LINEAR ALGEBRA OF PASCAL MATRICES
LINDSAY YATES
Abstract. The famous Pascal’s triangle appears in many areas of
mathematics, such as number theory, combinatorics and algebra.
Pascal matrices are derived from this triangle of binomial coeffi-
cients, which create simplistic matrices with interesting properties.
We explore properties of these matrices and the inverse of the Pascal
matrix plus the identity matrix times any positive integer. We
further consider a unique matrix called the Stirling matrix, which
can be factorized in terms of the Pascal matrix.
1. Introduction
The ancient arithmetic triangle, today known as Pascal’s triangle, is
an infinite numerical table represented in triangular form. The numbers
displayed in the triangle are called binomial coefficients,
n
k
, which
represent the number of ways of picking k unordered outcomes from n
possibilities. Each entry in the triangle is obtained by adding together
two entries from the row above: the one directly to the left and the
one directly to the right; this pattern can be seen in the image below.
The Pascal’s triangle has been known for over ten centuries. The set
of numbers that form the Pascal’s triangle were known before Blaise
Date: December 5, 2014.
1
LINEAR ALGEBRA OF PASCAL MATRICES 2
Pascal, although he is attributed with being the first one to publish
the information known about the triangle in his treatise, Trait´e du triangle
arithm´etique. The numbers originally arose from Indian studies
of combinatorics and the Greeks interest in figurate numbers. These
numbers were continually discussed by Islamic mathematicians during
the 10th century and in the 11th century by a Persian poet named
Omar Khayyam. They were also seen in China during the 13th century.
The Pascal’s triangle was officially published in Pascal’s treatise
soon after his death in 1665.
This triangle arises in many areas of mathematics such as algebra,
probability, and combinatorics. We were motivated by the Pascal’s
triangle prominence in the field of mathematics and its many applications,
in particular Pascal matrices. We wanted to further our studies
to consider the various properties and unique connections that Pascal
matrices has to other functions and number sequences.
2. Pascal Matrices
The Pascal’s triangle can be transcribed into a matrix containing
the binomial coefficients as its elements. We can form three types of
matrices: symmetric, lower triangular, and upper triangular, for any
integer n > 0.
The symmetric Pascal matrix of order n is defined by Sn = (sij ), where
sij =
i + j − 2
j − 1
for i, j = 1, 2, ...., n (1)
We can define the lower triangular Pascal matrix of order n by Ln =
(lij ), where
lij =
(
i−1
j−1
if i ≥ j
0 otherwise
(2)
The upper triangular Pascal matrix of order n is defined by Un = (uij ),
where
uij =
(
j−1
i−1
if j ≥ i
0 otherwise
(3)
LINEAR ALGEBRA OF PASCAL MATRICES 3
We notice that Un = (Ln)
T
, for any positive integer n.
For example, for n = 5 we have:
S5 =
1 1 1 1 1
1 2 3 4 5
1 3 6 10 15
1 4 10 20 35
1 5 15 35 70
L5 =
1 0 0 0 0
1 1 0 0 0
1 2 1 0 0
1 3 3 1 0
1 4 6 4 1
U5 =
1 1 1 1 1
0 1 2 3 4
0 0 1 3 6
0 0 0 1 4
0 0 0 0 1
These Pascal matrices have some interesting properties, which we present
next.
Theorem 2.1. [1] Let Sn be the symmetric Pascal matrix of order n
defined by (1), Ln be the lower triangular Pascal matrix of order n de-
fined by (2), and Un be the upper triangular Pascal matrix of order n
defined by (3), then Sn = LnUn.
Proof. Let Ln be the lower triangular Pascal matrix of order n defined
by (2) and Un be the upper triangular Pascal matrix of order n defined
by (3). By direct multiplication of matrices Ln and Un we obtain the
ij-th element of the product LnUn:
Xn
k=1
likukj =
Xn
k=1
liklkj , since Un = (Ln)
T
.
