (A) = 0. Moreover, by (2.11),
• if (θ2 , . . . , θd ) ∈ A then Z θ2 ,...,θd = R;
• if (θ2 , . . . , θd ) ∈ Ac = Rd−1 \A then there exists an index k1 ∈ {−N1 , . . . , N1 }
such that pk1 (θ2 , . . . , θd ) = 0, and so μ1 (Z θ2 ,...,θd ) = 0 by induction hypothesis.
Going back to (2.10), we see that
μd (Z ) = dθ2 . . . dθd dθ1 + dθ2 . . . dθd dθ1 = 0,
A Z θ2 ,...,θd Ac Z θ2 ,...,θd
pε ∞ ≤ fε ∞, f ε − pε ∞ < ε. (2.13)
Lemma 2.4 Let κ : [0, 1]d × [−π, π ]d → C be a measurable function. Then, there
exists a sequence {κm }m such that κm : [0, 1]d × [−π, π ]d → C is a function of the
form
Nm
κm (x, θ ) = a (m)
j (x) e
i j ·θ
, a (m)
j ∈ C ∞ ([0, 1]d ), N m ∈ Nd , (2.14)
j =−N m
and κm → κ a.e.
Proof The proof is essentially the same as [22, proof of Lemma 2.8]. The function
κ̃m = κχ{|κ|≤1/m} belongs to L ∞ ([0, 1]d × [−π, π ]d ) and converges to κ in measure.
Indeed, κ̃m → κ pointwise over [0, 1]d × [−π, π ]d , and the pointwise (a.e.) conver-
gence on a set of finite measure implies the convergence in measure [22, Lemma 2.4].
By [22, Lemma 2.2], the space generated by the trigonometric monomials
e2πi·x ei j ·θ = ei(2π1 x1 +...+2πd xd + j1 θ1 +...+ jd θd ) : , j ∈ Zd
μ2d {|κm − κ| > ε} ≤ μ2d ({|κm − κ̃m | > ε/2} ∪ {|κ̃m − κ| > ε/2})
≤ μ2d {|κm − κ̃m | > ε/2} + μ2d {|κ̃m − κ| > ε/2}
κm − κ̃m L 1
≤ + μ2d {|κ̃m − κ| > ε/2},
(ε/2)
and let
2.3 Multivariate Riemann-Integrable Functions 17
a i,n ∈ inf a(x), sup a(x) , i = 1, . . . , n.
x∈I i,n x∈I i,n
Then
n
a i,n χ Ii,n → a a.e. in [0, 1]d (2.15)
i=1
and
1
n
lim a i,n = a(x)dx. (2.16)
n→∞ N (n) i=1 [0,1]d
Proof The proof is essentially the same as [22, proof of Lemma 2.9]. Fix ε > 0 and let
x ∈ (0, 1]d be a continuity point of a. Then, there is a δ > 0 such that |a(y) − a(x)| ≤
ε whenever y ∈ [0, 1]d and |y − x|∞ ≤ δ. Take n ≥ (1/δ)1 and call I k,n the unique
hyperrectangle of the partition (0, 1]d = ni=1 I i,n containing x. For y ∈ I k,n , we
have y ∈ [0, 1]d and |y − x|∞ ≤ δ, hence |a(y) − a(x)| ≤ ε. It follows that
! n !
! !
! a χ (x) − a(x) ! = |ak,n − a(x)|
! i,n I i,n !
i=1
As a consequence, ni=1 a i,n χ Ii,n (x) → a(x) whenever x is a continuity point of a in
(0, 1]d . This implies (2.15), because a is Riemann-integrable and hence continuous
a.e. in [0, 1]d . Since
! n ! "
! ! 1
n n
! a i,n χ Ii,n !! ≤ a < ∞, a i,n = a i,n χ Ii,n ,
! ∞
N (n) i=1
i=1 [0,1]d i=1
(2.16) follows from (2.15) and from the dominated convergence theorem.
For the reader’s convenience, we report from [22, Sects. 2.4.1 and 2.4.3] several
matrix-norm inequalities that we shall use throughout the book. First, we recall the
expressions of the p-norms for p = 1, ∞:
m
m
|X |1 = max |xi j |, |X |∞ = max |xi j |, X ∈ Cm×m .
j=1,...,m i=1,...,m
i=1 j=1
18 2 Mathematical Background
XY 1 ≤ X p Y q, X, Y ∈ Cm×m . (2.18)
If X, Y are matrices of any dimension, say X ∈ Cm 1 ×m 2 and Y ∈ C1 ×2 , the tensor
(Kronecker) product of X and Y is the m 1 1 × m 2 2 matrix defined by
⎡ ⎤
x11 Y · · · x1m 2 Y
⎢ .. ⎥ ,
X ⊗ Y = xi j Y i=1,...,m 1 = ⎣ ... . ⎦
j=1,...,m 2
xm 1 1 Y · · · xm 1 m 2 Y
Tensor products and direct sums possess a lot of nice algebraic properties.
