2/5/22, 6:37 PM Method of Differences | Brilliant Math & Science Wiki

Method of Differences

The method of finite differences gives us a way to calculate a polynomial using its values at several consecutive points. T
often a good approach to finding the general term in a pattern, if we suspect that it follows a polynomial form.

Contents
Difference Table
Proof of Theorem

Applications

Additional Problems

Difference Table

Suppose we are given several consecutive integer points at which a polynomial is evaluated. What information does this
about the polynomial? To answer this question, we create the following table, known as the difference table:

n f (n) D1 (n)​ D2 (n)

​ D3 (n)
​ …
1
2 ​ ​

3
⋮

In the first column, we fill out the points at which the polynomial is evaluated (which I will assume is 1, 2, 3 …).
In the second column, we fill out the corresponding values of the polynomial at those points.
In the third column, we calculate the difference between two entries in the previous column. This is known as the first
difference and is given by D1 (n) ​ = f (n + 1) − f (n).
In the fourth column, we calculate the difference between two entries in the previous column. This is known as the sec
difference and is given by D2 (n) ​ = D1 (n + 1) − D1 (n).
​ ​

We continue building out subsequent columns of the table in this way, with Dk+1 (n) ​ = Dk (n + 1) − Dk (n).
​ ​

For example, if we are given that f (n) is a quadratic polynomial satisfying

​ f (1) = 4, ​ f (2) = 3, ​ f (3) = 4, ​ f (4) = 7, ​ f (5) = 12, ​

then the difference table is as follows:

n f (n) D1 (n) ​
D2 (n)​
D3 (n) ​ …
1 4 −1 2 0
2 3 1 2 0
3 ​
4 ​
3 ​
2 ​

4 7 5
5 12
⋮

https://brilliant.org/wiki/method-of-differences/#:~:text=The method of finite differences,it follows a polynomial form. 1/7

2/5/22, 6:37 PM Method of Differences | Brilliant Math & Science Wiki

Note that since we are not given f (6), we are unable to calculate D1 (5), D2 (4), D3 (3) or D4 (2). Now, there isn't (yet)
​ ​ ​ ​

reason why this table would ever end; if we are given infinitely many values, we can always calculate the differences of te
However, here is an interesting fact.

THEOREM

If a polynomial f (x) has degree k , then the k th difference is constant.

Furthermore, the k th difference is equal to k! times the leading coefficient of f (x). □ ​

See proof below.

In the above example, we see that D2 (1) = D2 (2) = D2 (3) = 2. Since f (x) is a quadratic polynomial, the above fact
​ ​

that D2 (n) = 2 for all n, and this allows us to fill in other entries in the table. We get that D2 (4) = 2 so D1 (5) = 7 and
​ ​ ​

f (6) = 19. These values are previously unknown to us, but the difference table allows us to calculate them without know
polynomial f (x)!

We can continue the difference table as follows:

n f (n) D1 (n)
​ D2 (n)
​ D3 (n) ​ …
1 4 −1 2 0
2 3 1 2 0
3 4 3 2 0
4 ​

7 ​

5 ​

2 ​

0 ​

5 12 7 2 0
6 19 9 2 0
⋮

EXAMPLE

If f (x) is a quadratic polynomial that satisfies f (1) = 3, f (2) = 9, f (3) = 19, what is f (4)?

We construct the difference table using the initial data, and obtain

n f (n) D1 (n)
​ D2 (n)
​

1 3 6 4
2 ​ 9 ​ 10 ​ ​

3 19
4

From the above theorem, we know that D2 (n) is a constant, and thus is equal to 4. This allows us to complete the table
​

follows:

n f (n) D1 (n)
​ D2 (n)
​

1 3 6 4
2 ​ 9 ​ 10 ​ 4 ​

3 19 14 4
4 33 18 4

Hence, f (4) = 33. □ ​

EXAMPLE

https://brilliant.org/wiki/method-of-differences/#:~:text=The method of finite differences,it follows a polynomial form. 2/7

2/5/22, 6:37 PM Method of Differences | Brilliant Math & Science Wiki

If g(x) is a cubic polynomial that satisfies

​ g(0) = 0, ​ g(1) = 3, g(2) = 2, ​ g(3) = 1, ​

what is g(5)?

