Sample Paper for the amsmath Package File name: testmath.tex

American Mathematical Society and Roland Campbell and Mark M. Dane and Jeremiah Jones

This is a test file containing extensive examples of mathematical constructs supported by the amsmath package.

Copyright 1995, 1999 American Mathematical Society, all rights reserved. Copying of this file is authorized only if either: (1) you make absolutely no changes to your copy, including name; OR (2) if you do make changes, you first rename it to some other name.

File renamed as testams.tex; meta data added for testing Tralics. Translation date: November 13, 2008.

Date: Version 2.0, 1999/11/15

chap1: First Author and Second Author

chap2: Third Author and Last Author

A communicator

A first translator and A second translator

Some dedicatory

Key words and phrases: latex, xml, html, math, sigma multipliers, strange duality.

(1991) 16H59

Sample Paper for the amsmath Package File name: testmath.tex

American Mathematical Society and Roland Campbell and Dane and Jeremiah Jones

A first thanks

A second thanks

Address: Department of Mathematics, Pennsylvania State University, Pittsburgh, Pennsylvania 13593


Current address: Atmospheric Research Station, Pala Lundi, Fiji

Email: DaneMark@ffr.choice


Address: Department of Philosophy, Freedman College, Periwinkle, Colorado 84320

Email: id739e@oseoi44 (Bitnet)

Short Table of Contents

1. Introduction
2. Enumeration of Hamiltonian paths in a graph
3. Main Theorem
4. Application
5. Secret Key Exchanges
6. Review
7. One-Way Complexity
8. Various font features of the amsmath package
9. Compound symbols and other features
1. appendix
1. Examples of multiple-line equation structures

1. Introduction

This paper contains examples of various features from AMS-LaTeX.

2. Enumeration of Hamiltonian paths in a graph

Let A=(a ij ) be the adjacency matrix of graph G. The corresponding Kirchhoff matrix K=(k ij ) is obtained from A by replacing in A each diagonal entry by the degree of its corresponding vertex; i.e., the ith diagonal entry is identified with the degree of the ith vertex. It is well known that


where K(i|i) is the ith principal submatrix of K.

1 \det\mathbf{K}(i|i)=\text{ the number of spanning trees of $G$},

Let C i(j) be the set of graphs obtained from G by attaching edge (v i v j ) to each spanning tree of G. Denote by C i = j C i(j) . It is obvious that the collection of Hamiltonian cycles is a subset of C i . Note that the cardinality of C i is k ii detK(i|i). Let X ^={x ^ 1 ,,x ^ n }.

2 $\wh X=\{\hat x_1,\dots,\hat x_n\}$

Define multiplication for the elements of X ^ by

(2.2)x ^ i x ^ j =x ^ j x ^ i ,x ^ i 2 =0,i,j=1,,n.

Let k ^ ij =k ij x ^ j and k ^ ij = ji k ^ ij . Then the number of Hamiltonian cycles H c is given by the relation [8]

(2.3) j=1 n x ^ j H c =1 2k ^ ij detK ^(i|i),i=1,,n.

The task here is to express (2.3) in a form free of any x ^ i , i=1,,n. The result also leads to the resolution of enumeration of Hamiltonian paths in a graph.

It is well known that the enumeration of Hamiltonian cycles and paths in a complete graph K n and in a complete bipartite graph K n 1 n 2 can only be found from first combinatorial principles [4]. One wonders if there exists a formula which can be used very efficiently to produce K n and K n 1 n 2 . Recently, using Lagrangian methods, Goulden and Jackson have shown that H c can be expressed in terms of the determinant and permanent of the adjacency matrix [3]. However, the formula of Goulden and Jackson determines neither K n nor K n 1 n 2 effectively. In this paper, using an algebraic method, we parametrize the adjacency matrix. The resulting formula also involves the determinant and permanent, but it can easily be applied to K n and K n 1 n 2 . In addition, we eliminate the permanent from H c and show that H c can be represented by a determinantal function of multivariables, each variable with domain {0,1}. Furthermore, we show that H c can be written by number of spanning trees of subgraphs. Finally, we apply the formulas to a complete multigraph K n 1 n p .

The conditions a ij =a ji , i,j=1,,n, are not required in this paper. All formulas can be extended to a digraph simply by multiplying H c by 2.

3. Main Theorem

Notation . For p,qP and nω we write (q,n)(p,n) if qp and A q,n =A p,n .

3 \begin{notation} For $p,q\in P$ and $n\in\omega$
4 ...
5 \end{notation}

Let B=(b ij ) be an n×n matrix. Let n={1,,n}. Using the properties of (2.2), it is readily seen that

Lemma 3.1.

(3.2) in jn b ij x ^ i = in x ^ i perB

where perB is the permanent of B.

Let Y ^={y ^ 1 ,,y ^ n }. Define multiplication for the elements of Y ^ by

(3.3)y ^ i y ^ j +y ^ j y ^ i =0,i,j=1,,n.

Then, it follows that

Lemma 3.4.

(3.5) in jn b ij y ^ j = in y ^ i detB.

Note that all basic properties of determinants are direct consequences of Lemma  3.4. Write

(3.6) jn b ij y ^ j = jn b ij (λ) y ^ j +(b ii λ i )y ^ i y ^


(3.7)b ii (λ) =λ i ,b ij (λ) =b ij ,ij.

Let B (λ) =(b ij (λ) ). By (3.5) and (3.6), it is straightforward to show the following result:

Theorem 3.8.

(3.9)detB= l=0 n I l n iI l (b ii λ i )detB (λ) (I l |I l ),

where I l ={i 1 ,,i l } and B (λ) (I l |I l ) is the principal submatrix obtained from B (λ) by deleting its i 1 ,,i l rows and columns.

Remark 3.1. Let M be an n×n matrix. The convention M(n|n)=1 has been used in (3.9) and hereafter.

Before proceeding with our discussion, we pause to note that Theorem 3.8 yields immediately a fundamental formula which can be used to compute the coefficients of a characteristic polynomial [10]:

Corollary 3.10. Write det(BxI)= l=0 n (1) l b l x l . Then

(3.11)b l = I l n detB(I l |I l ).


(3.12)K(t,t 1 ,,t n )=D 1 ta 12 t 2 a 1n t n a 21 t 1 D 2 ta 2n t n a n1 t 1 a n2 t 2 D n t,
6 \begin{pmatrix} D_1t&-a_{12}t_2&\dots&-a_{1n}t_n\\
7 -a_{21}t_1&D_2t&\dots&-a_{2n}t_n\\
8 \hdotsfor[2]{4}\\
9 -a_{n1}t_1&-a_{n2}t_2&\dots&D_nt\end{pmatrix}


(3.13)D i = jn a ij t j ,i=1,,n.


D(t 1 ,,t n )=δ δtdetK(t,t 1 ,,t n ) t=1 .


(3.14)D(t 1 ,,t n )= in D i detK(t=1,t 1 ,,t n ;i|i),

where K(t=1,t 1 ,,t n ;i|i) is the ith principal submatrix of K(t=1,t 1 ,,t n ).

Theorem  3.8 leads to

(3.15)detK(t 1 ,t 1 ,,t n )= In (1) I t nI iI t i jI (D j +λ j t j )detA (λt) (I ¯|I ¯).

Note that

(3.16)detK(t=1,t 1 ,,t n )= In (1) I iI t i jI (D j +λ j t j )detA (λ) (I ¯|I ¯)=0.

Let t i =x ^ i ,i=1,,n. Lemma  3.1 yields

(3) in a l i x i detK(t=1,x 1 ,,x n ;l|l)= in x ^ i In{l} (1) I perA (λ) (I|I)detA (λ) (I ¯{l}|I ¯{l}).
10 \begin{multline}
11 \biggl(\sum_{\,i\in\mathbf{n}}a_{l _i}x_i\biggr)
12 \det\mathbf{K}(t=1,x_1,\dots,x_n;l |l )\\
13 =\biggl(\prod_{\,i\in\mathbf{n}}\hat x_i\biggr)
14 \sum_{I\subseteq\mathbf{n}-\{l \}}
15 (-1)^{\envert{I}}\per\mathbf{A}^{(\lambda)}(I|I)
16 \det\mathbf{A}^{(\lambda)}
17 (\overline I\cup\{l \}|\overline I\cup\{l \}).
18 \label{sum-ali}
19 \end{multline}

By (2.3), (3.5), and (3.6), we have

Proposition 3.18.

(3.19)H c =1 2n l=0 n (1) l D l ,


(3.20)D l = I l n D(t 1 ,,t n )2| t i =0,ifiI l 1,otherwise,i=1,,n .

