Å A Š × ç ü ç î ¬tpç5t = Ú (!bç t æ) [u 3Ü/ 1Ü]
Part I. LINEAR ALGEBRA
1. (8 points) Let A be a 3 × 4 matrix over the real number field R, and let {(2, 3, 1, 0)} be a basis for the nullspace of A.
(a) What is the rank of A and the complete solution to Ax = 0?
Solution: The dimension of the nullspace is 1, so the rank of A is 4 − 1 = 3. The complete solution to Ax = 0 is x = c · (2, 3, 1, 0) for any constant c.
(b) Find a basis for the column space of AT.
Solution: The column space of AT is the row space of A. The nonzero rows of the row reduced echelon form
1 0 −2 0 0 1 −3 0
0 0 0 1
form a basis.
2. (a) (4 points) The linear transformation T : R2→ R2reflects a vector about the line y = −x and then projects that vector orthogonally onto the x-axis. Find the standard matrix for T .
Solution: T
·1 0
¸
=
·0 0
¸ and T
·0 1
¸
=
·−1 0
¸
so the matrix representation for T is
·0 −1
0 0
¸ .
(b) (4 points) Suppose T : R4→ R2 is a linear transformation with T (1, 0, 0, 1) = (2, 3) and T (0, 1, 1, 0) = (1, 5). Find T (6, 2, 2, 6).
Solution: Let v1= (1, 0, 0, 1) and v2= (0, 1, 1, 0). Then v = (6, 2, 2, 6) = 6v1+ 2v2. By linearity, T (v) = T (6v1+ 2v2) = 6 T (v1) + 2 T (v2) = 6 (2, 3) + 2 (1, 5) = (14, 28)
3. (8 points) Suppose the 3 × 3 matrix A over R has eigenvalues 0, 1, and 2 with eigenvectors v0, v1, v2, respectively.
(a) What is the trace of A − 2I?
Solution: A − 2I has eigenvalues −2, −1, 0 so its trace is −3.
(b) Solve the equation Ax = v1+ v2for x.
Solution: x = av0+ v1+12v2.
4. (16 points) Let V be the vector space R3over R. The following matrix is a projection matrix on V : P =211
1 2 −4
2 4 −8
−4 −8 16
.
(a) What subspace W of V does P project onto?
Solution: The projection matrix P projects onto the column space of P which is the line c · (1, 2, −4).
(b) What is the distance from that subspace W to b = (5, 4, −2)?
Solution: The vector from b to the subspace is
e = b − Pb =
5 4
−2
−2121
1 2
−4
=
4 2 2
and the distance is
kek =p
42+ 22+ 22= 2√ 6.
(c) What are the three eigenvalues of P?
Solution: Since P projects onto a line, its three eigenvalues are 0, 0, 1. The eigenvectors for 0 are vectors orthogonal to (1, 2, −4). The eigenvector for 1 is (1, 2, −4).
(d) If you solvedudt = −Pu (notice minus sign) starting from u(0), the solution u(t) approaches a steady state as t → ∞. Describe that limit vector u(∞)?
Å A Š × ç ü ç î ¬tpç5t = Ú (!bç t æ) [u 3Ü/ 2Ü]
Solution: The solution u(t) to the differential equation has the form u(t) = v1e−t+ v2where v1is in W and v2is in the orthogonal complement of W . Then u(∞) = v2, which is the projection of u(0) onto the orthogonal complement of W . That is, u(∞) = u(0) − P(u(0)).
5. (10 points) Let n denote a positive integer, V denote an n-dimensional vector space, and T denote a linear operator on V . Suppose v ∈ V is a nonzero vector such that Tkv = 0 for some positive integer k. Show that Tnv = 0.
Solution: Suppose k is the smallest positive integer such that Tkv = 0. The vectors v, T v, T2v, . . . , Tk−1v are linearly indepen- dent so k ≤ n:
Suppose c0v + c1T v + c2T2v + · · · + ck−1Tk−1v = 0. Applying Tk−1 to both sides we get c0Tk−1v = 0 and so c0= 0. Now applying Tk−2to both sides we get c1Tk−1v = 0 and so c1= 0. Continuing in this fashion, we see that cj= 0 for all j.
Part II. ADVANCED CALCULUS
6. (10 points) Let x1=π2 and suppose that xnare defined inductively for n = 2, 3, . . . by cos(xn+1) = sin(xxnn), 0 < xn+1< π2, for n = 1, 2, . . ..
(a) Prove that xn+1< xnfor n = 1, 2, . . .. Hint: You may find cos(x) =dxd sin(x) and the Mean Value Theorem useful.
Solution: By the Mean Value Theorem, there exists a 0 < c < xnsuch that sin xn= sin xn−sin 0 = cos c (xn−0) = cos c xn. This implies that xn+1= c < xn.
(b) Show that the sequence {xn}∞n=1is convergent.
Solution: Since the set {xn|n ∈ N} is nonempty and bounded from below by 0, l = inf{xn|n ∈ N} exists. The definition of l implies that for eachε> 0 there exists an xkfor some k ∈ N such that l +ε> xk. The result of part (a) says that {xn}∞n=1is decreasing which implies that l +ε> xk≥ xn> 0 for all n ≥ k. Therefore, we get |xn− l| <εfor all n ≥ k which means that lim
n→∞xn= l.
(c) Find explicitly the limit x of the sequence {xn}∞n=1.
Solution: Let f (x) = x cos x − sin x for x ∈ [0,π2]. To find all possible limits is equivalent to find zeros of f over [0,π2].
Note that f (0) = 0 and f0(x) = −x sin x < 0 for all x ∈ (0,π2], we can conclude that x = 0 is the only zero of f . Hence,
n→∞limxn= 0.
