Error estimate of Alternating Series (Proof)?

If the series ∑(-1)^n*(a(n)) converges to S, then the nth term of the sequence of partial sums of an alternating series approaches the sum of the series as n approaches ∞. The difference between the exact sum and the sum of the first n finite terms corresponds to the error in approximating the convergent sum by the nth partial sum.
How Can I proof that
|S-Sn|=∑(-1)^n+1a(n) from n+1 to ∞<an+1

1 Answer
Aug 31, 2017

Consider an alternating series:

#sum_(n=0)^oo (-1)^na_n#

with #a_n >=0# and satisfying the conditions:

#(1) " " lim_(n->oo) a_n = 0#

#(2) " " a_n >= a_(n+1)#

Based on Leibniz theorem the series is convergent and we have:

#sum_(n=0)^oo (-1)^na_n = L#

Consider now the sequence of partial sums of odd order:

#s_(2k+1) = sum_(j=0)^(2k+1) (-1)^ja_j = (a_0-a_1) + (a_2-a_3) + ... +(a_(2k)-a_(2k+1))#

as, based on #(2)#, #a_(2j) >= a_(2j+1)#, all the terms in the last sum are positive and then #s_(2n+1)# is a monotone increasing sequence.

But for a monotone increasing convergent sequence the limit is the upper bound, so that:

#s_(2k+1) <= L#

Similarly we have that the partial sums of even order:

#s_(2k) = sum_(j=0)^(2k) (-1)^ja_j = a_0 - (a_1 -a_2) - ... - (a_(2k-1)-a_(2k))#

form a monotone decreasing sequence with:

#s_(2k) >= L#

Now consider:

#abs(L-s_n) #

For #n# odd we have that #s_n <= L# and #s_(n+1) >= L#, then:

#abs (L-s_n) = L-s_n <= s_(n+1)-s_n = a_(n+1)#

while for #n# even we have #s_n >= L# and #s_(n+1) <= L#, and again:

#abs (L-s_n) = s_n-L <= s_n - s_(n+1) = a_(n+1)#

In both cases:

#abs (L-s_n) <= a_(n+1)#