We construct the difference table, using the initial data. Note that it doesn't matter what number we start with, as long
the difference is 1.

n g(n) D1 (n)
​ D2 (n)
​ D3 (n) ​

0 0 3 −4 4
1 ​
3 ​
−1 0 ​

2 2 −1
3 1

From the above, we know that D3 (n) is a constant, and thus it is equal to 4. This allows us to complete the table as foll
​

Note that we want to calculate g(5) and thus need to extend the table by more rows.

n g(n) D1 (n)
​ D2 (n)
​ D3 (n) ​

0 0 3 −4 4
1 3 −1 0 4
2 ​
2 ​
−1 4 ​
4 ​

3 1 3 8 4
4 4 11 12 4
5 15 23 16 4

Hence, g(5) = 15. □ ​

EXAMPLE

Construct the difference table for the function fk (n)​ = (n − 1) × (n − 2) × ⋯ × (n − k) for n = 1 to k + 1. Note that fk (n) is a polynomial of degree k
​

This is a specially chosen function. We easily see that

f (i) = {
0 i = 1 to k
i = k + 1.
​ ​

k!

This allows us to easily calculate the first difference column as

D1 (i) = {
0 i = 1 to k − 1
i = k.
​ ​

k!

Similarly, for the j th difference column, we have

Dj (i) = {
0 i = 1 to k − j
i = k − j + 1.
​ ​ ​

k!

To summarize the above, the difference table is

https://brilliant.org/wiki/method-of-differences/#:~:text=The method of finite differences,it follows a polynomial form. 3/7

2/5/22, 6:37 PM Method of Differences | Brilliant Math & Science Wiki

n fk (n)
​
D1 (n) ​
D2 (n)
​
… Dk−1 (n)​
Dk (n) ​

1 0 0 0 … 0 k!
2 0 0 0 … k!
3 ​
0 ​
0 ​
0 ​
… ​ ​ ​ ​

⋮
k 0 k!
k+1 k! □ ​

Proof of Theorem

THEOREM

If a polynomial f (x) has degree k , then the k th difference is constant.

Furthermore, the k th difference is equal to k! times the leading coefficient of f (x). □ ​

PROOF

Let's consider a general polynomial of degree k . We have

f (n) = ak nk + ak−1 nk−1 + … + a1 n + a0 .

​ ​ ​ ​

What is the value of D1 (n)? By definition, we have

​

k
D1 (n) = f (n + 1) − f (n) = ∑ ai [(n + 1)i − ni ] .
​ ​ ​

i=1

By the binomial theorem, we have

(n + 1)i − ni = ( )ni−1 + ( )ni−2 + ⋯ + ( )n1 + ( )n0 ,

i i i i
1 2 i−1
​ ​ ​ ​

implying that there are no terms of degree nk in D1 (n). Furthermore, the coefficient of nk−1 comes solely from the te
​

ak [(n + 1) − n ], and is equal to kak .

​
k k
​

We may likewise continue by induction to consider the second difference, third difference, etc. Each time, we see that t
degree of the polynomial decreases by 1. Hence, by the time we get to the k th difference, it is a polynomial of degree 0
Since the only polynomials of degree 0 are the constants, this implies Dk (n) is a constant polynomial. ​

What happens to the leading coefficient at each step? We see that it gets multiplied by the degree of the (current)
polynomial. Since the degree decreases by 1 at each step, we see that it gets multiplied by k × (k − 1) × ⋯ × 2 × 1
Dk
. Hence, we have Dk ​ = ak × k! or ak =
​ ​

k! . □
​

​ ​

If we have an entire row of the difference table, we can reconstruct the polynomial using the following theorem:

THEOREM

Let fk (n)​ = (n − 1) × (n − 2) × ⋯ × (n − k) as defined above.

If f (n) is a polynomial of degree k , then

https://brilliant.org/wiki/method-of-differences/#:~:text=The method of finite differences,it follows a polynomial form. 4/7

2/5/22, 6:37 PM Method of Differences | Brilliant Math & Science Wiki
k
Di (1)
f (n) = f (1) + ∑ × fi (n). □
​

​ ​ ​ ​

i=1
i!

In particular, the quadratic polynomial f in the first example satisfying f (1) = 3, f (2) = 9, f (3) = 19 is given by

f (n) = 3 + 1!6 (n − 1) +
​
4
2! (n
​
− 1)(n − 2)
= 2n2 + 1.
​ ​

The cubic polynomial g in the second example satisfying g(0) = 0, g(1) = 3, g(2) = 2, g(3) = 1 is given by

g(n) = 3 + −1 ​
0
1! (n − 1) + 2! (n − 1)(n − 2) +
4
3! (n​ − 1)(n − 2)(n − 3)
= 13 (2n3 − 12n2 + 19n).
​ ​

PROOF

As worked out in the above example, this theorem is true for the special case of the polynomial fk (n). Observe that fo ​

2 polynomials g(n) and h(n), we clearly have Di (g + h)(n) = Di g (n) + Di h(n), and Di (αg)(n) = αDi g (n). He
​ ​ ​ ​ ​

the theorem is true for any polynomial that can be written as a linear combination of fk (n). ​

Now, given any polynomial f (n) of degree k , we will show that there are unique coefficients {αi }ki=1 such that ​ ​

k
f (n) = f (1) + ∑ αi fi (n). ​ ​ ​

i=1

These coefficients can be obtained by setting n= 2 to find α1 , and then setting n = 3 to find α2 , so on and so forth.
​ ​

get an identity, because both sides are polynomials of degree at most k , which agree on k + 1 values.