7. (a) (5 points) State what it means for a sequence { fn(x)}∞n=1of real valued functions on a set X ⊂ Rpto converge uniformly to a function f on X.
Solution: { fn}∞n=1is said to converge uniformly on X to f if for eachε> 0 there exists a K(ε) ∈ N such that for all n ≥ K(ε) and x ∈ X then | fn(x) − f (x)| <ε.
(b) (5 points) Let F : [0, 1] × [0, 1] → R be a continuous function on the unit square. Let fn(t) = F(1n,t) and f (t) = F(0,t).
Use the definition in part (a) to show that { fn(x)}∞n=1converges uniformly to f on the interval [0, 1].
Solution: Since F is continuous on [0, 1] × [0, 1], F is uniformly continuous there. This implies that for anyε> 0 and any t ∈ [0, 1], there exists aδ =δ(ε) > 0 such that if |x − 0| <δ then |F(x,t) − F(0,t)| <ε. Letting K(ε) = [δ1] + 1, where [δ1] is defined to be the greatest integer less than or equal to δ1, we note that n ≥ K(ε) > δ1 implies that 1n <δ. Thus, | fn(t) − f (t)| = |F(1n,t) − F(0,t)| <εfor all n ≥ K(ε). Thus { fn}∞n=1converges uniformly to f on [0, 1]
8. (10 points) For each integer k ≥ 1, let fk: R → R be differentiable satisfying | fk0(x)| ≤ 1, for all x ∈ R, and fk(0) = 0.
(a) For each x ∈ R, prove that the set { fk(x)}∞k=1is bounded.
Solution: For each x ∈ R, since | fk(x)| = | fk(x) − fk(0)| ≤ | fk0(ck) (x − 0)| ≤ |x|, where cklies between x and 0, the set { fk(x)}∞k=1is bounded.
(b) Use Cantor’s diagonal method to show that there is an increasing sequence n1< n2< n3< · · · of positive integers such that, for every x ∈ Q, we have { fnk(x)} is a convergent sequence of real numbers.
Å A Š × ç ü ç î ¬tpç5t = Ú (!bç t æ) [u 3Ü/ 3Ü]
Solution: Let Q = {x1, x2, · · · }. Since { fk(x1)} is bounded, we can extract a convergent subsequence, denoted { fk1(x1)}, out of { fk(x1)}. Next, the boundedness of { fk1(x2)} implies that we can extract a convergent subsequence, denoted { fk2(x2)}, out of { fk1(x2)}. Continuing this way, the boundedness of { fkj(xj+1)} implies that we can extract a convergent subsequence { fkj+1(xj+1)}, out of { fkj(xj+1)} for each j ≥ 1. Let fnk = fkkfor each k ≥ 1. Then { fnk} is a subsequence of { fn} and { fnk} converges at each xj∈ Q.
9. (10 points) A function f : R → R is said to be convex if for all x, y ∈ R,λ ∈ [0, 1], f (λx + (1 −λ)y) ≤λf (x) + (1 −λ) f (y).
Suppose that f : R → R is convex and that f00(x) exists for all x ∈ R.
(a) For any x < y and 0 < h < y − x, prove that
f (x + h) ≤y − x − h
y − x · f (x) + h
y − x· f (y), and f (y − h) ≤ h
y − x· f (x) +y − x − h y − x · f (y).
Solution: Setting x + h =λx + (1 −λ)y, we getλ =y−x−hy−x and 1 −λ=y−xh . Thus f (x + h) = f (y−x−hy−x x +y−xh y) ≤y−x−hy−x f (x) +y−xh f (y).
Setting y − h =βx + (1 −β)y, we getβ=y−xh and 1 −β =y−x−hy−x . Thus
f (y − h) = f (y−xh x +y−x−hy−x y) ≤y−xh f (x) +y−x−hy−x f (y).
(b) Prove that f0(x) ≤ f0(y) whenever x ≤ y.
Solution: Since
f0(x) − f0(y) = lim
h→0+
f (x+h)− f (x)
h − f (y−h)− f (y)
−h = lim
h→0+
f (x+h)+ f (y−h)− f (x)− f (y)
h ≤ lim
h→0+
f (x)+ f (y)− f (x)− f (y)
h = 0,
where we have used the result of part (a) in the last inequality. We have shown that f0(x) ≤ f0(y) whenever x ≤ y.
(c) Prove that f00(x) ≥ 0 for all x ∈ R.
Solution: The result of part (b) implies that f00(x) = lim
h→0+
f0(x+h)− f0(x)
h ≥ 0, we have f00(x) ≥ 0 for all x ∈ R.
10. (10 points) Let g : Rp→ Rpbelong to class C1(Rp), i.e. Dg(x) exists for all x ∈ Rpand the mapping x → Dg(x) is continuous.
Assume that there is a constant a such that kDg(x)k ≤ a < 1 for each x ∈ Rp.
(a) Show that the function f (x) = x + g(x) for x ∈ Rpsatisfies k f (x1) − f (x2) − (x1− x2)k ≤ akx1− x2k for all x1, x2∈ Rp. Solution: The Mean Value Theorem implies that there exists a z =λx1+ (1 −λ)x2∈ Rpfor someλ ∈ [0, 1] such that
k f (x1) − f (x2) − (x1− x2)k = kDg(z) · (x1− x2)k ≤ akx1− x2k.
(b) Show that f in part (a) is a bijection of Rpinto Rp.
Solution: Since D f = I + Dg and the eigenvalues of Dg are bounded by a < 1, D f is invertible everywhere. The Inverse Function Theorem implies that f is a local bijection. Since the result of part (a) says that f is a global one-to-one function of Rpinto Rp, f is a bijection of Rpinto Rp.