4. Application

We consider here the applications of Theorems 5.1 and  5.2 to a complete multipartite graph K n 1 n p . It can be shown that the number of spanning trees of K n 1 n p may be written

(4.1)T=n p2 i=1 p (nn i ) n i 1


(4.2)n=n 1 ++n p .

It follows from Theorems 5.1 and  5.2 that

(4.3)H c =1 2n l=0 n (1) l (nl) p2 l 1 ++l p =l i=1 p n i l i ·[(nl)(n i l i )] n i l i ·(nl) 2 j=1 p (n i l i ) 2 .
20 ... \binom{n_i}{l _i}\\


(4.4)H c =1 2 l=0 n1 (1) l (nl) p2 l 1 ++l p =l i=1 p n i l i ·[(nl)(n i l i )] n i l i 1l p n p [(nl)(n p l p )].

The enumeration of H c in a K n 1 n p graph can also be carried out by Theorem  7.15 or  7.23 together with the algebraic method of (2.2). Some elegant representations may be obtained. For example, H c in a K n 1 n 2 n 3 graph may be written

(4.5)H c =n 1 !n 2 !n 3 ! n 1 +n 2 +n 3 i n 1 in 2 n 3 n 1 +in 3 n 3 n 2 +i+n 1 1 in 2 1 n 3 n 1 +in 3 1 n 3 n 2 +i.

5. Secret Key Exchanges

Modern cryptography is fundamentally concerned with the problem of secure private communication. A Secret Key Exchange is a protocol where Alice and Bob, having no secret information in common to start, are able to agree on a common secret key, conversing over a public channel. The notion of a Secret Key Exchange protocol was first introduced in the seminal paper of Diffie and Hellman [1]. [1] presented a concrete implementation of a Secret Key Exchange protocol, dependent on a specific assumption (a variant on the discrete log), specially tailored to yield Secret Key Exchange. Secret Key Exchange is of course trivial if trapdoor permutations exist. However, there is no known implementation based on a weaker general assumption.

The concept of an informationally one-way function was introduced in [5]. We give only an informal definition here:

Definition 5.1. A polynomial time computable function f={f k } is informationally one-way if there is no probabilistic polynomial time algorithm which (with probability of the form 1k e for some e>0) returns on input y{0,1} k a random element of f 1 (y).

In the non-uniform setting [5] show that these are not weaker than one-way functions:

Theorem 5.1 ([5] (non-uniform)). The existence of informationally one-way functions implies the existence of one-way functions.

We will stick to the convention introduced above of saying “non-uniform” before the theorem statement when the theorem makes use of non-uniformity. It should be understood that if nothing is said then the result holds for both the uniform and the non-uniform models.

It now follows from Theorem 5.1 that

Theorem 5.2 (non-uniform). Weak SKE implies the existence of a one-way function.

More recently, the polynomial-time, interior point algorithms for linear programming have been extended to the case of convex quadratic programs [9], [13], certain linear complementarity problems [7], [11], and the nonlinear complementarity problem [6]. The connection between these algorithms and the classical Newton method for nonlinear equations is well explained in [7].

6. Review

We begin our discussion with the following definition:

Definition 6.1. A function H: n n is said to be B-differentiable at the point z if (i) H is Lipschitz continuous in a neighborhood of z, and (ii)  there exists a positive homogeneous function BH(z): n n , called the B-derivative of H at z, such that

lim v0 H(z+v)H(z)BH(z)v v=0.

The function H is B-differentiable in set S if it is B-differentiable at every point in S. The B-derivative BH(z) is said to be strong if

lim (v,v ' )(0,0) H(z+v)H(z+v ' )BH(z)(vv ' ) vv ' =0.

Lemma 6.1. There exists a smooth function ψ 0 (z) defined for z>12a satisfying the following properties:

  1. ψ 0 (z) is bounded above and below by positive constants c 1 ψ 0 (z)c 2 .

  2. If z>1, then ψ 0 (z)=1.

  3. For all z in the domain of ψ 0 , Δ 0 lnψ 0 0.

  4. If 12a<z<1a, then Δ 0 lnψ 0 c 3 >0.

We choose ψ 0 (z) to be a radial function depending only on r=z. Let h(r)0 be a suitable smooth function satisfying h(r)c 3 for 12a<z<1a, and h(r)=0 for z>1a 2. The radial Laplacian

Δ 0 lnψ 0 (r)=d 2 dr 2 +1 rd drlnψ 0 (r)

has smooth coefficients for r>12a. Therefore, we may apply the existence and uniqueness theory for ordinary differential equations. Simply let lnψ 0 (r) be the solution of the differential equation

d 2 dr 2 +1 rd drlnψ 0 (r)=h(r)

with initial conditions given by lnψ 0 (1)=0 and lnψ 0 ' (1)=0.

Next, let D ν be a finite collection of pairwise disjoint disks, all of which are contained in the unit disk centered at the origin in C. We assume that D ν ={zzz ν <δ}. Suppose that D ν (a) denotes the smaller concentric disk D ν (a)={zzz ν (12a)δ}. We define a smooth weight function Φ 0 (z) for zC ν D ν (a) by setting Φ 0 (z)=1 when z ν D ν and Φ 0 (z)=ψ 0 ((zz ν )/δ) when z is an element of D ν . It follows from Lemma 6.1 that Φ 0 satisfies the properties:

  1. Φ 0 (z) is bounded above and below by positive constants c 1 Φ 0 (z)c 2 .

  2. Δ 0 lnΦ 0 0 for all zC ν D ν (a), the domain where the function Φ 0 is defined.

  3. Δ 0 lnΦ 0 c 3 δ 2 when (12a)δ<zz ν <(1a)δ.

Let A ν denote the annulus A ν ={(12a)δ<zz ν <(1a)δ}, and set A= ν A ν . The properties ((2)) and ((3)) of Φ 0 may be summarized as Δ 0 lnΦ 0 c 3 δ 2 χ A , where χ A is the characteristic function of A.

Suppose that α is a nonnegative real constant. We apply Proposition 3.18 with Φ(z)=Φ 0 (z)e αz 2 . If uC 0 (R 2 ν D ν (a)), assume that D is a bounded domain containing the support of u and ADR 2 ν D ν (a). A calculation gives

D ¯u 2 Φ 0 (z)e αz 2 c 4 α D u 2 Φ 0 e αz 2 +c 5 δ 2 A u 2 Φ 0 e αz 2 .

The boundedness, property ((1)) of Φ 0 , then yields

D ¯u 2 e αz 2 c 6 α D u 2 e αz 2 +c 7 δ 2 A u 2 e αz 2 .

Let B(X) be the set of blocks of Λ X and let b(X)=B(X). If φQ X then φ is constant on the blocks of Λ X .

(6.2)P X ={φMΛ φ =Λ X },Q X ={φMΛ φ Λ X }.

If Λ φ Λ X then Λ φ =Λ Y for some YX so that

Q X = YX P Y .

Thus by Möbius inversion

P Y = XY μ(Y,X)Q X .

Thus there is a bijection from Q X to W B(X) . In particular Q X =w b(X) .

Next note that b(X)=dimX. We see this by choosing a basis for X consisting of vectors v k defined by

v i k =1ifiΛ k ,0otherwise.
21 \[v^{k}_{i}=
22 \begin{cases} 1 & \text{if $i \in \Lambda_{k}$},\\
23 0 &\text{otherwise.} \end{cases}
24 \]

Lemma 6.3. Let A be an arrangement. Then

χ(A,t)= BA (1) B t dimT(B) .

In order to compute R '' recall the definition of S(X,Y) from Lemma 3.1. Since HB, A H B. Thus if T(B)=Y then BS(H,Y). Let L '' =L(A '' ). Then

(6.4)R '' = HBA (1) B t dimT(B) = YL '' BS(H,Y) (1) B t dimY = YL '' BS(H,Y) (1) BA H t dimY = YL '' μ(H,Y)t dimY =χ(A '' ,t).

Corollary 6.5. Let (A,A ' ,A '' ) be a triple of arrangements. Then

π(A,t)=π(A ' ,t)+tπ(A '' ,t).

Definition 6.2. Let (A,A ' ,A '' ) be a triple with respect to the hyperplane HA. Call H a separator if T(A)L(A ' ).

Corollary 6.6. Let (A,A ' ,A '' ) be a triple with respect to HA.

  1. If H is a separator then

    μ(A)=μ(A '' )

    and hence

    μ(A)=μ(A '' ).
  2. If H is not a separator then

    μ(A)=μ(A ' )μ(A '' )


    μ(A)=μ(A ' )+μ(A '' ).

It follows from Theorem 5.1 that π(A,t) has leading term

(1) r(A) μ(A)t r(A) .

