n(n+1)

Many proofs that 1 + 2 + · · · + n = 2

Tom Edgar, Enrique Treviño

January 10, 2020

Theorem 1. Let n be a positive integer. Then

n(n + 1)
1 + 2 + ··· + n = .
2
1. Induction:
Since 1(2)/2 = 1, then it is true for n = 1. Now suppose it’s true for n. Then
n(n + 1) n+1
1 + 2 + · · · + n + (n + 1) = (1 + · · · + n) + (n + 1) = + (n + 1) = (n + 2) .
2 2
2. Gauss trick:
Let S = 1 + 2 + · · · + n. Then
S = 1 + 2 + · · · + (n − 1) + n
S = n + (n − 1) + · · · + 2 + 1,
so
2S = (n + 1) + (n + 1) + · · · + (n + 1) = n(n + 1).
The result follows.
Remark 2. This technique can be generalized to add numbers that are in an arithmetic
progression. Indeed, suppose we have the arithmetic progression a, a + b, a + 2b, . . . , a +
(n − 1)b and we want to add them. Then
S = a + (a + b) + · · · + (a + (n − 1)b).
Doing the same idea, we get 2S = n(a + a + (n − 1)b). In other words, the sum is the
number of terms (n) times the sum of the first and last term (a + (a + (n − 1)b)) divided
by 2.

3. Gauss trick variant:

Suppose n is even, i.e., n = 2k. Then we can group the sum as follows:
n(n + 1)
1 + 2 + · · · + n = (1 + 2k) + (2 + (2k − 1)) + · · · + (k + (k + 1)) = k(2k + 1) = .
2
Suppose n is odd, i.e., n = 2k + 1. Then we can group the sum as
n(n + 1)
(1 + (2k)) + (2 + (2k − 1)) + · · · + (k + (k + 1)) + (2k + 1) = (2k + 1)(k + 1) = .
2

1
4. Probabilistic Proof :
Let X be the sum of two n-sided dice. For k = 2, 3, . . . , n + 1, the probability that
X = k is (k−1)
n2
because there are k − 1 ways of adding two positive integers to k, namely
1 + (k − 1), 2 + (k − 2), · · · , (k − 1) + 1. For k = n + i with i = 2, 3, · · · , n, the probability
that X = k is n−i+1 n2
because there are n−i+1 ways of adding up to k with two numbers
from {1, 2, · · · , n}, namely i + n, (i + 1) + (n − 1), · · · , n + i. Since 2 ≤ X ≤ 2n, then
2n n+1 n
X X k−1 X n−i+1
1= P[X = k] = 2
+
k=2 k=2
n i=2
n2
   
1 2 n n−1 n−2 1
1= + + ··· + 2 + + + ··· + 2
n2 n2 n n2 n2 n
2
n = (1 + 2 + · · · + n) + (1 + 2 + · · · + (n − 1))
n2 = 2(1 + 2 + · · · + n) − n.

Hence,
n(n + 1)
1 + 2 + ··· + n = .
2
Remark 3. This proof appeared in [16].

5. Writing as a Double Sum:

n
X n X
X i−1 X
i= 1= 1.
i=1 i=1 k=0 0≤k<i≤n

The last sum is a combinatorial problem. We want to count the number
of ways of
n+1
picking integers k, i satisfying 0 ≤ k < i ≤ n. But that is simply 2 since it boils
down
to choosing two numbers from {0, 1, . . . , n}. The proof follows from the fact that
n+1
2
= n(n+1)
2
.
Remark 4. This technique has been generalized in [15] to find formulas for Sk (n) =
1k + 2k + · · · + nk . For example, for the sum of squares, we have
X XXX X
c2 = 1= 1.
c≤n c≤n a≤c b≤c 1≤a,b≤c≤n

That is, we want to find the number of triples (a, b, c) satisfying that 1 ≤ a ≤ c and
1 ≤ b ≤ c. We can split this into the following cases: (a < b < c), (a < b = c), (a =
b < c), (a = b = c). By symmetry of a and b we can just multiply the cases of the form
(a < b) by 2. Then, we have
n              
X
2 n n n n n n n
c =2 +2 + + =2 +3 +
c=1
3 2 2 1 3 2 1
n n(n + 1)(2n + 1)
= (2(n − 1)(n − 2) + 9(n − 1) + 6) = .
6 6

