Find the set of all functions that satisfy the inequality

Join the discussion
Registration is free. Start your own thread to ask a follow-up.
9 replies · 2K views
docnet
Messages
796
Reaction score
487
Homework Statement
please see below
Relevant Equations
please see below
Problem:

Find the set of all harmonic functions ##u(x,y,z)## that satisfy the following inequality in all of ##R^3##

$$|u(x,y,z)|\leq A+A(x^2+y^2+z^2)$$
where ##A## is a nonzero constant.

Work:

I removed the absolute value bars by re-writing the expression
$$-C-C(x^2+y^2+z^2)\leq u\leq C+C(x^2+y^2+z^2)$$

I am looking for a function ##u## that satisfies ##\Delta u = 0## in ##R^3##, because we can take ##A## as large as needed to satisfy the inequality for any ##u##

List of solutions: all harmonic functions, too many to list.

I suppose one obvious solution is obtained by setting ##\leq \Rightarrow =##

$$\pm u(x,y,z)=C+C(x^2+y^2+z^2)$$

I think the solution is supposed to be in abstract form to describe many types of functions. I don't understand why the term on the right side of the inequality is ##C+C(x^2+y^2+z^2)##. There must be a clue that I am oblivious to. Thank you in advance for any advice.
 
Physics news on Phys.org
Hi. I find it an interesting question.
u(x,y,z) has no singularity except infinite far ends.
As an obvious example say
[tex]u(x,y,z)=A e^{x+iy}[/tex], u is harmonic function,
[tex]|u|=A[/tex]
which satisfies the condition. Similar for ##Ae^{y+iz}## and ##Ae^{z+ix}##.
 
Last edited:
  • Like
Likes   Reactions: docnet
mitochan said:
Hi. I find it an interesting question.
u(x,y,z) has no singularity except infinite far ends.
As an obvious example say
[tex]u(x,y,z)=A e^{x+iy}[/tex], u is harmonic function,
[tex]|u|=A[/tex]
which satisfies the condition. Similar for ##Ae^{y+iz}## and ##Ae^{z+ix}##.

[itex]|Ae^{x + iy}| = Ae^x[/itex]. Does that satisfy the bound as [itex]x \to \infty[/itex]?

Consider solutions of the form [itex]u = e^{\mathbf{k} \cdot \mathbf{x}}[/itex] where [itex]\mathbf{k} = (k_1,k_2,k_3) \in \mathbb{C}^3[/itex]. This is harmonic if and only if [itex]k_1^2 + k_2^2 + k_3^2 = 0[/itex]. So either [itex]\mathbf{k} = 0[/itex] and [itex]u[/itex] is constant, or you can find a [itex]\mathbf{v} \in \mathbb{R}^3[/itex] such that [itex]\mathbf{k} \cdot \mathbf{v} > 0[/itex]. Consider [itex]|u(t\mathbf{v})|[/itex] as [itex]t \to \infty[/itex]. Does this satisfy the bound?

(EDIT: It is only necessary that [itex]\operatorname{Re}(\mathbf{k})\cdot \mathbf{v} > 0[/itex].)

The other solutions are polynonial solutions of the form [tex]u = a + \mathbf{b} \cdot \mathbf{x} + \mathbf{x}^T C \mathbf{x}[/tex] for some scalar [itex]a[/itex], vector [itex]\mathbf{b}[/itex] and symmetric matrix [itex]C[/itex].
 
Last edited:
  • Like
Likes   Reactions: docnet, mitochan and etotheipi
@pasmith

Thank you for your comment, may I ask a few questions about your comment?

pasmith said:
|Aex+iy|=Aex. Does that satisfy the bound as x→∞?

I think [itex]|Ae^{x + iy}|[/itex] does not satisfy the bound because exponential functions increase faster than quadratic functions as [itex]x \to \infty[/itex]. is this true?

pasmith said:
Consider solutions of the form ##u=e^{k⋅x}## where ##k=(k_1,k_2,k_3)∈C^3##. This is harmonic if and only if ##k_1^2+k_2^2+k_3^2=0##. So either ##k=0## and ##u## is constant, or you can find a ##v∈R^3## such that ##k⋅v>0##. Consider ##|u(tv)|## as ##t→∞##. Does this satisfy the bound?

Could you please briefly explain how ##Re(k)⋅v>0## makes a harmonic ##u##?

When ##u=e^{k⋅v}## where k and v are both vectors, isn't ##u## a constant which is included in the polynomial solution?

For ##|u(tv)|##, doesn't this restrict our domain from ##R^3## to lines spanned by ##v##? I'm pretty confused by this and I will be happy if you can give me any explanation at all. Thank you.
 
docnet said:
is this true?
Yea, it's true.[tex]k_1^2+k_2^2+k_3^2=0[/tex]
means at least one of them has real parts, say ##k_1##, which makes ##e^{k_1x_1} > A+A(x^2+y^2+z^2)## for large x_1 or -x_1.
 