The conclusion follows by comparing coefficients of the leading terms on both sides of the equation in Corollary 6.5. If H is a separator then r(A ' )<r(A) and there is no contribution from π(A ' ,t).

The Poincaré polynomial of an arrangement will appear repeatedly in these notes. It will be shown to equal the Poincaré polynomial of the graded algebras which we are going to associate with A. It is also the Poincaré polynomial of the complement M(A) for a complex arrangement. Here we prove that the Poincaré polynomial is the chamber counting function for a real arrangement. The complement M(A) is a disjoint union of chambers

M(A)= CCham(A) C.

The number of chambers is determined by the Poincaré polynomial as follows.

Theorem 6.7. Let A R be a real arrangement. Then

Cham(A R )=π(A R ,1).

We check the properties required in Corollary 6.6: (i) follows from π(Φ l ,t)=1, and (ii) is a consequence of Corollary 3.10.

Theorem 6.8. Let φ be a protocol for a random pair (X,Y). If one of σ φ (x ' ,y) and σ φ (x,y ' ) is a prefix of the other and (x,y)S X,Y , then

σ j (x ' ,y) j=1 =σ j (x,y) j=1 =σ j (x,y ' ) j=1 .

We show by induction on i that

σ j (x ' ,y) j=1 i =σ j (x,y) j=1 i =σ j (x,y ' ) j=1 i .

The induction hypothesis holds vacuously for i=0. Assume it holds for i1, in particular [σ j (x ' ,y)] j=1 i1 =[σ j (x,y ' )] j=1 i1 . Then one of [σ j (x ' ,y)] j=i and [σ j (x,y ' )] j=i is a prefix of the other which implies that one of σ i (x ' ,y) and σ i (x,y ' ) is a prefix of the other. If the ith message is transmitted by P X then, by the separate-transmissions property and the induction hypothesis, σ i (x,y)=σ i (x,y ' ), hence one of σ i (x,y) and σ i (x ' ,y) is a prefix of the other. By the implicit-termination property, neither σ i (x,y) nor σ i (x ' ,y) can be a proper prefix of the other, hence they must be the same and σ i (x ' ,y)=σ i (x,y)=σ i (x,y ' ). If the ith message is transmitted by P Y then, symmetrically, σ i (x,y)=σ i (x ' ,y) by the induction hypothesis and the separate-transmissions property, and, then, σ i (x,y)=σ i (x,y ' ) by the implicit-termination property, proving the induction step.

If φ is a protocol for (X,Y), and (x,y), (x ' ,y) are distinct inputs in S X,Y , then, by the correct-decision property, σ j (x,y) j=1 σ j (x ' ,y) j=1 .

Equation (6.4) defined P Y ´s ambiguity set S X|Y (y) to be the set of possible X values when Y=y. The last corollary implies that for all yS Y , the multiset(note: ) of codewords {σ φ (x,y):xS X|Y (y)} is prefix free.

7. One-Way Complexity

C ^ 1 (X|Y), the one-way complexity of a random pair (X,Y), is the number of bits P X must transmit in the worst case when P Y is not permitted to transmit any feedback messages. Starting with S X,Y , the support set of (X,Y), we define G(X|Y), the characteristic hypergraph of (X,Y), and show that

C ^ 1 (X|Y)=logχ(G(X|Y)).

Let (X,Y) be a random pair. For each y in S Y , the support set of Y, Equation (6.4) defined S X|Y (y) to be the set of possible x values when Y=y. The characteristic hypergraph G(X|Y) of (X,Y) has S X as its vertex set and the hyperedge S X|Y (y) for each yS Y .

We can now prove a continuity theorem.

Theorem 7.1. Let ΩR n be an open set, let uBV(Ω;R m ), and let

(7.2)T x u =yR m :y=u ˜(x)+Du Du(x),zforsomezR n

for every xΩS u . Let f:R m R k be a Lipschitz continuous function such that f(0)=0, and let v=f(u):ΩR k . Then vBV(Ω;R k ) and

(7.3)Jv=(f(u + )f(u ))ν u ·H n1 S u .

In addition, for D ˜u-almost every xΩ the restriction of the function f to T x u is differentiable at u ˜(x) and

(7.4)D ˜v=(f T x u )(u ˜)D ˜u D ˜u·D ˜u.

Before proving the theorem, we state without proof three elementary remarks which will be useful in the sequel.

Remark 7.1. Let ω:0,+0,+ be a continuous function such that ω(t)0 as t0. Then

lim h0 + g(ω(h))=Llim h0 + g(h)=L

for any function g:0,+R.

Remark 7.2. Let g:R n R be a Lipschitz continuous function and assume that

L(z)=lim h0 + g(hz)g(0) h

exists for every zQ n and that L is a linear function of z. Then g is differentiable at 0.

Remark 7.3. Let A:R n R m be a linear function, and let f:R m R be a function. Then the restriction of f to the range of A is differentiable at 0 if and only if f(A):R n R is differentiable at 0 and

(f Im(A) )(0)A=(f(A))(0).

We begin by showing that vBV(Ω;R k ) and


where K>0 is the Lipschitz constant of f. By (3.14) and by the approximation result quoted in §3, it is possible to find a sequence (u h )C 1 (Ω;R m ) converging to u in L 1 (Ω;R m ) and such that

lim h+ Ω u h dx=Du(Ω).

The functions v h =f(u h ) are locally Lipschitz continuous in Ω, and the definition of differential implies that v h Ku h almost everywhere in Ω. The lower semicontinuity of the total variation and (3.14) yield

(7.6)Dv(Ω)lim inf h+ Dv h (Ω)=lim inf h+ Ω v h dxKlim inf h+ Ω u h dx=KDu(Ω).

Since f(0)=0, we have also

Ω vdxK Ω udx;

therefore uBV(Ω;R k ). Repeating the same argument for every open set AΩ, we get (7.5) for every BB(Ω), because Dv, Du are Radon measures. To prove Lemma 6.1, first we observe that

(7.7)S v S u ,v ˜(x)=f(u ˜(x))xΩS u .

In fact, for every ε>0 we have

{yB ρ (x):v(y)f(u ˜(x))>ε}{yB ρ (x):u(y)u ˜(x)>ε/K},


lim ρ0 + {yB ρ (x):v(y)f(u ˜(x))>ε} ρ n =0

whenever xΩS u . By a similar argument, if xS u is a point such that there exists a triplet (u + ,u ,ν u ) satisfying (3.15), (3.16), then

(v + (x)v (x))ν v =(f(u + (x))f(u (x)))ν u ifxS v

and f(u (x))=f(u + (x)) if xS u S v . Hence, by (1.8) we get

Jv(B)= BS v (v + v )ν v dH n1 = BS v (f(u + )f(u ))ν u dH n1 = BS u (f(u + )f(u ))ν u dH n1

and Lemma 6.1 is proved.

To prove (7.7), it is not restrictive to assume that k=1. Moreover, to simplify our notation, from now on we shall assume that Ω=R n . The proof of (7.7) is divided into two steps. In the first step we prove the statement in the one-dimensional case (n=1), using Theorem 5.2. In the second step we achieve the general result using Theorem 7.1.

7.1. Step 1 Assume that n=1. Since S u is at most countable, (3.6) yields that D ˜v(S u S v )=0, so that (4.1) and (4.3) imply that Dv=D ˜v+Jv is the Radon-Nikodým decomposition of Dv in absolutely continuous and singular part with respect to D ˜u. By Theorem 5.2, we have

D ˜v D ˜u(t)=lim st + Dv(t,s) D ˜u(t,s),D ˜u D ˜u(t)=lim st + Du(t,s) D ˜u(t,s)

D ˜u-almost everywhere in R. It is well known (see, for instance, [12, 2.5.16]) that every one-dimensional function of bounded variation w has a unique left continuous representative, i.e., a function w ^ such that w ^=w almost everywhere and lim st w ^(s)=w ^(t) for every tR. These conditions imply

(7.8)u ^(t)=Du(,t),v ^(t)=Dv(,t)tR


(7.9)v ^(t)=f(u ^(t))tR.

Let tR be such that D ˜u(t,s)>0 for every s>t and assume that the limits in (4.4) exist. By (4.5) and (6.2) we get

v ^(s)v ^(t) D ˜u(t,s)=f(u ^(s))f(u ^(t)) D ˜u(t,s)=f(u ^(s))f(u ^(t)+D ˜u D ˜u(t)D ˜u(t,s)) D ˜u(t,s)+f(u ^(t)+D ˜u D ˜u(t)D ˜u(t,s))f(u ^(t)) D ˜u(t,s)

for every s>t. Using the Lipschitz condition on f we find

v ^(s)v ^(t) D ˜u(t,s)f(u ^(t)+D ˜u D ˜u(t)D ˜u(t,s))f(u ^(t)) D ˜u(t,s)Ku ^(s)u ^(t) D ˜u(t,s)D ˜u D ˜u(t).

