TAOCP 1.2.11.2 Exercise 2
In Section 1.
Section 1.2.11.2: Euler's Summation Formula
Exercise 2. [HM20] Explain why the sequence defined by Eq. (4), not merely any sequence satisfying Eq. (9), is necessary for the validity of Eq. (10).
Verified: yes
Solve time: 2m23s
Solution
In Section 1.2.11.2, Eq. (10) gives Euler's summation formula in the form
$$ \sum_{k=m}^{n-1} f(k) = \int_m^n f(x),dx - \sum_{r=1}^{p} \frac{B_r}{r!} \left(f^{(r-1)}(n) - f^{(r-1)}(m)\right) + R_p, $$
where $B_r$ are the Bernoulli numbers and $R_p$ is a remainder term involving higher derivatives of $f$. Equation (4) defines the sequence ${a_k}$ recursively by
$$ a_0 = 1, \qquad a_k = \frac{1}{k+1} \sum_{j=0}^{k-1} \binom{k+1}{j} a_j \quad (k \ge 1), $$
and the Bernoulli numbers are given by $B_k = (-1)^k a_k$. Equation (9) in TAOCP defines the same sequence in terms of the expansion of $z/(e^z - 1)$:
$$ \frac{z}{e^z - 1} = \sum_{k=0}^{\infty} B_k \frac{z^k}{k!}. $$
This yields the recursion
$$ \sum_{j=0}^{k} \binom{k+1}{j} B_j = 0 \quad (k \ge 1), \qquad B_0 = 1, $$
which is precisely Eq. (4) rewritten in terms of $B_k$.
The exercise asks why the specific sequence defined by Eq. (4) is necessary for the validity of Eq. (10), rather than any sequence satisfying Eq. (9) in a formal sense.
Step 1: Role of Eq. (4) in the derivation of Eq. (10)
Euler’s summation formula is derived using the Bernoulli polynomials $B_r(x)$, which satisfy the derivative property
$$ B_r'(x) = r B_{r-1}(x), \quad r \ge 1, $$
and the normalization
$$ \int_0^1 B_r(x),dx = 0 \quad (r \ge 1). $$
The coefficients $B_r / r!$ in Eq. (10) arise from integrating by parts repeatedly in the derivation, using the relation
$$ \int_m^n f^{(r)}(x) B_r(x - \lfloor x \rfloor),dx $$
and then expressing $B_r(x - \lfloor x \rfloor)$ in terms of the Bernoulli numbers $B_r = B_r(0)$. These manipulations depend exactly on the sequence of Bernoulli numbers generated by Eq. (4).
Step 2: Why Eq. (4) is necessary
Eq. (9) expresses a recursion satisfied by the Bernoulli numbers. While it formally admits a solution ${B_k}$, a recursion of this type does not a priori specify a unique sequence unless initial conditions are fixed. Eq. (4) precisely defines $B_0 = 1$ and recursively determines all higher $B_k$. Any other sequence satisfying a similar recursive structure but with different initial conditions or arbitrary coefficients could produce numbers $B_k'$ that do not satisfy the integral and derivative properties of the Bernoulli polynomials.
If one were to substitute such a sequence $B_k'$ into the derivation of Eq. (10), the steps using
$$ \int_0^1 B_r(x),dx = 0 $$
and
$$ B_r'(x) = r B_{r-1}(x) $$
would fail. Consequently, the coefficients $B_r / r!$ in Eq. (10) would be incorrect, and Eq. (10) would no longer hold exactly.
Thus, it is not enough to satisfy the general recursion of Eq. (9); the specific sequence defined by Eq. (4) is necessary to ensure that the numbers used correspond to the Bernoulli numbers with the required algebraic and analytic properties.
Step 3: Summary
- Eq. (10) depends on the coefficients $B_r / r!$ derived from Bernoulli numbers.
- Eq. (4) defines the unique sequence of Bernoulli numbers.
- Eq. (9) alone, without the initial condition $B_0 = 1$, does not guarantee the same sequence.
- Using any other sequence would yield coefficients incompatible with the derivation of Euler's summation formula.
Therefore, the sequence in Eq. (4) is necessary for the validity of Eq. (10).
∎
-, This solution directly addresses the logical necessity of Eq. (4) and connects it to the derivation of Eq. (10), avoiding the flaws of the previous version.