How to Typeset Equations in LATEX

Stefan M. Moser

21 June 2016
Version 4.5

Contents
1 Introduction 2

2 Single Equations: equation 3

3 Single Equations that are Too Long: multline 3

3.1 Case 1: The expression is not an equation . . . . . . . . . . . . . . . . . 5
3.2 Case 2: Additional comment . . . . . . . . . . . . . . . . . . . . . . . . 5
3.3 Case 3: LHS too long RHS too short . . . . . . . . . . . . . . . . . . 6
3.4 Case 4: A term on the RHS should not be split . . . . . . . . . . . . . . 6

4 Multiple Equations: IEEEeqnarray 7

4.1 Problems with traditional commands . . . . . . . . . . . . . . . . . . . . 7
4.2 Solution: basic usage of IEEEeqnarray . . . . . . . . . . . . . . . . . . . 9
4.3 A remark about consistency . . . . . . . . . . . . . . . . . . . . . . . . . 9
4.4 Using IEEEeqnarray for all situations . . . . . . . . . . . . . . . . . . . 11

5 More Details about IEEEeqnarray 11

5.1 Shift to the left: IEEEeqnarraynumspace . . . . . . . . . . . . . . . . . . 12
5.2 First line too long: IEEEeqnarraymulticol . . . . . . . . . . . . . . . . . 12
5.3 Line-break: unary versus binary operators . . . . . . . . . . . . . . . . . 13
5.4 Equation-numbering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
5.5 Page-breaks inside of IEEEeqnarray . . . . . . . . . . . . . . . . . . . . 17

6 Advanced Typesetting 18
6.1 IEEEeqnarraybox: general tables and arrays . . . . . . . . . . . . . . . . 18
6.2 Case distinctions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
6.3 Grouping numbered equations with a bracket . . . . . . . . . . . . . . . 22
6.4 Matrices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
6.5 Adapting the size of brackets . . . . . . . . . . . . . . . . . . . . . . . . 25
6.6 Framed equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
6.7 Fancy frames . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
6.8 Putting the QED correctly: proof . . . . . . . . . . . . . . . . . . . . . . 32
6.9 Putting the QED correctly: IEEEproof . . . . . . . . . . . . . . . . . . . 35
6.10 Double-column equations in a two-column layout . . . . . . . . . . . . . 37

1
How to Typeset Equations in LATEX 2

7 Emacs and IEEEeqnarray 39

8 Some Useful Definitions 40

9 Some Final Remarks and Acknowledgments 41

If you have only very limited time, simply read Section 4.2. If you have little more
time, read Sections 4 and 5.
This manual is written with the newest version of IEEEtran in mind.1

1 Introduction
LATEX is a very powerful tool for typesetting in general and for typesetting math in
particular. In spite of its power, however, there are still many ways of generating
better or less good results. This manual offers some tricks and hints that hopefully will
lead to the former. . .
Note that this manual does neither claim to provide the best nor the only solution.
Its aim is rather to give a couple of rules that can be followed easily and that will lead
to a good layout of all equations in a document. It is assumed that the reader has
already mastered the basics of LATEX.
The structure of this document is as follows. We introduce the most basic equation
in Section 2; Section 3 then explains some first possible reactions when an equation is
too long. The probably most important part is contained in Sections 4 and 5: there
we introduce the powerful IEEEeqnarray-environment that should be used in any case
instead of align or eqnarray.
In Section 6 some more advanced problems and possible solutions are discussed, and
Section 7 contains some hints and tricks about the editor Emacs. Finally, Section 8
makes some suggestions about special symbols that cannot be easily found in LATEX.
In the following any LATEX command will be set in typewriter font. RHS stands
for right-hand side, i.e., all terms on the right of the equality (or inequality) sign.
Similarly, LHS stands for left-hand side, i.e., all terms on the left of the equality sign.
To simplify our language, we will usually talk about equality. Obviously, the typesetting
does not change if an expression actually is an inequality.
This documents comes together with some additional files that might be helpful:

typeset_equations.tex: LATEX source file of this manual.

dot_emacs: commands to be included in the preference file of Emacs (.emacs)

(see Section 7).

IEEEtrantools.sty [2015/08/26 V1.5 by Michael Shell]: package needed for the

IEEEeqnarray-environment.

IEEEtran.cls [2015/08/26 V1.8b by Michael Shell]: LATEX document class pack-

age for papers in IEEE format.
1
IEEEtran.cls version 1.8b, and IEEEtrantools.sty version 1.5.

c Stefan M. Moser 21 June 2016, Version 4.5
How to Typeset Equations in LATEX 3

IEEEtran_HOWTO.pdf [2015/08]: official manual of the IEEEtran-class. The part

about IEEEeqnarray is found in Appendix F.
Note that IEEEtran.cls and IEEEtrantools.sty is provided automatically by any
up-to-date LATEX-distribution.

2 Single Equations: equation

The main strength of LATEX concerning typesetting of mathematics is based on the
package amsmath. Every current distribution of LATEX will come with this package
included, so you only need to make sure that the following line is included in the
header of your document:

\usepackage{amsmath}

Throughout this document it is assumed that amsmath is loaded.

Single equations should be exclusively typed using the equation-environment:

\begin{equation}
a = b + c a=b+c (1)
\end{equation}

In case one does not want to have an equation number, the *-version is used:

\begin{equation*}
a = b + c a=b+c
\end{equation*}

All other possibilities of typesetting simple equations have disadvantages:

The displaymath-environment offers no equation-numbering. To add or to re-
move a * in the equation-environment is much more flexible.
Commands like $$...$$, \[...\], etc., have the additional disadvantage that
the source code is extremely poorly readable. Moreover, $$...$$ is faulty: the
vertical spacing after the equation is too large in certain situations.
We summarize:

Unless we decide to rely exclusively on IEEEeqnarray (see the discussion in

Sections 4.3 and 4.4), we should only use equation (and no other environment)
to produce a single equation.

3 Single Equations that are Too Long: multline

If an equation is too long, we have to wrap it somehow. Unfortunately, wrapped equa-
tions are usually less easy to read than not-wrapped ones. To improve the readability,
one should follow certain rules on how to do the wrapping:

c Stefan M. Moser 21 June 2016, Version 4.5
How to Typeset Equations in LATEX 4

1. In general one should always wrap an equation before an equality sign or

an operator.

2. A wrap before an equality sign is preferable to a wrap before any operator.

3. A wrap before a plus- or minus-operator is preferable to a wrap before a

multiplication-operator.

4. Any other type of wrap should be avoided if ever possible.

The easiest way to achieve such a wrapping is the use of the multline-environment:2

\begin{multline}
a + b + c + d + e + f
+ g + h + i
a+b+c+d+e+f +g+h+i
\\
= j + k + l + m + n = j + k + l + m + n (2)
\end{multline}

The difference to the equation-environment is that an arbitrary line-break (or also

multiple line-breaks) can be introduced. This is done by putting a \\ at those places
where the equation needs to be wrapped.
Similarly to equation* there also exists a multline*-version for preventing an
equation number.
However, in spite of its ease of use, often the IEEEeqnarray-environment (see Sec-
tion 4) will yield better results. Particularly, consider the following common situation:

\begin{equation}
a = b + c + d + e + f
+ g + h + i + j
+ k + l + m + n + o + p a = b+c+d+e+f +g+h+i+j+k+l+m+n+o+p
\label{eq:equation_too_long} (3)
\end{equation}

Here the RHS is too long to fit on one line. The multline-environment will now yield
the following:

\begin{multline}
a = b + c + d + e + f
+ g + h + i + j \\ a=b+c+d+e+f +g+h+i+j
+ k + l + m + n + o + p +k+l+m+n+o+p (4)
\end{multline}

This is of course much better than (3), but it has the disadvantage that the equality
sign loses its natural stronger importance over the plus operator in front of k. A better
solution is provided by the IEEEeqnarray-environment that will be discussed in detail
in Sections 4 and 5:
2
As a reminder: it is necessary to include the amsmath-package for this command to work!

c Stefan M. Moser 21 June 2016, Version 4.5
How to Typeset Equations in LATEX 5

\begin{IEEEeqnarray}{rCl}
a & = & b + c + d + e + f
+ g + h + i + j \nonumber\\
a=b+c+d+e+f +g+h+i+j
&& +\> k + l + m + n + o + p
\label{eq:dont_use_multline} +k+l+m+n+o+p (5)
\end{IEEEeqnarray}

In this case the second line is horizontally aligned to the first line: the + in front of k is
exactly below b, i.e., the RHS is clearly visible as contrast to the LHS of the equation.
Also note that multline wrongly forces a minimum spacing on the left of the first
line even if it has not enough space on the right, causing a noncentered equation. This
can even lead to the very ugly typesetting where the second line containing the RHS
of an equality is actually to the left of the first line containing the LHS:

\begin{multline}
a + b + c + d + e + f + g
+ h + i + j \\ a+b+c+d+e+f +g+h+i+j
= k + l + m + n + o + p = k+l+m+n+o+p+q+r+s+t+u
+ q + r + s + t + u (6)
\end{multline}

For this reason we give the following rule:

The multline-environment should exclusively be used in the four specific

situations described in Sections 3.13.4.