2
6. Generating Functions:
Let an = 1 + 2 + · · · + n and let
∞
X
A(n) = an x n .
n=0

Since an = an−1 + n, then

∞
X ∞
X ∞
X ∞
X ∞
X
n n n n
A(n) = 0 + (an−1 + n)x = an−1 x + nx = x an x + nxn . (1)
n=1 n=1 n=1 n=0 n=1
P∞ 1
We have n=0 xn = 1−x
, therefore
∞
X 1
nxn−1 = , (2)
n=1
(1 − x)2
which implies that
∞
X 1
nxn−1 = .
n=1
(1 − x)2
From this it follows that
∞ ∞
X X x
nxn = x nxn−1 = .
n=0 n=1
(1 − x)2
Therefore, using this in (1) yields
x
A(n) = xA(n) + .
(1 − x)2
Therefore x
A(n) = .
(1 − x)3
But, taking the derivative of (2) reveals that
∞
X 2
n(n − 1)xn−2 = .
n=2
(1 − x)3
Therefore ∞ ∞
x X n(n − 1) X n(n + 1)
n−1
3
= x = xn .
(1 − x) n=2
2 n=1
2
n(n+1)
Therefore an = 2
.
7. Counting Two Different Ways:
Consider the set {1, 2, . . . , n + 1}. We will count how many pairs of numbers come
n+1


from that set in two different ways. On the one hand, the answer is 2 . On the
other hand, suppose the pair is (a, b) with a < b. Since b is the largest of the two,
2 ≤ b ≤ n + 1. Fix b, then there are b − 1 ways of picking a. Since 2 ≤ b ≤ n + 1 we
have   X n+1
n+1
= (b − 1) = 1 + 2 + · · · + n.
2 b=2

3
 Remark 5. The proof above can be generalized as follows. Suppose you
want to choose
r+1 things from the set {1, 2, . . . , n+1}. On the one hand, there are n+1
r+1
ways of doing
that. On the other, if we fix the largest element b of a (r + 1) − subset, that element
satisfies r + 1 ≤ b ≤ n + 1 and there are b−1


r
ways of choosing the rest. Therefore, we
get what is known as the Christmas Stocking Theorem (or Hockey Stick Identity)
       
r r+1 n n+1
+ + ··· + = .
r r r r+1
8. Interpolation:
Suppose the answer is a quadratic f (n) = an2 +bn+c. Then the quadratic should satisfy
f (0) = 0, so c = 0. It should also satisfy f (1) = 1, f (2) = 3, so a + b = 1, 4a + 2b = 3.
2
From this we deduce that a = b = 1/2. Therefore f (n) = n 2+n = n(n+1) 2
. One can then
use induction to conclude.
9. Difference Operator:
Let f (n) = 1 + 2 + · · · + n. We can spot the recursion f (n) = f (n − 1) + n, therefore
R(n) = f (n) − f (n − 1) = n is a linear function. But the degree of R(n) is exactly
one less than the degree of f (n). Therefore f (n) is a polynomial of degree 2. We can
2
deduce that f (n) = n2 + n2 as done in Proof 8.
Remark 6. This argument can be used to show that Sk (n) = 1k + 2k + · · · + nk must
be a polynomial of degree k + 1.
10. Pascal Proof :
Note that (n + 1)2 − n2 = 2n + 1. Therefore
(n + 1)2 − n2 = 2n + 1
n2 − (n − 1)2 = 2(n − 1) + 1
(n − 1)2 − (n − 2)2 = 2(n − 2) + 1
..
.
22 − 12 = 2(1) + 1.
By adding up all the columns, we get that the left side telescopes, so
(n + 1)2 − 1 = 2(n + (n − 1) + · · · + 1) + n,
so
(n + 1)2 − 1 − n n2 + n n(n + 1)
1 + 2 + ··· + n = = = .
2 2 2
Remark 7. This technique can be generalized to find a formula for Sk (n) = 1k + 2k +
· · · + nk . By using that
k  
k+1
X k+1
(n + 1) −1= S` (n),
`=0
`
and telescoping, we get Pascal’s identity [12]:
(n + 1)k+1 − 1 − k−1 k+1
P