By (7.5), the function sD ˜u(t,s) is continuous and converges to 0 as st. Therefore Remark 7.1 and the previous inequality imply

D ˜v D ˜u(t)=lim h0 + f(u ^(t)+hD ˜u D ˜u(t))f(u ^(t)) hD ˜u-a.e.inR.

By (4.4), u ^(x)=u ˜(x) for every xRS u ; moreover, applying the same argument to the functions u ' (t)=u(t), v ' (t)=f(u ' (t))=v(t), we get

D ˜v D ˜u(t)=lim h0 f(u ˜(t)+hD ˜u D ˜u(t))f(u ˜(t)) hD ˜u-a.e.inR

and our statement is proved.

7.2. Step 2 Let us consider now the general case n>1. Let νR n be such that ν=1, and let π ν ={yR n :y,ν=0}. In the following, we shall identify R n with π ν ×R, and we shall denote by y the variable ranging in π ν and by t the variable ranging in R. By the just proven one-dimensional result, and by Theorem 3.8, we get

lim h0 f(u ˜(y+tν)+hD ˜u y D ˜u y (t))f(u ˜(y+tν)) h=D ˜v y D ˜u y (t)D ˜u y -a.e.inR

for H n1 -almost every yπ ν . We claim that

(7.10)D ˜u,ν D ˜u,ν(y+tν)=D ˜u y D ˜u y (t)D ˜u y -a.e.inR

for H n1 -almost every yπ ν . In fact, by (3) and (3.20) we get

π ν D ˜u y D ˜u y ·D ˜u y dH n1 (y)= π ν D ˜u y dH n1 (y)=D ˜u,ν=D ˜u,ν D ˜u,ν·D ˜u,ν= π ν D ˜u,ν D ˜u,ν(y+·ν)·D ˜u y dH n1 (y)

and (6.2) follows from (3.14). By the same argument it is possible to prove that

(7.11)D ˜v,ν D ˜u,ν(y+tν)=D ˜v y D ˜u y (t)D ˜u y -a.e.inR

for H n1 -almost every yπ ν . By (6.2) and (6.4) we get

lim h0 f(u ˜(y+tν)+hD ˜u,ν D ˜u,ν(y+tν))f(u ˜(y+tν)) h=D ˜v,ν D ˜u,ν(y+tν)

for H n1 -almost every yπ ν , and using again (3.15), (3.16) we get

lim h0 f(u ˜(x)+hD ˜u,ν D ˜u,ν(x))f(u ˜(x)) h=D ˜v,ν D ˜u,ν(x)

D ˜u,ν-a.e. in R n .

Since the function D ˜u,ν/D ˜u is strictly positive D ˜u,ν-almost everywhere, we obtain also

lim h0 f(u ˜(x)+hD ˜u,ν D ˜u(x)D ˜u,ν D ˜u,ν(x))f(u ˜(x)) h=D ˜u,ν D ˜u(x)D ˜v,ν D ˜u,ν(x)

D ˜u,ν-almost everywhere in R n .

Finally, since

(7)D ˜u,ν D ˜uD ˜u,ν D ˜u,ν=D ˜u,ν D ˜u=D ˜u D ˜u,νD ˜u-a.e.inR n D ˜u,ν D ˜uD ˜v,ν D ˜u,ν=D ˜v,ν D ˜u=D ˜v D ˜u,νD ˜u-a.e.inR n

and since both sides of (7.9) are zero D ˜u-almost everywhere on D ˜u,ν-negligible sets, we conclude that

lim h0 fu ˜(x)+hD ˜u D ˜u(x),νf(u ˜(x)) h=D ˜v D ˜u(x),ν,

D ˜u-a.e. in R n . Since ν is arbitrary, by Remarks 7.2 and 7.3 the restriction of f to the affine space T x u is differentiable at u ˜(x) for D ˜u-almost every xR n and (7.2) holds.□

It follows from (3.14), (3.15), and (3.16) that

(7.13)D(t 1 ,,t n )= In (1) I1 I iI t i jI (D j +λ j t j )detA (λ) (I ¯|I ¯).

Let t i =x ^ i , i=1,,n. Lemma 1 leads to

(7.14)D(x ^ 1 ,,x ^ n )= in x ^ i In (1) I1 IperA (λ) (I|I)detA (λ) (I ¯|I ¯).

By (2.3), (3.14), and (7.14), we have the following result:

Theorem 7.15.

(7.16)H c =1 2n l=1 n l(1) l1 A l (λ) ,


(7.17)A l (λ) = I l n perA (λ) (I l |I l )detA (λ) (I ¯ l |I ¯ l ),I l =l.

It is worth noting that A l (λ) of (7.17) is similar to the coefficients b l of the characteristic polynomial of (3.11). It is well known in graph theory that the coefficients b l can be expressed as a sum over certain subgraphs. It is interesting to see whether A l , λ=0, structural properties of a graph.

We may call (7.16) a parametric representation of H c . In computation, the parameter λ i plays very important roles. The choice of the parameter usually depends on the properties of the given graph. For a complete graph K n , let λ i =1, i=1,,n. It follows from (7.17) that

(7.18)A l (1) =n!,ifl=10,otherwise.

By (7.16)

(7.19)H c =1 2(n1)!.

For a complete bipartite graph K n 1 n 2 , let λ i =0, i=1,,n. By (7.17),

(7.20)A l =n 1 !n 2 !δ n 1 n 2 ,ifl=20,otherwise.

Theorem  7.15 leads to

(7.21)H c =1 n 1 +n 2 n 1 !n 2 !δ n 1 n 2 .

Now, we consider an asymmetrical approach. Theorem 3.8 leads to

(7)detK(t=1,t 1 ,,t n ;l|l)= In{l} (1) I iI t i jI (D j +λ j t j )detA (λ) (I ¯{l}|I ¯{l}).

By (2.3) and (3) we have the following asymmetrical result:

Theorem 7.23.

(7.24)H c =1 2 In{l} (1) I perA (λ) (I|I)detA (λ) (I ¯{l}|I ¯{l})

which reduces to Goulden–Jackson´s formula when λ i =0,i=1,,n [10].

8. Various font features of the amsmath package

8.1. Bold versions of special symbols In the amsmath package \boldsymbol is used for getting individual bold math symbols and bold Greek letters—everything in math except for letters of the Latin alphabet, where you´d use \mathbf. For example,

25 A_\infty + \pi A_0 \sim
26 \mathbf{A}_{\boldsymbol{\infty}} \boldsymbol{+}
27 \boldsymbol{\pi} \mathbf{A}_{\boldsymbol{0}}

looks like this:

A +πA 0 A +πA 0

8.2. “Poor man´s bold” If a bold version of a particular symbol doesn´t exist in the available fonts, then \boldsymbol can´t be used to make that symbol bold. At the present time, this means that \boldsymbol can´t be used with symbols from the msam and msbm fonts, among others. In some cases, poor man´s bold (\pmb) can be used instead of \boldsymbol:

x y|y z
28 \[\frac{\partial x}{\partial y}
29 \pmb{\bigg\vert}
30 \frac{\partial y}{\partial z}\]

So-called “large operator” symbols such as ∑ and ∏ require an additional command, \mathop, to produce proper spacing and limits when \pmb is used. For further details see The TeXbook.

i<Biodd κ κF(r i ) i<Biodd κ κ(r i )
31 \[\sum_{\substack{i<B\\\text{$i$ odd}}}
32 \prod_\kappa \kappa F(r_i)\qquad
33 \mathop{\pmb{\sum}}_{\substack{i<B\\\text{$i$ odd}}}
34 \mathop{\pmb{\prod}}_\kappa \kappa(r_i)
35 \]

9. Compound symbols and other features

9.1. Multiple integral signs \iint, \iiint, and \iiiint give multiple integral signs with the spacing between them nicely adjusted, in both text and display style. \idotsint gives two integral signs with dots between them.

(9.1) A f(x,y)dxdy A f(x,y,z)dxdydz A f(w,x,y,z)dwdxdydz... A f(x 1 ,,x k )

9.3. Over and under arrows Some extra over and under arrow operations are provided in the amsmath package. (Basic LaTeX provides \overrightarrow and \overleftarrow).