3.1 Case 1: The expression is not an equation

If the expression is not an equation, i.e., there is no equality sign, then there exists no
RHS or LHS and multline offers a nice solution:

\begin{multline}
a + b + c + d + e + f \\ a+b+c+d+e+f
+ g + h + i + j + k + l \\
+ m + n + o + p + q +g+h+i+j+k+l
\end{multline} +m+n+o+p+q (7)

3.2 Case 2: Additional comment

If there is an additional comment at the end of the equation that does not fit on the
same line, then this comment can be put onto the next line:

\begin{multline}
a + b + c + d
= e + f + g + h, \quad \\
a + b + c + d = e + f + g + h,
\text{for } 0 \le n
\le n_{\textnormal{max}} for 0 n nmax (8)
\end{multline}

c Stefan M. Moser 21 June 2016, Version 4.5
How to Typeset Equations in LATEX 6

3.3 Case 3: LHS too long RHS too short

If the LHS of a single equation is too long and the RHS is very short, then one cannot
break the equation in front of the equality sign as wished, but one is forced to do it
somewhere on the LHS. In this case one cannot nicely keep the natural separation of
LHS and RHS anyway and multline offers the best (of bad) solutions:

\begin{multline}
a + b + c + d + e + f
+ g \\+ h + i + j a+b+c+d+e+f +g
+ k + l = m +h+i+j+k+l =m (9)
\end{multline}

3.4 Case 4: A term on the RHS should not be split

The following is a special (and rather rare) case: the LHS would be short enough and/or
the RHS long enough in order to wrap the equation in a way as shown in (5), i.e., this
usually would call for the IEEEeqnarray-environment. However, a term on the RHS is
an entity that we rather would not split, but it is too long to fit:3

\begin{multline}
h^{-}(X|Y) \le \frac{n+1}{e}
- h(X|Y)
n+1
\\ h (X|Y ) h(X|Y )
+ \int p(y) \log \left( e
  !
E |X|2 Y = y
Z
\frac{\mathsf{E}\big[|X|^2
+ p(y) log dy (10)
\big| Y=y\big]}{n} n
\right) \dd y
\end{multline}

In this example the integral on the RHS is too long, but should not be split for read-
ability.
Note that even in this case it might be possible to find different, possibly better
solutions based on IEEEeqnarray-environment:

\begin{IEEEeqnarray}{rCl}
\IEEEeqnarraymulticol{3}{l}{
h^{-}(X|Y)
}\nonumber\\\quad h (X|Y )
& \le & \frac{n+1}{e} n+1
- h(X|Y) \nonumber\\ h(X|Y )
e
&& + \int p(y) \log \left(   !
E |X|2 Y = y
Z
\frac{\mathsf{E}\big[|X|^2 + p(y) log dy
\big| Y=y\big]}{n} n
\right) \dd y (11)
\nonumber\\*
\end{IEEEeqnarray}

3
For a definition of \dd, see Section 8.

c Stefan M. Moser 21 June 2016, Version 4.5
How to Typeset Equations in LATEX 7

4 Multiple Equations: IEEEeqnarray

In the most general situation, we have a sequence of several equalities that do not fit
onto one line. Here we need to work with horizontal alignment in order to keep the
array of equations in a nice and readable structure.
Before we offer our suggestions on how to do this, we start with a few bad examples
that show the biggest drawbacks of common solutions.

4.1 Problems with traditional commands

To group multiple equations, the align-environment4 could be used:

\begin{align}
a & = b + c \\
a=b+c (12)
& = d + e
\end{align} =d+e (13)

While this looks neat as long as every equation fits onto one line, this approach does
not work anymore once a single line is too long:

\begin{align}
a & = b + c \\
& = d + e + f + g + h + i a=b+c (14)
+ j + k + l \nonumber\\ =d+e+f +g+h+i+j+k+l
& + m + n + o \\ +m+n+o (15)
& = p + q + r + s
=p+q+r+s (16)
\end{align}

Here + m should be below d and not below the equality sign. Of course, one could
add some space by, e.g., \hspace{...}, but this will never yield a precise arrangement
(and is bad programming style!).
A better solution is offered by the eqnarray-environment:

\begin{eqnarray}
a & = & b + c \\
& = & d + e + f + g + h + i a = b+c (17)
+ j + k + l \nonumber\\ = d+e+f +g+h+i+j+k+l
&& +\> m + n + o \\ +m+n+o (18)
& = & p + q + r + s
= p+q+r+s (19)
\end{eqnarray}

The eqnarray-environment, however, has a few very severe disadvantages:

The spaces around the equality signs are too big. Particularly, they are not the
same as in the multline- and equation-environments:
4
The align-environment can also be used to group several blocks of equations beside each other.
However, also for this situation, we recommend to use the IEEEeqnarray-environment with an argument
like, e.g., {rCl+rCl}.

c Stefan M. Moser 21 June 2016, Version 4.5
How to Typeset Equations in LATEX 8

\begin{eqnarray}
a & = & a = a
\end{eqnarray} a = a=a (20)

The expression sometimes overlaps with the equation number even though there
would be enough room on the left:

\begin{eqnarray}
a & = & b + c
\\
& = & d + e + f + g + h^2 a = b+c (21)
+ i^2 + j = 2 2
d + e + f + g + h + i + j(22)
\label{eq:faultyeqnarray}
\end{eqnarray}

The eqnarray-environment offers a command \lefteqn{...} that can be used

when the LHS is too long:

\begin{eqnarray}
\lefteqn{a + b + c + d
+ e + f + g + h}\nonumber\\ a+b+c+d+e+f +g+h
& = & i + j + k + l + m
\\ = i+j+k+l+m (23)
& = & n + o + p + q + r + s = n+o+p+q+r+s (24)
\end{eqnarray}

Unfortunately, this command is faulty: if the RHS is too short, the array is not
properly centered:

\begin{eqnarray}
\lefteqn{a + b + c + d
+ e + f + g + h}
a+b+c+d+e+f +g+h
\nonumber\\
& = & i + j = i+j (25)
\end{eqnarray}

Moreover, it is very complicated to change the horizontal alignment of the equality

sign on the second line.

Thus:

NEVER ever use the eqnarray-environment!

To overcome these problems we recommend the IEEEeqnarray-environment.

c Stefan M. Moser 21 June 2016, Version 4.5
How to Typeset Equations in LATEX 9

4.2 Solution: basic usage of IEEEeqnarray

The IEEEeqnarray-environment is a very powerful command with many options. In
this manual we will only introduce some of the most important functionalities. For
more information we refer to the official manual.5 First of all, in order to be able to
use the IEEEeqnarray-environment, one needs to include the package6 IEEEtrantools.
Include the following line in the header of your document:

\usepackage{IEEEtrantools}

The strength of IEEEeqnarray is the possibility of specifying the number of columns

in the equation array. Usually, this specification will be {rCl}, i.e., three columns, the
first column right-justified, the middle one centered with a little more space around
it (therefore we specify capital C instead of lower-case c) and the third column left-
justified:

\begin{IEEEeqnarray}{rCl}
a & = & b + c
\\
& = & d + e + f + g + h a=b+c (26)
+ i + j + k \nonumber\\ =d+e+f +g+h+i+j+k
&& +\> l + m + n + o +l+m+n+o (27)
\\
=p+q+r+s (28)
& = & p + q + r + s
\end{IEEEeqnarray}

However, we can specify any number of needed columns. For example, {c} will give only
one column (which is centered) or {rCll} will add a fourth, left-justified column, e.g.,
for additional specifications. Moreover, beside l, c, r, L, C, R for math mode entries,
there also exists s, t, u for left, centered, and right text mode entries, respectively. We
can even add additional spacing by . and / and ? and " in increasing order.7 More
details about the usage of IEEEeqnarray will be given in Section 5.
Note that in contrast to eqnarray the spaces around the equality signs are correct.

4.3 A remark about consistency

There are three more issues that have not been mentioned so far, but that might cause
inconsistencies when all three environments, equation, multline, and IEEEeqnarray,
are used intermixedly:
multline allows for an equation starting on top of a page, while equation and
IEEEeqnarray try to put a line of text first, before the equation starts. Moreover,
the spacing before and after the environment is not exactly identical for equation,
multline, and IEEEeqnarray.
5
The official manual IEEEtran HOWTO.pdf is distributed together with this short introduction. The
part about IEEEeqnarray can be found in Appendix F.
6
This package is also distributed together with this manual, but it is already included in any up-
to-date LATEX distribution. Note that if a document uses the IEEEtran-class, then IEEEtrantools is
loaded automatically and must not be included separately.
7
For examples of spacing, we refer to Section 6.1. More spacing types can be found in the examples
given in Sections 5.3 and 6.8, and in the official manual.

c Stefan M. Moser 21 June 2016, Version 4.5
How to Typeset Equations in LATEX 10

equation uses an automatic mechanism to move the equation number onto the
next line if the expression is too long. While this is convenient, sometimes the
equation number is forced onto the next line, even if there was still enough space
available on the line:

\begin{equation}
a = \sum_{k=1}^n\sum_{\ell=1}^n n X
n
\sin \bigl(2\pi \, b_k \, X

a= sin 2 bk c` dk e` fk g` h
c_{\ell} \, d_k \, e_{\ell} \,
k=1 `=1
f_k \, g_{\ell} \, h \bigr) (29)
\end{equation}

With IEEEeqnarray the placement of the equation number is fully under our
control:

\begin{IEEEeqnarray}{c}
a = \sum_{k=1}^n\sum_{\ell=1}^n
\sin \bigl(2\pi \, b_k \,
c_{\ell} \, d_k \, e_{\ell} \, n X
X n


f_k \, g_{\ell} \, h \bigr) a= sin 2 bk c` dk e` fk g` h (30)
\IEEEeqnarraynumspace k=1 `=1
\label{eq:labelc1}
\end{IEEEeqnarray}

or

\begin{IEEEeqnarray}{c}
a = \sum_{k=1}^n\sum_{\ell=1}^n
\sin \bigl(2\pi \, b_k \, n X
n
c_{\ell} \, d_k \, e_{\ell} \,
X

a= sin 2 bk c` dk e` fk g` h
f_k \, g_{\ell} \, h \bigr) k=1 `=1
\nonumber\\* (31)
\label{eq:labelc2}
\end{IEEEeqnarray}

equation forces the equation number to appear in normal font, even if the equa-
tion is within an environment8 of different font:

\textbf{\textit{\color{red} This is our main result:

This is our main result:
\begin{equation} a=b+c (32)
a = b + c
\end{equation}}}

IEEEeqnarray respects the settings of the environment:

8
A typical example of such a situation is an equation inside of a theorem that is typeset in italic
font.

c Stefan M. Moser 21 June 2016, Version 4.5
How to Typeset Equations in LATEX 11

\textbf{\textit{\color{red} This is our main result:

