3

As a follow-up to this related question, I'd like to know under what circumstances, if any, $\Delta x$, $\delta x$ and $dx$ all mean the same thing, and under what circumstances they can all be said to be approximately equivalent in a reasonably valid way.

For bonus points, it would also be nice to know where the inexact differential, $\text{đ}x$, can be used in place of one of the other symbols.

The reason that I ask this is that all of these symbols are commonly used as notation in thermodynamics and stat mech textbooks, and are often interchanged in ways that can be confusing to follow. In the math used in physics it's common to elide exactly correct statements with approximately correct ones, and it would be nice to have guidance when it comes to picking this sort of thing apart.

tel
  • 1,913
  • when you write $\delta x$, do you mean $\partial x$? – crf Apr 12 '14 at 02:51
  • @crf no. Here's an example of how $\delta x$ is used in the first chapter of Kardar's Statistical Physics of Particles: $\delta E \leq T\delta S + \bf{J}\cdot \delta \bf{x}$ – tel Apr 12 '14 at 03:09
  • 1
    @crf all I know for sure about $\delta x$ is "something something calculus of variations, waves-hands" – tel Apr 12 '14 at 03:11
  • 1
    $\Delta x$ always means a finite but usually small amount of change. $d x$ means an infinitesimal change (in the Leibniz sense of infinitesimal), you can regard it as some sort of limit as $\Delta x \to 0$. $\delta x$ is usually used like $\Delta x$ but for variation of a function instead of a number. – achille hui Apr 12 '14 at 03:12
  • @achillehui thanks, that makes it clearer. What I'd really like to know though is when can one be legitimately substituted for another – tel Apr 12 '14 at 03:35
  • They are the same thing when you are wearing a physicist's hat. $dx$ are definitely different from the other two when you are wearing a mathematician's hat. – achille hui Apr 12 '14 at 03:39
  • @tel Why? Why do you want to know when one can be legitimately substituted for another?.. And what if the answer is that they can't be substituted one for another? – Rustyn Apr 12 '14 at 03:40
  • 1
    @Rustyn if you could explain why they can't be legitimately substituted for one another, that would possibly be the best thing of all. I asked this question on the math SE instead of the physics SE on purpose. What I'd really like is some kind of exact comparison of the meaning of these various symbols to use as a guide for understanding these types of substitutions in the physics literature. – tel Apr 12 '14 at 03:49

1 Answers1

4

These letters are frequently used informally.

One thing to keep in mind, though, is that $\mathrm{d}x$ and $\Delta x$ generally refer to opposite sorts of things, although unfortunately the distinction is not often made clear.

$\mathrm{d}x$ is a gadget that tells you how the variable $x$ varies; in particular, it is not a number. If we have a relationship between three variables $z = f(x,y)$, then we know that we can express how $z$ varies in terms of how $x$ and $y$ vary:

$$\mathrm{d}z = f_1(x,y) \mathrm{d}x + f_2(x,y) \mathrm{d}y$$

$\Delta x$, however, tends to refer to a specific variation in $x$. This is actually a bad notation since it isn't just about $x$; e.g. if we are using $x,y$ coordinates on the plane, $\Delta x$ refers to a 'displacement' in which $x$ varies while $y$ is held constant. But if we were using $x, \bar{y}$ coordinates on the plane (where $\bar{y} = x + y$), then $\Delta x$ would refer to a 'displacement' in which in $x$ varies while $\bar{y}$ constant. These are two very different directions to be displaced.

But let's assume we have that problem sorted out. $\Delta x$ often means an actual difference in the value of $x$ at two different points. Going back to the previous example of $z = f(x,y)$, we have the differential approximation

$$ \Delta z \approx f_1(x,y) \Delta x + f_2(x,y) \Delta y$$

Pay attention to the difference between the previous version: the previous one had nothing to do with an actual displacement: it is describing a feature of how variations in $x,y,z$ are related in all possible ways they are allowed to vary. This one, however, it's an approximation between actual numbers coming from a single displacement.


Sometimes, $\Delta x$ means a differential geometry-style infinitesimal: i.e. a (tangent) vector (whereas $\mathrm{d}x$ would be a covector in this style; i.e. a vector of the opposite variance). In this case, $\Delta x$ and $\mathrm{d}x$ can be combined together to produce the number $1$, and $\Delta x$ combined with $\mathrm{d}y$ produces zero (assuming one particular way of sorting out the ambiguity of what $\Delta x$ implies we're supposed to be holding constant). It is more common, however, to use notations like $\frac{\partial}{\partial x}$ (or $\partial_x$) in this setting, though.

This notation is suggestive, because we can write equations like

$$f(P + \Delta P) = f(P) + f'(P) \Delta P$$

$P + \Delta P$ is a suggestive way to refer to a particular vector ($\Delta P$) located at the point $P$. With this meaning, the equation above becomes a literally true statement about applying differentiable functions to vectors, and this gives us a fairly direct way to think of vectors as referring to points "infinitesimally close" to the ordinary points.


Sometimes, $\Delta x$ is just an elaborate symbol for another variable that doesn't really have anything to do with $x$. e.g. we might see the formula for the derivative written as

$$ f'(x) := \lim_{\Delta x \to 0} \frac{f(x + \Delta x) - f(x)}{\Delta x} $$

which doesn't mean anything different from

$$ f'(x) := \lim_{h \to 0} \frac{f(x + h) - f(x)}{h} $$

Another example is the Taylor series formula

$$ f(x + \Delta x) = \sum_{n=0}^{\infty} f^{(n)}(x) \frac{(\Delta x)^n}{n!} $$

The use of the symbol $\Delta x$ rather than some other symbol is simply to help remind the reader how we are going to use it in formulas.