enter image description here

Evalute the definite integral:

$$\int_{0}^{3} (x^2+1) d[x]$$

$[x] -$ is integer part of the $x$.

Is the solution correct?

  • 1
    $\begingroup$ You did not evaluate the definite integral but the indefinite integral. $\endgroup$ – Mohammad Zuhair Khan Dec 26 '17 at 17:00
  • $\begingroup$ I get answer $17$ $\endgroup$ – MathUser Dec 26 '17 at 17:02
  • $\begingroup$ Me too assuming your evaluation was right. $\endgroup$ – Mohammad Zuhair Khan Dec 26 '17 at 17:06

The result depends on the definition of "integrating with respect to $[x]$."

One possible definition of the integral at stake is as follows: $[x]$ defines a measure on the real line. The measure is concentrated on the integers. For example $[x](\{2\})=1$, $[x](\{1,2\})=2$, but $[x]((1,2))=0$. So $$\int_A f(x)d[x]=\sum_{i\in A}f(i).$$ In our special case

$$\int_0^3 x^2+1\ d[x]=\int_{[0,3]}x^2+1\ d[x]=1+2+5+10=18.$$

If we mean integrating over $[0,3]$ when we write $\int_0^3$.

Or the result is $17$ if we have in mind $(0,3]$ when we write $\int_0^3$.

Interpretation as a Lebesgue-Stieltjes integral

According to the formal definition of the Lebesgue-Stieltjes integral the result is $17$. Why? Because the measure generated by the "integrator" is defined as follows: First we define the measure for an interval the following way:



Interpretation as a Riemann-Stieltjes integral:

By the definition we need a partition of the integrating domain $$\mathbb P=\{x_0=0<x_1<x_2<\cdots<x_n=3\}$$ then the approximating sum is

$$S(\mathbb P,x^2+1,[x])=\sum_{i=0}^{n}(c_i^2+1)([x_{i+1}]-[x_i])$$ where $c_i\in [x_{i+1},x_i].$ The Riemann-Stieltjes is said to be existing if (roughly speaking) the limit of $S$ exists when the length of even the wider interval in the partition above tends to zero.

The limit exists and it equals $17$ (not $18$) because the "integrator differences" $[x_{i+1}]-[x_i]$ are all zeros if there is no integer in the interval $[x_i,x_{i+1}]$ and they are all one otherwise. Within $[0,3]$ there are four integers: $0,1,2,3$. $0$, however, even if it is in $[0,x_1]$, does not count because the corresponding "integrator increment" $[x_{1}]-[0]=0$ (if the partition is dense enough), while $3$ contributes since $[3]-[x_{n-1}]=1$ (if the partition is dense enough. So contribute $1$ and $2$.

The OP formally applies the method of integration by part with the following choice $$\frac{d[x]}{dx}=v', x^2+1=u, \text{ and }\int_0^3 uv'\ dx=[uv]_0^3-\int_0^3u'v\ dx.$$ From here, following the formal process, we get $$\int_0^3 x^2+1\ d[x]=\big[[x](1+x^2)\big]_0^3-2\int_0^3x[x]\ dx.$$

So far so good. But then there is a sign mistake (circled in red) and then I cannot follow the reasoning.

Integrating by part

The method exists:

$$\int_0^3(x^2+1)\ d[x]=(3^2+1)[3]-(0^2+1)[0]-\int_0^3 [x]\ df(x)=$$ $$=30-\int_0^3[x]2x\ dx=30-2\left(\int_0^1 0\ x\ dx+\int_1^2 \ 1\ x \ dx+\int_2^3\ 2 x\ dx\right)=$$ $$=30-2\left(0+\frac32+5\right)=17.$$


Now, I know that this is what the OP did. But there was a sign mistake unfortunately. Thanks to the OP; I learned a lot.

  • $\begingroup$ I think the first method is correct. Is it right to apply integration by parts here? $\endgroup$ – Rick Dec 27 '17 at 2:15
  • 2
    $\begingroup$ @Rick: en.wikipedia.org/wiki/… $\endgroup$ – zoli Dec 27 '17 at 8:57
  • $\begingroup$ wow that's nice. I didn't know about that Stieltjes part before. $\endgroup$ – Rick Dec 27 '17 at 14:56

Your answer should have been a real number. Notice that the integer part $[x]$ is constant on $(0,1)$ and $(1,2)$ and $(2,3)$ which result in $d[x]=0$ on these intervals. Therefore the value of your definite integral is $0$.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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