This is our main result:
\begin{IEEEeqnarray}{c} a=b+c (33)
a = b + c
\end{IEEEeqnarray}}}

If this is undesired, one can change the behavior of IEEEeqnarray to behave9 like
equation:

\renewcommand{\theequationdis}{{\normalfont (\theequation)}}
\renewcommand{\theIEEEsubequationdis}{{\normalfont (\theIEEEsubequation)}}

\textbf{\textit{\color{red}
This is our main result: This is our main result:
\begin{IEEEeqnarray}{rCl}
a=b+c (34)
a & = & b + c \\
& = & d + e \IEEEyesnumber =d+e (35a)
\IEEEyessubnumber
\end{IEEEeqnarray}}}

4.4 Using IEEEeqnarray for all situations

As seen above, there might be reason to rely on IEEEeqnarray exclusively in all situ-
ations.
To replace an equation-environment we use IEEEeqnarray with only one column
{c}, see (30) and (31) above.
Emulating multline is slightly more complicated: we implement IEEEeqnarray
with only one column {l}, use \IEEEeqnarraymulticol10 after the line-break(s) to
adapt the column type of the new line, and manually add some shift:

\begin{IEEEeqnarray*}{l}
a + b + c + d + e + f
\\ \qquad
+\> g + h + i + j + k + l a+b+c+d+e+f
\qquad \\
\IEEEeqnarraymulticol{1}{r}{ +g+h+i+j+k+l
+\> m + n + o + p + q } +m+n+o+p+q (36)
\IEEEyesnumber
\end{IEEEeqnarray*}

5 More Details about IEEEeqnarray

In the following we will describe how we use IEEEeqnarray to solve the most common
situations.
9
For an explanation of the subnumbering, see Section 5.4.
10
For a more detailed explanation of this command, see Section 5.2.

c Stefan M. Moser 21 June 2016, Version 4.5
How to Typeset Equations in LATEX 12

5.1 Shift to the left: IEEEeqnarraynumspace

If a line overlaps with the equation number as in (22), the command

\IEEEeqnarraynumspace

can be used. It has to be added in the corresponding line and makes sure that the
whole equation array is shifted by the size of the equation numbers (the shift depends
on the size of the number!). Instead of

\begin{IEEEeqnarray}{rCl}
a & = & b + c
\\
& = & d + e + f + g + h a=b+c (37)
+ i + j + k + m = d + e + f + g + h + i + j + k + m(38)
\\ =l+n+o (39)
& = & l + n + o
\end{IEEEeqnarray}

we get

\begin{IEEEeqnarray}{rCl}
a & = & b + c
\\
& = & d + e + f + g + h a=b+c (40)
+ i + j + k + m = d + e + f + g + h + i + j + k + m (41)
\IEEEeqnarraynumspace\\ =l+n+o (42)
& = & l + n + o
\end{IEEEeqnarray}

Note that if there is not enough space on the line, this shift will force the numbers
to cross the right boundary of the text. So be sure to check the result!

The boundary of the text can be

seen from this text above the
equation array. The number is The boundary of the text can be seen from
clearly beyond it: this text above the equation array. The num-
\begin{IEEEeqnarray}{rCl} ber is clearly beyond it:
a & = & d + e + f + g + h
+ i + j + k + l + m + n a = d + e + f + g + h + i + j + k + l + m + n(43)
\IEEEeqnarraynumspace
\end{IEEEeqnarray}

In such a case one needs an additional wrapping.

5.2 First line too long: IEEEeqnarraymulticol

If the LHS is too long and as a replacement for the faulty \lefteqn{}-command,
IEEEeqnarray offers the \IEEEeqnarraymulticol-command, which works in all situ-
ations:

c Stefan M. Moser 21 June 2016, Version 4.5
How to Typeset Equations in LATEX 13

\begin{IEEEeqnarray}{rCl}
\IEEEeqnarraymulticol{3}{l}{
a + b + c + d + e + f
+ g + h a+b+c+d+e+f +g+h
}\nonumber\\* \quad
& = & i + j =i+j (44)
\\ =k+l+m (45)
& = & k + l + m
\end{IEEEeqnarray}

The usage is identical to the \multicolumns-command in the tabular-environment.

The first argument {3} specifies that three columns shall be combined into one, which
will be left-justified {l}. We usually add a * to the line-break \\ to prevent a page-
break at this position.
Note that by adapting the \quad-command one can easily adapt the depth of the
equation signs,11 e.g.,

\begin{IEEEeqnarray}{rCl}
\IEEEeqnarraymulticol{3}{l}{
a + b + c + d + e + f
+ g + h
}\nonumber\\* \qquad\qquad a+b+c+d+e+f +g+h
& = & i + j =i+j (46)
\label{eq:label45} =k+l+m (47)
\\
& = & k + l + m
\end{IEEEeqnarray}

Note that \IEEEeqnarraymulticol must be the first command in a cell. This is usu-
ally no problem; however, it might be the cause of some strange compilation errors. For
example, one might put a \label-command on the first line inside12 of IEEEeqnarray,
which is OK in general, but not OK if it is followed by the \IEEEeqnarraymulticol-
command.

5.3 Line-break: unary versus binary operators

If an equation is split onto two or more lines, LATEX interprets the first + or as a
sign instead of an operator. Therefore, it is necessary to add an additional space \>
between the operator and the term: instead of

\begin{IEEEeqnarray}{rCl}
a & = & b + c
\\
& = & d + e + f + g + h a=b+c (48)
+ i + j + k \nonumber\\ =d+e+f +g+h+i+j+k
&& + l + m + n + o +l + m + n + o (49)
\\
=p+q+r+s (50)
& = & p + q + r + s
\end{IEEEeqnarray}
11
I think that one quad is the distance that looks good in most cases.
12
I strongly recommend to put each label at the end of the corresponding equation; see Section 5.4.

c Stefan M. Moser 21 June 2016, Version 4.5
How to Typeset Equations in LATEX 14

we should write

\begin{IEEEeqnarray}{rCl}
a & = & b + c
\\
& = & d + e + f + g + h a=b+c (51)
+ i + j + k \nonumber\\
=d+e+f +g+h+i+j+k
&& +\> l + m + n + o
\label{eq:add_space} +l+m+n+o (52)
\\ =p+q+r+s (53)
& = & p + q + r + s
\end{IEEEeqnarray}

(Compare the space between + and l!)

Attention: The distinction between the unary operator (sign) and the binary
operator (addition/subtraction) is not satisfactorily solved in LATEX.13 In some cases
LATEX will automatically assume that the operator cannot be unary and will therefore
add additional spacing. This happens, e.g., in front of

an operator name like \log, \sin, \det, \max, etc.,

an integral \int or sum \sum,

a bracket with adaptive size using \left and \right (this is in contrast to normal
brackets or brackets with fixed size like \big( ).

This decision, however, might be faulty. E.g., it makes perfect sense to have a unary
operator in front of the logarithm:

\begin{IEEEeqnarray*}{rCl"s}
\log \frac{1}{a}
& = & -\log a 1
& (binary, wrong) \\ log = log a (binary, wrong)
a
& = & -{\log a} = log a (unary, correct)
& (unary, correct)
\end{IEEEeqnarray*}

In this case, you have to correct it manually. Unfortunately, there is no clean way of
doing this. To enforce a unary operator, enclosing the expression following the unary
operator and/or the unary operator itself into curly brackets {...} will usually work.
For the opposite direction, i.e., to enforce a binary operator (as, e.g., needed in (52)),
the only option is to put in the correct space \> manually.14
In the following example, compare the spacing between the first minus-sign on the
RHS and b (or log b):
13
The problem actually goes back to TEX.
14
This spacing command adds the flexible space medmuskip = 4mu plus 2mu minus 4mu.

c Stefan M. Moser 21 June 2016, Version 4.5
How to Typeset Equations in LATEX 15

\begin{IEEEeqnarray*}{rCls}
a & = & - b - b - c
& (default unary) \\
& = & {-} {b} - b - c
& (default unary, no effect) \\ a = b b c (default unary)
& = & -\> b - b - c = b b c (default unary, no effect)
& (changed to binary) \\
=bbc (changed to binary)
& = & - \log b - b - d
& (default binary) \\ = log b b d (default binary)
& = & {-} {\log b} - b - d = log b b d (changed to unary)
& (changed to unary) \\ = log b bd (changed d to unary)
& = & - \log b - b {-} d
& (changed $-d$ to unary)
\end{IEEEeqnarray*}

We learn:

Whenever you wrap a line, quickly check the result and verify that the spacing
is correct!

5.4 Equation-numbering
While IEEEeqnarray assigns an equation number to all lines, the starred version
IEEEeqnarray* suppresses all numbers. This behavior can be changed individually
per line by the commands

\IEEEyesnumber and \IEEEnonumber (or \nonumber).

