Mathematical Reflections

Problem O111 by Titu Andreescu

Prove that, for each integer n ≥ 0, the number
      2       2
n n 2 n n n 2 n
+2 +2 + ... +2 +2 + ...
0 2 4 1 3 5
is triangular.

Solution by Darij Grinberg. First, we consider a more general setting:

Theorem
  1. Let q be a real. Define a sequence (f0 , f1 , f2 , ...) by fn =
P n k
q for every n ∈ N, and define a sequence (g0 , g1 , g2 , ...) by gn =
k∈N 2k 
n
q k for every n ∈ N. 1 Then, for every n ∈ N, we have the
P
k∈N 2k + 1
matrix equality    n
fn qgn 1 q
= (1)
gn f n 1 1
and the equalities
fn2 − qgn2 = (1 − q)n ; (2)
2fn gn = g2n . (3)
For any a ∈ N and b ∈ N, we have
fa+b = fa fb + qga gb ; (4)
ga+b = fa gb + ga fb . (5)

Remark. Here, N means the set {0, 1, 2, ...} .

Proof of Theorem 1. We will prove (1) by induction on n:
Induction base. The definitions of f0 and g0 yield
X 0  X  1, if k = 0;
k
f0 = q = · qk
2k 0, if k 6= 0
k∈N k∈N
     
0 1, if 2k = 0; 1, if k = 0;
since = =
2k 0, if 2k 6= 0 0, if k 6= 0
= 1 · q0 = 1 · 1 = 1
  
1 n k n
q k are well-defined, for the following reason:
P P
Here, the sums q and
k∈N
P 2k k∈N 2k + 1
A sum of the form a (k) (where a : N → R is some map) is well-defined (without using analysis)
k∈N
 finitely many k ∈ N satisfy a (k) = 0. But this condition is easily verified
if all but  for our sum
P n k n
q (in fact, all k ∈ N \ {0, 1, ..., n} satisfy k > n, thus 2k ≥ k > n, thus = 0, thus
k∈N 2k 2k
   
n k n k
q = 0; thus, all but finitely many k ∈ N satisfy q = 0) and (similarly) for our sum
2k  2k
n
q k . Similar arguments can show that all other sums of the form
P P
a (k) that we will
k∈N 2k + 1 k∈N
encounter in our solution will be well-defined.

1
and
 
  
X  0 1, if 2k + 1 = 0;
0 X
 since 2k + 1 = = 0, 
g0 = qk = 0 · qk 0, if 2k + 1 6= 0
2k + 1
k∈N k∈N because k ∈ N yields 2k + 1 6= 0
= 0.
Hence,        0
f0 qg0 1 q·0 1 0 1 q
= = = .
g0 f 0 0 1 0 1 1 1
In other words, (1) holds for n = 0. This completes the induction base.
Induction step. Let N ∈ N. Assume that (1) holds for n = N. We have to show
that (1) holds for n = N + 1 as well.
The recurrence of the binomial coefficients shows that
     
N +1 N N
= + for every a ∈ Z. (6)
a a a−1
 
n
(Here, we are using the convention that = 0 if a < 0.)
a
Since (1) holds for n = N, we have
   N
fN qgN 1 q
= . (7)
gN fN 1 1
But the definitions of fN +1 and gN +1 yield
X N + 1 X  N   N 
k
fN +1 = q = + qk
k∈N
2k k∈N
2k 2k − 1
       
N +1 N N
as = + (by (6), applied to a = 2k)
2k 2k 2k − 1
X N  X N  X N  X N 
k k k
= q + q = q + qk
k∈N
2k k∈N
2k − 1 k∈N
2k k∈N;
2k − 1
k≥1
 P P 
here we replaced the sign by an sign, since the addend for k = 0 is zero
 k∈N k∈N; 
    k≥1   
 N N N 
= (as = = 0 for k = 0)
2k − 1 2·0−1 −1
X N  X N

= qk + q k+1
k∈N
2k k∈N |
2 (k + 1) − 1 |{z}
 {z  } =q·qk
N
=
2k + 1
(here we substituted k + 1 for k in the second sum)
X N  X N 
k
= q +q q k = fN + qgN = 1 · fN + q · gN
k∈N
2k k∈N
2k + 1
| {z } | {z }
=fN =gN
(by the definition of fN ) (by the definition of gN )

2
and
X N + 1 X  N   N

k
gN +1 = q = + qk
k∈N
2k + 1 k∈N |
2k + 1 2k + 1 − 1
 {z   }
N N
= +
2k
  + 1 2k 
N N
= +
2k 2k + 1
       
