Solutions To Tutorial 2 (Week 3) : Lecturers: Daniel Daners and James Parkinson
Solutions To Tutorial 2 (Week 3) : Lecturers: Daniel Daners and James Parkinson
Copyright
c 2012 The University of Sydney 1
2. Find the tangent plane to the surface z cos x − sin y + z 2 tan−1 x = 1 at (0, π, 1).
Solution: The surface is implicitly defined by f (x, y, z) = 1. The gradient vector
is
z2
−1
∇f (x, y, z) = −z sin x + , − cos y, cos x + 2z tan x ,
1 + x2
and so ∇f (0, π, 1) = (1, 1, 1). Therefore we are looking for the equation of the
plane through (0, π, 1) with normal vector (1, 1, 1), and so the equation is
x + (y − π) + (z − 1) = 0, or x + y + z = π + 1.
4. Find the equation of the tangent plane to the surface x3 − y 2 + xyz = 0 at the
point (1, 1, 0). Do this in two ways:
(a) by thinking of the surface as a level surface g(x, y, z) = 0, and
Solution: A surface defined by an equation g(x, y, z) = C is a level surface
of the function g. If a = (a, b, c) is a point on this surface then the gradient
vector of g at (a, b, c) is normal to the tangent plane of this level surface
at (a, b, c). So the tangent plane has equation
∇g(a) · (x − a) = 0.
In this case
∇g = (3x2 + yz, −2y + xz, xy).
Evaluating at (a, b, c) = (1, 1, 0), we get ∇g(a) = (3, −2, 1). So the plane
has equation
∂f 2x3 + y 2 ∂f x3 + y 2
=− and =
∂x x2 y ∂y xy 2
2
Using the formula for the tangent plane to a graph we see that
∂f ∂f
z = f (1, 1) + (1, 1)(x − 1) + (1, 1)(y − 1)
∂x ∂y
= 0 − 3(x − 1) + 2(y − 1)
= −3x + 2y + 1.
5. Find all points on the circle x2 + y 2 = 1 where the circle is tangent to a level curve
of the function x2 + 6xy + y 2 .
Solution: At a point with the required property ∇(x2 + y 2 − 1) = (2x, 2y) must
be a scalar multiple of ∇(x2 + 6xy + y 2 ) = (2x + 6y, 2y + 6x). Hence we need to
solve the system
For fixed λ the first two equations are a linear system for x, y. One obvious solution
is (x, y) = (0, 0), but this does not satisfy the third equation. To ensure the
existence of a non-trivial solution of the linear system we need that the determinant
of the coefficient matrix is zero. Hence we require that (λ − 1)2 − 9λ2 = 0. Solving
this quadratic equation for λ we get λ1 = −1/2 and λ2 = 1/4. Substituting
λ1 = −1/2 into √ the first equation we get x = −y, and so from the last equation
x√ = −y = ± 2/2. If λ2 = 1/4 then x = y, and by the last equation x = y =
2 2 2 2
± 2/2. Hence the contour
√ lines
√ of x +6xy+y √ are√tangent to the circle x +y = 1
at the four points ±( 2/2, − 2/2) and ±( 2/2, 2/2).
fx = −2 sin(x + y 2 ) = 0
fy = −4y sin(x + y 2 ) − 2y = 0.
3
Solution: We have
f (x, y) = 2 cos(x + y 2 ) − y 2 + 3
fx (x, y) = −2 sin(x + y 2 )
fy (x, y) = −4y sin(x + y 2 ) − 2y
fxx (x, y) = −2 cos(x + y 2 )
fyy = −4 sin(x + y 2 ) − 8y 2 cos(x + y 2 ) − 2
fxy (x, y) = −4y cos(x + y 2 ).
Let’s choose the critical point (0, 0). Thus the Taylor polynomial is
(c) Use your answer from (b) to have a guess about the nature of the critical
point (0, 0) (is it a local max, min, or saddle?). Now use Taylor’s Theorem
to prove your claim.
Solution: I’m guessing that (0, 0) is going to be a local maximum. This
is because
R2 (x, 0)
lim = 0.
(x,y)→(0,0) x2 + y 2
Thus
|R2 (x, 0)| < (x2 + y 2 ) whenever 0 < kxk < δ.
Therefore if 0 < kxk < δ we have
So f (x, y) < f (0, 0) for all (x, y) near (0, 0) (with (x, y) 6= (0, 0)), and so
f (0, 0) is a (strict) local maximum.
4
(d) Now apply the second derivatives test to get the same answer (hopefully!).
Solution: The Hessian matrix at (0, 0) is
fxx (0, 0) fyx (0, 0) −2 0
Hf (0, 0) = = .
fxy (0, 0) fyy (0, 0) 0 −2
7. Find all critical points of f (x, y, z) = x2 +y 2 +z 2 +2xy +2yz −4z +x and determine
their nature using the second derivatives test.
