6

A few of us over on MITx have noticed that $\int f(x) dx $ is appearing as $\int dx f(x)$.

It's not the maths of it that worries me. It's just I recently read a justification (analytical?) of the second form somewhere but can't recall it or where I saw it.

Can anyone give me a reference?

AppliedImagination
  • 149
  • 2
    It's just notation. – gspr Jun 21 '12 at 13:46
  • 1
    It's just a different notational convention (common in physics). – anon Jun 21 '12 at 13:46
  • Yes, but there was a reason for it being a preferred notation... – AppliedImagination Jun 21 '12 at 13:47
  • 1
    One "justification" I've heard is that if you consider integration as a function taking, say, integrable functions to real number, then writing $\int \mathrm{d}x f$ sort of looks like "applying the function $\int \mathrm{d}x$ to $f$". I don't like the argument myself, but it's one I've heard at least. – gspr Jun 21 '12 at 13:48
  • I've "reasoned" similarly to myself: $\int dx$ is an operator; putting the dx after the operand function is just using it as a bracket... But I don't think that's the justification I saw. – AppliedImagination Jun 21 '12 at 13:53
  • 1
    You can also think of integrating as a kind of sum over these infinitesimal products $f(x)dx$. So then interchanging the factors seems natural, i.e. then both orders are regarded as equivalent. Especially if you have many different variables and dimensions it is prudent to see at the beginning with respect to which you are integration. So you don't first have to match all integral signs with their respective $d$s. – bodo Jun 21 '12 at 13:53
  • @canaaerus, that's a good, practical point I hadn't thought of. – AppliedImagination Jun 21 '12 at 13:57
  • 1
    The argument I heard and which is quite convincing is that in physics $f(x)$ can have very long form. So in order not to forget about $dx$ we write it first :) – qoqosz Jun 21 '12 at 14:04
  • Brilliant! You're not a physicist are you? :-) – AppliedImagination Jun 21 '12 at 14:06
  • @AppliedImagination actually I study physics :) – qoqosz Jun 21 '12 at 14:17
  • 2
    See http://math.stackexchange.com/questions/128108. – joriki Jun 21 '12 at 14:24
  • 1
    see also: http://math.stackexchange.com/q/5230 – t.b. Jun 21 '12 at 14:30
  • 1
    Thanks, t.b. and joriki. I'd turned up these posts, but neither of them or their answers give a rationale for preferring one version or another, and that's what I'm after. – AppliedImagination Jun 21 '12 at 14:38

1 Answers1

14

The second form sometimes makes it easier for the reader to match variables of integrations with their limits. Compare $$ \int_0^1\int_{-\infty}^{\infty}\int_{-\eta}^{\eta}\int_{0}^{|t|} \Big\{\text{some long and complicated formula here}\Big\}\,ds\,dt\,d\zeta\,d\eta $$ and $$ \int_0^1 d\eta\int_{-\infty}^{\infty}d\zeta\int_{-\eta}^{\eta}dt\int_{0}^{|t|} ds\,\Big\{\text{some long and complicated formula here}\Big\} $$