Last edited:
  • Like
Likes   Reactions: docnet
docnet said:
@pasmith

Thank you for your comment, may I ask a few questions about your comment?

I think [itex]|Ae^{x + iy}|[/itex] does not satisfy the bound because exponential functions increase faster than quadratic functions as [itex]x \to \infty[/itex]. is this true?

Yes.

Could you please briefly explain how ##Re(k)⋅v>0## makes a harmonic ##u##?

When ##u=e^{k⋅v}## where k and v are both vectors, isn't ##u## a constant which is included in the polynomial solution?

For ##|u(tv)|##, doesn't this restrict our domain from ##R^3## to lines spanned by ##v##? I'm pretty confused by this and I will be happy if you can give me any explanation at all. Thank you.

The condition for [itex]e^{\mathbf{k} \cdot \mathbf{x}}[/itex] to be harmonic is that [itex]\sum k_i^2 = 0[/itex].

You are asked to find functions which satisfy [itex]|u(x,y,z)| \leq A(1 + x^2 + y^2 + z^2)[/itex] for all [itex](x,y,z) \in \mathbb{R}^3[/itex]. If this bound fails to hold at even a single point, [itex]u[/itex] is not one of the functions you are looking for.

In this case, if [itex]\mathbf{k}[/itex] is not zero then you can find a (non-zero) [itex]\mathbf{v} \in \mathbb{R}^3[/itex] such that [itex]\operatorname{Re}(\mathbf{k}) \cdot \mathbf{v} > 0[/itex]. This guarantees that [tex] |u(\mathbf{v}t)| = |\exp(\operatorname{Re}(\mathbf{k}) \cdot \mathbf{v} t)| > <br /> A(1 + \|\mathbf{v}\|^2t^2)[/tex] as [itex]t \to \infty[/itex]. This means that [itex]u(x,y,z) \leq A(1 + x^2 + y^2 + z^2)[/itex] doesn't hold on all of [itex]\mathbb{R}^3[/itex] for this choice of [itex]u[/itex].
 
  • Like
Likes   Reactions: docnet
pasmith said:
The other solutions are polynonial solutions of the form for some scalar , vector and symmetric matrix .

Following this way for ##ax^2,by^2,cz^2## terms to be harmonic
[tex]a+b+c=0[/tex]
for general u(x,y,z) of
[tex]u(x,y,z)=ax^2+by^2+cz^2+dxy+eyz+fzx+g[/tex]
In order to satisfy the magnitude condition roughly
[tex]|h|\leq A, h=\{a,b,c,d,e,f,g\}[/tex]
I have not gone into detailed investigation.
 
[EDIT]
I forgot x,y,z terms
[tex]u(x,y,z)=A(ax^2+by^2+cz^2+dxy+eyz+fzx+gx+hy+iz+j)[/tex]
where
[tex]a+b+c=0[/tex]
Rough estimation is
[tex]|k| \leq 1, k=\{a,b,c,d,e,f,g,h,i,j\}[/tex]
 
For simple 1 dimension case ##u(x)=A(ax^2+bx+c)##, a=0,
[tex]|bx+c|\leq 1+x^2[/tex]
[tex]b^2+c^2 \leq 1[/tex]

Two dimension case ##u(x,y)=A(ax^2-ay^2+bxy+cx+dy+e)##,
[tex]|ax^2-ay^2+bxy+cx+dy+e|<1+x^2+y^2[/tex]
Is is not easy.
 
Last edited:
  • Like
Likes   Reactions: docnet
pasmith said:
The other solutions are polynonial solutions of the form u=a+b⋅x+xTCx for some scalar a, vector b and symmetric matrix C.

May I ask how you need C to be symmetric?

I believe the requirement on C can be reduced to the following:

##u(x_1,x_2,x_3)=a+\vec{b}⋅x+x^tCx##

where ##x^t=(x_1, x_2, x_3)## and ##C∈Mat(3;R)## with ##Tr(C)=0##

mitochan said:
For simple 1 dimension case ##u(x)=A(ax^2+bx+c)##, a=0,
[tex]|bx+c|\leq 1+x^2[/tex]
[tex]b^2+c^2 \leq 1[/tex]

Two dimension case ##u(x,y)=A(ax^2-ay^2+bxy+cx+dy+e)##,
[tex]|ax^2-ay^2+bxy+cx+dy+e|<1+x^2+y^2[/tex]
Is is not easy.

Thank you, your posts are amazing. I believe you can avoid the work of setting the limits on coefficients, because you can make the assumption that A can be taken arbitrarily large, larger than any constant. so the solution is simpler. The cross-terms can be expressed by the matrix C, so we avoid doing even more writing.