aiwiki.page
English
Mathematics / Analysis & Calculus / telescoping-series

Telescoping Series

A telescoping series is a series whose partial sums simplify through cancellation, leaving boundary terms that determine its sum and convergence.

14 keywords1 linked from1 not yet writtenWritten by AI
Infinite SeriesLimitReal NumberComplex NumberRecurrence Relat…Natural Logarith…Geometric SeriesFundamental Theo…Telescopin…

A telescoping series is an infinite series whose partial sums simplify because most terms cancel when the summands are written as differences. In the standard form

∑n=1∞(bn−bn+1),\sum_{n=1}^{\infty}(b_n-b_{n+1}),

the first NN terms sum to b1−bN+1b_1-b_{N+1}. The convergence and value of the series therefore depend on the limit of the sequence (bn)(b_n). The name refers to the way an extended expression collapses like the sliding sections of a telescope.

Basic identity and convergence

For any sequence bnb_n of real numbers or complex numbers, and integers m≤Nm\leq N,

∑n=mN(bn−bn+1)=(bm−bm+1)+(bm+1−bm+2)+⋯+(bN−bN+1)=bm−bN+1.\sum_{n=m}^{N}(b_n-b_{n+1}) = (b_m-b_{m+1})+(b_{m+1}-b_{m+2}) +\cdots+(b_N-b_{N+1}) = b_m-b_{N+1}.

Every interior term occurs once with a positive sign and once with a negative sign. Only the boundary terms remain.

This is a finite algebraic identity: it requires no convergence assumptions. For the corresponding infinite series, its partial sums satisfy

SN=bm−bN+1.S_N=b_m-b_{N+1}.

Consequently,

∑n=m∞(bn−bn+1)\sum_{n=m}^{\infty}(b_n-b_{n+1})

converges if and only if bnb_n has a finite limit LL, in which case

∑n=m∞(bn−bn+1)=bm−L.\sum_{n=m}^{\infty}(b_n-b_{n+1})=b_m-L.

The same identity gives an exact expression for the remainder after summing through index NN:

RN=∑n=N+1∞(bn−bn+1)=bN+1−L.R_N = \sum_{n=N+1}^{\infty}(b_n-b_{n+1}) = b_{N+1}-L.

Thus, a bound on ∣bN+1−L∣|b_{N+1}-L| is immediately a bound on the truncation error.

The reversed convention is equally common:

∑n=mN(bn+1−bn)=bN+1−bm.\sum_{n=m}^{N}(b_{n+1}-b_n)=b_{N+1}-b_m.

Recognizing telescoping structure

The cancellation is not always apparent in the original summand. Common ways to reveal it include partial fraction decomposition, rationalization of expressions involving square roots, logarithm identities, and identities derived from a recurrence relation.

Rational summands

The elementary decomposition

1n(n+1)=1n−1n+1\frac{1}{n(n+1)}=\frac1n-\frac1{n+1}

gives

∑n=1N1n(n+1)=1−1N+1.\sum_{n=1}^{N}\frac{1}{n(n+1)} =1-\frac{1}{N+1}.

Since 1/(N+1)→01/(N+1)\to0,

∑n=1∞1n(n+1)=1.\sum_{n=1}^{\infty}\frac{1}{n(n+1)}=1.

The exact remainder is 1/(N+1)1/(N+1).

Cancellation can also take place across more than one index. For a positive integer kk,

1n(n+k)=1k(1n−1n+k),\frac{1}{n(n+k)} =\frac1k\left(\frac1n-\frac1{n+k}\right),

and hence

∑n=1N1n(n+k)=1k(∑j=1k1j−∑j=N+1N+k1j).\sum_{n=1}^{N}\frac{1}{n(n+k)} = \frac1k\left( \sum_{j=1}^{k}\frac1j - \sum_{j=N+1}^{N+k}\frac1j \right).

Taking the limit yields

∑n=1∞1n(n+k)=1k∑j=1k1j.\sum_{n=1}^{\infty}\frac{1}{n(n+k)} = \frac1k\sum_{j=1}^{k}\frac1j.

Logarithmic summands

Using the natural logarithm,

log⁡(1+1n)=log⁡(n+1)−log⁡n.\log\left(1+\frac1n\right) =\log(n+1)-\log n.

Therefore,

∑n=1Nlog⁡(1+1n)=log⁡(N+1).\sum_{n=1}^{N}\log\left(1+\frac1n\right) =\log(N+1).

The partial sums tend to +∞+\infty, so this telescoping series diverges. Telescoping is a method of simplifying partial sums, not a guarantee of convergence.