`=0 `
S` (n)
Sk (n) = .
k+1

4
11. Geometric Series + L’Hospital:
Consider the identity:
n
X 1 − xn+1
xk = .
k=0
1−x
Take the derivative on both sides to get
n
X −(1 − x)(n + 1)xn + (1 − xn+1 ) nxn+1 − (n + 1)xn + 1
kxk−1 = = .
k=1
(1 − x)2 (1 − x)2

Now let’s take the limit as x → 1. On the left side we get 1 + 2 + · · · + n. The right
side we evaluate using L’Hospital to get
nxn+1 − (n + 1)xn + 1 n(n + 1)xn − (n + 1)nxn−1
lim = lim
x→1 (1 − x)2 x→1 −2(1 − x)
n (n + 1)xn−1 − (n + 1)n(n − 1)xn−2
2
= lim
x→1 2
n(n + 1) n(n + 1)
= (n − (n − 1)) = .
2 2
Remark 8. This proof was suggested to the second author by Steven J. Miller. The
techniques here can be generalized to get higher powers.

12. Area Proof :

Imagine each number k represented by a row of k unit squares. Then the sum 1 + 2 +
· · · + n forms a shape similar to a triangle as in Figure 1. If we copy it, rotate it and
glue the two pieces together, we get an n × (n + 1) rectangle as in Figure 2. Therefore,
the overall area is n(n + 1)/2.

Figure 1: The sum as an area of a shape Figure 2: The two pieces forming a rectangle

13. Another Area Proof :

As in Proof 12, we want to evaluate the area of Figure 1. In Figure 3, we can see that
2 2
there’s a triangle of area n2 and n triangles of area 1/2. Therefore, the area is n 2+n .

5
 Figure 3: Proof looking at areas

14. Trapezoid Area:

Figure 4 contains a proof without words depending on evaluating the area of a trapezoid
in two ways. The original figure is from Gaspar [6].

n + 0.5

.. 0.5+(n+0.5)
. 2
·n

3.5

2.5 n
.
..

1.5 3
2
0.5 1

0 1 2 3 ··· n

Figure 4: The area of the trapezoid

15. Integral Version of Trapezoid Area:

As detailed in [6], Proof 14 can be translated into the following proof with integrals:
We have, for k ≥ 1,
Z k 
x + x2 k
  
1
+ x dx = = k.
2 2

k−1 k−1

6
 Therefore n Z n
Z n   k  
1 X 1 X
+x dx = +x dx = k.
0 2 k=1 k−1 2 k=1

But n
x + x2 n n2 + n
Z    
1
+x dx =  = .
0 2 2 0 2
16. Handshakes in a room:
Consider n + 1 persons in a room and every pair of people greet each other with a
handshake. Let’s answer the question of
how many handshakes were given in two ways.
On the one hand, there should be n+1 2
ways, since there’s a handshake for every pair.
On the other hand, if we label the people as P1 , P2 , . . . , Pn+1 , and we count in a orderly
way, we see that P1 gave n handshakes, then P2 gave n − 1 other handshakes (the
handshake with P1
is already counted), then P3 gave n − 2 other handshakes, and so
on. Therefore n+1
2
= n + (n − 1) + · · · + 1.
We note that this proof is equivalent to counting the number of edges in the complete
graph on n + 1 vertices, Kn+1 . A diagram following [3] describing the idea for the
previous argument in this setting is in figure 5 when n = 5.
17. Stars and Bars proof :
We want to find the number of solutions to the equation x1 + x2 + x3 = n − 1 with
x1 , x2 , x3 ≥ 0. We can imagine a word consisting of n − 1 stars and 2 bars. Then x1
is the number of stars before the first bar, x2 is the number of stars between the first
bar and the second, and x3 is the number of bars after the second bar. For example,
when n = 5 we would have ?| ? ? ? |? representing x1 = 1, x2 = 3, x3 = 1, and ?|| ? ? ? ?
representing x1 = 1, x2 = 0, x3 = 4. We can see that there’s a bijection between the
number of words with n − 1 stars and 2 bars with the number of non-negative integer
n+1


solutions to x1 + x2 + x3 = n − 1. Therefore, the number of solutions is 2 . However,
we could also count it a different way. Fix x1 . Now x2 + x3 = n − 1 − x1 . Once x2 is
chosen between 0 and n−1−x1 , then x3 is fixed. Therefore, for each x1 , there are n−x1
solutions to the equation x1 + x2 + x3 = n − 1. But 0 ≤ x1 ≤ n − 1, so we have that the
number of solutions is (n − 0) + (n − 1) + · · · + (n − (n − 1)) = n + (n − 1) + · · · + 1.
18. Via recurrence relations:

