Take the 2-minute tour ×
Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. It's 100% free, no registration required.

I'm helping a student through a course in mathematics. In the course text, we came across the following problem concerning the Lagrange multiplier technique. Given a differentiable function with continuous partial derivatives $f:\mathbb{R}\to\mathbb{R}:(x,y)\mapsto f(x,y)$ that has to be extremized and a constraint given by an implicit relation $g(x,y)=0$ with $g$ likewise having continuous partial derivatives. The Lagrange multiplier technique looks for points $(x^*,y^*,\lambda^*)$ such that

$$\nabla f(x^*,y^*)=\lambda^* \nabla g(x^*,y^*)$$

$\lambda$ is the so-called Lagrange multiplier.

The technique rests upon the fact that the gradient of the constraint is different from the zero vector: $\nabla g(x^*,y^*) \neq 0$. Then, the text says that if that condition is not fulfilled, it is possible to not have solutions of the system of equations resulting from the technique while there can actually be a constrained extremum.

I have tried to construct such an example, but until now unsuccessfully as every attempt I make with a $\nabla g(x^*,y^*) = 0$ turns out to have a solution. Does anyone know of a nice counterexample?

And does one know of a counterexample when there are $n$ variables and $m$ constraints with $n>m>1$ ? For the latter, the condition becomes that the gradients of the constraints have to be linearly independent. So the question is: can we find an example where there is no solution for the Lagrange multiplier method, but there is an extremum and the gradients of the constraints are linearly dependent?

share|improve this question
add comment

2 Answers

up vote 4 down vote accepted

Here is an example, albeit with two constraints: We want to find the minimum of the function $$f(x,y,z):=y\ ,$$ given the constraints $$F(x,y,z):=x^6-z=0\ ,\qquad G(x,y,z):=y^3-z=0\ .$$ The constraints define the curve $$\gamma:\quad x\mapsto (x,x^2, x^6)\qquad(-\infty<x<\infty)\ ,$$ and it is easy to see that the minimum of $f\restriction \gamma$ is taken at the origin. On the other hand $$\nabla f(0,0,0)=(0,1,0)\ ;\quad\nabla F(0,0,0)=\nabla G(0,0,0)=(0,0,-1)\ ;$$ whence $\nabla f(0,0,0)$ is not a linear combination of $\nabla F(0,0,0)$ and $\nabla G(0,0,0)$. This means that using Lagrange's method the origin would not have shown up as a conditionally stationary point.

share|improve this answer
    
Thanks, I see now that if I had kept an eye on the gradients, I would have found an example much more easily. –  Raskolnikov May 20 '12 at 16:53
add comment

In the Lagrange multiplier technique, what you really need is a point $(x^*,y^*)$ where $\nabla f(x^*,y^*)$ and $\nabla g(x^*,y^*)$ are linearly dependent. If the text says otherwise, they are wrong. This is slightly different from what you wrote down. Try $g(x,y)=x^2+y^2$ and $f(x,y) = (x-1)^2=0$, which has a constrained minimum at $(1,0)$. This doesn't satisfy $n > m >1$, but it shouldn't be to hard to modify it so it does.

share|improve this answer
    
Thanks, this is a great example for the case with one constraint and 2 variables. –  Raskolnikov May 20 '12 at 16:53
add comment

Your Answer

 
discard

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.