Solution: We need to solve the equation ∇f (x, y, z) = 0. So we have the
equations
2x + 2y + 1 = 0
2y + 2x + 2z = 0
2z + 2y − 4 = 0.
The first two equations imply that z = 12 . Then the third equation implies that
y = 23 , and the first implies that x = −2. So the only critical point is (x, y, z) =
(−2, 32 , 21 ). The Hessian matrix is
2 2 0
3 1
Hf (−2, , ) = 2 2 2 .
2 2
0 2 2
√ √
It has eigenvalues λ = 2 − 2 2, 2, 2 + 2 2. Since λmin < 0 and λmax > 0 we see
that Hf (−2, 32 , 21 ) is indefinite, and so (−2, 32 , 21 ) is a saddle point.
8. The temperature in Nemo’s tank at the point (x, y, z) is governed by the equation
T (x, y, z) = x2 + 2y 2 − xyz. Nemo is at (x, y, z) = (1, 2, −1). It is way too cold!
(T (1, 2, −1) = 11 degrees). Which way should he swim?
Solution: Nemo decides to swim in the direction of the gradient vector, because
he remembers that this vector points in the direction of the most rapid rate of
increase in temperature. He carefully calculates that
5
Questions for extra practice
Solution: Let x = (x, y) and a = (a, b). The third order Taylor Polynomial
should be
T3 (x; a) = T2 (x; a) + cubic terms,
where T2 (x; a) = f (a) + ∇f (a) · (x − a) + 12 (x − a)T Hf (a)(x − a), and
cubic terms = c30 (x − a)3 + c21 (x − a)2 (y − b) + c12 (x − a)(y − b)2 + c03 (y − b)3
Therefore we set
1
T3 (x; a) = T2 (x; a) + fxxx (a)(x − a)3 + 3fxxy (a)(x − a)2 (y − b)
6
+ 3fxyy (a)(x − a)(y − b)2 + fyyy (a)(y − b)3 .
Let R3 (x, a) = f (x) − T3 (x, a). Taylor’s Theorem says that if the third order
partials exist around x = a and are continuous at x = a then
R3 (x, a)
lim = 0.
x→a kx − ak3
6
10. Find a point on the surface
√ x2 /9 + y 2 /4 + z 2 = 1 at which the tangent plane is
perpendicular to (1, 1, 3). Write down an equation of the tangent plane.
Solution: Set f (x, y, z) = x2 /9+y 2 /4+z 2 −1. Then ∇f (x, y, z) = (2/9x, 1/2y, 2z)
is perpendicular to the surface x2 /9 + y 2 /4 + z 2 = 1 at (x, y, z). We need to find a
point (x,
√ y, z) satisfying f (x, y, z) = 0 and for which ∇f (x, y, z) = (2/9x, 1/2y, 2z) =
λ(1, 1, 3) for some λ ∈ R. Hence we need to solve the following system of equa-
tions:
√ x2 y 2
2x = 9λ, y = 2λ, 2z = 3λ, + + z 2 = 1.
9 4
If we solve the first three equations for x, y and z, respectively, and substitute them
into the last we get
9 3
1 = λ2 + λ2 + λ2 = 4λ2 .
4 4
Solving for λ we get λ = ±1/2. Hence the √ tangent plane to the surface√ x2 /9 +
y 2 /4 + z 2√= 1 is perpendicular to (1, 1, 3) at the points (9λ/2, √ 2λ, 3λ/2) =
±(9/4, 1,√ 3/4). An equation √ for the tangent plane at (9/4,√1, 3/4) is therefore
x + y + 3z = 4, at −(9/4, 1, 3/4) an equation is x + y + 3z = −4.
4
11. Find all critical points of the function f (x, y) = −xy + x
+ y2 .
Solution: We need to find the points for which ∇f (x, y) = (−y − 4/x2 , −x −
2/y 2 ) = (0, 0), that is,
Note that x 6= 0 and y 6= 0 as otherwise the function is not defined. From the first
equation y = −4/x2 . Substituting this into the second equation we get x+2x4 /16 =
x(1 + x3 /8) = 0. As x 6= 0 we have 1 + x3 /8 = 0, so x3 = −8 and therefore x = −2.
Now y = −4/x2 = −4/4 = −1. Hence (−2, −1) is the only critical point of f .
(f · g)′ = (f1 g1 + · · · + fn gn )′
= (f1 g1 )′ + · · · + (fn gn )′
= f1′ g1 + · · · + fn′ gn + f1 g1′ + · · · + fn gn′ = f ′ · g + f · g ′ .