(i) Associativity: for all matrices X , Y , Z ,
(X ⊗ Y ) ⊗ Z = X ⊗ (Y ⊗ Z ),
(X ⊕ Y ) ⊕ Z = X ⊕ (Y ⊕ Z ).
(X 1 ⊗ Y1 )(X 2 ⊗ Y2 ) = (X 1 X 2 ) ⊗ (Y1 Y2 ),
(X 1 ⊕ Y1 )(X 2 ⊕ Y2 ) = (X 1 X 2 ) ⊕ (Y1 Y2 ).
(X ⊗ Y )∗ = X ∗ ⊗ Y ∗ , (X ⊗ Y )T = X T ⊗ Y T
(X ⊕ Y )∗ = X ∗ ⊕ Y ∗ , (X ⊕ Y )T = X T ⊕ Y T .
(iv) Bilinearity (of tensor products): for each fixed matrix X , the application
Y → X ⊗ Y
is linear on C1 ×2 for all 1 , 2 ∈ N; for each fixed matrix Y , the application
X → X ⊗ Y
see [22, Exercise 2.5]. In particular, for all X ∈ Cm×m , Y ∈ C× , and 1 ≤ p ≤ ∞,
we have
X ⊗Y p = X p Y p, (2.25)
! ! p p 1/ p
( X p + Y p ) , if 1 ≤ p < ∞,
X ⊕Y = ! ( X p , Y p )! p =
p
max( X ∞ , Y ∞ ), if p = ∞,
(2.26)
20 2 Mathematical Background
and
d
rank(X i − Yi )
rank(X 1 ⊗ · · · ⊗ X d − Y1 ⊗ · · · ⊗ Yd ) ≤ N (m) , (2.31)
i=1
mi
rank(X 1 ⊗ · · · ⊗ X d − Y1 ⊗ · · · ⊗ Yd )
d "
= rank Y1 ⊗ · · · ⊗ Yi−1 ⊗ (X i − Yi ) ⊗ X i+1 ⊗ · · · ⊗ X d
i=1
d
≤ rank(Y1 ⊗ · · · ⊗ Yi−1 ⊗ (X i − Yi ) ⊗ X i+1 ⊗ · · · ⊗ X d )
i=1
d
= rank(Y1 ⊗ · · · ⊗ Yi−1 )rank(X i − Yi )rank(X i+1 ⊗ · · · ⊗ X d )
i=1
Another random document with
no related content on Scribd:
Bathe the eye frequently with warm milk and water, and apply,
every night at bedtime, a warm white-bread poultice.
No medicine is required; but if the child be gross, keep him for a
few days from meat, and let him live on bread and milk and
farinaceous puddings.
243. If a child have large bowels, what would you recommend as
likely to reduce their size?
It ought to be borne in mind that the bowels of a child are larger in
proportion than those of an adult. But, if they be actually larger than
they ought to be, let them be well rubbed for a quarter of an hour at a
time night and morning, with soap liniment, and then apply a broad
flannel belt. “A broad flannel belt worn night and day, firm but not
tight, is very serviceable.”[244] The child ought to be prevented from
drinking as much as he has been in the habit of doing; let him be
encouraged to exercise himself well in the open air; and let strict
regard be paid to his diet.
244. What are the best aperients for a child?
If it be actually necessary to give him opening medicine, one or
two teaspoonfuls of syrup of senna, repeated if necessary, in four
hours, will generally answer the purpose; or, for a change, one or two
teaspoonfuls of castor oil may be substituted. Lenitive electuary
(compound confection of senna) is another excellent aperient for the
young, it being mild in its operation, and pleasant to take; a child
fancying it is nothing more than jam, and which it much resembles
both in appearance and in taste. The dose is half or one teaspoonful
early in the morning occasionally. Senna is an admirable aperient for
a child, and is a safe one, which is more than can be said of many
others. It is worthy of note that “the taste of senna may be concealed
by sweetening the infusion,[245] adding milk, and drinking as ordinary
tea, which, when thus prepared, it much resembles.”[246] Honey, too,
is a nice aperient for a child—a teaspoonful ought to be given either
by itself, or spread on a slice of bread.