(9.3)ψ δ (t)E t h =ψ δ (t)E t h ψ δ (t)E t h =ψ δ (t)E t h ψ δ (t)E t h =ψ δ (t)E t h
36 \begin{align*}
37 \overrightarrow{\psi_\delta(t) E_t h}&
38 =\underrightarrow{\psi_\delta(t) E_t h}\\
39 \overleftarrow{\psi_\delta(t) E_t h}&
40 =\underleftarrow{\psi_\delta(t) E_t h}\\
41 \overleftrightarrow{\psi_\delta(t) E_t h}&
42 =\underleftrightarrow{\psi_\delta(t) E_t h}
43 \end{align*}

These all scale properly in subscript sizes:

AB axdx
44 \[\int_{\overrightarrow{AB}} ax\,dx\]

9.5. Dots Normally you need only type \dots for ellipsis dots in a math formula. The main exception is when the dots fall at the end of the formula; then you need to specify one of \dotsc (series dots, after a comma), \dotsb (binary dots, for binary relations or operators), \dotsm (multiplication dots), or \dotsi (dots after an integral). For example, the input

45 Then we have the series $A_1,A_2,\dotsc$,
46 the regional sum $A_1+A_2+\dotsb$,
47 the orthogonal product $A_1A_2\dotsm$,
48 and the infinite integral
49 \[\int_{A_1}\int_{A_2}\dotsi\].


Then we have the series A 1 ,A 2 ,, the regional sum A 1 +A 2 +, the orthogonal product A 1 A 2 , and the infinite integral

A 1 A 2

9.6. Accents in math Double accents:

H ^ ^C ˇ ˇT ˜ ˜A ´ ´G ` `D ˙ ˙D ¨ ¨B ˘ ˘B ¯ ¯V
50 \[\Hat{\Hat{H}}\quad\Check{\Check{C}}\quad
51 \Tilde{\Tilde{T}}\quad\Acute{\Acute{A}}\quad
52 \Grave{\Grave{G}}\quad\Dot{\Dot{D}}\quad
53 \Ddot{\Ddot{D}}\quad\Breve{\Breve{B}}\quad
54 \Bar{\Bar{B}}\quad\Vec{\Vec{V}}\]

This double accent operation is complicated and tends to slow down the processing of a LaTeX file.

9.7. Dot accents \dddot and \ddddot are available to produce triple and quadruple dot accents in addition to the \dot and \ddot accents already available in LaTeX:

55 \[\dddot{Q}\qquad\ddddot{R}\]

9.8. Roots In the amsmath package \leftroot and \uproot allow you to adjust the position of the root index of a radical:

56 \sqrt[\leftroot{-2}\uproot{2}\beta]{k}

gives good positioning of the β:

k β

9.9. Boxed formulas The command \boxed puts a box around its argument, like \fbox except that the contents are in math mode:

57 \boxed{W_t-F\subseteq V(P_i)\subseteq W_t}
W t FV(P i )W t .

9.10. Extensible arrows \xleftarrow and \xrightarrow produce arrows that extend automatically to accommodate unusually wide subscripts or superscripts. The text of the subscript or superscript are given as an optional resp. mandatory argument: Example:

0 ζ αF×[n1] 0 α(b)E 0 b
58 \[0 \xleftarrow[\zeta]{\alpha} F\times\triangle[n-1]
59   \xrightarrow{\partial_0\alpha(b)} E^{\partial_0b}\]

9.11. \overset, \underset, and \sideset Examples:

X *X *X b a
60 \[\overset{*}{X}\qquad\underset{*}{X}\qquad
61 \overset{a}{\underset{b}{X}}\]

The command \sideset is for a rather special purpose: putting symbols at the subscript and superscript corners of a large operator symbol such as ∑ or ∏, without affecting the placement of limits. Examples:

**** k ' 0im E i βx
62 \[\sideset{_*^*}{_*^*}\prod_k\qquad
63 \sideset{}{'}\sum_{0\le i\le m} E_i\beta x
64 \]

9.12. The \text command The main use of the command \text is for words or phrases in a display:

y=y ' ifandonlyify k ' =δ k y τ(k)
65 \[\mathbf{y}=\mathbf{y}'\quad\text{if and only if}\quad
66 y'_k=\delta_k y_{\tau(k)}\]

9.13. Operator names The more common math functions such as log, sin, and lim have predefined control sequences: \log, \sin, \lim. The amsmath package provides \DeclareMathOperator and \DeclareMathOperator* for producing new function names that will have the same typographical treatment. Examples:

f =ess sup xR n f(x)
67 \[\norm{f}_\infty=
68 \esssup_{x\in R^n}\abs{f(x)}\]
meas 1 {uR + 1 :f * (u)>α}=meas n {xR n :f(x)α}α>0.
69 \[\meas_1\{u\in R_+^1\colon f^*(u)>\alpha\}
70 =\meas_n\{x\in R^n\colon \abs{f(x)}\geq\alpha\}
71 \quad \forall\alpha>0.\]

\esssup and \meas would be defined in the document preamble as

72 \DeclareMathOperator*{\esssup}{ess\,sup}
73 \DeclareMathOperator{\meas}{meas}

The following special operator names are predefined in the amsmath package: \varlimsup, \varliminf, \varinjlim, and \varprojlim. Here´s what they look like in use:

(9.13)lim ¯ n Q(u n ,u n u # )0lim ̲ n a n+1 /a n =0lim (m i λ ·) * 0lim pS(A) A p 0
74 \begin{align}
75 &\varlimsup_{n\rightarrow\infty}
76        \mathcal{Q}(u_n,u_n-u^{\#})\le0\\
77 &\varliminf_{n\rightarrow\infty}
78   \left\lvert a_{n+1}\right\rvert/\left\lvert a_n\right\rvert=0\\
79 &\varinjlim (m_i^\lambda\cdot)^*\le0\\
80 &\varprojlim_{p\in S(A)}A_p\le0
81 \end{align}

9.15. \mod and its relatives The commands \mod and \pod are variants of \pmod preferred by some authors; \mod omits the parentheses, whereas \pod omits the `mod´ and retains the parentheses. Examples:

(9.15)xy+1(modm 2 )xy+1modm 2 xy+1(m 2 )
82 \begin{align}
83 x&\equiv y+1\pmod{m^2}\\
84 x&\equiv y+1\mod{m^2}\\
85 x&\equiv y+1\pod{m^2}
86 \end{align}

9.17. Fractions and related constructions The usual notation for binomials is similar to the fraction concept, so it has a similar command \binom with two arguments. Example:

(9.18) γΓ C I γ =2 k k 12 k1 +k 22 k2 ++(1) l k l2 kl ++(1) k =(21) k =1
87 \begin{equation}
88 \begin{split}
89 [\sum_{\gamma\in\Gamma_C} I_\gamma&
90 =2^k-\binom{k}{1}2^{k-1}+\binom{k}{2}2^{k-2}\\
91 &\quad+\dots+(-1)^l\binom{k}{l}2^{k-l}
92 +\dots+(-1)^k\\
93 &=(2-1)^k=1
94 \end{split}
95 \end{equation}

There are also abbreviations

96 \dfrac        \dbinom
97 \tfrac        \tbinom

for the commonly needed constructions

98 {\displaystyle\frac ... }   {\displaystyle\binom ... }
99 {\textstyle\frac ... }      {\textstyle\binom ... }

The generalized fraction command \genfrac provides full access to the six TeX fraction primitives:

(9.17)\over:n+1 2\overwithdelims:n+1 2\atop:n+1 2\atopwithdelims:n+1 2\above:n+1 2\abovewithdelims:n+1 2
100 \text{\cn{over}: }&\genfrac{}{}{}{}{n+1}{2}&
101 \text{\cn{overwithdelims}: }&
102   \genfrac{\langle}{\rangle}{}{}{n+1}{2}\\
103 \text{\cn{atop}: }&\genfrac{}{}{0pt}{}{n+1}{2}&
104 \text{\cn{atopwithdelims}: }&
105   \genfrac{(}{)}{0pt}{}{n+1}{2}\\
106 \text{\cn{above}: }&\genfrac{}{}{1pt}{}{n+1}{2}&
107 \text{\cn{abovewithdelims}: }&
108   \genfrac{[}{]}{1pt}{}{n+1}{2}

9.20. Continued fractions The continued fraction

(9.21)1 2+1 2+1 2+1 2+1 2+

can be obtained by typing

109 \cfrac{1}{\sqrt{2}+
110  \cfrac{1}{\sqrt{2}+
111   \cfrac{1}{\sqrt{2}+
112    \cfrac{1}{\sqrt{2}+
113     \cfrac{1}{\sqrt{2}+\dotsb
114 }}}}}

Left or right placement of any of the numerators is accomplished by using \cfrac[l] or \cfrac[r] instead of \cfrac.

9.22. Smash In amsmath there are optional arguments t and b for the plain TeX command \smash, because sometimes it is advantageous to be able to `smash´ only the top or only the bottom of something while retaining the natural depth or height. In the formula X j =(1/λ j )X j ' \smash[b] has been used to limit the size of the radical symbol.

