# Appropriate Notation: $\equiv$ versus $:=$

With respect to assignments/definitions, when is it appropriate to use $\equiv$ as in

$$M \equiv \max\{b_1, b_2, \dots, b_n\}$$

which I encountered in my analysis textbook as opposed to the "colon equals" sign, where this example is taken from Terence Tao's blog :

$$S(x, \alpha):= \sum_{p\le x} e(\alpha p)$$

Is it user-background dependent, or are there certain circumstances in which one is more appropriate than the other?

-
The colon-equals should be used, if at all, only for definitions. I don’t use it; I think that it’s always pretty clear from context when an equals sign is a definition and when it’s a statement. (On the very rare occasions when I use a special symbol, I use $\triangleq$.) –  Brian M. Scott Aug 13 '12 at 15:20
Thanks for weighing in Prof. @BrianM.Scott –  KingOliver Aug 13 '12 at 15:25
As the book of Walter Rudin teaches, the best idea is always to write in the correct language what you are doing. Too many symbols mean a waste of time. Use words whenever possible. –  Siminore Aug 13 '12 at 15:41
@BrianM.Scott: The problem is, sometimes you're not sure if the author is making a local definition, or just using a symbol you're not familiar with (or a previous local definition you have since forgotten) and affirming an equality, which can be quite confusing. I sometimes wish people used $:=$ more often. In any event, the symbol is also used for variable assignment in Pascal and pseudocode (and maybe others, I'm not much of a programmer). –  tomasz Aug 13 '12 at 16:02
@AnonymousCoward see here: tex.stackexchange.com/questions/4216/how-to-typeset-correctly –  KingOliver Aug 14 '12 at 12:33

An "equality by definition" is a directed mental operation, so it is nonsymmetric to begin with. It's only natural to express such an equality by a nonsymmetric symbol such as $:=\, .\$ Seeing a formula like $e=\lim_{n\to\infty}\left(1+{1\over n}\right)^n$ for the first time a naive reader would look for an $e$ on the foregoing pages in the hope that it would then become immediately clear why such a formula should be true.

On the other hand, symbols like $=$, $\equiv$, $\sim$ and the like stand for symmetric relations between predeclared mathematical objects or variables. The symbol $\equiv$ is used , e.g., in elementary number theory for a "weakened" equality (equality modulo some given $m$), and in analysis for a "universally valid" equality: An "identity" like $\cos^2 x+\sin^2 x\equiv1$ is not meant to define a solution set (like $x^2-5x+6=0$); instead, it is expressing the idea of "equal for all $x$ under discussion".

-
"enforced equality" - i.e., an identity as opposed to a mere equation. –  Ｊ. Ｍ. Aug 13 '12 at 16:10
I guess the latter is what people have in mind when using $\equiv$ for definitions: The name is, by definition, equivalent to the expression in all and any circumstances. –  celtschk Aug 13 '12 at 16:58
@ J.M. and @celtschk: My schoolboy's English left me alone: What I actually meant was "universal". See my edit. –  Christian Blatter Aug 13 '12 at 17:14
Which doesn't affect the statement in my comment. –  celtschk Aug 13 '12 at 17:19
Whether "universal" or "enforced", it still works out. I was just pointing out the more customary synonym. :) –  Ｊ. Ｍ. Aug 14 '12 at 3:49

$x:=y$ means $x$ is defined to be $y$.

The notation $\equiv$ is also (sometimes) used to mean that, but it also have other uses such as $4\equiv0$ (mod 2).

I encountered $:=$ a lot more than $\equiv$ , and it is my personal favourite.

There is also the notation $\overset{\Delta}{=}$ to mean "equal by definition"

By the way, some people also use the notaion $x=:y$ to mean $y$ is defined to be $x$

-
Yeah, $:=$ and $=:$ have the benefit of being clearly asymmetrical, so there's no room for confusion and you can use them in reverse when needed... –  tomasz Aug 13 '12 at 15:59
@tomasz - I agree, that is one of the reasons this is my favourite. –  Belgi Aug 13 '12 at 16:05

It's entirely up to the whim of the author. Other symbols that can mean the same thing are $\triangleq$ and $=_{def}$. I think that only a minority of authors use any special notation, however; the majority just use a regular equals sign.

-
Sometimes, I even see $\stackrel{\text{def}}{=}$... –  Ｊ. Ｍ. Aug 13 '12 at 15:54
I prefer that notation, as the delta equals symbol seems to not be universal, and the colon equals symbol can get munged with assignment in Maple/Pascal, but this symbol is unambiguous. –  Arkamis Aug 13 '12 at 16:03
But many languages use $=$ for assignment. Should I therefore avoid $=$ in equations because it could be mistaken for a C/Java/Perl/Python/... assignment? –  celtschk Aug 13 '12 at 17:29

The notation $x:= y$ is preferred as $\equiv$ has another meaning in modular arithmetic (though it is almost always clear from context as to which is meant). However, there is one big advantage to using the $:=$. That is, it is not graphically symmetric and hence allows for strings such as $$y:= f(x) \leq g(x) =: L$$ where here we are defining both $y$ and $L$. This statement would be much more cumbersome using $\equiv$, and it would not make sense if one simply wrote $$y \equiv f(x) \leq g(x) \equiv L.$$

-

Upvoting the other answers and comments... and: conventions vary. The only way to know with reasonable confidence is from context. However, one can't know whether an author "believes in" setting context. The most important criterion may be whether or not one is tracking things well enough to reasonably infer which equalities are assignments, and which are assertions. If this seems to be an issue, likely one should back-track a bit, anyway.

In practice, assignment will be clear because the left-hand side is a single symbol (even if composed of several marks), and is appearing for the first time. The first-time-appearance criterion is obviously more easily applied if all first-time appearances are highlighted by a consistent convention (rather than being buring in-line, without emphasis or fanfare).

-

The $\equiv$ symbol has different standard meanings in different contexts:

• Congruence in number theory, and various generalizations;
• Geometric congruence;
• Equality for all values of the variables, as opposed to an equation in which one seeks the values that make the equation true;
• $x$ is defined to be $y$;
• probably a bunch of others.

But I suspect "$:=$" is not used for anything other than definitions.

So the latter at least avoids ambiguity. But if you're reading something written by someone who doesn't see it that way, you still want to understand what is being said, so you should be aware of usage conventions that you might reasonably consider less than optimal.

-