For subnumbering the corresponding commands

\IEEEyessubnumber and \IEEEnosubnumber

are available. These four commands only affect the line on which they are invoked,
however, there also exist starred versions

\IEEEyesnumber*, \IEEEnonumber*,
\IEEEyessubnumber*, \IEEEnosubnumber*

that will remain active over several lines until another starred command is invoked.
Consider the following example.

c Stefan M. Moser 21 June 2016, Version 4.5
How to Typeset Equations in LATEX 16

\begin{IEEEeqnarray*}{rCl}
a & = & b_{1} \\ a = b1
& = & b_{2} \IEEEyesnumber\\ = b2 (54)
& = & b_{3} \\
= b3
& = & b_{4} \IEEEyesnumber*\\
& = & b_{5} \\ = b4 (55)
& = & b_{6} \\ = b5 (56)
& = & b_{7} \IEEEnonumber\\ = b6 (57)
& = & b_{8} \\
= b7
& = & b_{9} \IEEEnonumber*\\
& = & b_{10} \\ = b8 (58)
& = & b_{11} \IEEEyessubnumber*\\ = b9
& = & b_{12} \\ = b10
& = & b_{13} \IEEEyesnumber\\
= b11 (58a)
& = & b_{14} \\
& = & b_{15} = b12 (58b)
\end{IEEEeqnarray*} = b13 (59)
(\ldots some text\ldots) = b14 (59a)
\begin{IEEEeqnarray}{rCl}
= b15 (59b)
\label{eq:bad_placement}
a&= & b_{16} \IEEEyessubnumber*\\ (. . . some text. . . )
& = & b_{17} \\
& = & b_{18} \IEEEyesnumber a = b16 (59c)
\IEEEyessubnumber*\\
= b17 (59d)
& = & b_{19} \\
& = & b_{20} \IEEEnosubnumber*\\ = b18 (60a)
& = & b_{21} \\ = b19 (60b)
& = & b_{22} \nonumber\\ = b20 (61)
& = & b_{23}
= b21 (62)
\end{IEEEeqnarray}
(\ldots more text\ldots) = b22
\begin{IEEEeqnarray}{rCl} = b23 (63)
\IEEEyesnumber\IEEEyessubnumber*
a & = & b_{24} \\ (. . . more text. . . )
& = & b_{25}
\label{eq:good_placement}\\ a = b24 (64a)
& = & b_{26} = b25 (64b)
\end{IEEEeqnarray} = b26 (64c)

Note that the behavior in the line 13 (i.e., the line containing b13 ) is probably un-
wanted: there the command \IEEEyesnumber temporarily switches to a normal equa-
tion number (implicitly resetting the subnumbers), but in the subsequent line the
\IEEEyessubnumber* from line 11 takes control again, i.e., subnumbering is reactivated.
The correct way of increasing the number and start directly with a new subnumber is
shown in line 18 and in line 24. Also note that the subnumbering works even across
different IEEEeqnarray-environments, as can be seen in line 16.
The best way of understanding the numbering behavior is to note that in spite of
the eight different commands, there are only three different modes:

1. No equation number (corresponding to \IEEEnonumber).

c Stefan M. Moser 21 June 2016, Version 4.5
How to Typeset Equations in LATEX 17

2. A normal equation number (corresponding to \IEEEyesnumber): the equation

counter is incremented and then displayed.

3. An equation number with subnumber (corresponding to \IEEEyessubnumber):

only the subequation counter is incremented and then both the equation and the
subequation numbers are displayed. (Attention: If the equation number shall be
incremented as well, which is usually the case for the start of a new subnumbering,
then also \IEEEyesnumber has to be given!)

The understanding of the working of these three modes is also important when us-
ing labels to refer to equations. Note that the label must always be given after the
\IEEEyesnumber or \IEEEyessubnumber command as only then the counters have been
increased to the correct value. Otherwise, a label can produce an undesired output:
e.g., the label eq:bad_placement in line 16 points15 (wrongly) to (60).
We learn:

Labels should always be put as last command right in front of the line-break \\
or the end of the equation it belongs to.

Besides preventing unwanted results, this rules also increases the readability of the
source code and prevents a compilation error in the situation of an \IEEEeqnarraymul
ticol-command after a label-definition. A correct example is shown in (64b).

5.5 Page-breaks inside of IEEEeqnarray

By default, amsmath does not allow page-breaks within multiple equations, which usu-
ally is too restrictive, particularly, if a document contains long equation arrays. This
behavior can be changed by putting the following line into the document header:

\interdisplaylinepenalty=xx

Here, xx is some number: the larger this number, the less likely it is that an equation
array is broken over to the next page. So, a value 0 fully allows page-breaks, a value
2500 allows page-breaks, but only if LATEX finds no better solution, or a value 9999
basically prevents page-breaks.16
15
To understand this, note that when the label-command was invoked, subnumbering was deactiv-
ated. So the label only refers to a normal equation number. However, no such number was active there
either, so the label is passed on to line 18 where the equation counter is incremented for the first time.
16
I usually use a value 1000 that in principle allows page-breaks, but still asks LATEX to check if there
is no other way.

c Stefan M. Moser 21 June 2016, Version 4.5
How to Typeset Equations in LATEX 18

6 Advanced Typesetting
In this section we address a couple of more advanced typesetting problems and tools.

6.1 IEEEeqnarraybox: general tables and arrays

The package IEEEtrantools also provides the environment IEEEeqnarraybox. This is
basically the same as IEEEeqnarray but with the difference that it can be nested within
other structures. Therefore it does not generate a full equation itself nor an equation
number. It can be used both in text-mode (e.g., inside a table) or in math-mode (e.g.,
inside an equation). Hence, IEEEeqnarraybox is a replacement both for array and
tabular.17
This is a silly table:
\begin{center}
\begin{IEEEeqnarraybox}{t.t.t} This is a silly table:
\textbf{Item} &
\textbf{Color} & Item Color Number
\textbf{Number} \\
cars green 17
cars & green & 17 \\
trucks & red & 4 \\ trucks red 4
bikes & blue & 25 bikes blue 25
\end{IEEEeqnarraybox}
\end{center}

Note that t in the argument of IEEEeqnarraybox stands for centered text and . adds
space between the columns. Further possible arguments are s for left text, u for right
text, v for a vertical line, and V for a vertical double-line. More details can be found in
Tables IV and V on page 18 in the manual IEEEtran_HOWTO.pdf.
Another example:18

\begin{equation}
P_U(u) = \left\{ \,
\begin{IEEEeqnarraybox}[][c]{l?s}
\IEEEstrut
0.1 & if $u=0$, \\
0.3 & if $u=1$, \\ 0.1
if u = 0,
PU (u) = 0.3 if u = 1, (65)
0.6 & if $u=2$.
0.6 if u = 2.

\IEEEstrut
\end{IEEEeqnarraybox}
\right.
\label{eq:example_left_right1}
\end{equation}

Here ? is a large horizontal space between the columns, and \IEEEstrut adds a tiny
space above the first and below the bottom line. Moreover, note that the second
optional argument [c] makes sure that the IEEEeqnarraybox is vertically centered.
17
In case one does not want to let IEEEeqnarraybox to detect the mode automatically, but to force one
of these two modes, there are two subforms: IEEEeqnarrayboxm for math-mode and IEEEeqnarrayboxt
for text-mode.
18
For another way of generating case distinctions, see Section 6.2.

c Stefan M. Moser 21 June 2016, Version 4.5
How to Typeset Equations in LATEX 19

The other possible values for this option are [t] for aligning the first row with the
surrounding baseline and [b] for aligning the bottom row with the surrounding baseline.
Default is [b], i.e., if we do not specify this option, we get the following (in this case
unwanted) result:

\begin{equation*}
P_U(u) = \left\{ \,
0.1 if u = 0,

\begin{IEEEeqnarraybox}{l?s}

0.3 if u = 1,

0.1 & if $u=0$, \\

0.3 & if $u=1$, \\ PU (u) = 0.6 if u = 2.
0.6 & if $u=2$.



\end{IEEEeqnarraybox}

\right.
\end{equation*}

We also dropped \IEEEstrut here with the result that the curly bracket is slightly too
small at the top line.
Actually, these manually placed \IEEEstrut commands are rather tiring. More-
over, when we would like to add vertical lines in a table, a first naive application of
IEEEeqnarraybox yields the following:

\begin{equation*}
\begin{IEEEeqnarraybox}
{cc;v;ccc}
D_1 & D_2 & & X_1 & X_2 & X_3 D1 D2 X1 X2 X3
\\\hline 0 0 +1 +1 +1
0 & 0 && +1 & +1 & +1\\ 0 1 +1 1 1
0 & 1 && +1 & -1 & -1\\ 1 0 1 +1 1
1 & 0 && -1 & +1 & -1\\
1 1 1 1 +1
1 & 1 && -1 & -1 & +1
\end{IEEEeqnarraybox}
\end{equation*}

We see that IEEEeqnarraybox makes a complete line-break after each line. This is of
course unwanted. Therefore, the command \IEEEeqnarraystrutmode is provided that
switches the spacing system completely over to struts:

\begin{equation*}
\begin{IEEEeqnarraybox}[
\IEEEeqnarraystrutmode
]{cc;v;ccc}
D1 D2 X1 X2 X3
D_1 & D_2 & & X_1 & X_2 & X_3
0 0 +1 +1 +1
\\\hline
0 1 +1 1 1
0 & 0 && +1 & +1 & +1\\
1 0 1 +1 1
0 & 1 && +1 & -1 & -1\\
1 1 1 1 +1
1 & 0 && -1 & +1 & -1\\
1 & 1 && -1 & -1 & +1
\end{IEEEeqnarraybox}
\end{equation*}

The strutmode also easily allows to ask for more air between each line and thereby
eliminating the need of manually adding \IEEEstrut:

c Stefan M. Moser 21 June 2016, Version 4.5
How to Typeset Equations in LATEX 20

\begin{equation*}
\begin{IEEEeqnarraybox}[
\IEEEeqnarraystrutmode
\IEEEeqnarraystrutsizeadd{3pt}
{1pt} D1 D2 X1 X2 X3
]{cc/v/ccc} 0 0 +1 +1 +1
D_1 & D_2 & & X_1 & X_2 & X_3
\\\hline 0 1 +1 1 1
0 & 0 && +1 & +1 & +1\\ 1 0 1 +1 1
0 & 1 && +1 & -1 & -1\\ 1 1 1 1 +1
1 & 0 && -1 & +1 & -1\\
1 & 1 && -1 & -1 & +1
\end{IEEEeqnarraybox}
\end{equation*}

Here the first argument of \IEEEeqnarraystrutsizeadd{3pt}{1pt} adds space above

into each line, the second adds space below into each line.

6.2 Case distinctions

Case distinctions can be generated using IEEEeqnarraybox as shown in Section 6.1.
However, in the standard situation the usage of cases is simpler and we therefore
recommend to use this:

\begin{equation}
P_U(u) =
\begin{cases}
0.1 & \text{if } u=0,
\\ 0.1 if u = 0,

PU (u) = 0.3 if u = 1, (66)
0.3 & \text{if } u=1,
0.6 if u = 2.