115 $X_j=(1/\sqrt{\smash[b]{\lambda_j}})X_j'$

Without the use of \smash[b] the formula would have appeared thus: X j =(1/λ j )X j ' , with the radical extending to encompass the depth of the subscript j.

9.23. The `cases´ environment `Cases´ constructions like the following can be produced using the cases environment.

(9.24)P rj =0ifrjisodd,r!(1) (rj)/2 ifrjiseven.
116 \begin{equation} P_{r-j}=
117   \begin{cases}
118     0&  \text{if $r-j$ is odd},\\
119     r!\,(-1)^{(r-j)/2}&  \text{if $r-j$ is even}.
120   \end{cases}
121 \end{equation}

Notice the use of \text and the embedded math.

9.25. Matrix Here are samples of the matrix environments, \matrix, \pmatrix, \bmatrix, \Bmatrix, \vmatrix and \Vmatrix:

122 \begin{matrix}
123 \vartheta& \varrho\\\varphi& \varpi
124 \end{matrix}\quad
125 \begin{pmatrix}
126 \vartheta& \varrho\\\varphi& \varpi
127 \end{pmatrix}\quad
128 \begin{bmatrix}
129 \vartheta& \varrho\\\varphi& \varpi
130 \end{bmatrix}\quad
131 \begin{Bmatrix}
132 \vartheta& \varrho\\\varphi& \varpi
133 \end{Bmatrix}\quad
134 \begin{vmatrix}
135 \vartheta& \varrho\\\varphi& \varpi
136 \end{vmatrix}\quad
137 \begin{Vmatrix}
138 \vartheta& \varrho\\\varphi& \varpi
139 \end{Vmatrix}

To produce a small matrix suitable for use in text, use the smallmatrix environment.

140 \begin{math}
141   \bigl( \begin{smallmatrix}
142       a&b\\ c&d
143     \end{smallmatrix} \bigr)
144 \end{math}

To show the effect of the matrix on the surrounding lines of a paragraph, we put it here: abcd and follow it with enough text to ensure that there will be at least one full line below the matrix.

\hdotsfor{number} produces a row of dots in a matrix spanning the given number of columns:

W(Φ)=ϕ (ϕ 1 ,ε 1 )00ϕk n2 (ϕ 2 ,ε 1 )ϕ (ϕ 2 ,ε 2 )0ϕk n1 (ϕ n ,ε 1 )ϕk n2 (ϕ n ,ε 2 )ϕk nn1 (ϕ n ,ε n1 )ϕ (ϕ n ,ε n )
145 \[W(\Phi)= \begin{Vmatrix}
146 \dfrac\varphi{(\varphi_1,\varepsilon_1)}&0&\dots&0\\
147 \dfrac{\varphi k_{n2}}{(\varphi_2,\varepsilon_1)}&
148 \dfrac\varphi{(\varphi_2,\varepsilon_2)}&\dots&0\\
149 \hdotsfor{5}\\
150 \dfrac{\varphi k_{n1}}{(\varphi_n,\varepsilon_1)}&
151 \dfrac{\varphi k_{n2}}{(\varphi_n,\varepsilon_2)}&\dots&
152 \dfrac{\varphi k_{n\,n-1}}{(\varphi_n,\varepsilon_{n-1})}&
153 \dfrac{\varphi}{(\varphi_n,\varepsilon_n)}
154 \end{Vmatrix}\]

The spacing of the dots can be varied through use of a square-bracket option, for example, \hdotsfor[1.5]{3}. The number in square brackets will be used as a multiplier; the normal value is 1.

9.27. The \substack command The \substack command can be used to produce a multiline subscript or superscript: for example

155 \sum_{\substack{0\le i\le m\\ 0<j<n}} P(i,j)

produces a two-line subscript underneath the sum:

(9.28) 0im0<j<n P(i,j)

A slightly more generalized form is the subarray environment which allows you to specify that each line should be left-aligned instead of centered, as here:

(9.29) 0im0<j<n P(i,j)
156 \sum_{\begin{subarray}{l}
157         0\le i\le m\\ 0<j<n
158       \end{subarray}}
159  P(i,j)

9.30. Big-g-g delimiters Here are some big delimiters, first in \normalsize:

E y 0 t ε L x,y x (s) ϕ (x) d s
160 \[\biggl(\mathbf{E}_{y}
161   \int_0^{t_\varepsilon}L_{x,y^x(s)}\varphi(x)\,ds
162   \biggr)
163 \]

and now in \Large size:

E y 0 t ε L x,y x (s) ϕ (x) d s

164 {\Large
165 \[\biggl(\mathbf{E}_{y}
166   \int_0^{t_\varepsilon}L_{x,y^x(s)}\varphi(x)\,ds
167   \biggr)
168 \]}

1. appendix

1. Examples of multiple-line equation structures

Note: Starting on this page, vertical rules are added at the margins so that the positioning of various display elements with respect to the margins can be seen more clearly.

1.1. Split The split environment is not an independent environment but should be used inside something else such as equation or align.

If there is not enough room for it, the equation number for a split will be shifted to the previous line, when equation numbers are on the left; the number shifts down to the next line when numbers are on the right.

(1.2)f h,ε (x,y)=εE x,y 0 t ε L x,y ε (εu) ϕ(x)du=hL x,z ϕ(x)ρ x (dz)+h[1 t ε E y 0 t ε L x,y x (s) ϕ (x) d s t ε L x,z ϕ (x) ρ x (dz)=+h[+1 t ε E y 0 t ε L x,y x (s) ϕ (x) d s E x,y 0 t ε L x,y ε (εs) ϕ (x) d s]=hL ^ x ϕ(x)+hθ ε (x,y),

Some text after to test the below-display spacing.

169 \begin{equation}
170 \begin{split}
171 f_{h,\varepsilon}(x,y)
172 &=\varepsilon\mathbf{E}_{x,y}\int_0^{t_\varepsilon}
173 L_{x,y_\varepsilon(\varepsilon u)}\varphi(x)\,du\\
174 &= h\int L_{x,z}\varphi(x)\rho_x(dz)\\
175 &\quad+h\biggl[\frac{1}{t_\varepsilon}\biggl(\mathbf{E}_{y}
176   \int_0^{t_\varepsilon}L_{x,y^x(s)}\varphi(x)\,ds
177   -t_\varepsilon\int L_{x,z}\varphi(x)\rho_x(dz)\biggr)\\
178 &\phantom{{=}+h\biggl[}+\frac{1}{t_\varepsilon}
179   \biggl(\mathbf{E}_{y}\int_0^{t_\varepsilon}L_{x,y^x(s)}
180     \varphi(x)\,ds -\mathbf{E}_{x,y}\int_0^{t_\varepsilon}
181    L_{x,y_\varepsilon(\varepsilon s)}
182    \varphi(x)\,ds\biggr)\biggr]\\
183 &=h\wh{L}_x\varphi(x)+h\theta_\varepsilon(x,y),
184 \end{split}
185 \end{equation}

Unnumbered version:

f h,ε (x,y)=εE x,y 0 t ε L x,y ε (εu) ϕ(x)du=hL x,z ϕ(x)ρ x (dz)+h[1 t ε E y 0 t ε L x,y x (s) ϕ (x) d s t ε L x,z ϕ (x) ρ x (dz)=+h[+1 t ε E y 0 t ε L x,y x (s) ϕ (x) d s E x,y 0 t ε L x,y ε (εs) ϕ (x) d s]=hL ^ x ϕ(x)+hθ ε (x,y),

Some text after to test the below-display spacing.

186 \begin{equation*}
187 \begin{split}
188 f_{h,\varepsilon}(x,y)
189 &=\varepsilon\mathbf{E}_{x,y}\int_0^{t_\varepsilon}
190 L_{x,y_\varepsilon(\varepsilon u)}\varphi(x)\,du\\
191 &= h\int L_{x,z}\varphi(x)\rho_x(dz)\\
192 &\quad+h\biggl[\frac{1}{t_\varepsilon}\biggl(\mathbf{E}_{y}
193   \int_0^{t_\varepsilon}L_{x,y^x(s)}\varphi(x)\,ds
194   -t_\varepsilon\int L_{x,z}\varphi(x)\rho_x(dz)\biggr)\\
195 &\phantom{{=}+h\biggl[}+\frac{1}{t_\varepsilon}
196   \biggl(\mathbf{E}_{y}\int_0^{t_\varepsilon}L_{x,y^x(s)}
197     \varphi(x)\,ds -\mathbf{E}_{x,y}\int_0^{t_\varepsilon}
198    L_{x,y_\varepsilon(\varepsilon s)}
199    \varphi(x)\,ds\biggr)\biggr]\\
200 &=h\wh{L}_x\varphi(x)+h\theta_\varepsilon(x,y),
201 \end{split}
202 \end{equation*}

If the option centertags is included in the options list of the amsmath package, the equation numbers for split environments will be centered vertically on the height of the split:

(1.3)I 2 = 0 T ψ(t)u(a,t) γ(t) a dθ k(θ,t) a θ c(ξ)u t (ξ,t)dξdtC 6 f Ω S ˜ a, 1,0 W 2 (Ω,Γ l )u W 2 A ˜ (Ω;Γ r ,T).

