Hermite's identity explained

In mathematics, Hermite's identity, named after Charles Hermite, gives the value of a summation involving the floor function. It states that for every real number x and for every positive integer n the following identity holds:[1] [2]

n-1
\sum
k=0

\left\lfloorx+

k
n

\right\rfloor=\lfloornx\rfloor.

Proofs

Proof by algebraic manipulation

Split

x

into its integer part and fractional part,

x=\lfloorx\rfloor+\{x\}

. There is exactly one

k'\in\{1,\ldots,n\}

with

\lfloorx\rfloor=\left\lfloorx+

k'-1
n

\right\rfloor\lex<\left\lfloorx+

k'
n

\right\rfloor=\lfloorx\rfloor+1.

By subtracting the same integer

\lfloorx\rfloor

from inside the floor operations on the left and right sides of this inequality, it may be rewritten as

0=\left\lfloor\{x\}+

k'-1
n

\right\rfloor\le\{x\}<\left\lfloor\{x\}+

k'
n

\right\rfloor=1.

Therefore,

1-k'
n

\le\{x\}<1-

k'-1
n

,

and multiplying both sides by

n

gives

n-k'\len\{x\}<n-k'+1.

Now if the summation from Hermite's identity is split into two parts at index

k'

, it becomes
n-1
\begin{align} \sum
k=0

\left\lfloorx+

k
n

\right\rfloor &

k'-1
=\sum
k=0

\lfloor

n-1
x\rfloor+\sum
k=k'

(\lfloorx\rfloor+1)=n\lfloorx\rfloor+n-k'\\[8pt] &=n\lfloorx\rfloor+\lfloorn\{x\}\rfloor=\left\lfloorn\lfloorx\rfloor+n\{x\}\right\rfloor=\lfloornx\rfloor. \end{align}

Proof using functions

Consider the function

f(x)=\lfloorx\rfloor+\left\lfloorx+

1
n

\right\rfloor+\ldots+\left\lfloorx+

n-1
n

\right\rfloor-\lfloornx\rfloor

Then the identity is clearly equivalent to the statement

f(x)=0

for all real

x

. But then we find,

f\left(x+

1
n

\right)=\left\lfloorx+

1
n

\right\rfloor+\left\lfloorx+

2
n

\right\rfloor+\ldots+\left\lfloorx+1\right\rfloor-\lfloornx+1\rfloor=f(x)

Where in the last equality we use the fact that

\lfloorx+p\rfloor=\lfloorx\rfloor+p

for all integers

p

. But then

f

has period

1/n

. It then suffices to prove that

f(x)=0

for all

x\in[0,1/n)

. But in this case, the integral part of each summand in

f

is equal to 0. We deduce that the function is indeed 0 for all real inputs

x

.

Notes and References

  1. .
  2. .