Hence, this theorem is true for all polynomials f (n). □ ​

Applications

EXAMPLE

Find a quadratic polynomial which satisfies f (1) = 3, f (3) = 6, f (5) = 10.

Notice that we are given the values at 1, 3, 5, and hence cannot apply the method of finite differences directly. Instead,
will define a new polynomial given by f (2x − 1) = g(x). With this, we get that g(1) = f (1) = 3, g(2) = f (3) =
6, g(3) = f (5) = 10.

Creating the difference table for g , we get

n g(n) D1 (n)
​ D2 (n) ​ D3 (n) ​

1 3 3 1
2 6 4
​ ​ ​

3 10
3 1 (x+1)(x+2)
Hence, we obtain g(x) =3+ 1!
​ (x − 1) + 2!
(x
​ − 1)(x − 2) = 2
. Finally, since f (2x ​ − 1) = g(x), with th
y+1
substitution x = 2
, we get
​

https://brilliant.org/wiki/method-of-differences/#:~:text=The method of finite differences,it follows a polynomial form. 5/7

2/5/22, 6:37 PM Method of Differences | Brilliant Math & Science Wiki

(x + 1)(x + 2) ( y+1 y+1

2 + 1) ( 2 + 2) (y + 3)(y + 5)
f (y) = f (2x − 1) = g(x) = = = . □
​ ​

2 2 8
​ ​ ​ ​

EXAMPLE

Prove that there is no polynomial (of finite degree) which satisfies f (n) = 2n for all positive integers.

Construct the difference table for f (n) = 2n . Since D1 (n) = f (n + 1) − f (n) = 2n+1 − 2n = 2n = f (n), we see
​

D1 is exactly equal to f . This remains true for each k th difference column, which means that the k th difference column
​

never a constant. Hence there is no polynomial of degree k that satisfies f (n) = 2n . □ ​

EXAMPLE

Prove that there is no polynomial (of finite degree) which satisfies f (n) = Fn for all positive integers, where Fn is the nth Fibonacci Number.
​ ​

Construct the difference table for f (n) = Fn . Since D1 (n) = f (n + 1) − f (n) = Fn+1 − Fn = Fn−1 , we see that
​ ​ ​ ​ ​

D1 (n) is exactly equal to Fn−1 .

​ ​

Similarly, D2 (n) ​ = D1 (n + 1) − D1 (n) = Fn − Fn−1 = Fn−2 .

​ ​ ​ ​ ​

In general, Dk (n) ​ = Fn−k . ​

In particular, this shows that the k th difference column is never a constant. Hence there is no polynomial of degree k t
satisfies f (n) = Fn . □
​ ​

Additional Problems

TRY IT YOURSELF

A polynomial f (x) has degree 8 and f (i) = 2i for i = Submit your answer

0, 1, 2, 3, 4, 5, 6, 7, 8.

Find f (9).

TRY IT YOURSELF

https://brilliant.org/wiki/method-of-differences/#:~:text=The method of finite differences,it follows a polynomial form. 6/7

2/5/22, 6:37 PM Method of Differences | Brilliant Math & Science Wiki

Let f (x) be a quintic polynomial such that Insufficient data

f (1) =1 12
f (2) =1
f (3) =2 8
f (4) =3
​ ​

13
f (5) =5
f (6) = 8.

Determine f (7).

Note: Many people are answering this incorrectly because they think it is the Fibonacci
sequence, but this problem is asking about a quintic polynomial that passes through those
points. That does not necessarily mean the next term behaves as the Fibonacci sequence
would.

TRY IT YOURSELF

1
P (x) is a cubic polynomial, and for x = 1,  2,  3,  4, P (x) = 1+x+x2 ​ . Submit your answer

For some positive coprime integers a and b, P (5) = − ab . ​

Find the value of a + b.

This problem is a part of <Christmas Streak 2017> series.

TRY IT YOURSELF

Submit your answer

f (0) = 1
f (4) − f (2) = 2 ​ ​

f (6) − f (4) = 3

Given that f (x) is a quadratic polynomial satisfying the three equations above, what is the
value of f (8)?

1) Use the method of finite difference to find the general term of the following sequence:

0, 6, 34, 102, 228, … .

2) What comes next?

1, 2, 4, 7, 11, 16, ____

Cite as:
Method of Differences.
Brilliant.org.
Retrieved
18:33, February 5, 2022,
from
https://brilliant.org/wiki/method-of-differences/

https://brilliant.org/wiki/method-of-differences/#:~:text=The method of finite differences,it follows a polynomial form. 7/7