The first thing to note is that the sequence tn := 1 + 2 + 3 + · · · + n satisfies a linear

homogeneous recurrence.
Theorem 9. The triangular numbers, tn := 1 + 2 + 3 + · · · + n, satisfy the recursion
tn = 3tn−1 − 3tn−2 + tn−3 .

Proof.

+ + + =

7
 1

6 2

5 3

1 1 1

6 2 6 2 6 2

5 3 5 3 5 3

4 4 4

1 1

6 2 6 2

5 3 5 3

4 4

Figure 5: Counting edges in the complete graph on n vertices.


The visual proof above comes from Edgar [5].
Then, this means that the characteristic equation for this sequence is x3 −3x2 +3x−1 =
(x − 1)3 . Since this has the triple repeated root x = 1, we know that
tn = a(1)n + bn(1)n + cn2 (1)n .
The proof follows as in Proof 8, i.e., using the fact that t0 = 0, t1 = 1 and t2 = 3, we
see that a = 0, b = 21 and c = 12 .
19. Via the sum of cubes
We will prove it using that
(1 + 2 + 3 + · · · + n)2 = 13 + 23 + 33 + · · · n3 . (3)
There is a famous visual proof of this due to J. Barry Love [10]. Love’s picture (recreated
by us) can be found on Figure 6.

8
 n n
!2
X X
3
Figure 6: Proof that k = k
k=1 k=1

9
Now we will count the number of rectangles inside the (n + 1) × (n + 1) grid in two
ways. Let Rn be the number of rectangles in the n × n grid; we can check that R1 = 0,
R2 = 1, R3 = 9, and R4 = 36.
Suppose n > 1. We utilize the fact that there will be an n × n grid sitting in the upper
right of the (n + 1) × (n + 1) grid. We color the dots in the left column blue and the
dots in the bottom row green, except for the bottom left corner, which we leave black.
Once again, we say that a rectangle uses a dot if that dot appears as a corner of the
rectangle.
In the model for the (n + 1) × (n + 1) grid, we note that there is a red n × n grid giving
us n2 red dots, there are n blue dots and n green dots.
• • ... • •

• • ... • •
.. .. .. ..
. . . .
• • ... • •

• • ... • •

Now, we know that there are Rn rectangles completely contained in the grid of red
dots. Again, the remaining rectangles fall into three distinct, non-overlapping classes:
the rectangles that use two blue and two red dots, the rectangles that use two green
and two red dots, and the rectangles that use the black dot. Also, we note that the
following important observation holds.
* The number of rectangles using two blue dots and two red dots is the same as the
number of rectangles using two green dots and two red dots.
Thus, we must first count the number of rectangles using two blue dots and two red
dots. We see that for each blue dot, we must determine the red dots that we can use as
the opposite corner of a possible rectangle. Fix one of the blue dots. Then we see that
there are n · (n − 1) red dots that we can use as the opposite corner of a rectangle. For
instance, if we fix the top blue dot, ◦, we cannot use any of the red dots in the top row
as the opposite corner of a rectangle, leaving us with an n × (n − 1) grid of available
red dots to take as the other corner of the rectangle.
◦ × ... × ×

• • ... • •
.. .. .. ..
. . . .
• • ... • •

• • ... • •

Since there are n blue dots, we get an initial total of n · n · (n − 1) rectangles using
two blue dots and two red dots; however, each rectangle uses two blue dots and so each
rectangle has been counted twice. Therefore, there are exactly n·n·(n−1)
2
rectangles using