N +1 N N
as = + (by (6), applied to a = 2k + 1)
2k + 1 2k + 1 2k + 1 − 1
X  N   N 
= + qk
k∈N
2k 2k + 1
X N  X N 
k
= q + q k = fN + gN = 1 · fN + 1 · gN .
k∈N
2k k∈N
2k + 1
| {z } | {z }
=fN =gN
(by the definition of fN ) (by the definition of gN )

Hence,
     
fN +1 qgN +1 1 · fN + q · gN q (1 · fN + 1 · gN ) 1 · fN + q · gN 1 · qgN + q · fN
= =
gN +1 fN +1 1 · fN + 1 · gN 1 · fN + q · gN 1 · fN + 1 · gN 1 · qgN + 1 · fN
     N  N +1
1 q fN qgN 1 q 1 q 1 q
= = = .
1 1 gN fN 1 1 1 1 1 1
| {z }
 N
1 q 
=
1 1
(by (7))

In other words, (1) holds for n = N + 1. This completes the induction step. Thus, the
induction proof is complete, so that (1) is proven for all n ∈ N.
Now, (2) follows from
   n 
2 2 fn qgn 1 q
fn − qgn = fn · fn − qgn · gn = det = det (by (1))
gn fn 1 1
  n
1 q
= det = (1 · 1 − q · 1)n = (1 − q)n .
1 1

3
 For any a ∈ N and b ∈ N, we have
   a+b
fa+b qga+b 1 q
= (by (1), applied to n = a + b)
ga+b fa+b 1 1
 a  b
1 q 1 q
= ·
1 1 1 1
| {z }

| {z }

fa qga  fb qgb 
= =
ga f a gb f b
(by (1), applied (by (1), applied
to n=a) to n=b)
     
fa qga fb qgb fa · fb + qga · gb fa · qgb + qga · fb
= · =
ga f a gb f b ga · fb + fa · gb ga · qgb + fa · fb
 
fa fb + qga gb q (fa gb + ga fb )
= .
fa gb + ga fb fa fb + qga gb

Thus, fa+b = fa fb + qga gb and ga+b = fa gb + ga fb , so that (4) and (5) are proven. For
every n ∈ N, we have

g2n = gn+n = fn gn + gn fn (by (5))

= 2fn gn ,

and (3) follows.

Altogether, we have now proven Theorem 1.
From now on, we set q = 2. Then,
 2n
2 2
f2n − qg2n = 1 − q  (by (2), applied to 2n instead of n)
|{z}
=2
2n
= (1 − 2) = (−1)2n = 1 (since 2n is even) (8)

for every n ∈ N. Hence,

1 1 2
fn2 gn2 = (fn gn )2 = (2fn gn )2 = g2n (by (3))
4 4  

1 1 2 1 1 
 2 2 2 


= · · qg2n = · · f2n − f2n − qg2n
4 q 4 q

 | {z }
|{z} =1
1 (by (8))
=
2
(since q=2)
1 1 2

1 2
1
= · · f2n −1 = f2n − 1 = (f2n − 1) (f2n + 1)
4 2 8 8  
1 f2n − 1 f2n + 1 1 f2n − 1 f2n − 1
= · · = · · +1
2 2 2 2 2 2

4
 f2n − 1
for every n ∈ N. Since ∈ Z for every n ∈ N (since
2
 
   
 2n P 2n k 
20 + 2 −1
   
P 2n k P 2n k
2·0 k∈N; 2k

q −1 2 −1
f2n − 1 k∈N 2k k∈N 2k k≥1
= = =
2  2  2 2
 
P 2n k 
1 + 2 −1

k∈N; 2k      
k≥1 2n 0 2n 0
= since 2 = 2 =1·1=1
2 2·0 0
 
P 2n k
2
k∈N; 2k
k≥1
X 2n
= = 2k−1 ∈ Z
2 k∈N;
2k
k≥1

), this yields that fn2 gn2 is a triangular number for every n ∈ N. This is exactly what
the problem asked us to prove.

Remark. Theorem 1 could be
proved more quickly using the binomial formula
√
n √ n
applied to 1 + q and 1 − q . However, such a proof would fail if we replace
R by a field of characteristic 2 and q is a square in that field. The proof given above
works over any field and for any q. (Then again, from a deeper viewpoint, it is just a
straightforward elementarization of the proof using the binomial formula.)