= f ′ × g + f × g′.
a c
13. Let A = be a 2 × 2 real symmetric matrix. Show that:
c b
7
(a) A is positive definite if and only if det(A) > 0 and a > 0.
Solution: The characteristic equation of the matrix A is λ2 − (a + b)λ +
(ab−c2 ). Therefore by the root-coefficient relations the eigenvalues λmax and
λmin satisfy
λmin + λmax = a + b and λmin λmax = ab − c2 .
Suppose that A is positive definite. Then λmin , λmax > 0, and so a + b > 0
and ab − c2 > 0. Thus det(A) = ab − c2 > 0, and since ab > c2 we know that
a, b 6= 0 and that they have the same sign, and hence a > 0 by the inequality
a + b > 0.
Conversely, if det(A) = ab − c2 > 0 and a > 0 then b > 0 (by ab > c2 ) and
thus λmax + λmin = a + b > 0 and λmax λmin = ab − c2 > 0. Therefore λmin
and λmax are nonzero and have the same sign, and therefore λmin , λmax > 0
since λmin + λmax > 0. Hence A is positive definite.
(b) A is negative definite if and only if det(A) > 0 and a < 0.
Solution: This is very similar to (a).
(c) A is indefinite if and only if det(A) < 0.
Solution: Suppose that A is indefinite. Then λmin < 0 and λmax > 0.
Hence ab − c2 = λmin λmax < 0, and so det(A) < 0. Conversely, if det(A) < 0
then ab − c2 < 0, and so λmin λmax = ab − c2 < 0, and so λmin and λmax have
opposite signs, and so A is indefinite.
This checks the ‘quick version’ of the 2nd derivatives test for functions f : R2 → R.
14. Suppose that f : R2 → R has continuous second order partial derivaties. Suppose
that z = f (u, v) and that u = x + y and v = x − y. Show that
∂z ∂z ∂z 2 ∂z 2 ∂ 2z ∂ 2z ∂ 2z
= − and = − .
∂x ∂y ∂u ∂v ∂x∂y ∂u2 ∂v 2
8
15. A function f : RN → R is radial if f (x) depends only on kxk, i.e., if f (x) = g(kxk),
where g(r) is some function, which we shall assume to be differentiable. Show that
if f is radial then ∇f (x) is a scalar multiple of x.
p
Solution: Write f (x) = g x21 + · · · + x2N . Then by the Chain Rule,
∂f ′
q
2 2
∂ q
=g x1 + · · · + xN x21 + · · · + x2N
∂xj ∂xj
′
q xj g ′ (kxk)
=g x21 + · · · + x2N p 2 = xj .
x1 + · · · + x2N kxk
g ′ (kxk)
Hence ∇f (x) = kxk
x.
Challenging problems
16. Let f : R2 → R and assume that the mixed partial derivatives fxy and fyx both
exist in a disc around (a, b). Let h, k > 0, and let
∆(h, k) = f (a + h, b + k) − f (a, b + k) − f (a + h, b) + f (a, b).
(a) Write ∆(h, k) as F (a + h) − F (a) where F (x) = f (x, b + k) − f (x, b). Use
the Mean Value Theorem twice to prove that
∆(h, k)
= fxy (α, β)
hk
where α = α(h, k) and β = β(h, k) with (α, β) → (a, b) as (h, k) → (0, 0).
Solution: We have
F ′ (x) = fx (x, b + k) − fx (x, b).
By the Mean Value Theorem, F (a + h) − F (a) = F ′ (α)h for some α between
a and a + h. Hence
∆(h, k) = F (a + h) − F (a) = F ′ (α)h = fx (α, b + k) − fx (α, b) h
Now apply the Mean Value Theorem again, this time to ϕ(y) = fx (α, y).
Thus ϕ(b + k) − ϕ(b) = ϕ′ (β)k for some β between b and b + k. But
∂
ϕ′ (y) = fx (α, y) = fxy (α, y),
∂y
and so ∆(h, k) = fxy (α, β)hk, where (α, β) → (a, b) as (h, k) → (0, 0).
(b) Argue analogously to show that
∆(h, k)
= fyx (γ, δ)
hk
where γ = γ(h, k) and δ = δ(h, k) with (γ, δ) → (a, b) as (h, k) → (0, 0).
Solution: The same reasoning, starting with ∆(h, k) = G(b + k) − G(b)
where G(y) = f (a + h, y) − f (a, y).
(c) Deduce that if fxy and fyx are continuous at (a, b) then fxy (a, b) = fyx (a, b).
Solution: By parts (a) and (b) we have fxy (α, β) = fyx (γ, δ) where (α, β)
and (γ, δ) tend to (a, b) as (h, k) → (0, 0). Therefore by the assumed conti-
nuity of fxy and fyx at (a, b) we have fxy (a, b) = fyx (a, b).