OU blog

Personal Blogs

Zafar Bakhromov

Resolving a Confusion About Implicit Function Theorem and Manifolds

Visible to anyone in the world

"How do I show that a unit sphere is a smooth submanifold of double-struck cap r cubed ?"

That is the question I was pondering about when trying to solve exercise 3.1.1 of the book "Vector Calculus, Linear Algebra and Differential Forms - A Unified Approach" by John and Barbara Burke Hubbard. The problem asks for a proof of the above question, but trying to prove that the unit sphere is a submanifold of double-struck cap r cubed from the definition of a manifold is very cumbersome. For reference, this is how a manifold embedded in double-struck cap r super n is defined as in the said book:

A subset cap m subset of or equals double-struck cap r super n is a smooth k -dimensional manifold if locally it is the graph of a cap c super one mapping  bold f expressing n minus k variables as functions of the other k variables.

Here, "locally" means that every point bold x element of cap m has a neighborhood cap u subset of or equals double-struck cap r super n such that cap m intersection cap u (the part of cap m in cap u ) is the graph of a mapping expressing n minus k of the coordinates of each point in cap m intersection cap u in terms of the other k .

So, in order to prove that the unit sphere cap s squared colon equals left curly bracket left parenthesis x comma y comma z right parenthesis vertical line sum with 3 summands x squared plus y squared plus z squared equals one right curly bracket is a manifold in double-struck cap r cubed , I would have to specify

  • a "chunk" of the (some variable, some variable) plane, such as cap d sub x comma y colon equals left curly bracket left parenthesis x comma y right parenthesis vertical line x squared plus y squared less than one right curly bracket
  • a "part" of the (remaining variable)-axis, such as double-struck cap r sub z super plus colon equals left curly bracket z vertical line z greater than zero right curly bracket
  • a function giving the graph of cap s squared in the open neighborhood "(chunk) direct product (part of axis)". In this case, cap d sub x comma y multiplication double-struck cap r sub z super plus is the open neighborhood we're looking for, and cap s squared intersection left parenthesis cap d sub x comma y multiplication double-struck cap r sub z super plus right parenthesis is the graph of the function Square root of one minus x squared minus y squared .

Quite tedious, but this is not over. Since this method only specifies an open neighborhood of double-struck cap r cubed when (in this case) z not equals zero , we have to also do the same thing using all of cap d sub x comma y comma cap d sub y comma z comma cap d sub x comma z and double-struck cap r sub x super plus comma double-struck cap r sub x super minus comma double-struck cap r sub y super plus comma double-struck cap r sub y super minus comma double-struck cap r sub z super plus comma double-struck cap r sub z super minus ., and the functions:

prefix plus minus of Square root of one minus x squared minus y squared comma prefix plus minus of Square root of one minus y squared minus z squared comma prefix plus minus of Square root of one minus x squared minus z squared

Wow.

The implicit function theorem, on the other hand, gives a neat way to check whether

cap m colon equals cap f super negative one of zero equals left curly bracket x element of cap u vertical line cap f of x equals zero comma cap u colon open right curly bracket

is a smooth manifold. We just have to check that the derivative of cap f is surjective at every point on cap m . In this case,

cap s squared equals cap f super negative one of zero

where sum with 3 summands cap f of x comma y comma z colon equals x squared plus y squared plus z squared minus one , so since

multirelation left square bracket cap d times cap f of x comma y comma z right square bracket equals left square bracket two times x two times y two times z right square bracket not equals left square bracket zero zero zero right square bracket

by the implicit function theorem we get that cap s squared is a smooth manifold in double-struck cap r cubed .

Looking back, the thing I was first misunderstanding was that, I mistakengly believed "I have to find a local representation of cap m at every point first, then I can use the implicit function theorem". But doing so would defeat the whole purpose of the implicit function theorem, because by that point I would have already proved that cap m is a manifold!

Also, I was having some difficulties grasping why we have to check the z equals zero case for cap s squared discussed earlier, why just being globally describable as two graphs is not enough to show that cap s squared is a manifold. First, the z equals zero case does not define an open neighborhood; it's just going to be a two dimensional plane. But at a deeper level, it taught me that manifold structure is a local condition at every point. This is exactly why the implicit function theorem is so powerful; we just have to check one condition and it automatically guarantees that a local representation as a graph exists at every point!

In the end, what felt like a technical shortcut turned out to be something deeper: the implicit function theorem is not a computational trick, but a precise machine that guarantees local manifold structure from a single global condition.

Permalink Add your comment
Share post