# Unable to put equation number for a matrix written using figure command - IEEE Transaction

I am using IEEE Transaction template which is in a double column format.

I have a large matrix which dows not fit in a single column. So, I used the figure option which spanned the matrix in both the columns.

However, I cannot put an Equation number. I need to refer this matrix in the text body. I have put the \label but in the text, there is no equation number.

If I don't explicitly put \nonumber, the latex throws compile error.

Is there a way so that the matrix gets assigned an equation number? The following code display the matrix at the botto of the page. I want to display it where cited near the text in equation format, having an equation number.

The first sentence on the right column is not aligned with the first sentence in the left column. Also how to put a label to the equation so that I can refer it? What is the correct way acceptable in IEEE journal or in most journals for formatting these kind of matrices?

• Aside: \mathbf{\theta^* - \hat{\theta^*}} cannot possibly work. I assume the following is roughly what you want: \boldsymbol{\theta}^*-\hat{\boldsymbol{\theta}}^*
– Mico
Jun 3 '17 at 19:30

First of all your example lacked a MWE. Second of all I invite you to read the amsmath documentation: those environment are really a blast.

You will see there that in align* the * has a different meaning, ie "non-numbered" environment, w.r.t align.

Here you don't need the align and can go with a simple equation. I also used the boldsymbol in the examples to render the bold theta, thaks to Mico

          \documentclass{IEEEtran}
\usepackage{amsmath,mathtools,amsfonts,amssymb,bm}
\begin{document}
\begin{figure*}
$$A(\boldsymbol{\theta}^* - \hat{\boldsymbol{\theta}^*}) > \text{constant} \begin{psmallmatrix} {E[A]}^{-1} & 0 &0 & 0 & 0 \\[3ex] 0 & {E[B]}^{-1} & 0 & 0 & 0 \\[3ex] 0 & 0 & \bigl[E[{(\mathbf{h}^H\Re{\mathbf{s'}_n}){(\mathbf{h}^H\Re{\mathbf{s'}_n})^H}]\bigr]}^{-1} & 0 & 0\\[3ex] 0 & 0 & 0 & 2\sum_{n=0}^{N-1}{E[{(\mathbf{h}^H\Im{\mathbf{s'}_n}){(\mathbf{h}^H\Im{\mathbf{s'}_n})}^H}]}^{-1} & 0 \\[3ex] 0 & 0 & 0 & 0 & D \end{psmallmatrix}$$
\end{figure*}
\end{document}


Since your example occupies much of the page I would suggest to use placeholders for those big formulas and put them below:

\documentclass{IEEEtran}
\usepackage{amsmath,mathtools,amsfonts,amssymb,bm}
\begin{document}
\begin{figure*}
$$A(\boldsymbol{\theta}^* - \hat{\boldsymbol{\theta}^*}) > K \begin{pmatrix} E[A]^{-1} & 0 &0 & 0 & 0 \\ 0 & E[B]^{-1} & 0 & 0 & 0 \\ 0 & 0 & \alpha & 0 & 0\\ 0 & 0 & 0 &\beta & 0 \\ 0 & 0 & 0 & 0 & D \end{pmatrix}$$
\begin{align*}
\text{where:}& \\
K&~\text{is a constant}\\
\alpha&=\bigl[E[{(\mathbf{h}^H\Re{\mathbf{s'}_n}){(\mathbf{h}^H\Re{\mathbf{s'}_n})}^H}]\bigr]^{-1}\\
\beta&=2\sum_{n=0}^{N-1}{E[{(\mathbf{h}^H\Im{\mathbf{s'}_n}){(\mathbf{h}^H\Im{\mathbf{s'}_n})}^H}]}^{-1}
\end{align*}
\end{figure*}
\end{document}


Here's a solution that uses a pmatrix environment along with the bm package. Note that I've replaced

A(\mathbf{\theta^* - \hat{\theta^*}})


(which, in any case, doesn't work) with

A(\bm{\theta}^* - \hat{\bm{\theta}}^*)


I've also gotten rid of lots of unneeded pairs of curly braces.

\documentclass{IEEEtran}
\usepackage{amsmath,amssymb,bm}
\begin{document}
\begin{figure*}
\setlength\arraycolsep{4pt} % default: 5pt
$$A(\bm{\theta}^* - \hat{\bm{\theta}}^*) > \text{constant} \begin{pmatrix} E[A]^{-1} & 0 & 0 & 0 & 0 \\[1.5ex] 0 & E[B]^{-1} & 0 & 0 & 0 \\[1.5ex] 0 & 0 & \Bigl[E\bigl[(\mathbf{h}^H\Re{\mathbf{s}'_n}) (\mathbf{h}^H\Re{\mathbf{s}'_n})^H\bigr] \Bigr]^{-1} & 0 & 0\\[1.5ex] 0 & 0 & 0 & 2\sum_{n=0}^{N-1}E\bigl[(\mathbf{h}^H\Im{\mathbf{s}'_n}) (\mathbf{h}^H\Im{\mathbf{s}'_n})^H\,\bigr]^{-1} & 0 \\[1.5ex] 0 & 0 & 0 & 0 & D \end{pmatrix}$$
\end{figure*}
\end{document}


First I suggest you use the strip environment, from cuted (sttools bundle), so your equation won't float.

