A quick guide to help you write professional LaTeX docs.

This is a quick guide to help you write $\LaTeX$ in a professional way. Throughout this tutorial, I assume you're familiar with the basic $\LaTeX$ syntax and know how to do all the basic stuff, e.g., compiling, using itemize, enumerate environments, etc.

I'm not writing this guide to tell you how to write $\LaTeX$ correctly, instead, I'm trying to convey to you that you should write $\LaTeX$ just like how I write it so someone with lots of experience with $\LaTeX$ can't nitpick you.

You may notice that lots of commands I tend to use are rather long (usually means it's in plain $\TeX$ style), this is because I don't need to type out all these: I use hsnips in particular, which allows me to type a few self-defined characters and expend automatically. For a more advanced $\LaTeX$ setup like this, please see VSCode-LaTeX-Inkscape.

Also, this is the best way to see what's the difference compared to other $\TeX$ commands. If you want, you're welcome to just define your own command and then use it instead.

• Abstract
• Disclaimer
• Table of Content
• The Basic
  • Newline \\
  • Paragraph \par
• Math Environment
  • Math mode
  • Text in Math mode
  • Symbols
    • Semantic
    • Misused
    • Functions
    • Else
  • Brackets, Parentheses, etc
    • Automatic Sizing
    • Manual Sizing
  • Spacing
  • Limits
• Else
  • Continued Fractions
  • Self-defined Commands
  • Text Spacing
  • Quote
• Further Reading

I often see things like

This is one paragraph.\\

This is another.

The command \\ will produce a line break, i.e., will end the current line and start a new one. This is different from a paragraph, as the start of paragraphs is usually indented. So if we write

This is one paragraph.\\ This is still in the paragraph!

This is another.

the output will be

  This is one paragraph.
This is still in one paragraph!
  This is another.

Notice the indention at the start of the two paragraphs, while there is no indention on that new line since we only introduce a line break, which will not start a new paragraph. So basically, $\LaTeX$ will wrap text in adjacent lines as if they were part of the same paragraph while treating \n (i.e., newline symbol) as a sign for starting a new paragraph. So, with the following input:

This is 
one paragraph

This is another.

the output will be like this:

  This is one paragraph.
  This is another.

Hence, while \\ will force a new line and if you already have a blank line between paragraphs, \\ is redundant. And I don't recommend you manually insert a line break in your paragraph to do auto-adjusting.

Another confusing command is \par. I have seen something like

\par This is one paragraph.

\par This is another.

Although this has the same output as above, this is indeed wrong. The \par command is used to end a paragraph, not to start one. So \par is the same thing as leaving a blank line as we mentioned. If you really want to use \par, this is something you should write

This is one paragraph.\par This is another.

and the output will be

  This is one paragraph.
  This is another.

The reason why you can theoretically use \par to start a paragraph is that, this command will not be counted multiple times: If you are already in a new paragraph, \par will do nothing. For example,

This is one paragraph.\par\par\par\par This is another.

produces

  This is one paragraph.
  This is another.

And if you put \par in front,

\par This is one paragraph.\par This is another.

Since at the beginning of the document, we're already in a new paragraph, so the first \par will do nothing, result in the equivalent input:

This is one paragraph.\par This is another.

Hence this will produce the desired output. But again, I suggest you to use blank line to start new paragraph.

As you know, $...$ and $$...$$ are two commonly used commands to write a mathematical expression. For example, $e^{i\pi} + 1 = 0$ produces $e^{i\pi} + 1 = 0$, while $$e^{i\pi} + 1 = 0$$ produces

$$e^{i\pi} + 1 = 0.$$

Both $...$ and $$...$$ are fine but this is why you should avoid them: they are $\TeX$ commands, which are old and causing problems. A better way to write math equations is to use \(...\) and \[...\], which is supported by modern $\LaTeX$.

While equation* environment is also an good option for unnumbered equations, but I tend to keep things simple and unified. For numbered equations, equation and align environment should be used.

When you need to have some texts in your math equation, please use \text{...} to wrap the texts around. For example,

\[
  x_i \geq 0 \text{ for all } i,
\]

which produces

$$x_i \geq 0 \text{ for all }i$$

is a good example. Notice that you should always indent both sides of your text inside \text{}, otherwise you'll have something like

$$x_i\geq 0\text{for all}i,$$

which is not desirable.

This is because $\LaTeX$ will neglect any spacing in math mode. In other words, if you write something like $x y$, you'll just get $x y$ instead of $x\; y$. More about spacing in Math mode later.

Also, you should always think twice when choosing between \text{} or \mathrm{}. They render the same output, but it's always a good habit to keep your source code clean and semantically correct. A quick guide is that when writing text, use \text{}, when writing math shorthand, use \mathrm{} instead. For example, if a variable $u_{\text{up}}$ has a flag up, you should write u_{\text{up}} instead of u_{\mathrm{up}} since up is a text. But if you're doing an integral, say

$$\int x\,\mathrm{d}x,$$

you should write \int x \,\mathrm{d}x instead of \int x \,\text{d}x.