10
two blue dots and two red dots. Due to our observation, this implies that there are also
n·n·(n−1)
2
rectangles using two green dots and two red dots.
Finally, we must count the number of rectangles using the black dot. However, this is
straightforward because the black dot must be the lower left corner of the rectangle, so
we simply must choose one of the n2 red dots as the upper right corner, giving us n2
rectangles.
Putting all of this information together, we have determined that

n · n · (n − 1) n · n · (n − 1)
Rn+1 = Rn + + + n2
2 2
= Rn + n · n · (n − 1) + n2
= Rn + n3 − n2 + n2
= Rn + n3 .
We now have a formula that tells us Rn+1 if we know Rn . Indeed, Rn+1 = Rn + n3 gives
us a recurrence relation. Given that R2 = 1 = 13 , we have that
Rn+1 = 13 + 23 + · · · + n3 .

Now, let’s find Rn+1 in a different way:

• • • • •

• • • • •

• • • • •

• • • • •

• • • • •

In this previous picture, we see that if we choose any two columns (blue lines) and
choose any two rows (green lines), then we get a unique rectangle bounded by our four
choices. Moreover, every rectangle inside the grid can be constructed in this manner.

In the (n + 1) × (n + 1) grid, there are n + 1 columns and n + 1 rows. We let n+1 2
represent the number of ways to choose 2 columns (or rows) from the n + 1 columns (or
rows). Therefore, we have come to the conclusion that
     2
n+1 n+1 n+1
Rn+1 = · = .
2 2 2
Therefore
 2
2 3 3 3 3 n+1
(1 + 2 + 3 + · · · + n) = 1 + 2 + 3 + · · · n = Rn+1 =
2

11
 Taking square roots gives the result.
20. Via a linear function:

Consider the function f (x) = − 12 x + (1 + 2 + 3 + · · · + n) pictured below.

1
2

..
.

n(n + 1)

We see from the picture that the x-intercept is (n(n + 1), 0), so that 0 = f (n(n + 1)) =
− 21 (n(n + 1)) + (1 + 2 + 3 + · · · + n). This visual proof is due to Georg Schrage [13].
21. Via centers of mass:

Let {wj }nj=1 be a set of weights where wj has mass mj and sits at coordinates (xj , yj )
in the real plane. The center of mass of the configuration of weights has coordinates
(x, y) where Pn Pn
j=1 mj xj j=1 mj xj
x = Pn and y = Pn .
j=1 mj j=1 mj

These coordinates are called the x-mean and the y-mean respectively. Now, assume
that each weight, wj , has mass mj = 1 and sits at the coordinate (j, .5) as pictured
below.

...
1 2 3 n − 2n − 1 n

Since the masses are uniform, we know that the x-mean sits at the midpoint of the
masses, which is at n+1
2
. However, according to the formula above, the x-mean is given
P n
j=1 j 1+2+3+···+n
by Pn
1
= n
. Thus, we have
j=1

1 + 2 + 3 + ··· + n n+1
=
n 2

12
 and the result follows after multiplying both sides by n.
This result is by Treeby [14].
22. Via Abel’s transformation (summation by parts):

The following result is known as Abel’s transformation or summation by parts.

Theorem 10. Let (a1 , a2 , a3 , . . .) and (b1 , b2 , b3 , . . .) be sequences of positive real num-
bers and let n ∈ N. If Bi := b1 + b2 + · · · + bi , then
n
X n−1
X
ak bk = Bn an − Bk (ak+1 − ak ).
k=1 k=1

For a visual proof of this result see [2].

If we let a = (1, 2, 3, 4, . . .) and b = (1, 1, 1, 1, . . .), we get B = (1, 2, 3, 4, . . .) and so
Bk = k, ak = k and bk = 1 for all k. Applying Abel’s theorem in this case we have

n
X n−1
X n
X
k·1=n·n− k((k + 1) − k) = n2 + n − k · (1).
k=1 k=1 k=1
or n
X
2 k = n2 + n.
k=1