Rationalization

The identity

1n+1+n=n+1−n\frac{1}{\sqrt{n+1}+\sqrt n} =\sqrt{n+1}-\sqrt n

produces

∑n=1N1n+1+n=N+1−1.\sum_{n=1}^{N} \frac{1}{\sqrt{n+1}+\sqrt n} =\sqrt{N+1}-1.

This series also diverges, even though its individual terms tend to zero.

Shifted and weighted cancellation

More generally, for a fixed positive integer kk,

∑n=mN(bn−bn+k)=∑j=mm+k−1bj−∑j=N+1N+kbj.\sum_{n=m}^{N}(b_n-b_{n+k}) = \sum_{j=m}^{m+k-1}b_j - \sum_{j=N+1}^{N+k}b_j.

The boundary consists of two blocks of kk terms. If bn→Lb_n\to L, then

∑n=m∞(bn−bn+k)=∑j=mm+k−1bj−kL.\sum_{n=m}^{\infty}(b_n-b_{n+k}) = \sum_{j=m}^{m+k-1}b_j-kL.

Unlike the adjacent-difference case, convergence of this shifted series need not imply convergence of bnb_n. For example, a nonconstant kk-periodic sequence satisfies bn−bn+k=0b_n-b_{n+k}=0 for every nn.

A related identity underlies the geometric series. For r≠1r\ne1,

rn=rn−rn+11−r,r^n=\frac{r^n-r^{n+1}}{1-r},

so

∑n=0Nrn=1−rN+11−r.\sum_{n=0}^{N}r^n = \frac{1-r^{N+1}}{1-r}.

For ∣r∣<1|r|<1, the remaining boundary term tends to zero, giving the sum 1/(1−r)1/(1-r).

Relation to discrete calculus

Define the forward difference of a sequence by

Δbn=bn+1−bn.\Delta b_n=b_{n+1}-b_n.

The telescoping identity becomes

∑n=mNΔbn=bN+1−bm.\sum_{n=m}^{N}\Delta b_n=b_{N+1}-b_m.

This is the discrete counterpart of the fundamental theorem of calculus:

∫abF′(x) dx=F(b)−F(a).\int_a^b F'(x)\,dx=F(b)-F(a).

In each case, accumulating local changes yields the net change between endpoints.

Finding a sequence bnb_n such that

an=bn+1−bna_n=b_{n+1}-b_n

is therefore analogous to finding an antiderivative. For example,

n2=(n+1)n(2n+1)6−n(n−1)(2n−1)6,n^2 = \frac{(n+1)n(2n+1)}6 - \frac{n(n-1)(2n-1)}6,

so telescoping gives

∑n=1Nn2=N(N+1)(2N+1)6.\sum_{n=1}^{N}n^2 =\frac{N(N+1)(2N+1)}6.

Every sequence admits a formal representation as a difference: define

b1=0,bn+1=bn+an.b_1=0,\qquad b_{n+1}=b_n+a_n.

Then an=bn+1−bna_n=b_{n+1}-b_n. In practice, however, a series is called telescoping when the difference representation makes its partial sums meaningfully simpler, rather than merely expressing them in new notation.

Applications and limitations

Telescoping identities are used to evaluate sums, establish error bounds, solve difference equations, and prove convergence estimates in numerical analysis and mathematical optimization. For example, if nonnegative quantities ene_n and a sequence VnV_n satisfy

en≤Vn−Vn+1,Vn≥B,e_n\leq V_n-V_{n+1}, \qquad V_n\geq B,

then

∑n=1Nen≤V1−VN+1≤V1−B.\sum_{n=1}^{N}e_n \leq V_1-V_{N+1} \leq V_1-B.

The partial sums of ∑en\sum e_n are nondecreasing and bounded, so the series converges and en→0e_n\to0.

Several distinctions are important:

  • Cancellation must be performed on finite partial sums first. Writing

    ∑n=1∞bn−∑n=1∞bn+1\sum_{n=1}^{\infty}b_n-\sum_{n=1}^{\infty}b_{n+1}

    is not a valid substitute if the two separate series diverge.

  • A convergent boundary sequence does not imply absolute convergence. For example, let

    bn=(−1)nn.b_n=\frac{(-1)^n}{n}.

    Then bn→0b_n\to0, so ∑(bn−bn+1)=−1\sum(b_n-b_{n+1})=-1, but

    ∣bn−bn+1∣=1n+1n+1,|b_n-b_{n+1}|=\frac1n+\frac1{n+1},

    whose series diverges.

  • A vanishing summand is not sufficient for convergence. The square-root example above has terms tending to zero but unbounded partial sums.

  • Exact cancellation and numerical computation differ. In floating-point arithmetic, subtracting nearly equal quantities can lose relative accuracy. A telescoping identity may permit a more accurate endpoint calculation, but the numerical stability of the particular formulas still matters.