Physics Stack Exchange is a question and answer site for active researchers, academics and students of physics. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

Are the following operations O.K.? This is related to the Feynman parameter trick.

$$F:= \int_0^1 \mathrm{d}x\int_0^{1-x}\mathrm{d}y \frac{1}{f(x,y)+\mathrm{i}\epsilon}.$$ Now using

$$\frac{1}{A+i\epsilon} = PV\frac{1}{A}-i\pi\delta(A),$$ where $PV$ denotes the Cauchy Principal Value, we get (taking only the imaginary part):

$$\Im{F} = -\pi \int_0^1 \mathrm{d}x\int_0^{1-x}\mathrm{d}y\, \delta(f(x,y)) .$$

The trouble I got is that the zeros of $f(x,y)$ which I call $y^{\pm}$ seems to be outside integration range and hence the delta should yield zero. BUT here's what's funny: when I ignore all this and just perform the formal calculations (assuming I do it correctly) namely; replacing $\delta(f(x,y))$ with

$$\frac{1}{\bigl\vert \partial f/\partial y\bigr\vert_{y=y^{\pm}}}\times(\delta(y-y^-)+\delta(y-y^+)),\ \ \ (1)$$

(where $|\partial f/\partial y|_\pm $ are equal) and assuming that $y^{\pm}\in[0,1-x]$ (which seems to be false) the two deltas just give $1+1 = 2$. Then the result seems to be correct, or at least it agrees with what I have calculated the same thing using a totally different method.

Could this all just be a coincidence? I mean shouldn't the deltas produce zero if $y^{\pm}\notin[0,1-x]$, or I'm I using the wrong formula $(1)$?

share|cite|improve this question
Actually that's a much better title than we usually get. More informative and specific (to a point) is generally better when it comes to titles. – David Z Mar 10 '13 at 23:29
Eq. (1) only works if $|\partial f/\partial y|$ is the same for both $y^\pm$, otherwise you can't pull it out of the brackets. Don't know if this helps your problem. – Michael Brown Mar 11 '13 at 0:50
For readers unfamiliar with the equation in the question title, learn more at this Wikipedia page :-D – Steve B Mar 11 '13 at 0:56
Sorry i should add that it is the same for both. – The Noob Mar 11 '13 at 10:30
@The Noob: If you would like the community to help solve your apparent paradox, you would have to give the explicit form of $f(x,y)$. Right now, it is hard to conclude anything other than what have already been said by you in the question formulation (v4). – Qmechanic Mar 13 '13 at 17:28

If you are sure that $f$ is continuous and does not vanish in the integration domain, it is by no means necessary making use of regularization theory of distributions. Consider the initial integral:

$$F:= \int_0^1 \mathrm{d}x\int_0^{1-x}\mathrm{d}y \frac{1}{f(x,y)+\mathrm{i}\epsilon}.$$ It can be re-written as: $$F:= \int_{T} \frac{1}{f(x,y)+\mathrm{i}\epsilon} \mathrm{d}x\mathrm{d}y,$$ where $T$ is the closed triangle: $$ T := \{(x,y) \in [0,1]\times [0,1] \:|\: 0 \leq y \leq 1-x\}\:.$$

If $f(x,y)$ is continuous on $T$ and does not vanish therein, the function $$[0,1]\times T \ni (\epsilon, x,y) \mapsto \left|\frac{1}{f(x,y)+\mathrm{i}\epsilon}\right|$$ is continuous and thus bounded. Let us call $M\geq 0$ its maximum. We can conclude that $$\left|\frac{1}{f(x,y)+\mathrm{i}\epsilon}\right| \leq M\quad \mbox{for every $\epsilon \in [0,1]$ and $(x,y)\in T$.} $$ As $T$ has finite measure, the constant function $T \ni (x,y) \mapsto M$ has finite integral. We can thus apply Lebesgue's dominated convergence theorem, that permits to swap the symbol of limit with that of integral and getting this way: $$\lim_{\epsilon \to 0^+} \int_0^1 \mathrm{d}x\int_0^{1-x}\mathrm{d}y \frac{1}{f(x,y)+\mathrm{i}\epsilon} = \int_T \lim_{\epsilon \to 0^+}\frac{1}{f(x,y)+\mathrm{i}\epsilon} \mathrm{d}x\mathrm{d}y = \int_T \frac{1}{f(x,y)} \mathrm{d}x\mathrm{d}y $$

share|cite|improve this answer
Obviously, in my answer, I supposed $f$ to be real valued. – Valter Moretti Dec 11 '13 at 20:47

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.