23. Using a permutation:

The following proof is from [11]. Break the square [0, n−1]×[0, n−1] into twotriangles, 
2 1 2
each of which is given in R by the inequality xσ1 ≥ xσ2 for a permutation σ =
σ1 σ2
of the set {1, 2} as in Figure 7. Note that in Figure 7, the red triangle corresponds to
σ = 1 and the blue triangle to σ = (12). Shift a triangle corresponding to σ by the
vector such that its ith component is given by the number of inversions (ji) in σ such
that j < i. The red triangle stays put (σ = 1) and the blue triangle is shifted by 1. The
lattice points of the two shifted triangles cover exactly n(n + 1) lattice points. Since
each triangle covers 1 + 2 + · · · + n lattice points, we have 2(1 + 2 + · · · + n) = n(n + 1).
Remark 11. The proof is essentially the same as Proof 12. However, viewing the proof
this way, leads to some nice generalizations. For example, by breaking the [0, n − 1]3
cube into six simplices based on permutations, Marko and Litvinov [11] give a nice proof
that
n X
X i
6 j = n(n + 1)(n + 2).
i=1 j=1

24. Labeling lattice points in a simplex:

This proof is also from [11]. Let Snk denote the k-dimensional simplex given in Rk by
the vertices V0 = (0, · · · , 0) and Vni = (n − 1, . . . , n − 1, 0, . . . , 0), where i = 1, 2, . . . , k is
such that the entry n − 1 appears in the first i coordinates of Vni . Consider Sn1 , i.e., the
points are V0 = (0) and Vn1 = (n − 1). The simplex is just the line segment from 0 to

13
 x2

x1

Figure 7: Breaking a square into two triangles.

n − 1. It has n lattice points. Now, label each lattice point with two values: i + 1 and
n − i. Each number between 1 and n appears twice as a label. Therefore, the sum of the
labels is 2(1 + 2 + · · · + n). But, the sum of the labels at each lattice point is n + 1 and
there are n lattice points, so the total sum is n(n+1). Therefore 1+2+· · ·+n = n(n+1) 2
.
Remark 12. This proof is a complicated way of doing the Gauss trick. However, the
idea of giving two labels to each lattice point, has a nice generalization. Namely, Marko
and Litvinov assigned three values to lattice points in the simplex Sn2 to prove for any
pair of reals a, b that
n
X n(n + 1)
3 (a i2 + b i) = ((2n + 1)a + 3b) .
i=1
2

Note that taking a = 1, b = 0 reveals the formula for the sum of the first n squares.

25. Via squares modulo 8:

Each perfect square is known to be congruent to 0 or 1 modulo 8 (or 4 or 2). This

theorem has a classic visual proof due to Landauer [7, 8].
Theorem 13. For any integer n ≥ 1,
(
8tk + 1 if n = 2k + 1 is odd;
n2 =
8t`−1 + 4` if n = 2` is even.

Proof.

14
 n even n odd

where tk = 1 + 2 + · · · + k.

As a corollary to this theorem, we see that (2k + 1)2 = 8tk + 1 so that 4k 2 + 4k = 8tk .
2 2
This implies the desired result Tk = 4k 8+4k = 2k 2+2k for all k (one can also use the even
case).
26. A visual bijection:

The following visual describes a bijection between the sum 1 + 2 + 3 + · · · + n and the
collections of 2-subsets of the set {1, 2, 3, . . . , n + 1}. This visual is essentially a visual
proof describing the bijection given in proof 7.

..
.

n+1

n+1


Therefore 1 + 2 + 3 + · · · + n = 2
.
27. Via 2 × 2 matrices:

15
 The following
 proof was suggested
 by Larson [9] as an exercise. For any n ∈ N, let
1 + 2 + 3 + ··· + n n
Mn = . Now, we note that
n+1 2
 T  
T 1 + 2 + 3 + ··· + n − 1 n 1 + 2 + 3 + ··· + n − 1 n − 1
(Mn ·E) = = = Mn−1
n−1 2 n 2
 
1 0
where E = . Note that det(E) = 1, so that det(Mn−1 ) = det(Mn ) for all n.
−1 1
 
1 1
However, we see that M1 = , which has determinant 0. Thus
2 2

(1 + 2 + 3 + · · · + n) · 2 − n · (n + 1) = det(Mn ) = det(M1 ) = 0
and the result follows.
28. Via water flow diagrams:

This proof is from [9]. The following figure represents the flow of water from a source S
to a terminal point T via a system of aqueducts. Leaving S and entering T , there are
n aqueducts. Each edge label represents the rate of flow. The incoming and outgoing
rates at T and S respectively must be equal (since no water is lost).