Some text after to test the below-display spacing.

Use of split within align:

(1.1)I 1 = Ω gRudΩC 3 Ω a x g(ξ,t)dξ 2 dΩ 1/2 × Ω u x 2 +1 k a x cu t dξ 2 cΩ 1/2 C 4 fS ˜ a, 1,0 W 2 (Ω,Γ l )u W 2 A ˜ (Ω;Γ r ,T).I 2 = 0 T ψ(t)u(a,t) γ(t) a dθ k(θ,t) a θ c(ξ)u t (ξ,t)dξdtC 6 f Ω S ˜ a, 1,0 W 2 (Ω,Γ l )u W 2 A ˜ (Ω;Γ r ,T).

Some text after to test the below-display spacing.

203 \begin{align}
204 \begin{split}\abs{I_1}
205   &=\left\lvert \int_\Omega gRu\,d\Omega\right\rvert\\
206 &\le C_3\left[\int_\Omega\left(\int_{a}^x
207   g(\xi,t)\,d\xi\right)^2d\Omega\right]^{1/2}\\
208 &\quad\times \left[\int_\Omega\left\{u^2_x+\frac{1}{k}
209   \left(\int_{a}^x cu_t\,d\xi\right)^2\right\}
210   c\Omega\right]^{1/2}\\
211 &\le C_4\left\lvert \left\lvert f\left\lvert \wt{S}^{-1,0}_{a,-}
212   W_2(\Omega,\Gamma_l)\right\rvert\right\rvert
213   \left\lvert \abs{u}\overset{\circ}\to W_2^{\wt{A}}
214   (\Omega;\Gamma_r,T)\right\rvert\right\rvert.
215 \end{split}\label{eq:A}\\
216 \begin{split}\abs{I_2}&=\left\lvert \int_{0}^T \psi(t)\left\{u(a,t)
217   -\int_{\gamma(t)}^a\frac{d\theta}{k(\theta,t)}
218   \int_{a}^\theta c(\xi)u_t(\xi,t)\,d\xi\right\}dt\right\rvert\\
219 &\le C_6\left\lvert \left\lvert f\int_\Omega
220   \left\lvert \wt{S}^{-1,0}_{a,-}
221   W_2(\Omega,\Gamma_l)\right\rvert\right\rvert
222   \left\lvert \abs{u}\overset{\circ}\to W_2^{\wt{A}}
223   (\Omega;\Gamma_r,T)\right\rvert\right\rvert.
224 \end{split}
225 \end{align}

Unnumbered align, with a number on the second split:

(1.1)I 1 = Ω gRudΩC 3 Ω a x g(ξ,t)dξ 2 dΩ 1/2 =× Ω u x 2 +1 k a x cu t dξ 2 cΩ 1/2 C 4 fS ˜ a, 1,0 W 2 (Ω,Γ l )u W 2 A ˜ (Ω;Γ r ,T).I 2 = 0 T ψ(t)u(a,t) γ(t) a dθ k(θ,t) a θ c(ξ)u t (ξ,t)dξdtC 6 f Ω S ˜ a, 1,0 W 2 (Ω,Γ l )u W 2 A ˜ (Ω;Γ r ,T).

Some text after to test the below-display spacing.

226 \begin{align*}
227 \begin{split}\abs{I_1}&=\left\lvert \int_\Omega gRu\,d\Omega\right\rvert\\
228   &\le C_3\left[\int_\Omega\left(\int_{a}^x
229   g(\xi,t)\,d\xi\right)^2d\Omega\right]^{1/2}\\
230 &\phantom{=}\times \left[\int_\Omega\left\{u^2_x+\frac{1}{k}
231   \left(\int_{a}^x cu_t\,d\xi\right)^2\right\}
232   c\Omega\right]^{1/2}\\
233 &\le C_4\left\lvert \left\lvert f\left\lvert \wt{S}^{-1,0}_{a,-}
234   W_2(\Omega,\Gamma_l)\right\rvert\right\rvert
235   \left\lvert \abs{u}\overset{\circ}\to W_2^{\wt{A}}
236   (\Omega;\Gamma_r,T)\right\rvert\right\rvert.
237 \end{split}\\
238 \begin{split}\abs{I_2}&=\left\lvert \int_{0}^T \psi(t)\left\{u(a,t)
239   -\int_{\gamma(t)}^a\frac{d\theta}{k(\theta,t)}
240   \int_{a}^\theta c(\xi)u_t(\xi,t)\,d\xi\right\}dt\right\rvert\\
241 &\le C_6\left\lvert \left\lvert f\int_\Omega
242   \left\lvert \wt{S}^{-1,0}_{a,-}
243   W_2(\Omega,\Gamma_l)\right\rvert\right\rvert
244   \left\lvert \abs{u}\overset{\circ}\to W_2^{\wt{A}}
245   (\Omega;\Gamma_r,T)\right\rvert\right\rvert.
246 \end{split}\tag{\theequation$'$}
247 \end{align*}

1.6. Multline Numbered version:

(1.6) a b a b [f(x) 2 g(y) 2 +f(y) 2 g(x) 2 ] 2 f (x) g (x) f (y) g (y) d xdy= a b g (y) 2 a b f 2 + f (y) 2 a b g 2 2 f (y) g (y) a b f gdy

To test the use of \label and \ref, we refer to the number of this equation here: (1.6).

248 \begin{multline}\label{eq:E}
249 \int_a^b\biggl\{\int_a^b[f(x)^2g(y)^2+f(y)^2g(x)^2]
250  -2f(x)g(x)f(y)g(y)\,dx\biggr\}\,dy \\
251  =\int_a^b\biggl\{g(y)^2\int_a^bf^2+f(y)^2
252   \int_a^b g^2-2f(y)g(y)\int_a^b fg\biggr\}\,dy
253 \end{multline}

Unnumbered version:

a b a b [f(x) 2 g(y) 2 +f(y) 2 g(x) 2 ] 2 f (x) g (x) f (y) g (y) d xdy= a b g (y) 2 a b f 2 + f (y) 2 a b g 2 2 f (y) g (y) a b f gdy

Some text after to test the below-display spacing.

254 \begin{multline*}
255 \int_a^b\biggl\{\int_a^b[f(x)^2g(y)^2+f(y)^2g(x)^2]
256  -2f(x)g(x)f(y)g(y)\,dx\biggr\}\,dy \\
257  =\int_a^b\biggl\{g(y)^2\int_a^bf^2+f(y)^2
258   \int_a^b g^2-2f(y)g(y)\int_a^b fg\biggr\}\,dy
259 \end{multline*}

1.8. Gather Numbered version with \notag on the second line:

(1.8)D(a,r){zC:za<r},seg(a,r){zC:z=a,za<r},c(e,θ,r){(x,y)C:xe<ytanθ,0<y<r},C(E,θ,r) eE c(e,θ,r).
260 \begin{gather}
261 D(a,r)\equiv\{z\in\mathbf{C}\colon \abs{z-a}<r\},\\
262 \seg(a,r)\equiv\{z\in\mathbf{C}\colon
263 \Im z= \Im a,\ \abs{z-a}<r\},\notag\\
264 c(e,\theta,r)\equiv\{(x,y)\in\mathbf{C}
265 \colon \abs{x-e}<y\tan\theta,\ 0<y<r\},\\
266 C(E,\theta,r)\equiv\bigcup_{e\in E}c(e,\theta,r).
267 \end{gather}

Unnumbered version.

D(a,r){zC:za<r},seg(a,r){zC:z=a,za<r},c(e,θ,r){(x,y)C:xe<ytanθ,0<y<r},C(E,θ,r) eE c(e,θ,r).

Some text after to test the below-display spacing.

268 \begin{gather*}
269 D(a,r)\equiv\{z\in\mathbf{C}\colon \abs{z-a}<r\},\\
270 \seg (a,r)\equiv\{z\in\mathbf{C}\colon
271 \Im z= \Im a,\ \abs{z-a}<r\},\\
272 c(e,\theta,r)\equiv\{(x,y)\in\mathbf{C}
273  \colon \abs{x-e}<y\tan\theta,\ 0<y<r\},\\
274 C(E,\theta,r)\equiv\bigcup_{e\in E}c(e,\theta,r).
275 \end{gather*}

1.9. Align Numbered version:

(1.9)γ x (t)=(costu+sintx,v),γ y (t)=(u,costv+sinty),γ z (t)=costu+α βsintv,β αsintu+costv.