Most of the time, when you're writing a math symbol with your direct keyboard input, there's a corresponding command in $\LaTeX$ as well. Here is a short list of them:

• $\colon$ (\colon): Instead of writing f: X \to Y for $f:X\to Y$, write f\colon X \to Y instead.

• $\ldots$ (\ldots): This is a straightforward one. Instead of writing a_1, a_2, ... for $a_1, a_2, \ldots$, write a_2, a_2, \ldots instead.

  There's something called \cdots also, which is my personal preferred one. If you really want to know about ..., there are actually five types of them:

  Code	Output	Comment
  A_1, A_2, \dotsc	$A_1, A_2, \dotsc$	for dots with commas
  A_1 + \dotsb + A_N	$A_1 + \dotsb + A_N$	for dots with binary operators/relations
  A_1 \dotsm A_N	$A_1 \dotsm A_N$	for multipication dots
  \int_{a}^{b} \dotsi	$\int_{a}^{b} \dotsi$	for dots with integrals
  A_1\dotso A_N	$A_1\dotso A_N$	for other dots

  The above is the conventions suggested by American Mathematical Society. If you don't want to follow them strictly, just choose one of them and stick with it.

• $\coloneqq$, $\eqqcolon$ (\coloneqq, \eqqcolon): Another direct one. When you define a new symbol such as let $y\coloneqq x_1-x_2$, write y \coloneqq x_1 - x_2 instead of y := x_1 - x_2.

  You need to put \usepackage{mathtools} in your header, namely you need the mathtools package.

• $\gg$, $\ll$ (\gg, \ll): Use \gg for much greater than and \ll for much less than instead of directly using >> and <<. The former ones produce $\gg$, $\ll$, while the latter ones produce $>>$ and $<<$.

Also, some symbols are by default equivalent to your direct keyboard output.

• $\prime$ (\prime): x' is equivalent to x^\prime.

  There exists some pathological examples like ${x^\prime}^\prime$ v.s. ${x'}^{'}$. But for usual cases, $x'$ is equivalent to $x^\prime$, and indeed if you write x'', $\LaTeX$ renders this as x^{\prime\prime} as $x''$ and so on. I personally prefer to use ^\prime since in this way I get a full control of my output.

• $\ast$ (\ast): * is equivalent to \ast.

  There are also something called \star, which produces $\star$. In some cases, $x^\star$ may be desired. For me, I prefer to use \ast whenever I want to render $x^\ast$. This is because I generally regard * as an operator, which can mean convolution for example. And to mark a variable, I think ^\ast is a better touch.

There are also commands which produce very similar output (or even exactly the same), but with different semantics. To keep the source code clean, we mention some of them.

• \rightarrow v.s. \to ( $\rightarrow$ ): I use \to for mapping, e.g. $f\colon X\to Y$, while \rightarrow for all other cases.

• \leftarrow v.s. \gets ( $\leftarrow$ ): I use \gets when writing pseudocode, while \leftarrow for all other cases.

• \Rightarrow v.s. \implies ( $\Rightarrow$ v.s. $\implies$ ): I use \implies when writing proofs, while \Rightarrow for all other cases.

• \Leftarrow v.s. \impliedby ( $\Leftarrow$ v.s. $\impliedby$ ): I use \impliedby when writing proofs, while \Leftarrow for all other cases.

• \Leftrightarrow v.s. \iff ( $\Leftrightarrow$ v.s. $\iff$ ): Similarly, I use \iff when writing proofs, while \Leftrightarrow for all other cases.

  Since \implies, \impliedby, and also \iff are quite long, so I redefined them into their corresponding ones for a more compact look. But in the code, I still type \implies when writing proof. To redefine them, put

  \let\implies\Rightarrow
  \let\impliedby\Leftarrow
  \let\iff\Leftrightarrow
  

  into your preamble. This makes \implies, \impliedby and \iff shorter and in my opinion has a better look.

I sometimes see things like $\cup_{i=1}^{\infty} X_i$ instead of $\bigcup\nolimits_{i=1}^\infty X_i$. The latter one is preferred since \cup is a binary operator, so it should only be used when you want to write something like $A\cup B$. If you want to perform such an operation multiple times as in our example, use \bigcup instead. Below are some examples.

There are lots of functions that have their own commands, e.g., $\lg (x)$, $\log (x)$, $\sin (x)$, $\cos (x)$, etc. Please don't write lg(x), which produces $lg(x)$. Instead, write \lg(x) for $\lg(x)$. Similarly, the same for \log(x),\sin(x),\cos(x)` and so on.