Second you don't need to use psmallmatrix, which makes its contents hard to read. Playing with \arraycolsep, you can use the normalpmatrix. Otherwise, one can use the \medmath command from nccmath, which types its contents in about 80 % of \displaystyle. Below is a demo of both.

Last, I removed a number of useless { } pairs.

\documentclass{IEEEtran}
\usepackage{mathtools, nccmath}
\DeclareMathOperator{\RE}{Re}
\DeclareMathOperator{\IM}{Im}
\usepackage{cuted}
\usepackage{lipsum}

\begin{document}
\lipsum[1-3]
\begin{strip}
\vspace*{-\stripsep}
$$\label{eq:1} \setlength\arraycolsep{2pt} A(\mathbf{\theta^* - \hat{\theta^*}}) > \text{constant}\cdot\! \begin{pmatrix} E[A]^{-1} & 0 &0 & 0 & 0 \\[1ex] 0 &\! E[B]^{-1}\! & 0 & 0 & 0 \\[1ex] 0 & 0 &\!\! \biggl[E\bigl[(\mathbf{h}^H\RE{\mathbf{s'}_n}){(\mathbf{h}^H\RE{\mathbf{s'}_n})}^H\bigr]\biggr]^{-1}\!\! & 0 & 0\\[1.5ex] 0 & 0 & 0 & \!\!\sum\limits_{n=0}^{N-1}E\bigl[(\mathbf{h}^H\IM{\mathbf{s'}_n}){(\mathbf{h}^H\IM{\mathbf{s'}_n})}^H\bigr]^{-1} & 0 \\[1ex] 0 & 0 & 0 & 0 & D \end{pmatrix}$$
\lipsum[4-5]
$$\label{eq:2} A(\mathbf{\theta^* - \hat{\theta^*}}) > \text{constant}\cdot\! \medmath{\begin{pmatrix} E[A]^{-1} & 0 &0 & 0 & 0 \\[1ex] 0 & E[B]^{-1} & 0 & 0 & 0 \\[1ex] 0 & 0 &\biggl[E\bigl[(\mathbf{h}^H\RE{\mathbf{s'}_n}){(\mathbf{h}^H\RE{\mathbf{s'}_n})}^H\bigr]\biggr]^{-1} & 0 & 0\\[1.5ex] 0 & 0 & 0 & \sum\limits_{n=0}^{N-1}E\bigl[(\mathbf{h}^H\IM{\mathbf{s'}_n}){(\mathbf{h}^H\IM{\mathbf{s'}_n})}^H\bigr]^{-1} & 0 \\[1ex] 0 & 0 & 0 & 0 & D \end{pmatrix}}$$
\vspace*{-\stripsep}
\end{strip}
\lipsum[1-3]

\end{document}


• The equation does not go at the bottom of the page, and It looks neat and also saves a lot of space. Thank you for this solution which I prefer over the others. However, I do find some problems which I have put in the updates part of the Question. Visually, on close inspection the colums are not coming aligned ? Also, waht is the utility of the command \setlength\arraycolsep{2pt} Please let me know how to rectify these.
– SKM
Jun 3 '17 at 20:29
• The value of the horizontal distance between columns is decreased to 2pt (the default is 6pt, if I remember well). This doesn't make the matrix look too tight, because there are large entries which never are on the same row. For the labels, I've added a label to each of the equations in my code. Last point: I have my columns bottom-aligned, but what you obtain might depend on the contents of the two columns above. Just in case, tryy the directive \raggedbottom in your preamble. Otherwise, could post a small code reproducing the problem? Jun 3 '17 at 20:43
• I have included the code. I have several such large matrices in my article and for each I have used your approach. And the same problem of alignment of the column above the equation arises. Please see my updates. thank you
– SKM
Jun 3 '17 at 21:55
• @SKM: There seems to be a problem with IEEEtran.cls: the same code, with article.cls and twocolumn` produces no misalignment. I'll try to understand why tomorrow, and if I can find a workaround. Jun 3 '17 at 23:51
• Ok, thanks for looking at the problem. I will wait for your answer :)
– SKM
Jun 4 '17 at 3:37