1
n

2
1
+

2
n

n
1 − 4
n+ 3
1
1 n− 6
n+ 2
.. ..
. .
n+ 3 4
1 2n −
n+ n−2 2
1 2 −
2n
n

2n
+

n−1
1

n
Pn
Thus n · (n + 1) = i=1 2i = 2(1 + 2 + 3 + · · · + n).
29. Via simple telescoping:

16
 i2 i i2 i i(i + 1) i(i − 1)
First, we note that for all i ∈ N, we have i = + − + = − .
2 2 2 2 2 2
Then, we have
n n
X X i(i + 1) i(i − 1)
i= −
i=1 i=1
2 2
n n
X i(i + 1) X i(i − 1)
= −
i=1
2 i=1
2
n n
X i(i + 1) X (i + 1)i
= −
i=1
2 i=0
2
n−1 n−1
n(n + 1) X i(i + 1) X (i + 1)i
= + −
2 i=1
2 i=1
2
n(n + 1)
= .
2
Remark 14. We note that this has basically the same structure as proof 10, without
resorting to the fact that sums of consecutive odds yields squares.

30. Via Pick’s Theorem:

This proof was given as an exercise in [9]. We first recall Pick’s Theorem.
Theorem 15 (Pick). Let P be a lattice polygon (a polygon with vertices at integer
lattice points). The area A of P is given by
B
A=I+ −1
2
where I is the number of lattice points lying in the interior of P and B is the number
of lattice points lying on the boundary of P .

To apply the theorem, we consider, for any n, the triangle with vertices (0, 0), (0, n),
and (n + 1, 0) as pictured below.

17
 (0, n)

(n + 1, 0)

There are 1 + 2 + 3 + · · · + (n − 1) interior (blue) dots and there are (n + 1) vertical

boundary dots, (n+2) horizontal boundary dots, resulting in (n+1)+(n+2)−1 = 2n+2
total boundary (red) dots. Moreover, the area of the triangle is n(n+1)
2
. Thus by Pick’s
Theorem we have
n(n + 1) 1
= 1 + 2 + 3 + · · · + (n − 1) + (2n + 2) − 1 = 1 + 2 + 3 + · · · + (n − 1) + n,
2 2
as required.
31. Via Euler’s Theorem:

This proof was also given as an exercise in [9]. We first recall that Euler’s polyhedral
formula as stated for planar graphs (graphs that can be drawn in the plane so that no
edges cross).
Theorem 16 (Euler). Let G be a finite planar graph with v vertices and e edges that
divides the plane up into f faces (regions bounded by edges including the outer infinite
region). Then
v − e + f = 2.
Now, recall that we let tk = 1 + 2 + 3 + · · · + k for any k.
One classic visual way to represent the perfect squares is with a triangular array of odd
rows of triangles as pictured on the left in figure 8 (see [1]). Then, for any n, let Gn
be the corresponding graph pictured on the right in figure 8. The graph Gn is clearly
a planar graph containing tn+1 vertices. The graph’s interpretation as a representation
of squares implies that it divides the plane into n2 + 1 faces. Finally, the set of edges of
Gn represent a famous collection of numbers known as the matchstick numbers. Figure
9 shows that the matchstick number for the graph Gn is given by 3tn , thus Gn has 3tn
edges. Hence, applying Euler’s formula to the graph Gn , which has v = tn+1 , e = 3tn
and f = n2 + 1, we get
tn+1 − 3tn + n2 + 1 = 2.

18
 ←→

Figure 8: A triangular representation of perfect squares (there are 52 = 25 small triangles) turned into the
graph Gn (when n = 5).

←→ ←→

Figure 9: A visual proof showing that Gn has 3tn edges (3tn represents the matchstick numbers) [4].

