Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. 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

Please ,how to find that the general solution of $u'''(t)=e(t) , t\in [0,1]$ is given by $u(t)=c_0+c_1t+c_2 t^2 +\frac12 \int_0^t (t-s)^2 e(s) ds$

$e:(0,1)\rightarrow \mathbb{R}$, and $e\in L(0,1)$.

i think that it is general homogeneous solution + particular solution ,

the general homogeneous is $c_0+c_1 t+ c_2 t^2$ , but i dont know how to finde that the paricular solution is $\frac12 \int_0^t (t-s)^2 e(s)ds $


Thank you

share|cite|improve this question

You can find the particular solution by doing an integration by parts.

$\frac12 \int_0^t (t-s)^2 e(s)ds = [0-0] -2* \frac12 \int_0^t (t-s) E(s)ds $ where E(s) is a primitive of $e(s)$ such that $E(0)=0$

Similarly we have : $-\int_0^t (t-s) E(s)ds = - ( [0-0] - \int_0^t F(s)ds)$ where $F(s)$...

And u(t)=$\int_0^t F(s)ds$ is a solution of $u'''(t)=e(t)$

EDIT, to be clearer :

u(t)=$\int_0^t u'(s)ds + u(0)$

u(t)=$-\int_0^t (t-s) u''(s)ds + u'(0)t + u(0)$

$u(t)=\frac12 \int_0^t (t-s)^2 u'''(s) ds + u''(0)/2*t^2 + u'(0)t + u(0) = \frac12 \int_0^t (t-s)^2 e(s) ds + u''(0)/2*t^2 + u'(0)t + u(0) $

share|cite|improve this answer
But if i dont know that the particular solution is $\frac12 \int_0^t (t-s)^2 e(s)ds$ , how to do ? – Vrouvrou May 22 '13 at 19:19
please @gvo thank you – Vrouvrou May 22 '13 at 19:26
You can use integration by parts from u(t) and replace F with u', E with u'' and then e with u''' in my equations, and go from the last one to the first. The choice of (t-s) as a primitive of 1 would come from fact that it worth 0 when s=t. – gvo May 22 '13 at 22:02
in the last line the integral is from 0 to t. please what i must write in the beginig : $\displaystyle u(t)=\int_0^t \int_0^{\alpha}\int_0^{\beta} u'''(s) ds d{\beta}d{\alpha}$ ?? thank you – Vrouvrou May 23 '13 at 15:28
Corrected, bad copy paste. I don't really understand your question, but it seems related to the answer of anon. – gvo May 24 '13 at 9:54

It suffices to see that the strange-looking integral is a particular solution. You probably came up with the following type of expression, or at least you should agree it is an obvious way to go:

$$u_p(t)=\int_0^t\int_0^v\int_0^u e(s)dsdudv.\tag{1}$$

Now this is equal to

$$\int_0^te(s)\cdot{\rm Area}(\{(u,v):s\le u\le v\le t\})ds=\int_0^t\frac{(t-s)^2}{2}e(s)ds. \tag{2}$$

How did we get this? First off, the region of points $(u,v)$ such that $s\le u\le v\le t$ (where $s$ and $t$ are fixed) is a right triangle (try plotting some examples to see this) with sides each of length $t-s$, so its area is $\frac{1}{2}(t-s)^2$. But the more substantial formula is the following:

$$\iint\cdots\int_D f(x_0)dx_0dx_1\cdots dx_n=\int f(x_0)\cdot{\rm Vol}(\{(x_0,x_1,\cdots,x_n)\in D\})dx_0 \tag{3}$$

(under suitable hypotheses most likely). This is a "continuous" generalization of a discrete version

$$\sum_{(x,y)\in A}f(x)=\sum_{x\in X}f(x)\,\#\{(x,y)\in A\} \tag{4}$$

(Where $A\subseteq X\times Y$.)

Anyway I presume most of the above is not relevant to you. If you want a way to derive $(2)$ from scratch without knowing ahead of time what to look for, you will need to familiarize yourself with a technique to change double integrals like $\int_0^v\int_0^u e(s)dsdu$ into single integrals (using by-parts integration), and apply it twice to go from $(1)$ to $(2)$.

If you just want to check that $(2)$ is a particular solution, then you can straight-up differentiate the given function three times and check that the result is $e(t)$, using the general formula

$$\frac{d}{dt}\int_0^t f(t,s)ds=f(t,t)+\int_0^tf_t(t,s)ds\tag{5}$$

(and this formula can be derived using the chain rule + fundamental theorem of calculus).