Some text after to test the below-display spacing.

276 \begin{align}
277 \gamma_x(t)&=(\cos tu+\sin tx,v),\\
278 \gamma_y(t)&=(u,\cos tv+\sin ty),\\
279 \gamma_z(t)&=\left(\cos tu+\frac\alpha\beta\sin tv,
280   -\frac\beta\alpha\sin tu+\cos tv\right).
281 \end{align}

Unnumbered version:

(1.9)γ x (t)=(costu+sintx,v),γ y (t)=(u,costv+sinty),γ z (t)=costu+α βsintv,β αsintu+costv.

Some text after to test the below-display spacing.

282 \begin{align*}
283 \gamma_x(t)&=(\cos tu+\sin tx,v),\\
284 \gamma_y(t)&=(u,\cos tv+\sin ty),\\
285 \gamma_z(t)&=\left(\cos tu+\frac\alpha\beta\sin tv,
286   -\frac\beta\alpha\sin tu+\cos tv\right).
287 \end{align*}

A variation:

(1.9)x=yby()x ' =y ' by()x+x ' =y+y ' byAxiom1.

Some text after to test the below-display spacing.

288 \begin{align}
289 x& =y && \text {by (\ref{eq:C})}\\
290 x'& = y' && \text {by (\ref{eq:D})}\\
291 x+x' & = y+y' && \text {by Axiom 1.}
292 \end{align}

1.13. Align and split within gather When using the align environment within the gather environment, one or the other, or both, should be unnumbered (using the * form); numbering both the outer and inner environment would cause a conflict.

Automatically numbered gather with split and align*:

(1.13)ϕ(x,z)=zγ 10 xγ mn x m z n =zMr 1 xMr (m+n) x m z n ζ 0 =(ξ 0 ) 2 ,ζ 1 =ξ 0 ξ 1 ,ζ 2 =(ξ 1 ) 2 ,

Here the split environment gets a number from the outer gather environment; numbers for individual lines of the align* are suppressed because of the star.

293 \begin{gather}
294 \begin{split} \varphi(x,z)
295 &=z-\gamma_{10}x-\gamma_{mn}x^mz^n\\
296 &=z-Mr^{-1}x-Mr^{-(m+n)}x^mz^n
297 \end{split}\\[6pt]
298 \begin{align*}
299 \zeta^0 &=(\xi^0)^2,\\
300 \zeta^1 &=\xi^0\xi^1,\\
301 \zeta^2 &=(\xi^1)^2,
302 \end{align*}
303 \end{gather}

The *-ed form of gather with the non-*-ed form of align.

ϕ(x,z)=zγ 10 xγ mn x m z n =zMr 1 xMr (m+n) x m z n ζ 0 =(ξ 0 ) 2 ,ζ 1 =ξ 0 ξ 1 ,ζ 2 =(ξ 1 ) 2 ,

Some text after to test the below-display spacing.

304 \begin{gather*}
305 \begin{split} \varphi(x,z)
306 &=z-\gamma_{10}x-\gamma_{mn}x^mz^n\\
307 &=z-Mr^{-1}x-Mr^{-(m+n)}x^mz^n
308 \end{split}\\[6pt]
309 \begin{align} \zeta^0&=(\xi^0)^2,\\
310 \zeta^1 &=\xi^0\xi^1,\\
311 \zeta^2 &=(\xi^1)^2,
312 \end{align}
313 \end{gather*}

1.15. Alignat Numbered version:

(1.15)V i =v i q i v j ,X i =x i q i x j ,U i =u i ,forij;V j =v j ,X j =x j ,U j u j + ij q i u i .

Some text after to test the below-display spacing.

314 \begin{alignat}{3}
315 V_i & =v_i - q_i v_j, & \qquad X_i & = x_i - q_i x_j,
316  & \qquad U_i & = u_i,
317  \qquad \text{for $i\ne j$;}\label{eq:B}\\
318 V_j & = v_j, & \qquad X_j & = x_j,
319   & \qquad U_j & u_j + \sum_{i\ne j} q_i u_i.
320 \end{alignat}

Unnumbered version:

(1.15)V i =v i q i v j ,X i =x i q i x j ,U i =u i ,forij;V j =v j ,X j =x j ,U j u j + ij q i u i .

Some text after to test the below-display spacing.

321 \begin{alignat*}3
322 V_i & =v_i - q_i v_j, & \qquad X_i & = x_i - q_i x_j,
323  & \qquad U_i & = u_i,
324  \qquad \text{for $i\ne j$;} \\
325 V_j & = v_j, & \qquad X_j & = x_j,
326   & \qquad U_j & u_j + \sum_{i\ne j} q_i u_i.
327 \end{alignat*}

The most common use for alignat is for things like

(1.15)x=yby()x ' =y ' by()x+x ' =y+y ' byAxiom1.

Some text after to test the below-display spacing.

328 \begin{alignat}{2}
329 x& =y && \qquad \text {by (\ref{eq:A})}\label{eq:C}\\
330 x'& = y' && \qquad \text {by (\ref{eq:B})}\label{eq:D}\\
331 x+x' & = y+y' && \qquad \text {by Axiom 1.}
332 \end{alignat}

1.19. Additions for Tralics Added a \nocite{fre:cichon}

df dx df dx df dx d 2 f dx 2 d 2 f dx 2 d 2 f dx 2 df dx x 0 df dx x 0 df dx x 0 d 2 g dy 2 0 d 2 g dy 2 0 d 2 g dy 2 0

f x= 2 f x 2 =p V T = 3 f xy 2 = 2 f x 2 = 5 f x 2 y 3 z

f x= 2 f x 2 =p V T = 3 f xy 2 = 2 f x 2 = 5 f x 2 y 3 z


[1] W. Diffie, E. Hellman. New directions in cryptography. in « IEEE Transactions on Information Theory », number 5, volume 22, 1976, pages 644–654.

[2] D. H. Fremlin. Cichon´s diagram. 1983/1984, presented at the Séminaire Initiation à l´Analyse, G. Choquet, M. Rogalski, J. Saint Raymond, at the Université Pierre et Marie Curie, Paris, 23e année.

[3] I. P. Goulden, D. M. Jackson. The enumeration of directed closed Euler trails and directed Hamiltonian circuits by Langrangian methods. in « European J. Combin », volume 2, 1981, pages 131-212.

[4] F. Harary, E. M. Palmer. Graphical enumeration. Academic Press, 1973.

[5] R. Impagliazzo, L. Levin, M. Luby. Pseudo-random generation from one-way functions. in « Proc. 21st STOC », ACM, pages 12–24, New York, 1989.

[6] M. Kojima, S. Mizuno, A. Yoshise. A new continuation method for complementarity problems with uniform p-functions. Tech. Report, number B-194, Tokyo Inst. of Technology, Tokyo, 1987, Dept. of Information Sciences.

[7] M. Kojima, S. Mizuno, A. Yoshise. A polynomial-time algorithm for a class of linear complementarity problems. Tech. Report, number B-193, Tokyo Inst. of Technology, Tokyo, 1987, Dept. of Information Sciences.

[8] C. J. Liu, Yutze Chow. On operator and formal sum methods for graph enumeration problems. in « SIAM J. Algorithms Discrete Methods », volume 5, 1984, pages 384–438.

[9] R. D. Monteiro, I. Adler. Interior path following primal-dual algorithms, part II: Quadratic programming. Working paper, Dept. of Industrial Engineering and Operations Research, August, 1987.

[10] M. Marcus, H. Minc. A survey of matrix theory and matrix inequalities. in « Complementary Series in Math. », volume 14, 1964, pages 21-48.

[11] S. Mizuno, A. Yoshise, T. Kikuchi. Practical polynomial time algorithms for linear complementarity problems. Technical report, number Tech. Report 13, Tokyo Inst. of Technology, Dept. of Industrial Engineering and Management, Tokyo, April, 1988.

[12] E. M. Stein. Singular integrals and differentiability properties of functions. Princeton Univ. Press, Princeton, N.J., 1970.

[13] Y. Ye. Interior algorithms for linear, quadratic and linearly constrained convex programming. Ph. D. Thesis, Stanford Univ., Palo Alto, Calif, July, 1987, Dept. of Engineering–Economic Systems, unpublished.


Note 1. A multiset allows multiplicity of elements. Hence, {0,01,01} is prefix free as a set, but not as a multiset.