Is There a Formal Proof of the Multivariable Chain Rule?

  • Context: Undergrad 
  • Thread starter Thread starter ice109
  • Start date Start date
  • Tags Tags
    Multivariable Proof
Click For Summary

Discussion Overview

The discussion centers around the existence and formal proof of the multivariable chain rule in calculus. Participants explore various approaches to proving the rule, including references to specific texts and personal insights into the rigor and detail of the proofs.

Discussion Character

  • Exploratory, Technical explanation, Debate/contested

Main Points Raised

  • One participant requests a formal proof of the multivariable chain rule.
  • Another participant suggests a proof from Spivak's "Calculus on Manifolds" and proposes a method involving the definition of the derivative.
  • Several participants express confusion regarding the applicability of the proposed proof to multivariable contexts.
  • There is a sentiment among participants that a more rigorous or detailed proof exists, with some suggesting that the current proof lacks sufficient rigor.
  • A detailed informal derivation of the chain rule is presented, outlining the relationships between functions and their derivatives, but it remains unclear if this satisfies all participants' criteria for rigor.

Areas of Agreement / Disagreement

Participants do not reach a consensus on the sufficiency or clarity of the proofs discussed. Multiple competing views on the rigor and applicability of the proposed methods remain unresolved.

Contextual Notes

Some participants express uncertainty about the definitions and conditions under which the chain rule is applied, indicating a need for clarity on differentiability and the nature of the mappings involved.

ice109
Messages
1,708
Reaction score
6
can someone link/show me a formal proof of the multivariable chain rule?
 
Physics news on Phys.org
There's a cute proof of this in Spivak's "Calculus on Manifolds". If you want to try it yourself, try defining the function fd(x) = f(x + h) - f(x) - df(h) and similarly for g, pretty much the numerator in the definition of the derivative, then evaluating the limit in the definition of the derivative for [itex]g\circ f[/itex], where you let the linear function in the limit be dg(f(x))df(x).
 
what? how is this multavariable?
 
ice109 said:
what? how is this multavariable?

let x be a vector x=(x1,x2,...,xn)
 
There must be a better proof of the chain rule that that.
 
ObsessiveMathsFreak said:
There must be a better proof of the chain rule than that.

yea i thought of that and thought the same thing; that there's got to be a more rigorous proof.
 
ice109 said:
yea i thought of that and thought the same thing; that there's got to be a more rigorous proof.

more rigorous than very rigorous
interesting
maybe you mean more detailed
that I can provide

Chain rule
let
f:E->F
g:F->G
with E (or a subset) open in F
F (or a subset) open in G
and f differentiable at x
g differentiable at f(x)
then
(g◦f)' exist with
(g◦f)'=[g'◦f][f']
or in more full notation
[g(f(x))]'=[g'(f(x)][f'(x)]
notes on mappings
f:E->F
g:F->G
f':E->L(E,F)
g':F->L(F,G)
(g◦f):E->G
(g◦f)':E->L(E,G)
g'◦f:E->(F,G)
[g'◦f][f']:E->L(E,G)
where L(E,F) is a space of linear mappings from E to F
so all is as it should be
thus derivatives are linear mappings
Δx=dx
Δf:=f(x+dx)-f(x)
df:=f'(x)dx
Δf=df+o(dx) (f differentiable)
informal derivation
dg(f(x))=g'(f(x))df(x)+o(df(x))=g'(f(x))f'(x)dx+o(dx)+f'(x)o(dx)
more formal
let
Δf=df+|dx|rf
rf=(Δf-df)/|dx|
so lim rf=0
now
lim rf=lim rg=0 (f,g differentiable)
Δ[g(f(x))]=g'(f(x))Δf(x)+|Δf|rg(f(x))
=g'(f(x))f'(x)dx+|dx|r(f)+|dx||f'(x)dx/|dx|+rf|rg)
=g'(f(x))f'(x)dx+|dx|{rf+|f'(x)dx/|dx|+rf|rg}
we now need only
lim {r(f)+|f'(x)dx/|dx|+rf(x)|rg(f(x))}=0
which is clear from
rf(x)->0
rg(f(x))->0
and
|f'(x)dx/|dx|+rf|<|f'(x)|+|rf|<∞
ie bounded near x
where |f'(x)| is the norm induced on linear maps by the norm on vectors
 
Last edited:

Similar threads

  • · Replies 10 ·
Replies
10
Views
3K
  • · Replies 5 ·
Replies
5
Views
2K
  • · Replies 10 ·
Replies
10
Views
3K
  • · Replies 21 ·
Replies
21
Views
6K
  • · Replies 2 ·
Replies
2
Views
1K
  • · Replies 2 ·
Replies
2
Views
2K
  • · Replies 4 ·
Replies
4
Views
3K
  • · Replies 8 ·
Replies
8
Views
3K
  • · Replies 14 ·
Replies
14
Views
2K
  • · Replies 6 ·
Replies
6
Views
4K