\\
0.6 & \text{if } u=2.
\end{cases}
\end{equation}

For more complicated examples we do need to rely on IEEEeqnarraybox:

c Stefan M. Moser 21 June 2016, Version 4.5
How to Typeset Equations in LATEX 21

\begin{equation}
\left.
\begin{IEEEeqnarraybox}[
\IEEEeqnarraystrutmode
\IEEEeqnarraystrutsizeadd{2pt}
{2pt}
][c]{rCl}
x & = & a + b\\
y & = & a - b
\end{IEEEeqnarraybox}
\, \right\}
\iff
\left\{ \, x y
a = +
)
x=a+b

\begin{IEEEeqnarraybox}[ 2 2
(67)
\IEEEeqnarraystrutmode y =ab x y
b=


\IEEEeqnarraystrutsizeadd{7pt} 2 2
{7pt}
][c]{rCl}
a & = & \frac{x}{2}
+ \frac{y}{2}
\\
b & = & \frac{x}{2}
- \frac{y}{2}
\end{IEEEeqnarraybox}
\right.
\label{eq:example_left_right2}
\end{equation}

For case distinctions with equation numbers, the package

\usepackage{cases}

provides by far the easiest solution:

\begin{numcases}{|x|=}
x & for $x \geq 0$,
for x 0,

\\ x (68)
|x| =
-x & for $x < 0$. x for x < 0. (69)
\end{numcases}

Note the differences to the usual cases-environment:

The left-hand side must be typeset as compulsory argument to the environment.
The second column is not in math-mode but directly in text-mode.
For subnumbering we can use the corresponding subnumcases-environment:

\begin{subnumcases}{P_U(u)=}
0.1 & if $u=0$,

\\
0.1
if u = 0, (70a)
0.3 & if $u=1$, PU (u) = 0.3 if u = 1, (70b)
\\
0.6 if u = 2. (70c)

0.6 & if $u=2$.
\end{subnumcases}

c Stefan M. Moser 21 June 2016, Version 4.5
How to Typeset Equations in LATEX 22

6.3 Grouping numbered equations with a bracket

Sometimes, one would like to group several equations together with a bracket. We have
already seen in Section 6.1 how this can be achieved by using IEEEeqnarraybox inside
of an equation-environment:

\begin{equation}
\left\{
\begin{IEEEeqnarraybox}[
\IEEEeqnarraystrutmode
\IEEEeqnarraystrutsizeadd{2pt}
{2pt} (
x = f (x, u)
][c]{rCl} (71)
\dot{x} & = & f(x,u) x + x = h(x)
\\
x+\dot{x} & = & h(x)
\end{IEEEeqnarraybox}
\right.
\end{equation}

The problem here is that since the equation number is provided by the equation-
environment, we only get one equation number. But here, a number for each equation
would make much more sense.
We could again rely on numcases (see Section 6.2), but then we have no way of
aligning the equations horizontally:

\begin{numcases}{}
\dot{x} = f(x,u) 
\\ x = f (x, u) (72)
x+\dot{x} = h(x) x + x = h(x) (73)
\end{numcases}

Note that misusing the second column of numcases is not an option either:

\ldots very poor typesetting:

\begin{numcases}{}
\dot{x} & $\displaystyle . . . very poor typesetting:
= f(x,u)$ 
\\ x = f (x, u) (74)
x+\dot{x} & $\displaystyle x + x = h(x) (75)
= h(x)$
\end{numcases}

The problem can be solved using IEEEeqnarray: we define an extra column on

the most left that will only contain the bracket. However, as this bracket needs to
be far higher than the line where it is defined, the trick is to use \smash to make its
true height invisible to IEEEeqnarray, and then design its height manually using
the \IEEEstrut-command. The number of necessary jots depends on the height of the
equation and needs to be adapted manually:

c Stefan M. Moser 21 June 2016, Version 4.5
How to Typeset Equations in LATEX 23

\begin{IEEEeqnarray}{rrCl}
& \dot{x} & = & f(x,u)
\\*
\smash{\left\{
\IEEEstrut[8\jot]
x = f (x, u) (76)
\right.} x + x = h(x) (77)
& x+\dot{x} & = & h(x)

x+x
= g(x) (78)
\\*
& x+\ddot{x} & = & g(x)
\end{IEEEeqnarray}

The star in \\* is used to prevent the possibility of a page-break within the structure.
This works fine as long as the number of equations is odd and the total height of the
equations above the middle row is about the same as the total height of the equations
below. For example, for five equations (this time using subnumbers for a change):

\begin{IEEEeqnarray}{rrCl}
\IEEEyesnumber\IEEEyessubnumber*
& a_1 + a_2 & = & f(x,u)
\\*
& a_1 & = & \frac{1}{2}h(x)
\\*

a1 + a2 = f (x, u) (79a)
\smash{\left\{ 1


a1 = h(x) (79b)


\IEEEstrut[16\jot]


2
\right.} b = g(x, u) (79c)
& b & = & g(x,u) h(x)



\\*

y = (79d)
10



& y_{\theta} & = & b2 + a = g(x, u)

(79e)
2
\frac{h(x)}{10}
\\*
& b^2 + a_2 & = & g(x,u)
\end{IEEEeqnarray}

However, if the heights of the equations differ greatly or if the number of equations is
even, we get into a problem:

Bad example:
\begin{IEEEeqnarray}{rrCl}
& a_1 + a_2 & = &
\sum_{k=1}^{\frac{M}{2}} f_k(x,u)
\label{eq:uneven1} Bad example:
\\* M
\smash{\left\{
X2

a + a = fk (x, u) (80)

\IEEEstrut[15\jot]

1 2


\right.}

k=1

& b & = & g(x,u) b = g(x, u) (81)

\label{eq:uneven2} y = h(x) (82)




\\*




& y_{\theta} & = & h(x)
\label{eq:uneven3}
\end{IEEEeqnarray}

or

c Stefan M. Moser 21 June 2016, Version 4.5
How to Typeset Equations in LATEX 24

Bad example:
\begin{IEEEeqnarray}{rrCl}
& \dot{x} & = & f(x,u) Bad example:
\\*

\smash{\left\{
x = f (x, u)
(83)
\IEEEstrut[8\jot]
\right.} \nonumber
y = h(x) (84)

\\*
& y_{\theta} & = & h(x)
\end{IEEEeqnarray}

To solve this issue, we need manual tinkering. The basic idea is to use a hidden row
at a place of our choice. To make the row hidden, we need to manually move down
the row above the hidden row, and to move up the row below, both by about half the
usual line spacing:

\begin{IEEEeqnarray}{rrCl}
& \dot{x} & = & f(x,u)
\\*[-0.625\normalbaselineskip]
% start invisible row
\smash{\left\{ (
\IEEEstrut[6\jot] x = f (x, u) (85)
\right.} \nonumber x + x = h(x) (86)
% end invisible row
\\*[-0.625\normalbaselineskip]
& x+\dot{x} & = & h(x)
\end{IEEEeqnarray}

In the case of an odd, but unequally sized number of equations, we can put the bracket
on an individual row anywhere and then moving it up or down depending on how we
need it. The example (80)(82) with the three unequally sized equations now looks as
follows:

\begin{IEEEeqnarray}{rrCl}
& a_1 + a_2 & = &
\sum_{k=1}^{\frac{M}{2}} f_k(x,u)
\\*[-0.1\normalbaselineskip] M
\smash{\left\{
2
X
a + a = fk (x, u) (87)

\IEEEstrut[12\jot]

1 2
\right.} \nonumber k=1
\\*[-0.525\normalbaselineskip]

b = g(x, u) (88)

& b & = & g(x,u)
y = h(x) (89)
\\*
& y_{\theta} & = & h(x)
\end{IEEEeqnarray}

Note how we can move the bracket up and down by changing the amount of shift in
both \\*[...\normalbaselineskip]-commands: if we add +2 to the first and 2 to
the second command (which makes sure that in total we have added 2 2 = 0), we get:

c Stefan M. Moser 21 June 2016, Version 4.5
How to Typeset Equations in LATEX 25

\begin{IEEEeqnarray}{rrCl}
& a_1 + a_2 & = &
\sum_{k=1}^{\frac{M}{2}} f_k(x,u)
\\*[1.9\normalbaselineskip]
M
\smash{\left\{ 2
X
\IEEEstrut[12\jot] a1 + a2 = fk (x, u) (90)
\right.} \nonumber
k=1
b = g(x, u) (91)

\\*[-2.525\normalbaselineskip]


& b & = & g(x,u) y = h(x) (92)
\\*



& y_{\theta} & = & h(x)

\end{IEEEeqnarray}

6.4 Matrices
Matrices could be generated by IEEEeqnarraybox, however, the environment pmatrix
is easier to use:

\begin{equation}
\mathsf{P} =
\begin{pmatrix}
p_{11} & p_{12} & \ldots
& p_{1n} \\
p11 p12 ... p1n

p_{21} & p_{22} & \ldots p21 p22 ... p2n
& p_{2n} \\ P= . (93)

.. .. ..
\vdots & \vdots & \ddots .. . . .
& \vdots \\ pm1 pm2 ... pmn
p_{m1} & p_{m2} & \ldots
& p_{mn}
\end{pmatrix}
\end{equation}

Note that it is not necessary to specify the number of columns (or rows) in advance.
More possibilities are bmatrix (for matrices with square brackets), Bmatrix (curly
brackets), vmatrix (|), Vmatrix (k), and matrix (no brackets at all).