There are two symbols I want to mention, the empty set symbol and l (ell).

• You can either write \emptyset or \varnothing to denote an empty set, which produces $\emptyset$ and $\varnothing$. I prefer the latter one, but choose whatever you want.
• On the other hand, it seems like not everyone knows the command \ell for producing a nice looking $\ell$, and instead, they simply type l, which produces $l$. Not sure whether they're intended, but this is something worth mentioning.

The most painful thing when I read a $\LaTeX$ document is when seeing something like

$$N \coloneqq \vert\sum\limits_{j=1}^\infty(\sum\limits_{i=1}^\infty X_{ij})\vert$$

with the source code being

\[
  N \coloneqq \vert \sum\limits_{j=1}^\infty ( \sum\limits_{i=1}^\infty X_{ij} ) \vert.
\]

The size of the absolute value and the parenthesis are still in the default size, while the formula being wrapped is much higher than the default size. Before we talk about how to resolve this sizing issue, we should first see the common commands which will cause this kind of problem.

• $\lvert \ldots \rvert$ (\lvert ... \rvert): This is a fun one since we often use this to denote the absolute value like $\lvert x \rvert$. In this case, write \lvert x \rvert instead of |x|.

  Indeed, there are something called \vert, which is synonym to |, and amsmath recommends to use \lvert ...\rvert for absolute value.

• $\lVert \ldots \rVert$ (\lVert ... \rVert or \| ... \|): For norm, write \lVert x\rVert instead of ||x|| for $\lVert x\rVert$.

  Notice that you can also write \Vert x \Vert or \| x |\ for $\Vert x\Vert$, though I still prefer \lVert x \rVert from the very same reason with \lvert ... \rvert.

• $\lbrack \ldots \rbrack$ (\lbrack ... \rbrack or [ ... ]): They are indeed equivalent, so I prefer [ ... ] for simplicity.

• $\langle\ldots\rangle$ (\langle ... \rangle): Please don't use <x, y> for things like inner product, use \langle x, y \rangle instead to produce $\langle x, y\rangle$.

• $\{\ldots\}$ (\{ ... \}): The usual set notation.

Let's see how we can fix this.

To automatically resize the brackets, and parentheses, we use \left...\right... to do this. For the above example, the resized formula should be

\[
  N \coloneqq \left\vert \sum\limits_{j=1}^\infty \left( \sum\limits_{i=1}^\infty X_{ij} \right) \right\vert
\]

which produces

$$N \coloneqq \left\vert\sum\limits_{j=1}^\infty\left(\sum\limits_{i=1}^\infty X_{ij}\right)\right\vert.$$

Interestingly enough, though this is already powerful, there are more commands that can be utilized. They are \left./\right. and \middle. A typical usage for \left. or \right. is when you want to automatically resize an operator which only appears on one side. For example:

$$\left.\frac{x^2}{2}\right\vert_0^1$$ with the source code being

\left. \frac{x^2}{2} \right|_0^1

And for \middle, you might have encountered the following situation:

$$\Delta^n\coloneqq \left\{(t_0, \ldots, t_n)\in\mathbb{R}^{n+1}\; |\; t_i\geq 0, \sum_{i=0}^{n}t_i=1\right\}$$

We see that $|$ is not being resized together with $\{ \}$. To do this, we add a \middle before | and get

$$\Delta^n\coloneqq \left\{(t_0, \ldots, t_n)\in\mathbb{R}^{n+1}\;\middle|\; t_i\geq 0, \sum_{i=0}^{n}t_i=1\right\}$$

as we desired.

Sometimes the automatic sizing may tend to be too big since it's trying to wrap everything inside, for example, $\vert \hat{x}^{(n)}i\vert$ is way better than $\left\vert \hat{x}^{(n)}{i}\right\vert$ while the latter uses \left\vert ... \right\vert and the former simply uses \vert ... \vert. In this case, uses \big, \Big, \bigg, and \Bigg instead of \left and \right as modifiers. For example,

\[
  ( \big( \Big( \bigg( \Bigg(
\]

produces

$$( \big( \Big( \bigg( \Bigg($$

A particularly important use case is that, when you use \underbrace, the automatic resizing will be much larger than expected. For example,

$$\mathbb{E}\left\lbrack \underbrace{\prod_{i=0}^\infty X_i}\right\rbrack\text{ v.s. }\mathbb{E}\bigg[ \underbrace{\prod_{i=0}^{\infty} X_i}\bigg],$$

where I use \left[ ... \right] on the left and \bigg[ ... \bigg] on the right. Notice that we even haven't written anything under the braces, and the left one is already ugly.

I have seen someone tried

$$\mathbb{E}\underbrace{\left\lbrack \prod_{i=0}^\infty X_i\right\rbrack}$$

to avoid the auto-resizing issue. Please don't do that, just don't.