Going from $(1)$ to $(2)$ with by-parts integration: Alright, first let's look at

$$\int_0^v\int_0^u e(s)dsdu. \tag{6}$$

Use by-parts ($X=\int_0^u e(s)ds$ and $Y=u$) to get

$$\int_0^v XdY=[XY]_0^v-\int_0^v YdX=v\int_0^v e(s)ds-\int_0^v ue(u)du=\int_0^v(v-s)e(s)ds. \tag{7}$$

Thus (using by-parts again)

$$\int_0^t\int_0^v\int_0^ue(s)dsdudv=\int_0^t\int_0^v(v-s)e(s)dsdv \tag{8}$$

$$=\int_0^tv\int_0^ve(s)dsdv-\int_0^t\int_0^vse(s)dsdv \tag{9}$$

($dY=vdv$ and $X=\int_0^ve(s)ds$ in the first integral, same by-parts as $(6)$-$(7)$ in second integral)

$$=\left[\frac{t^2}{2}\int_0^te(s)ds-\int_0^t\frac{v^2}{2}e(v)dv\right]-\left[\int_0^t (t-s)se(s)ds\right] \tag{10}$$

$$=\int_0^t\frac{t^2-s^2-2(t-s)s}{2}e(s)ds=\int_0^t\frac{(t-s)^2}{2}e(s)ds. \tag{11}$$

Going from $(1)$ to $(2)$ with reparametrization: The region of integration in ${\bf R}^3$ is

$$D=\{(s,u,v):0\le s\le u\le v\le t\}. \tag{12}$$


$$\int_0^t\int_0^v\int_0^ue(s)dsdudv=\iiint_D e(s)dV=\int_0^t\int_s^t\int_u^te(s)dvduds \tag{13}$$

$$=\int_0^t e(s) \left(\int_s^t\int_u^t 1dvdu\right)ds=\int_0^t\frac{(t-s)^2}{2}e(s)ds. \tag{14}$$

Simply checking that the integral expression is a particular solution: differentiating once,

$$\frac{d}{dt}\int_0^t\frac{(t-s)^2}{2}e(s)ds=\frac{(t-t)^2}{2}e(t)+\int_0^t(t-s)e(s)ds. \tag{15}$$

Differentiating a second time,

$$\frac{d}{dt}\int_0^t(t-s)e(s)ds=(t-t)e(t)+\int_0^te(s)ds. \tag{16}$$

Differentiating a third time,

$$\frac{d}{dt}\int_0^te(s)ds=e(t). \tag{17}$$

Hence $u(t)=\int_0^t\frac{(t-s)^2}{2}e(s)ds$ satisfies $u'''(t)=e(t)$.

Proof of $(5)$: let $G(u,t)$ be such that $\frac{dG}{dt}(u,t)=f(u,t)$. Then

$$\frac{d}{dt}\int_0^tf(t,s)ds=\frac{d}{dt}G(t,t)=\frac{dG}{d\,{\small\rm 1st\,coord}}(t,t)+\frac{dG}{d\,{\small\rm 2nd\,coord}}(t,t) \tag{18}$$

$$=\int_0^tf_u(u,s)|_{u=t}ds+f(u,t)|_{u=t}=\int_0^tf_t(t,s)ds+f(t,t). \tag{19}$$

share|cite|improve this answer
Nice, I really like this geometrical way to solve this integral. – gvo May 24 '13 at 9:49
@gvo please do you know how to passe from (1) to (2) ? – Vrouvrou May 24 '13 at 11:43
We use Fubini to obtein (2) ??? please – Vrouvrou May 24 '13 at 16:22
@Vrouvrou I have updated my answer with how to go from (1) to (2) using repeated by-parts integration and reparametrization, as well as with how to differentiate (2) three times and get e(t) hence proving it is a particular solution. – anon May 24 '13 at 19:27
thank you , i just don't understand (7) what is dX please – Vrouvrou May 24 '13 at 20:56

The particular solution

$$u(t)=\frac{1}{2}\int^{1}_{0}(t-s)^2e(s)ds $$

is the Green's function solution $G(t,s)$ for this problem.

share|cite|improve this answer
ok ,and so what is the relation with my problem ? , please thank you – Vrouvrou May 22 '13 at 19:48

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.