6.5 Adapting the size of brackets

LATEX offers the functionality of brackets being automatically adapted to the size of the
expression they embrace. This is done using the pair of directives19 \left and \right:

\begin{equation} !
n
a = \log \left( 1 + X
a = log 1 + bk (94)
\sum_{k=1}^n b_k \right)
k=1
\end{equation}
19
Unfortunately, the left/right command pair has a weakness: in certain situations the chosen
(1)
bracket size is slightly too big. For example, if expressions
  with larger superscripts like a are typeset
in displaystyle or, like here, in footnotes, we get a(1) . I suggest to choose the bracket size manually
(big) in these cases: a(1) .

c Stefan M. Moser 21 June 2016, Version 4.5
How to Typeset Equations in LATEX 26

The brackets do not need to be round, but can be of various types, e.g.,

\begin{equation*}
\left\| \left(
\left[ \left\{ \left|
\left\lfloor \left\lceil  
 1 
\frac{1}{2}


 2 


\right\rceil \right\rfloor
\right| \right\} \right]
\right) \right\|
\end{equation*}

It is important to note that \left and \right always must occur as a pair, but as
we have just seen they can be nested. Moreover, the brackets do not need to match:

\begin{equation*}  
\left( \frac{1}{2}, 1 \right] 1
,1 R
\subset \mathbb{R} 2
\end{equation*}

One side can even be made invisible by using a dot instead of a bracket (\left. or
\right.). We have already seen such examples in (65) or (67).
For an additional element in between a \left-\right pair that should have the
same size as the surrounding brackets, the command \middle is available:

\begin{equation}   
H\left(X \, \middle| \, Y
H X  (95)
\frac{Y}{X} \right) X
\end{equation}

Here both the size of the vertical bar and of the round brackets are adapted according
Y
to the size of X .
Unfortunately, \left-\right pairing cannot be done across a line-break. So, if we
wrap an equation using multline or IEEEeqnarray, we cannot have a \left on one
side and the corresponding \right on the other side of the \\. In a first attempt, we
might try to fix this by introducing a \right. on the left of \\ and a \left. on the
right of \\ as shown in the following example:

\begin{IEEEeqnarray}{rCl}
a & = & \log \left( 1 \right.
\nonumber\\
a = log (1
&& \qquad \left. + \> 
\frac{b}{2} \right) b
+ (96)
\label{eq:wrong_try} 2
\end{IEEEeqnarray}

As can be seen from this example, this approach usually does not work, because the
sizes of the opening and closing brackets do not match anymore. In the example (96),
the opening bracket adapts its size to 1, while the closing bracket adapts its size to
b
2.

c Stefan M. Moser 21 June 2016, Version 4.5
How to Typeset Equations in LATEX 27

There are two ways to try to fix this. The by far easier way is to choose the bracket
size manually:

\begin{IEEEeqnarray}{rCl}
a & = & \log \bigg( 1 
\nonumber\\ a = log 1
&& \qquad +\> 
b
\frac{b}{2} \bigg) + (97)
\end{IEEEeqnarray} 2

There are four sizes available: in increasing order \big, \Big, \bigg, and \Bigg. This
manual approach will fail, though, if the expression in the brackets requires a bracket
size larger than \Bigg, as shown in the following example:

\begin{IEEEeqnarray}{rCl}
a & = & \log \Bigg( 1
\nonumber\\ a = log 1
&& \qquad + \sum_{k=1}^n
\frac{e^{1+\frac{b_k^2}{c_k^2}}} b2
k
n 1+ 2 !
{1+\frac{b_k^2}{c_k^2}} X e ck
+ b2k
(98)
\Bigg) \label{eq:sizecorr1} 1+
k=1 c2k
\end{IEEEeqnarray}

For this case we need a trick: since we want to rely on a

\left( ... \right. \\ \left. ... \right)

construction, we need to make sure that both pairs are adapted to the same size. To
that goal we define the following command in the document header:

\newcommand{\sizecorr}[1]{\makebox[0cm]{\phantom{$\displaystyle #1$}}}

We then pick the larger of the two expressions on either side of \\ (in (98) this is the
term on the second line) and typeset it a second time also on the other side of the
line-break (inside of the corresponding \left-\right pair). However, since we do not
actually want to see this expression there, we put it into \sizecorr{} and thereby
make it both invisible and of zero width. In the example (98) this looks as follows:

c Stefan M. Moser 21 June 2016, Version 4.5
How to Typeset Equations in LATEX 28

\begin{IEEEeqnarray}{rCl}
a & = & \log \left(
% copy-paste from below, invisible
\sizecorr{
\sum_{k=1}^n
\frac{e^{1+\frac{b_k^2}{c_k^2}}}
a = log 1

{1+\frac{b_k^2}{c_k^2}}
}
b2

% end copy-paste n 1+ k
c2
1 \right. \nonumber\\
X e k
+ b2k
(99)
&& \qquad \left. + \sum_{k=1}^n k=1 1 + c2
k
\frac{e^{1+\frac{b_k^2}{c_k^2}}}
{1+\frac{b_k^2}{c_k^2}}
\right) \label{eq:sizecorr2}
\end{IEEEeqnarray}

Note how the expression inside of \sizecorr{} does not actually appear, but is used
for computing the correct bracket size.

6.6 Framed equations

To generate equations that are framed, one can use the \boxed{...}-command. Un-
fortunately, this usually will yield a too tight frame around the equation:

\begin{equation}
\boxed{
a = b + c a=b+c (100)
}
\end{equation}

To give the frame a little bit more air we need to redefine the length-variable
\fboxsep. We do this in a way that restores its original definition afterwards:

\begin{equation}
\newlength{\fboxstore}
\setlength{\fboxstore}{\fboxsep}
\setlength{\fboxsep}{6pt}
\boxed{ a=b+c (101)
a = b + c
}
\setlength{\fboxsep}{\fboxstore}
\end{equation}

Note that the \newlength-command must be given only once per document. To ease
ones life, we recommend to define a macro for this in the document header:

\newlength{\eqboxstorage}
\newcommand{\eqbox}[1]{
\setlength{\eqboxstorage}{\fboxsep}
\setlength{\fboxsep}{6pt}
\boxed{#1}
\setlength{\fboxsep}{\eqboxstorage}
}

c Stefan M. Moser 21 June 2016, Version 4.5
How to Typeset Equations in LATEX 29

Now the framed equation can be produced as follows:

\begin{equation}
\eqbox{
a = b + c a=b+c (102)
}
\end{equation}

In case of multline or IEEEeqnarray this approach does not work because the
boxed{...} command does not allow line-breaks or similar. Therefore we need to rely
on IEEEeqnarraybox for boxes around equations on several lines:

\begin{equation}
\eqbox{
\begin{IEEEeqnarraybox}{rCl}
a & = & b + c
\\
& = & d + e + f + g + h a=b+c
+ i + j + k \\ =d+e+f +g+h+i+j+k
&& +\> l + m + n + o +l+m+n+o
\\ =p+q+r+s (103)
& = & p + q + r + s
\end{IEEEeqnarraybox}
}
\end{equation}

Some comments:

The basic idea here is to replace the original IEEEeqnarray command by a

IEEEeqnarraybox and then wrap everything into an equation-environment.

The equation number is produced by the surrounding equation-environment. If

we would like to have the equation number vertically centered, we need to center
the IEEEeqnarraybox:

\begin{equation}
\eqbox{
\begin{IEEEeqnarraybox}[][c]{rCl}
a & = & b + c + d + e
+ f + g + h
\\ a=b+c+d+e+f +g+h
&& +\> i + j + k + l
+i+j+k+l+m+n (104)
+ m + n
\\ +o+p+q
&& +\> o + p + q
\end{IEEEeqnarraybox}
}
\end{equation}

in constrast to

c Stefan M. Moser 21 June 2016, Version 4.5
How to Typeset Equations in LATEX 30

\begin{equation}
\eqbox{
\begin{IEEEeqnarraybox}{rCl}
a & = & b + c + d + e
+ f + g + h
\\ a=b+c+d+e+f +g+h
&& +\> i + j + k + l
+i+j+k+l+m+n
+ m + n
\\ +o+p+q (105)
&& +\> o + p + q
\end{IEEEeqnarraybox}
}
\end{equation}

When changing the IEEEeqnarray into a IEEEeqnarraybox, be careful to delete

any remaining \nonumber or \IEEEnonumber commands inside of the IEEEeqnar
raybox! Since IEEEeqnarraybox does not know equation numbers anyway, any
remaining \nonumber command will leak through and prevent equation to put
a number!

\begin{equation}
\eqbox{
\begin{IEEEeqnarraybox}{rCl}
a & = & b + c + d + e
a=b+c+d+e+f +g+h
+ f + g + h \nonumber\\
&& +\> i + j + k + l +i+j+k+l
\end{IEEEeqnarraybox}
}
\end{equation}

6.7 Fancy frames

Fancier frames can be produced using the mdframed package. Use the following com-
mands in the header of your document:20

\usepackage{tikz}
\usetikzlibrary{shadows} %defines shadows
\usepackage[framemethod=tikz]{mdframed}

Then we can produce all kinds of fancy frames. We start by defining a certain style
(still in the header of your document):
20
The mdframed-package should be loaded after amsthm.sty.

c Stefan M. Moser 21 June 2016, Version 4.5
How to Typeset Equations in LATEX 31

\global\mdfdefinestyle{myboxstyle}{%
shadow=true,
linecolor=black,
shadowcolor=black,
shadowsize=6pt,
nobreak=false,
innertopmargin=10pt,
innerbottommargin=10pt,
leftmargin=5pt,
rightmargin=5pt,
needspace=1cm,
skipabove=10pt,
skipbelow=15pt,
middlelinewidth=1pt,
afterlastframe={\vspace{5pt}},
aftersingleframe={\vspace{5pt}},
tikzsetting={%
draw=black,
very thick}
}

These settings are quite self-explanatory. Just play around! Now we define different
types of framed boxes:

% framed box that can be broken at end of page

\newmdenv[style=myboxstyle]{whitebox}
\newmdenv[style=myboxstyle,backgroundcolor=black!20]{graybox}

% framed box that CANNOT be broken at end of page

\newmdenv[style=myboxstyle,nobreak=true]{blockwhitebox}
\newmdenv[style=myboxstyle,backgroundcolor=black!20,nobreak=true]{blockgraybox}

