You are on page 1of 4

An a priori estimate for the semi-linear problem

In this chapter we will rst introduce a modied decomposition, and then use it to proof an a priori estimate. Let us rst redene the decomposition in respect to dual cones in a way that it will be easier to continue working with it. (u)+ = (u) (u)+ = 0 for x for x (u) = (u)+ (u)+ = 0 for x for x

(1)

We have that (u)+ , (u) H2 () and that (u)+ 0, (u) 0. The right sides were chosen in a way that for a u H2 () and u 0 it follows that (u) = 0 and vice versa. In the rst case we have from the second order maximum principle that if u 0 u 0 With the denition of (u) it follows that (u) = (u) = 0 (3) (2)

and therefore (u) H2 () = 0. The second case follows the same way. Let us now dene f + := max(f, 0) and f f + := f with f the right side from the semi-linear BVP. With these we can introduce two functionals corresponding to linear boundary value problems of fourth order. I1 (u) =

(u)2 f + udx (4) (u) f udx


2

I2 (u) =

With the results from the previous chapter we get the existence of u1 and u2

with I1 (u1 ) = min I1 (u) with u1 0


uH2 ()

I2 (u2 ) = min I2 (u) with u2 0.


uH2 ()

Theorem 1. If u H2 () fulls the Euler-Lagrange equation of the semilinear problem, then it holds that u2 u u1 . (5)

Proof. Since u H2 () and u1/2 H2 (), we have that ( u u1/2 ) H2 (). We can decompose ( u u1/2 ) in respect to dual cones ( u u1/2 ) = ( u u1/2 )+ + ( u u1/2 ) . In the next steps we will use that (u)+ u, (u) u and that (u)+/ = (u)/+ . We can write 0 ( u u1 )+ =
2 H2 ()

= ( u u1 )

2 H2 ()

( u u1 )

2 H2 ()

|( u u1 )|2 dx |( u u1 )|2 dx +

( u u1 ) ( u u1 ) dx ( u u1 ) (( u u1 ))+ dx ( u u1 )+ ( u u1 )dx

|( u u1 )2 |dx

(( u u1 ))(( u u1 ) ( u u1 ) )dx (2 ( u u1 ))(( u u1 ) + ( u u1 ) )dx

(6)

= =

(f u |u |p1 f + )(( u u1 )+ )dx (f u |u |p1 )(( u u1 )+ )dx

Suppose u u1 0. Then we have that ( u u1 )+ 0 f 0 u |u |p1 0 . Now we have 0 ( u u1 )+


2 H2 ()

(7)

and it follows that ( u u1 )+ 0. This also means that ( u u1 )+ = 0 = (( u u1 ))+ and we can write ( u u1 ) = |( u u1 )|. (8)

Since ( u u1 ) 0, we write |( u u1 )| = ( u u1 ) and arrive at ( u u1 ) 0. With ( u u1 ) H2 () = 0 we get that u = u1 almost everywhere. We supposed that u u1 and this yields us an upper bound for u . The lower bound follows similarly. We can write 0 ( u u2 ) =
2 H2 ()

= ( u u2 )

2 H2 ()

( u u2 )+

2 H2 ()

|( u u2 )|2 dx + |( u u2 )|2 dx +

( u u2 )+ (( u u2 )+ )dx ( u u2 )+ (( u u2 )) dx ( u u2 )+ ( u u2 )dx

|( u u2 )|2 dx +

(( u u2 ))(( u u2 ) + ( u u2 )+ )dx (2 ( u u2 ))(( u u2 ) + ( u u2 )+ )dx

(9)

= =

(f u |u |p1 f )(2( u u2 )+ + ( u u2 ) )dx (f + u |u |p1 )(2( u u2 )+ + ( u u2 ) )dx

Suppose u u2 0. Then we have that ( u u2 ) 0 ( u u2 )+ = 0 f+ 0 u |u |p1 0 . Now we have 0 ( u u2 )


2 H2 ()

(10)

and it follows that ( u u2 ) 0. This also means that ( u u2 ) = 0 = (( u u2 )) and with the same reasoning as above, we arrive at u = u2 a.e. (11)

We have found that u also has a lower bound. Since c1 := u1 L () < and c2 := u2 L () < we arrive at the desired a priori estimate. u
L ()

< max(c1 , c2 )

(12)

You might also like