Another use case is that \left( k g(x) \right) produces $\left( k g(x) \right)$, while \left and \right produce the same size delimiters as those nested within it. In this case, we can use \big( k g(x) \big), which produces $\big( k g(x) \big)$ to further distinguish the nested parentheses.

As we have seen before, when I type an indefinite integral, we have something like

$$\int x\,\mathrm{d}x$$

with the source code being \int x\,\mathrm{d}x. Notice that there's a \, before $\mathrm{d}x$, which gives us a small indent. since if we don't have this \,, we will end up with

$$\int x\mathrm{d}x,$$

where $x\mathrm{d}x$ is now a single entity rather than two independent ones. $\LaTeX$ provides several such commands to give you a small indent.

While \quad and \qquad are commonly used for spacing in every scenario, the above only works in math environments.

When using inline math environment, you'll often see $\sum\nolimits_{i=1}^\infty x_i$, which is the default behavior when typing \sum_{i=1}^\infty x_i. But if you type the same thing in display math mode, you'll get

$$\sum_{i=1}^\infty x_i$$

instead. You can indeed put the subscript and superscript below/on the summation symbol in inline math mode like $\sum\limits_{i=1}^\infty x_i$ by using \limits followed by \sum, i.e., \sum\limits_{i=1}^\infty x_i. This \limits command can be used in various scenarios, for examples, $\prod$, $\lim$, $\coprod$, or $\bigcup$ and $\bigcap$.

This indeed works for all big commands like $\int$, $\iint$, $\iiint$, $\iiiint$, $\oint$, $\idotsint$, $\bigodot$, $\bigoplus$, $\bigotimes$, $\bigvee$, $\bigwedge$, $\bigsqcup$, $\biguplus$. For example, $\bigoplus\nolimits_{i=1}^\infty G_i$ v.s. $\bigoplus\limits_{i=1}^\infty G_i.$

But notice that \limits works differently with integral signs: It'll put the upper-bound directly on top of the integral sign, and the same is done for the lower-bound. For example:

$$\int_{a}^{b} x\,\mathrm{d}x\text{ v.s. }\int\limits_{a}^{b} x\,\mathrm{d}x.$$

Notice that if you want to apply this to all your integral, please use \usepackage[intlimits]{amsmath} when loading amsmath package. This will only be applied when using integrals, since as mentioned, \limits with integral is treated differently.

One important thing that relates to sizing but is neglected by lots of people is the way of handling continued fractions. If we use \frac{}{} throughout, we'll have something like

$$x = a_0 + \frac{1}{a_1 + \frac{1}{a_2 + \frac{1}{a_3 + \frac{1}{a_4} } } }$$

with the code being

\[
  x = a_0 + \frac{1}{a_1 + \frac{1}{a_2 + \frac{1}{a_3 + \frac{1}{a_4} } } }
\]

which is ugly. Instead, we use \cfrac{}{}, where that extra c stands for continued. In this case, we have

$$x = a_0 + \cfrac{1}{a_1 + \cfrac{1}{a_2 + \cfrac{1}{a_3 + \cfrac{1}{a_4} } } },$$

with the code being

\[
  x = a_0 + \cfrac{1}{a_1 + \cfrac{1}{a_2 + \cfrac{1}{a_3 + \cfrac{1}{a_4} } } }.
\]

We see that with \cfrac{}{}, the equation is spaced more equally in the vertical direction, hence it's clearer.

Sometimes you may want to define your operators. For example, while there is a default \ker for producing the kernel of a function like $\ker(f)$, there is no default \im for the image of a function. To do this, we should use

\DeclareMathOperator{\im}{Im}

instead of

\newcommand{\im}{\mathrm{Im}}

since we'll have some spacing issues if we go with the latter one.

Unlike spacing in math mode, when writing text, there are just a few things to note. The $\LaTeX$ compiler uses a simple algorithm to determine whether you're going to end a sentence. Specifically, it looks at ., and see whether the character before it is a lowercase alphabet. So for example, when you write

  This is common, e.g. A, B, and C.

Then the compiler will think you're going to end the sentence right after e.g., so it will create an extra spacing for you. Other common situations are w.r.t. (with respect to), w.p. (with probability), and w.h.p. (with high probability). The way to fix this is to add \ after ., i.e., you may write

  This is common, e.g.\ A, B, and C.

In this way, you force the compiler to create a normal spacing after the period, fixing the problem.

It is typical that when you write a quote in $\LaTeX$, you will write something like

  "this is a quote"

in $\LaTeX$, this will render as $\text{"this is a quote"}$, where we get two backward on both the beginning of the quote and the end of the quote. Instead, you should write

  ``this is a quote''

which renders as $\text{``this is a quote''}$.

1. CS 209-Mathematical Writing
2. Wikibooks-LaTeX/Mathematics
3. Yiaho Liu: Introduction to LaTeX