% invisible box that CANNOT be broken at end of page

\newmdenv[nobreak=true,hidealllines=true]{blockbox}

As the name suggests, the graybox adds a gray background color into the box, while the
background in whitebox remains white. Moreover, blockwhitebox creates the same
framed box as whitebox, but makes sure that whole box is typeset onto one single
page, while the regular whitebox can be split onto two (or even more) pages.
Examples:

\begin{whitebox}
\begin{IEEEeqnarray}[
\vspace{-\baselineskip}
]{rCl} a=b+c (106)
a & = & b + c
=d+e (107)
\\
& = & d + e
\end{IEEEeqnarray}
\end{whitebox}

or

c Stefan M. Moser 21 June 2016, Version 4.5
How to Typeset Equations in LATEX 32

\begin{graybox}
\begin{theorem}
This is a fancy theorem: Theorem 1. This is a fancy theorem:
we know by now that we know by now that
\begin{equation}
a = b + c. a = b + c. (108)
\end{equation}
\end{theorem}
\end{graybox}

Note that in the former example, we have removed some space above the equation
(that is automatically added by IEEEeqnarray) in order to have proper spacing. In
the latter example we have assumed that the theorem-environment has been defined
in the header:
\usepackage{amsthm}
\newtheorem{theorem}{Theorem}

6.8 Putting the QED correctly: proof

The package amsthm that we have used in Section 6.7 to generate a theorem actually
also defines a proof-environment:

\begin{proof}
This is the proof of some
Proof. This is the proof of some theorem.
theorem. Once the proof is
Once the proof is finished, a white box is put
finished, a white box is put
at the end to denote QED.
at the end to denote QED.
\end{proof}

The QED-symbol should be put on the last line of the proof. However, if the last line
is an equation, then this is done wrongly:

\begin{proof}
This is a proof that ends Proof. This is a proof that ends with an equa-
with an equation: (bad) tion: (bad)
\begin{equation*} a = b + c.
a = b + c.
\end{equation*}
\end{proof}

In such a case, the QED-symbol must be put by hand using the command \qedhere:

\begin{proof}
This is a proof that ends
with an equation: (correct) Proof. This is a proof that ends with an equa-
\begin{equation*} tion: (correct)
a = b + c. \qedhere
a = b + c.
\end{equation*}
\end{proof}

Unfortunately, this correction does not work for IEEEeqnarray:

c Stefan M. Moser 21 June 2016, Version 4.5
How to Typeset Equations in LATEX 33

\begin{proof}
This is a proof that ends
Proof. This is a proof that ends with an equa-
with an equation array: (wrong)
tion array: (wrong)
\begin{IEEEeqnarray*}{rCl}
a & = & b + c \\ a=b+c
& = & d + e. \qedhere
\end{IEEEeqnarray*} = d + e.
\end{proof}

The reason for this is the internal structure of IEEEeqnarray: it always puts two
invisible columns at both sides of the array that only contain a stretchable space.
Thereby, IEEEeqnarray ensures that the equation array is horizontally centered. The
\qedhere-command should actually be put outside this stretchable space, but this does
not happen as these columns are invisible to the user.
There is, however, a very simple remedy: we explicitly define these stretching
columns ourselves!

\begin{proof}
This is a proof that ends
Proof. This is a proof that ends with an equa-
with an equation array: (correct)
tion array: (correct)
\begin{IEEEeqnarray*}{+rCl+x*}
a & = & b + c \\ a=b+c
& = & d + e. & \qedhere
\end{IEEEeqnarray*} = d + e.
\end{proof}

Here, the + in {+rCl+x*} denotes a stretchable space, one on the left of the equations
(which, if not specified, will be done automatically by IEEEeqnarray) and one on the
right of the equations. But now on the right, after the stretching column, we add
an empty column x. This column will only be needed on the last line for putting
the \qedhere-command. Finally, we specify a *. This is a null-space that prevents
IEEEeqnarray to add another unwanted +-space.
In case of equation numbering, we have a similar problem. If you compare

\begin{proof}
This is a proof that ends with Proof. This is a proof that ends with a
a numbered equation: (bad) numbered equation: (bad)
\begin{equation}
a = b + c. a = b + c. (109)
\end{equation}
\end{proof}

with

\begin{proof}
This is a proof that ends with Proof. This is a proof that ends with a
a numbered equation: (correct) numbered equation: (correct)
\begin{equation}
a = b + c. \qedhere a = b + c. (110)
\end{equation}
\end{proof}

c Stefan M. Moser 21 June 2016, Version 4.5
How to Typeset Equations in LATEX 34

you notice that in the (correct) second version the  is much closer to the equation
than in the first version.
Similarly, the correct way of putting the QED-symbol at the end of an equation
array is as follows:

\begin{proof}
This is a proof that ends
Proof. This is a proof that ends with an equa-
with an equation array: (correct)
tion array: (correct)
\begin{IEEEeqnarray}{rCl+x*}
a & = & b + c \\ a=b+c (111)
& = & d + e. \label{eq:star}
= d + e. (112)
\\* &&& \qedhere\nonumber
\end{IEEEeqnarray}
\end{proof}

which contrasts with the poorer version:

\begin{proof}
This is a proof that ends Proof. This is a proof that ends with an equa-
with an equation array: (bad) tion array: (bad)
\begin{IEEEeqnarray}{rCl}
a=b+c (113)
a & = & b + c \\
& = & d + e. = d + e. (114)
\end{IEEEeqnarray}
\end{proof}

Note that we use a starred line-break in (112) to prevent a page-break just before the
QED-sign.
We would like to point out that equation does not handle the \qedhere-command
correctly in all cases. Compare the following:

\begin{proof}
This is a bad example for the
usage of \verb+\qedhere+ in Proof. This is a bad example for the usage of
combination with \verb+equation+: \qedhere in combination with equation:
\begin{equation}
X
a = \sum_{\substack{x_i\\ a= f (xi ). (115)
|x_i|>0}} f(x_i). xi
|xi |>0
\qedhere
\end{equation}
\end{proof}

with the much better solution:

\begin{proof}
This is the corrected example Proof. This is the corrected example using
using \verb+IEEEeqnarray+: IEEEeqnarray:
\begin{IEEEeqnarray}{c+x*}
X
a = \sum_{\substack{x_i\\ a= f (xi ). (116)
|x_i|>0}} f(x_i). xi
\\* & \qedhere\nonumber |xi |>0

\end{IEEEeqnarray}
\end{proof}

c Stefan M. Moser 21 June 2016, Version 4.5
How to Typeset Equations in LATEX 35

You notice how the  in the bad example is far too close the equation number and is
actually inside the mathematical expression. A similar problem also occurs in the case
of no equation number.
Hence:

We recommend not to use \qedhere in combination with equation, but

exclusively with IEEEeqnarray.

6.9 Putting the QED correctly: IEEEproof

IEEEtrantools also provides its own proof-environment that is slightly more flexible
than the proof of amsthm: IEEEproof. Note that under the IEEEtran-class, amsthm is
not permitted and therefore proof is not defined, i.e., one must use IEEEproof.
IEEEproof offers the command \IEEEQEDhere that produces the QED-symbol right
at the place where it is invoked and will switch off the QED-symbol at the end.

\begin{IEEEproof}
This is a short proof:
Proof: This is a short proof:
\begin{IEEEeqnarray}{rCl+x*}
a & = & b + c \\ a=b+c (117)
& = & d+ e \label{eq:qed}
=d+e (118)
\\* &&& \nonumber\IEEEQEDhere
\end{IEEEeqnarray}
\end{IEEEproof}

So, in this sense \IEEEQEDhere plays the same role for IEEEproof as \qedhere for
proof. Note, however, that their behavior is not exactly equivalent: \IEEEQEDhere
always puts the QED-symbol right at the place it is invoked and does, e.g., not move
it to the end of the line inside of a list or an equation*. So, for example, inside of a
list, an additional \hfill is needed:

\begin{IEEEproof}
A proof containing a list and Proof: A proof containing a list and two
two QED-symbols: QED-symbols:
\begin{enumerate}
\item Fact one.\IEEEQEDhere 1. Fact one.
\item Fact two.\hfill\IEEEQEDhere
\end{enumerate} 2. Fact two.
\end{IEEEproof}

Unfortunately, \hfill will not work inside an equation. To get the behavior of
\qedhere there, one needs to use \IEEEQEDhereeqn instead:

c Stefan M. Moser 21 June 2016, Version 4.5
How to Typeset Equations in LATEX 36

\begin{IEEEproof}
Placed directly behind math:
\begin{equation*} Proof: Placed directly behind math:
a = b + c. \hfill\IEEEQEDhere
\end{equation*} a = b + c.
Moved to the end of line:
Moved to the end of line:
\begin{equation*}
a = b + c. \IEEEQEDhereeqn a = b + c.
\end{equation*}
\end{IEEEproof}

\IEEEQEDhereeqn even works in situations with equation numbers, however, in contrast

to \qedhere it does not move the QED-symbol to the next line, but puts it in front of
the number:

\begin{IEEEproof}
Placed directly before the
equation number: Proof: Placed directly before the equa-
\begin{equation} tion number:
a = b + c. \IEEEQEDhereeqn a = b + c. (119)
\end{equation}
With some additional spacing: With some additional spacing:
\begin{equation}
a = b + c. \IEEEQEDhereeqn\; a = b + c. (120)
\end{equation}
\end{IEEEproof}

To get the behavior where the QED-symbol is moved to the next line, use the approach
based on IEEEeqnarray as shown in (118).
Once again:

We recommend not to use \IEEEQEDhere and \IEEEQEDhereeqn in combination

with equation, but to rely on \IEEEQEDhere and IEEEeqnarray exclusively.

Furthermore, IEEEproof offers the command \IEEEQEDoff to suppress the QED-

symbol completely; it allows to change the QED-symbol to be an open box as in
Section 6.8; and it allows to adapt the indentation of the proof header (default value
is 2\parindent). The latter two features are shown in the following example:

\renewcommand{\IEEEproofindentspace}{0em}
\renewcommand{\IEEEQED}{\IEEEQEDopen}
\begin{IEEEproof}
Proof: Proof without indentation and an
Proof without
open QED-symbol.
indentation and an
open QED-symbol.
\end{IEEEproof}

The default QED-symbol can be reactivated again by redefining \IEEEQED to be \IEEEQ

EDclosed.

c Stefan M. Moser 21 June 2016, Version 4.5
How to Typeset Equations in LATEX 37

We end this section by pointing out that IEEE standards do not allow a QED-
symbol and an equation put onto the same line. Instead one should follow the example
(118).