Some mothers are in the habit of giving their children jalap
gingerbread. I do not approve of it, as jalap is a drastic griping
purgative; besides, jalap is very nasty to take—nothing will make it
palatable.
Fluid magnesia—solution of the bicarbonate of magnesia—is a
good aperient for a child; and, as it has very little taste, is readily
given, more especially if made palatable by the addition either of a
little syrup or of brown sugar. The advantages which it has over the
old solid form are, that it is colorless and nearly tasteless, and never
forms concretions in the bowels, as the solid magnesia, if persevered
in for any length of time, sometimes does. A child two or three years
old may take one or two tablespoonfuls of the fluid, either by itself or
in his food, repeating it every four hours until the bowels be opened.
When the child is old enough to drink the draught off immediately,
the addition of one or two teaspoonfuls of lemon-juice, to each dose
of the fluid magnesia, makes a pleasant effervescing draught, and
increases its efficacy as an aperient.
Bran-bread[247] and treacle will frequently open the bowels; and as
treacle is wholesome, it may be substituted for butter when the
bowels are inclined to be costive. A roasted apple, eaten with raw
sugar, is another excellent mild aperient for a child. Milk gruel—that
is to say, milk thickened with oatmeal—forms an excellent food for
him, and often keeps his bowels regular, and thus (which is a very
important consideration) supersedes the necessity of giving him an
aperient. An orange (taking care he does not eat the peel or the pulp),
or a fig after dinner, or a few Muscatel raisins, will frequently
regulate the bowels.
Stewed prunes is another admirable remedy for the costiveness of
a child. The manner of stewing them is as follows: Put a pound of
prunes in a brown jar, add two tablespoonfuls of raw sugar, then
cover the prunes and the sugar with cold water; place them in the
oven, and let them stew for four hours. A child should every morning
eat half a dozen or a dozen of them, until the bowels be relieved,
taking care that he does not swallow the stones.
A suppository is a mild and ready way of opening the bowels of a
child. When he is two or three years old and upwards, a candle
suppository is better than a soap suppository. The way of preparing
it is as follows: Cut a piece of dip-tallow candle—the length of three
inches—and insert it as you would a clyster pipe, about two inches up
the fundament, allowing the remaining inch to be in sight, and there
let the suppository remain until the bowels be opened.
245. What are the most frequent causes of Protrusion of the lower
bowel?
The too common and reprehensible practice of a parent
administering frequent aperients, especially calomel and jalap, to her
child. Another cause, is allowing him to remain for a quarter of an
hour or more at a time on his chair; this induces him to strain, and to
force the gut down.
246. What are the remedies?
If the protrusion of the bowel have been brought on by the abuse
of aperients, abstain for the future from giving them; but if medicine
be absolutely required, give the mildest—such as either syrup of
senna or castor oil—and the less of those the better.
If the external application of a purgative will have the desired
effect, it will, in such cases, be better than the internal
administration of aperients. Dr. Merriman’s Purgative Liniment[248]
is a good one for the purpose. Let the bowels be well rubbed every
night and morning, for five minutes at a time, with the liniment.
A wet compress to the bowels will frequently open them, and will
thus do away with the necessity of giving an aperient—a most
important consideration. Fold a napkin in six thicknesses, soak it in
cold water, and apply it to the bowels, over which put either a thin
covering or sheet of gutta-percha, or a piece of oiled silk; keep it in its
place with a broad flannel roller, and let it remain on the bowels for
three or four hours, or until they be opened.
Try what diet will do, as opening the bowels by a regulated diet is
far preferable to the giving of aperients. Let him have either bran-
bread, Robinson’s Patent Groats made into gruel with new milk, or
Du Barry’s Arabica Revalenta, or a slice of Huntley and Palmer’s
lump gingerbread. Let him eat stewed prunes, stewed rhubarb,
roasted apples, strawberries, raspberries, the inside of grapes and
gooseberries, figs, etc. Give him early every morning a draught of
cold water.
Let me, again, urge you not to give aperients in these cases, or in
any case, unless you are absolutely compelled. By following my
advice you will save yourself an immense deal of trouble, and your
child a long catalogue of misery. Again, I say, look well into the
matter, and whenever it be practicable, avoid purgatives.
Now, with regard to the best manner of returning the bowel, lay
the child upon the bed on his face and bowels, with his hips a little
raised; then smear lard on the forefinger of your right hand (taking
care that the nail be cut close) and gently with your forefinger press
the bowel into its proper place.
