- #1

lavinia

Science Advisor

Gold Member

- 3,238

- 625

(xU[itex]_{x}[/itex]-yU[itex]_{y}[/itex]-U)/U =

(xV[itex]_{x}[/itex] - y V[itex]_{y}[/itex]+V)/V

- Thread starter lavinia
- Start date

- #1

lavinia

Science Advisor

Gold Member

- 3,238

- 625

(xU[itex]_{x}[/itex]-yU[itex]_{y}[/itex]-U)/U =

(xV[itex]_{x}[/itex] - y V[itex]_{y}[/itex]+V)/V

- #2

- 724

- 0

I'd like to know the answer myself. Does [itex]\frac{xU_x}{U}-\frac{xV_x}{V}=\frac{yU_y}{U}-\frac{yV_y}{V}[/itex] make it any easier? I've only dabbled in PDEs myself, but that seems like a separation of variables problem, when rephrased like that.

EDIT: Mathematica gives an error saying the answer is indeterminate, because there are more dependent variables than equations.

EDIT: Mathematica gives an error saying the answer is indeterminate, because there are more dependent variables than equations.

Last edited:

- #3

lavinia

Science Advisor

Gold Member

- 3,238

- 625

U = x V = K/x sort of works but doesn't allow the origin or the y axis.I'd like to know the answer myself. Does [itex]\frac{xU_x}{U}-\frac{xV_x}{V}=\frac{yU_y}{U}-\frac{yV_y}{V}[/itex] make it any easier? I've only dabbled in PDEs myself, but that seems like a separation of variables problem, when rephrased like that.

EDIT: Mathematica gives an error saying the answer is indeterminate, because there are more dependent variables than equations.

Also the answer is a little trivial.

- #4

MathematicalPhysicist

Gold Member

- 4,362

- 241

x d/dx ln (U) -y d/dy ln(U) -1 = xd/dx ln (V) -y d/dy ln(V) +1

x d/dx (ln(U/V))=y d/dy (ln(U/V))+2

Now make a guess of a function of the type U/V = exp(F(x,y))

to get:

x d/dx F = y d/dy F +2

so you have x d/dx F -y d/dy F =2

d/dx (xF) - d/dy (yF) = 2

Don't see how to solve this, though.

p.s the derivatives above are partial btw.

- #5

Ben Niehoff

Science Advisor

Gold Member

- 1,879

- 162

I would rewrite this as

(xU[itex]_{x}[/itex]-yU[itex]_{y}[/itex]-U)/U =

(xV[itex]_{x}[/itex] - y V[itex]_{y}[/itex]+V)/V

[tex]x \partial_x (\log U) - y \partial_y (\log U) - 1 = x \partial_x (\log V) - y \partial_y (\log V) + 1[/tex]

or

[tex]x \partial_x (\log U - \log V) - y \partial_y (\log U - \log V) = 2[/tex]

Define some function [itex]e^f = U/V[/itex] and you have

[tex]x \partial_x f - y \partial_y f = 2[/tex]

which should be easy to solve by characteristics. Mathematica gives

[tex]f = 2 \log x + C x y[/tex]

for arbitrary C. Then U and V can be any functions satisfying

[tex]\frac{U}{V} = x^2 e^{C x y}[/tex]

- #6

lavinia

Science Advisor

Gold Member

- 3,238

- 625

What approach would you recommend to learning more about differential equations? Right now i am learning differential geometry.

- #7

Ben Niehoff

Science Advisor

Gold Member

- 1,879

- 162

- #8

MathematicalPhysicist

Gold Member

- 4,362

- 241

Sometime I wonder why I even bother answering. :-/

- #9

lavinia

Science Advisor

Gold Member

- 3,238

- 625

I thank you for this solution and feel that I should explain where this equation came from.

To learn some differential geometry I posed the question of what vector fields on a surface can be tangent to geodesics for some Riemannian metric. Computation led to the condition that a unit length vector field V is tangent to geodesics if its Lie bracket with iV, the unit vector field orthogonal to it, is a multiple of iV (and that this multiple satisfies a differential equation along the geodesic that relates it to the Gauss curvature of the metric).

For the vector field xd/dx - yd/dy in a neighborhood of the origin in the plane, one gets the differential equation in this post. I think your solution shows that the field of hyperbolas y = k/x can be geodesics and that iV can be chosen to be x[itex]^{2}[/itex]d/dx + e[itex]^{-xy}[/itex]d/dy.

So it seems that a vector field of index -1 around a singularity can be a geodesic flow.

I think the Gauss curvature is -1 for this example but I haven't yet done the ugly calculation.

To learn some differential geometry I posed the question of what vector fields on a surface can be tangent to geodesics for some Riemannian metric. Computation led to the condition that a unit length vector field V is tangent to geodesics if its Lie bracket with iV, the unit vector field orthogonal to it, is a multiple of iV (and that this multiple satisfies a differential equation along the geodesic that relates it to the Gauss curvature of the metric).

For the vector field xd/dx - yd/dy in a neighborhood of the origin in the plane, one gets the differential equation in this post. I think your solution shows that the field of hyperbolas y = k/x can be geodesics and that iV can be chosen to be x[itex]^{2}[/itex]d/dx + e[itex]^{-xy}[/itex]d/dy.

So it seems that a vector field of index -1 around a singularity can be a geodesic flow.

I think the Gauss curvature is -1 for this example but I haven't yet done the ugly calculation.

Last edited:

- Last Post

- Replies
- 2

- Views
- 4K

- Replies
- 4

- Views
- 3K

- Last Post

- Replies
- 4

- Views
- 3K

- Replies
- 4

- Views
- 1K

- Replies
- 2

- Views
- 2K

- Replies
- 4

- Views
- 2K

- Replies
- 2

- Views
- 1K

- Last Post

- Replies
- 5

- Views
- 2K

- Replies
- 1

- Views
- 2K

- Replies
- 3

- Views
- 1K