Then, Xn
k=1
likljk =
Xn
k=1
i − 1
k − 1
j − 1
k − 1
=
X
j
k=1
i − 1
k − 1
j − 1
k − 1
=
=
X
j
k=1
i − 1
k − 1
j − 1
j − k
, since lik = 0 for k>j.
The Vandermonde identity says that:
Xn
t=0
m
t
n
n − t
=
m + n
n
, for any m,n,t ∈ N (4)
Let m = i − 1, n = j − 1, and t = k − 1 in (4).
Then, X
j
k=1
i − 1
k − 1
j − 1
j − k
=
i + j − 2
j − 1
= sij , the entries of the symmetric
Pascal matrix Sn. Hence, Sn = LnUn.
LINEAR ALGEBRA OF PASCAL MATRICES 4
This result can be used to determine the determinant of the symmetric
Pascal matrix, Sn.
Corollary 2.2. If Sn is the symmetric Pascal matrix of order n defined
by (1), then det(Sn) = 1, for any positive integer n.
Proof. Let Sn be the symmetric Pascal matrix of order n defined by
(1). By Theorem 2.1, we know that Sn = LnUn, where Ln is the lower
triangular Pascal matrix of order n defined by (2) and Un is the upper
triangular Pascal matrix of order n defined by (3). Since Ln and Un
are triangular matrices, then det(Ln) = 1 and det(Un) = 1. It follows
that det(Sn) = det(LnUn) = det(Ln)det(Un) = 1.
Definition 2.3. [5] Let A and B be n × n matrices. We say that
A is similar to B if there is an invertible n × n matrix P such that
P
−1AP = B.
Theorem 2.4. [1] Let Sn be the symmetric Pascal matrix of order n
defined by (1), then Sn is similar to its inverse S
−1
n
.
This result shows the following property of the eigenvalues of Sn.
Corollary 2.5. [1] Let Sn be the symmetric Pascal matrix of order
n defined by (1). Then the eigenvalues of Sn are pairs of reciprocal
numbers.
Proof. Let Sn be the symmetric Pascal matrix of order n defined by
(1) and λ be an eigenvalue of Sn. Since the det(Sn) = 1, we know Sn
is invertible. It follows that λ 6= 0 and, λ
−1
is an eigenvalue of S
−1
n
.
Since Sn and S
−1
n
are similar by Theorem 2.4, then Sn and S
−1
n have
the same eigenvalues. Hence, λ and λ
−1 are eigenvalues of Sn, and the
eigenvalues of Sn are pairs of reciprocal numbers.
Remark 1. If n is odd, since the eigenvalues must come in pairs, one
of the eigenvalues must be equal to 1.
Example 2.6. The eigenvalues of the symmetric Pascal matrix, S2,
are λ1 =
3 + √
5
2
and λ2 =
3 −
√
5
2
, where λ1λ2 = 1 gives a reciprocal
pair.
Example 2.7. For n odd, let n = 3. Then the eigenvalues of the symmetric
Pascal matrix, S3, are λ1 = 4+√
15, λ2 = 4−
√
15, and λ3 = 1.
We note that λ1λ2 = 1 gives a reciprocal pair and λ3 = 1 is a selfreciprocal.
LINEAR ALGEBRA OF PASCAL MATRICES 5
In their paper, A Note on Pascal’s Matrix, Cheon, Kim, and Yoon
found an interesting factorization of the lower triangular Pascal matrix,
Ln.
Theorem 2.8. [4] Let Gk =
In−k OT
O Sk
be a matrix of order n, where Sk is the matrix of order k defined by:
sij =
(
1 if i ≥ j
0 j>i
for every k = 1, 2, ..., n. Then the lower triangular Pascal matrix of
order n can be written as: Ln = GnGn−1 · · · G1.
