Solution of the First HKUST Undergraduate Math Competition – Junior Level

∞
X X∞ Xn X∞
xj 1 n! n!
1. For all x ∈ R, ex = . So e = . For a positive integer n, In = ∈ Z and let an = .
j=0
j! j=0
j! j=0
j! j=n+1
j!
Then n sin(2πen!) = n sin(2πIn + 2πan) = n sin(2πan). Note

X ∞
1 1 1 1 1 1
≤ an = + + +··· ≤ = .
n+1 n + 1 (n + 1)(n + 2) (n + 1)(n + 2)(n + 3) (n + 1)k n
k=1

By sandwich theorem, an → 0 and nan → 1 as n → ∞. Using sin θ ∼ θ as θ → 0, we get

lim n sin(2πen!) = lim n sin(2πan) = lim 2πnan = 2π.

2. Subtracting the first row from each of the other rows, we get
 
3 1 1 1 ··· 1
 −2 3 0 0 ··· 0
 
 −2 0 4 0 ··· 0
Dn = det 
 −2 0 0 5 ··· 0.

 . .. .. .. .. .. 
 .. . . . . .
−2 0 0 0 ··· n

For 2 ≤ i ≤ n − 1, adding 2/(i + 1) times the i-th column to the first column, we get
 2 2 2 
3+ 3 + 4 + ···+ n 1 1 1 ··· 1
 0 3 0 0 ··· 0
   
 0 0 4 0 ··· 0
Dn = det   = n! 1 + 1 + 1 + · · · + 1 .
 0 0 0 5 ··· 0  2 3 n
 .. .. .. .. .. .. 
 . . . . . . 
0 0 0 0 ··· n

Dn 1 1 1
Now = 1 + + + · · · + diverges to +∞ by the p-test, hence it is unbounded.
n! 2 3 n

3. (Solution 1) Let S = {x ∈ [0, 1] : f(x) ≤ g(x)}. Now 0 ∈ S and S is bounded above by 1. Hence
w = sup S exists. Since f(0) < g(0) < g(1) < f(1) and f is continuous, we get 0 < w < 1. Since g is
monotone, g(w−) = lim g(x) and g(w+) = lim g(x) exist. Being supremum, there exists a sequence
x→w− x→w+
xn ∈ S converging to w. Since w > 0, we have f(w) = lim f(xn ) ≤ lim g(xn ) = g(w−). Next, take
n→∞ n→∞
a sequence yn ∈ (w, 1] converging to w. Now yn 6∈ S implies f(w) = lim f(yn ) ≥ lim g(yn ) = g(w+).
n→∞ n→∞
Finally, g(w−) ≥ f(w) ≥ g(w+) implies f(w) = g(w).

(Solution 2 due to Li Zhiming and Tai Ming Fung Philip) Assume for all w ∈ [0, 1], f(w) 6= g(w). We
will construct a sequence of nested intervals [an, bn] such that f(an ) < g(an ) < g(bn ) < f(bn ) by math
induction.
Let a1 = 0 and b1 = 1. We have f(a1 ) < g(a1 ) < g(b1 ) < f(b1 ). Suppose f(ak ) < g(ak ) < g(bk ) <
f(bk ). Let m = (ak +bk )/2. Since f(m) 6= g(m), either f(m) < g(m) or f(m) > g(m). In the former case,
let [ak+1, bk+1] = [m, bk ]. In the latter case, let [ak+1, bk+1] = [ak , m]. Since |ak − bk | = 1/2k−1 → 0, by
the nested interval theorem, ak and bk converge to some w ∈ [0, 1]. We are given that w 6= 0 or 1. Since
f is continuous and g is increasing, taking limit as k → ∞, we get f(w) ≤ g(w−) ≤ g(w+) ≤ f(w).
Since g(w−) ≤ g(w) ≤ g(w+), we get f(w) = g(w), a contradiction.

1
4. Fixing x and substituting u = xy in B, we get
Z 1Z 1 Z 1 Z x Z 1 Z 1 Z 1
xy uu uu
B= (xy) dy dx = du dx = dx du = − uu (ln u) du.
0 0 0 0 x 0 u x 0

Z 1 1
Then A − B = uu (1 + ln u) du = uu = 0. Therefore, A = B.
0 0+

5. Lemma If there exist M ∈ R and ε > 0 such that f (n) (x) > ε for all x ≥ M, then f is unbounded above.
Proof. Let cn−1 = f (n−1) (M ). Since f (n−1) (x) > cn−1 + εx for all x > M by the mean value theorem,
f (n−1) is unbounded above. Then there exists M 0 ∈ R such that f (n−1) (x) > ε for all x ≥ M 0 . Repeating
this n − 1 times more, we get f is unbounded above. This proved the lemma.

Now assume such a function f(x) exists. Consider

A(x) = f (1) (x) + f (2) (x) + f (3) (x), B(x) = f (4) (x) + f (5) (x) + · · · + f (12) (x),

C(x) = f (13) (x) + f (14)(x) + · · · + f (39) (x), D(x) = f (19) (x) + f (20) (x) + · · · + f (57) (x).

Let E(x) = A(x) + B(x) + C(x). We are given that 1 ≤ A(x), B(x), C(x) ≤ 3 (hence 3 ≤ E(x) ≤ 9) and
D(x) ≥ 1 for all real x. Now

D(x) = A(18) (x) + B (18) (x) + C (18)(x) = E (18)(x).

By the lemma, E is unbounded above, a contradiction to E(x) ≤ 9 for all real x.

6. (Solution 1 due to Li Siwei and Li Zhiming) Let {v1 , v2, . . . , vn} and {e1 , e2, . . . , en−1} be orthonormal
bases of V and E respectively. We will show there exists (c1 , c2, . . ., cn) ∈ Kn (K = R or C) which is
not (0, 0, . . ., 0) and v = c1v1 + c2 v2 + · · · + cnvn is orthogonal to e1 , e2, · · · , en−1. Then v is orthogonal
to E.
The conditions v 6= 0 and hv, ei i = c1 hv1 , eii + c2 hv2, ei i + · · · + cn hvn , eii = 0 for i = 1, 2, . . ., n − 1
is equivalent to the linear transformation T : Kn → Kn−1 defined by
   hv , e i hv2, e1 i ··· hvn , e1i  
c1 1 1 c1
 c2   hv1 , e2i hv2, e2 i ··· hvn , e2i   c2 
T  
 ...  =  .. .. .. ..  . 
  .. 
. . . .
cn hv1 , en−1i hv2 , en−1i · · · hvn, en−1i cn

has a null space not equal to {0}. This is the case because the range of T cannot be n-dimensional in
Kn−1. So such a v = c1v1 + c2v2 + · · · + cnvn exists.
(Solution 2) Let W = V ∩ E. Let V 0 be the orthogonal complement of W in V. Similarly, let E 0 be the
orthogonal complement of W in E. Since V 0 ∩ E 0 ⊆ V ∩ E ∩ W ⊥ = {0}, so V 0 ∩ E 0 = {0}.

Also, V 0 + E 0 ⊥ W and dim V 0 = dim E 0 + 1. So dim(V 0 + E 0 ) = dim V 0 + dim E 0 = 2(dim V 0 ) − 1,