Remember, if the above methods be observed, you cannot do the
slightest injury to the bowel, and the sooner it be returned the better
it will be for the child; for, if the bowel be allowed to remain long
down, it may slough or mortify, and death may ensue. The nurse,
every time he has a motion, must see that the bowel does not come
down, and if it does, she ought instantly to return it. Moreover, the
nurse should be careful not to allow the child to remain on his chair
more than two or three minutes at a time.
Another excellent remedy for the protrusion of the lower bowel is
to use every morning a cold salt and water sitz-bath. There need not
be more than a depth of three inches of water in the bath; a small
handful of table salt should be dissolved in the water; a dash of warm
water in the winter time must be added to take off the extreme chill,
and the child ought not to be allowed to sit in the bath for more than
one minute, or while the mother can count a hundred, taking care
the while to throw either a square of flannel or a small shawl over his
shoulders. The sitz-bath ought to be continued for months, or until
the complaint be removed. I cannot speak in too high praise of these
baths.
247. Do you advise me, every spring and fall, to give my child
brimstone to purify and sweeten his blood, and as a preventive
medicine?
Certainly not: if you wish to take away his appetite and to weaken
and depress him, give him brimstone! Brimstone is not a remedy fit
for a child’s stomach. The principal use and value of brimstone is as
an external application in itch, and as an internal remedy, mixed
with other laxatives, in piles—piles being a complaint of adults. In
olden times poor unfortunate children were dosed every spring and
fall with brimstone and treacle, to sweeten their blood! Fortunately
for the present race, there is not so much of that folly practiced, but
still there is room for improvement.
To dose a healthy child with physic is the grossest absurdity. No,
the less physic a delicate child has the better it will be for him, but
physic to a healthy child is downright poison!—and brimstone of all
medicines! It is both weakening and depressing to the system, and by
opening the pores of the skin and by relaxing the bowels, is likely to
give cold, and thus to make a healthy a sickly child. Sweeten his
blood! It is more likely to weaken his blood, and thus to make his
blood impure! Blood is not made pure by drugs, but by Nature’s
medicine: by exercise, by pure air, by wholesome diet, by sleep in a
well-ventilated apartment, by regular and thorough ablution.
Brimstone a preventive medicine? Preventive medicine—and
brimstone especially in the guise of a preventive medicine—is “a
mockery, a delusion, and a snare.”
248. When a child is delicate, and his body, without any
assignable cause is gradually wasting away, and the stomach
rejects all food that is taken, what plan can be adopted likely to
support his strength, and thus, probably, be the means of saving his
life?
I have seen, in such a case, great benefit to arise from half a
teacupful of either strong mutton broth, or of strong beef-tea, used as
an enema, every four hours.[249]
It should be administered slowly, in order that it may remain in
the bowel. If the child be sinking, either a dessertspoonful of brandy,
or half a wineglassful of port wine ought to be added to each enema.
The above plan ought only to be adopted if there be no diarrhœa. If
there be diarrhœa, an enema must not be used. Then, provided there
be great wasting away, and extreme exhaustion, and other remedies
having failed, it would be advisable to give, by the mouth, raw beef
of the finest quality, which ought to be taken from the hip-bone, and
should be shredded very fine. All fat and skin must be carefully
removed. One or two teaspoonfuls (according to the age of the child)
ought to be given every four hours. The giving of raw meat to
children in exhaustive diseases, such as excessive long-standing
diarrhœa, was introduced into practice by a Russian physician, a
Professor Weisse, of St. Petersburg. It certainly is, in these cases, a
most valuable remedy, and has frequently been the means of
snatching such patients from the jaws of death. Children usually take
raw meat with avidity and with a relish.
249. If a child be naturally delicate, what plan would you
recommend to strengthen him?
I should advise strict attention to the rules above mentioned, and
change of air—more especially, if it be possible, to the coast. Change
of air, sometimes, upon a delicate child, acts like magic, and may
restore him to health when all other means have failed. If a girl be
delicate, “carry her off to the farm, there to undergo the discipline of
new milk, brown bread, early hours, no lessons, and romps in the
hay-field.”[250] This advice is, of course, equally applicable for a
delicate boy, as delicate boys and delicate girls ought to be treated
alike. Unfortunately, in these very enlightened days! there is too
great a distinction made in the respective management and
treatment of boys and girls.
The best medicines for a delicate child will be the wine of iron and
cod-liver oil. Give them combined in the manner I shall advise when
speaking of the treatment of Rickets.
In diseases of long standing, and that resist the usual remedies,
there is nothing like change of air. Hippocrates, the father of
medicine, says:
“In longis morbis solum mutare.”
(In tedious diseases to change the place of residence.)