For example,
L4 =
1 0 0 0 0
1 1 0 0 0
1 1 1 0 0
1 1 1 1 0
1 1 1 1 1
1 0 0 0 0
0 1 0 0 0
0 1 1 0 0
0 1 1 1 0
0 1 1 1 1
1 0 0 0 0
0 1 0 0 0
0 0 1 0 0
0 0 1 1 0
0 0 1 1 1
1 0 0 0 0
0 1 0 0 0
0 0 1 0 0
0 0 0 1 0
0 0 0 1 1
=
1 0 0 0 0
1 1 0 0 0
1 2 1 0 0
1 3 3 1 0
1 4 6 4 1
To further our studies of the lower triangular Pascal matrix, we are
interested in studying the inverse of this matrix.
Theorem 2.9. [6] Let Ln be the lower triangular Pascal matrix of order
n defined by (2), then
L
−1
n = ((−1)i−j
lij ).
Proof. We will show that Ln · L
−1
n = In.
By direct multiplication of Ln and L
−1
n we get the ij-th element of the
product:
Xn
k=1
(−1)k−j
liklkj . (5)
LINEAR ALGEBRA OF PASCAL MATRICES 6
If i < j then the element (5) is zero and if i = j, then the element (5)
is 1.
We will show that for i > j, the element (5) is zero.
If i > j, the element is:
i − 1
j − 1
j − 1
j − 1
−
i − 1
j
j
j − 1
+ ... + ( − 1)i − j
i − 1
i − 1
i − 1
j − 1
=
=
i − 1
j − 1
i − j
i − j
−
i − j
i − j − 1
+ ... + ( − 1)i − j
i − j
0
= 0.
Hence, Ln · L
−1
n = In and L
−1
n = ((−1)i−j
lij ) is the inverse of Ln.
There is another unique way in which the inverse of the lower triangular
Pascal matrix can be written, using the Hadamard product of matrices.
Definition 2.10. [8] Let A, B be m × n matrices. The Hadamard
product of A and B is defined by:
[A ◦ B]ij = [A]ij [B]ij , for 1 ≤ i ≤ m, 1 ≤ j ≤ n.
Theorem 2.11. [8] Let τn be a n × n lower triangular matrix defined
below as:
τij =
(
(−1)i−j
if i ≥ j ≥ 1,
0 otherwise
The inverse of the lower triangular matrix can be found using the
Hadamard product:
L
−1
n = Ln ◦ τn
For example, if n = 4, then:
L
−1
4 =
1 0 0 0
−1 1 0 0
1 −2 1 0
−1 3 −3 1
=
1 0 0 0
1 1 0 0
1 2 1 0
1 3 3 1
◦
1 0 0 0
−1 1 0 0
1 −1 1 0
−1 1 −1 1
3. Inverse of the Pascal Matrix Plus An Integer
In this section we are going to describe the inverse of Ln +kIn where
Ln is the lower triangular matrix of order n defined by (2), In is the
identity matrix and k is a positive integer. We call Ln +kIn the Pascal
matrix plus an integer. First, we are considering the case for k = 1. By
direct computation of the inverse of Ln+In, we can observe that there is
LINEAR ALGEBRA OF PASCAL MATRICES 7
a close relation between the inverse of Ln+In and the Pascal matrix Ln.
For example, for n = 4,
L4 + I4 =
2 0 0 0
1 2 0 0
1 2 2 0
1 3 3 2
,
(L4 + I4)
−1 =
1
2
0 0 0
−1
4
1
2
0 0
0
−1
2
1
2
0
1
8
0
−3
4
1
2
=
1 0 0 0
1 1 0 0
1 2 1 0
1 3 3 1
◦
1
2
0 0 0
−1
4
1
2
0 0
0
−1
2
1
2
0
1
8
0
−3
4
1
2
In their paper, Explicit Inverse of the Pascal Matrix Plus One, S.L.
Yang and Z.K. Liu showed that the inverse of Ln +In is the Hadamard
product between Ln and a lower triangular matrix. We are going to
describe this unique lower triangular matrix next. For this we need to
define the Euler polynomials.
Euler