Equality of Mixed Partials (Clairaut’s Theorem)
In chapter 14, we encountered Clairaut’s Theorem, which says that when fxy and fyx are both continuous,
they will be equal. This was presented as a fact—something to believe only because I told you so. Such
teaching by dictatorship is always regrettable; a guiding principle of the discipline of mathematics is that it
is democratic and egalitarian—that anyone can challenge anyone else’s statement, and logic will judge who
is right. I had no choice, though; the justification of Clairaut’s Theorem would not have made sense to you
at the time.
It is now possible to close this gap: using the perspective of § 15.2, I can now back up the claim of Clairaut’s
Theorem with a proof. I will break off the main part as a lemma.
Lemma. Let fxy and fyx be continuous on rectangle R = [a, b] × [c, d]. Then
ZZ ZZ
fxy dA = fyx dA = f (b, d) − f (a, d) − f (b, c) + f (a, c).
R R
Proof . I will prove the assertion for fxy ; the proof for fyx will be left as an exercise.
ZZ Z b Z d
fxy dA = fxy (x, y) dy dx
a c
R
Z b y=d
(by the FTC) −→ = fx (x, y) dx
a y=c
Z b
= fx (x, d) − fx (x, c) dx
a
Z b Z b
= fx (x, d) dx − fx (x, c) dx
a a
(by the FTC) −→ = f (b, d) − f (a, d) − f (b, c) − f (a, c)
= f (b, d) − f (a, d) − f (b, c) + f (a, c).
Exercise. Prove that ZZ
fyx dA = f (b, d) − f (a, d) − f (b, c) + f (a, c).
R
This will complete the proof of the lemma.
Theorem. If fxy and fyx are continuous, then they are equal.
Proof by contradiction. Suppose they are not identically equal. Then at some point (a, b), they differ; say
fxy (a, b) − fyx (a, b) = h > 0.
Then by continuity, there is some small ∆x × ∆y rectangle, centered at (a, b), on which
h
fxy (x, y) − fyx (x, y) ≥ ,
2
so that ZZ ZZ
h h
fxy − fyx dA ≥ dA = (∆x)(∆y) > 0.
2 2
R R
But this contradicts the lemma: since fxy and fyx have equal integrals over the rectangle, necessarily also
ZZ ZZ ZZ
fxy − fyx dA = fxy dA − fyx dA = 0.
R R R