본문으로 바로 이동 / Skip to content

PROOF 43

Prove.

The chain rule

On differentiable functions.

A way to read this proof

First separate what is assumed from what is to be proved. At each step, ask which definition or earlier result permits the move. If it is unclear, return to that premise rather than skipping to the conclusion.

Definition D1

f is differentiable at x if the limit lim (h→0) [f(x+h) − f(x)]/h exists, and its value is written f′(x).

Proof.

(1)a − b means a + (−b), and F/h means F · h⁻¹. The difference quotient is defined only for h ≠ 0.convention
(2)Setting t = x + h, the same limit may also be written lim (t→x) [f(t) − f(x)]/(t − x).convention
(3)If a limit exists, its value is unique. Hence the notation f′(x) is justified.A2

∎

Definition D2

φ is continuous at u₀ if lim (u→u₀) φ(u) = φ(u₀).

Proof.

In ε-δ terms: for every ε > 0 there is η > 0 such that |u − u₀| < η implies |φ(u) − φ(u₀)| < ε. The case u = u₀ is included — there the left side is 0.A3, convention

∎

Definition D3

Set u₀ = f(x), and define φ(u) = [g(u) − g(u₀)]/(u − u₀) for u ≠ u₀, φ(u₀) = g′(u₀).

Proof.

(1)There may be h for which Δu = f(x+h) − f(x) equals 0, and then [g(u) − g(u₀)]/(u − u₀) is undefined.convention
(2)φ is the function obtained by assigning the value g′(u₀) at that one point, extending the domain to include u₀.(1), convention

∎

Lemma L1

If g is differentiable at u₀, then φ is continuous at u₀.

Proof.

(1)For u ≠ u₀, φ(u) = [g(u) − g(u₀)]/(u − u₀).D3
(2)Since g is differentiable at u₀, lim (u→u₀) [g(u) − g(u₀)]/(u − u₀) = g′(u₀)assumption, D1
(3)The limit considers only u with 0 < |u − u₀|. Hence the value in (2) does not depend on how φ(u₀) is defined.A3
(4)lim (u→u₀) φ(u) = g′(u₀) = φ(u₀)(1), (2), (3), D3
(5)Therefore φ is continuous at u₀.(4), D2

∎

Lemma L2

If lim (h→0) w(h) = u₀ and φ is continuous at u₀, then lim (h→0) φ(w(h)) = φ(u₀).

Proof.

(1)Let ε > 0.assumption
(2)Since φ is continuous at u₀, there is η > 0 such that |u − u₀| < η implies |φ(u) − φ(u₀)| < ε.(1), assumption, D2
(3)Since lim w = u₀, there is δ > 0 such that 0 < |h| < δ implies |w(h) − u₀| < η.(2), assumption, A3
(4)Therefore 0 < |h| < δ implies |φ(w(h)) − φ(u₀)| < ε.(2), (3)
(5)Line (2) holds also for h with w(h) = u₀. With a mere limit instead of continuity, such h could not be handled.(2), D2
(6)Therefore lim (h→0) φ(w(h)) = φ(u₀).(4), (5), A3

∎

Lemma L3

For every h, g(f(x+h)) − g(f(x)) = φ(f(x+h)) · [f(x+h) − f(x)].

Proof.

(1)Case f(x+h) ≠ u₀: φ(f(x+h)) = [g(f(x+h)) − g(u₀)] · [f(x+h) − u₀]⁻¹D3
(2)Multiplying both sides by f(x+h) − u₀ gives φ(f(x+h)) · [f(x+h) − f(x)] = g(f(x+h)) − g(f(x))(1), E, A5, A6, A7, A8
(3)Case f(x+h) = u₀: the left side is g(u₀) − g(u₀) = 0.A10
(4)The right side is φ(u₀) · 0 = 0, so the two sides agree.(3), A9, A10
(5)In both cases the equality holds.(2), (4)

∎

Theorem T1

If f is differentiable at x and g is differentiable at u₀ = f(x), then g∘f is differentiable at x and (g∘f)′(x) = g′(f(x)) · f′(x).

Proof.

(1)Suppose f and g are differentiable at x and u₀ respectively.assumption, D1
(2)If h ≠ 0, then [g(f(x+h)) − g(f(x))]/h = φ(f(x+h)) · [f(x+h) − f(x)]/hL3, E, A5
(3)Since f is differentiable at x, lim (h→0) f(x+h) = f(x) = u₀(1), A4
(4)Applying L1 and L2 to w(h) = f(x+h) gives lim (h→0) φ(f(x+h)) = φ(u₀) = g′(u₀)(3), L1, L2, D3
(5)lim (h→0) [f(x+h) − f(x)]/h = f′(x)(1), D1
(6)Taking limits on both sides of (2) gives lim (h→0) [g(f(x+h)) − g(f(x))]/h = g′(u₀) · f′(x)(2), (4), (5), E, A1
(7)Therefore (g∘f)′(x) = g′(f(x)) · f′(x).(6), D1

∎

Remark

Can A1 be removed from this proof?

Line (2) turns the difference quotient into a product of two pieces. Check what justifies the claim that when each piece has a limit, the product is the limit of the whole.