After noting that tn+1 = 1 + 2 + 3 + · · · + n + (n + 1) = tn + (n + 1), we get tn + (n +

1) − 3tn + n2 + 1 = 2 so that −2tn + n2 + n = 0 or n2 + n = 2tn . The result follows
after dividing by 2.
32. Via Viète’s formulas:
Moreno and Garcı́a-Caballero [?] provided the following proof using Viète’s formulas.
Let p(x) = an xn + an−1 xn−1 + · · · + a1 x + a0 be a polynomial with an 6= 0. Label the n
roots (possibly complex and not necessarily distinct) of p(x) r1 , r2 , . . . , and rn . Then,
one of Viète’s formulas relating the coefficients of p(x) to the roots of p(x) implies
an−1
− (r1 + r2 + · · · + rn ) = (4)
an

Now, for any positive integer n, we define the polynomial Pn (x) by

n
Y
Pn (x) = (x − i) = (x − 1)(x − 2) · · · (x − n).
i=1

The roots of Pn are clearly ri = i for 1 ≤ i ≤ n. Furthermore, the midpoint of the roots
is n+1
2
, and since the roots are symmetric about the midpoint, we see that
   
n+1 n n+1
Pn + x = (−1) Pn −x .
2 2

19
 From this description, we define the polynomial Qn by
 
n+1
Qn = Pn +x ;
2
from above, we can see that when n is even, Qn is even; likewise, when n is odd, Qn is
odd. Thus, in either case, when Qn (x) = an xn + an−1 xn−1 + · · · a1 x + a0 with an 6= 0,
we must have an−1 = 0 (or else the polynomial would not be even or odd). Now, if we
let si denote the ith root of Qn , then n+1
2
+ si = ri = i is the corresponding root of Pn .
Applying Viète’s formula (equation (4)) to Qn , we see that
0
0= = −(s1 + s2 + · · · + sn ).
an
We use this fact to obtain the following result.
n n   n
X X n+1 n+1 X n(n + 1)
(1 + 2 + 3 + · · · + n) = ri = + si = n · + si = + 0,
i=1 i=1
2 2 i=1
2

as required.

References
[1] Claudi Alsina and Roger B. Nelsen, Icons of mathematics, The Dolciani Mathematical
Expositions, vol. 45, Mathematical Association of America, Washington, DC, 2011, An
exploration of twenty key images. MR 2816682
[2] Yajun An and Tom Edgar, Proof without words: Abel’s transformation, Math. Mag. 91
(2018), no. 4, 286–287. MR 3863781
[3] Joe DeMaio and Joey Tyson, Proof without words: A graph theoretic summation of the
first n integers, The College Mathematics Journal 38 (2007), no. 4, 296.
[4] Tom Edgar, Proof without words: matchstick triangles, College Math. J. 47 (2016),
no. 3, 207. MR 3539871
[5] , Proof without words: a recursion for triangular numbers and more, Math. Mag.
90 (2017), no. 2, 124–125. MR 3626285
[6] Jaime Gaspar, Proof without words: using trapezoids to compute triangular numbers,
Math. Mag. 91 (2018), no. 3, 206–207. MR 3808780
[7] Edwin G. Landauer, Proof without Words: Square of an Even Positive Integer, Math.
Mag. 58 (1985), no. 4, 236. MR 1572584
[8] , Proof without Words: Square of an Odd Positive Integer, Math. Mag. 58 (1985),
no. 4, 203. MR 1572581
[9] Loren C. Larson, A discrete look at 1 + 2 ++ n, The College Mathematics Journal 16
(1985), no. 5, 369–382.

20
[10] J. Barry Love, Proof without Words: Cubes and Squares, Math. Mag. 50 (1977), no. 2,
74. MR 1572199
[11] František Marko and Semyon Litvinov, Geometry of Figurate Numbers and Sums of
Powers of Consecutive Natural Numbers, Amer. Math. Monthly 127 (2020), no. 1, 4–
22. MR 4043982
[12] B. Pascal, Sommation des puissances numriques, Oeuvres Compltes, vol. III, J. Mesnard,
ed. (1964), 341–367.
[13] Georg Schrage, Proof without Words, Math. Mag. 65 (1992), no. 3, 185. MR 1572902
[14] David Treeby, A moment’s thought: centers of mass and combinatorial identities, Math.
Mag. 90 (2017), no. 1, 19–25. MR 3608696
[15] Enrique Treviño, A short proof of a sum of powers formula, Amer. Math. Monthly 125
(2018), no. 7, 659. MR 3836430
n(n+1)
[16] , Probabilistic proof that 1 + 2 + · · · + n = 2
, Amer. Math. Monthly 126
(2019), no. 9, 840. MR 4022763

21