6.10 Double-column equations in a two-column layout

Many scientific publications are in a two-column layout in order to save space. This
means that the available width for the equations is considerably smaller than for a one-
column layout and will cause correspondingly more line-breaks. Then the advantages
of the IEEEeqnarray-environment are even more pronounced.
However, there are very rare situations when the breaking of an equation into two
or more lines will result in a very poor typesetting, even if IEEEeqnarray with all its
tricks is used. In such a case, a possible solution is to span an equation over both
columns. But the reader be warned:

Unless there is no other solution, we strongly discourage from the usage of

double-column equations in a two-column layout for aesthetic reasons and
because the LATEX code is rather ugly!

The trick is to use the figure-environment to create a floating object containing

the equation similarly to a included graphic. Concretely, we have to use figure* to
create a float that stretches over both columns. Since in this way the object becomes
floating, the equation numbering has to be carefully taken care of.
We explain the details using an example. We start by defining two auxiliary equation
counters:
\newcounter{storeeqcounter}
\newcounter{tempeqcounter}

The counter storeeqcounter will store the equation number that is assigned to the
floating equation, and the counter tempeqcounter will be used to restore the equation
counter to the correct number after it was temporarily set to the floating equations
number stored in storeeqcounter.
Note that if there are several floating equations in a document, each needs its own
unique definition of a storeeqcounter, i.e., one needs to introduce different names for
these counters (e.g., storeeqcounter_one, storeeqcounter_two, etc.). The counter
tempeqcounter can be reused for all floating equations.
Now, in the text where we will refer to the floating equation, we need to make sure
that the equation number is increased by one (i.e., at this place the equation numbering
will jump over one number, which is the number assigned to the floating equation), and
then we need to store this number for later use. This looks as follows:

c Stefan M. Moser 21 June 2016, Version 4.5
How to Typeset Equations in LATEX 38

\ldots and $a$ is given in

\eqref{eq:floatingeq}
%
%% Increase current equation
%% number and store it:
. . . and a is given in (121) on the top of this
\addtocounter{equation}{1}%
page/on top of Page 39.
\setcounter{storeeqcounter}%
{\value{equation}}%
%
on the top of this page/on top of
Page~\pageref{eq:floatingeq}.

Note that one must manually adapt the LATEX code to either the phrase on the top of
this page or the phrase on top of Page 39, depending on where the equation actually
appears.
Finally we typeset the floating equation:

\begin{figure*}[!t]
\normalsize
\setcounter{tempeqcounter}{\value{equation}} % temp store of current value
\begin{IEEEeqnarray}{rCl}
\setcounter{equation}{\value{storeeqcounter}} % number of this equation
a & = & b + c + d + e + f + g + h + i + j + k + l + m + n + o + p
\nonumber\\
&& +\> q + r + s + t + u + v + w + x + y + z + \alpha + \beta
+ \gamma + \delta + \epsilon
\label{eq:floatingeq}
\end{IEEEeqnarray}
\setcounter{equation}{\value{tempeqcounter}} % restore correct value
\hrulefill
\vspace*{4pt}
\end{figure*}

The exact location of this definition depends strongly on where the floating structure
should be appear, i.e., it might have to be placed quite far away from the text where the
equation is referred to.21 Note that this might need some trial and error, particularly
if there are other floating objects around to be placed by LATEX.
Be aware that due to a limitation of LATEX, double-column floating objects cannot
be placed at the bottom of pages, i.e., \begin{figure*}[!b] will not work correctly.
This can be corrected if we include the following line in the header of our document:

\usepackage{stfloats}

However, this package is very invasive and might cause troubles with other packages.22
21
It needs to be placed after the reference in the text, though, as otherwise the equation number
stored in storeeqcounter is not defined yet. This could again be fixed, but only if we set the equation
number (i.e., storeeqcounter) manually (ugly!!).
22
In particular, it cannot be used together with the package fixltx2e.sty. Luckily, the latter is not
needed anymore starting with TeXLive 2015.

c Stefan M. Moser 21 June 2016, Version 4.5
How to Typeset Equations in LATEX 39

a=b+c+d+e+f +g+h+i+j+k+l+m+n+o+p
+q+r+s+t+u+v+w+x+y+z+++++ (121)

7 Emacs and IEEEeqnarray

When working with Emacs, you can ease your life by defining a few new commands.
In the dot_emacs-file that comes together with this document the following commands
are defined:
Control-c i: Insert an IEEEeqnarray-environment (similar to Control-c Control-
e) with argument {rCl}.
Control-c o: As Control-c i, but the *-version.
Control-c b: Add a line-break at a specific place. This is very helpful in editing
too long lines. Suppose you have typed the following LATEX code:

\begin{IEEEeqnarray}{rCl}
a & = & b + c \\
& = & d + e + f + g + h + i a=b+c (122)
+ j + k + l + m + n + o =d+e+f +g+h+i+j+k+l+m+n
(123)
+o
\end{IEEEeqnarray}

After compiling you realize that you have to break the line before l. You now
just have to put the cursor on the +-sign in front of l and press Control-c b.
Then the line is wrapped there and also the additional space \> is added at the
right place:

\begin{IEEEeqnarray}{rCl}
a & = & b + c \\
& = & d + e + f + g + h + i a=b+c (124)
+ j + k \nonumber\\ =d+e+f +g+h+i+j+k
&& +\> l + m + n + o +l+m+n+o (125)
\end{IEEEeqnarray}

Control-c n: As Control-c b, but without adding the additional space \>.

Control-c Control-b: Remove a line-break (undo of Control-c b and Control-
c n). Position the cursor before the \nonumber and press Control-c Control-b.
Control-c m: Insert a \IEEEeqnarraymulticol-command. This is very helpful
when the LHS is too long. Suppose you have typed the following LATEX code:

\begin{IEEEeqnarray}{rCl}
a + b + c + d + e + f
+ g + h + i + j
a + b + c + d + e + f + g + h + i + j = k(126)
+l
& = & k + l \\
& = & m + n = m(127)
+n
\end{IEEEeqnarray}

c Stefan M. Moser 21 June 2016, Version 4.5
How to Typeset Equations in LATEX 40

After compiling you realize that the LHS is too long. You now just have to put
the cursor somewhere on the first line and type Control-c m. Then you get

\begin{IEEEeqnarray}{rCl}
\IEEEeqnarraymulticol{3}{l}{
a + b + c + d + e + f
+ g + h + i + j a+b+c+d+e+f +g+h+i+j
}\nonumber \\ \quad =k+l (128)
& = & k + l \\ =m+n (129)
& = & m + n
\end{IEEEeqnarray}

Finally, in the dot_emacs-file, settings are given that make IEEEeqnarray and
IEEEeqnarraybox known to Emacs LATEX-mode, reftex, and ispell. This way
many standard Emacs commands can be used as usual also in the context of
IEEEeqnarray. For example, Control-c ( will add an equation label.

8 Some Useful Definitions

There are a couple of mathematical symbols that cannot easily be found in LATEX-
symbol collections. In the following, a few such symbols are listed and a possible way
is proposed of how to define them.
Markov Chains: One of the existing customary ways to describe that three ran-
dom variables form a Markov chain is
\begin{equation*}
X \markov Y \markov Z Y (
X ( Z
\end{equation*}

is defined as a combination of \multimap (() and two

Here, the symbol (
minus-signs ():

\newcommand{\markov}{\mathrel{\multimap}\joinrel\mathrel{-}%
\joinrel\mathrel{\mkern-6mu}\joinrel\mathrel{-}}

For this definition to work, beside amsmath also the package amssymb needs to be
loaded.

Independence: To describe that two random variables are statistically independ-

ent, I personally prefer the following symbol:

\begin{equation*}
X \indep Y X
Y
\end{equation*}

Accordingly,

\begin{equation*}
X \dep Y X
6 Y
\end{equation*}

c Stefan M. Moser 21 June 2016, Version 4.5
How to Typeset Equations in LATEX 41

denotes that X and Y are not statistically independent.

These two symbols are created by two \bot () signs:

\newcommand{\indep}{\mathrel{\bot}\joinrel\mathrel{\mkern-5mu}%
\joinrel\mathrel{\bot}}
\newcommand{\dep}{\centernot\indep}

For this definition to work, beside amsmath also the package centernot needs to
be loaded.

Integration-d: The d in an integral is not a variable, but rather an operator. It

therefore should not be typeset italic d, but Roman d. Moreover, there should be
a small spacing before the operator:

\begin{equation*}
\int_a^b f(x) \dd x = \int_a^b Z b Z b x
\ln\left(\frac{x}{2}\right) f (x) dx = ln dx
\dd x a a 2
\end{equation*}

To make sure that this spacing always works out correctly, I recommend the
following definition:

\newcommand{\dd}{\mathop{}\!\mathrm{d}}

9 Some Final Remarks and Acknowledgments

The rules stated in this document are purely based on my own experience with
typesetting LATEX in my publications and on my some people might say unfortunate
habit of incorporating many mathematical expressions in there.
If you encounter any situation that seems to contradict the suggestions of this
document, then I would be very happy if you could send me a corresponding LATEX or
PDF file. As a matter of fact, any kind of feedback, criticism, suggestion, etc. is highly
appreciated! Write to

Thanks!
I would like to mention that during the writing and updating of this document
I profited tremendously from the help of Michael Shell, the author of IEEEtran. He
was always available for explanations when I got stuck somewhere. Moreover, I grate-
fully acknowledge the comments from (in alphabetical order) Helmut Bolcskei, Amos
Lapidoth, Edward Ross, Omar Scaglione, and Sergio Verd u.

Stefan M. Moser

c Stefan M. Moser 21 June 2016, Version 4.5