Considering the MRB constant which is approximated by
NSum[(-1)^n (n^(1/n) - 1), {n, 1, Infinity}, WorkingPrecision -> 70,
Method -> "AlternatingSigns"]
giving
0.18785964246206712024851793405427323005590309490013878617200468547342,
I wondered what happens when we change 1/n in n^(1/n) to 1 over other functions involving n?
Then I figured out that
Table[NSum[(-1)^n (1/n^(1/(n*10^x)) - 1), {n, 1, Infinity},
WorkingPrecision -> 70,
Method ->
"AlternatingSigns"] + (EulerGamma*Log[2] - Log[2]^2/2)*10^-x, {x,
1, 30}]
and also
Table[NSum[(-1)^n (n^(1/(n*10^x)) - 1), {n, 1, Infinity},
WorkingPrecision -> 70,
Method ->
"AlternatingSigns"] - (EulerGamma*Log[2] - Log[2]^2/2)*10^-x, {x,
1, 30}]
mantissa's converge to the same constant even though one formula uses the reciprocal of what the other uses!
{0.0002544931414649952541958458971418195836265633470373523204886572940\
13, 2.5213764023532449457379170196206807316529005454922927744915334919\
*10^-6, 2.\
519038430344263802098258061692809581811767812023594203581344434*10^-8,
2.51880480742043470745309496704890093429551288263349038480600225*10^\
-10, 2.5187814468698119802633255716680901574100656515276061645434502*\
10^-12, 2.\
518779110832166299320533338340144497540312267653607154108212*10^-14,
2.51877887722857589613401551895739497023827511567917103396240*10^-16,
2.5187788538682185974634313552273103698241994129327919890091*10^-18,
2.518778851532182885012852605041013681772455184206979797782*10^-20,
2.51877885129857931394195952567425580667348404161289030118*10^-22,
2.5187788512752189568366118656930887417812470506544970303*10^-24,
2.518778851272882921126094516174526112522881012060672865*10^-26,
2.51877885127264931755504295538746538935935766591293015*10^-28,
2.5187788512726259571979378010504072724396184685687105*10^-30,
2.518778851272623621162227285634117940301609670334958*10^-32,
2.51877885127262338755865623409266317188334844194942*10^-34,
2.5187788512726233641982991289385194366894777279525*10^-36,
2.518778851272623361862263418423105080586570333924*10^-38,
2.51877885127262336162865984737156364515044562412*10^-40,
2.5187788512726233616052994902664095016085871417*10^-42,
2.518778851272623361602963454555894087254542116*10^-44,
2.51877885127262336160272985098484254582037185*10^-46,
2.5187788512726233616027064906277373916892954*10^-48,
2.518778851272623361602704154592026876399594*10^-50,
2.51877885127262336160270392098845582610468*10^-52,
2.5187788512726233616027038976280987334158*10^-54,
2.518778851272623361602703895292063147553*10^-56,
2.51877885127262336160270389505846082303*10^-58,
2.5187788512726233616027038950351129312*10^-60,
2.518778851272623361602703895032901548*10^-62}.
This is for the same reason that
NSum[(-1)^n (-1 + 1/n^(10^-x/n) + (10^-x Log[n])/n), {n, 1, Infinity},
WorkingPrecision -> 70, Method -> "AlternatingSigns"] /. x -> 30
and
NSum[(-1)^n (-1 + n^(10^-x/n) - (10^-x Log[n])/n), {n, 1, Infinity},
WorkingPrecision -> 70, Method -> "AlternatingSigns"] /. x -> 30
give
2.51877885127262336160270389503276304518288666816092953082281396924871\
*10^-62
(For version 3.1 you will have to use x->27 or less.)
(I wonder what closed form starts with 2.518778851272623361602703895?)
While changing the sign of "1:"
NSum[(-1)^n (1 + n^(10^-x/n) - (10^-x Log[n])/n), {n, 1, Infinity},
WorkingPrecision -> 70, Method -> "AlternatingSigns"] /. x -> 30
gives a "divergent sum" of
-1.0000000000000000000000000000000000000000000000000000000000038475945
And another constant, with formulas behaving differently, I noticed is from
Table[NSum[(-1)^n (n^(1/(n + 10^x)) - 1), {n, 1, Infinity},
WorkingPrecision -> 70,
Method ->
"AlternatingSigns"] - (5 (Log[Pi] - Log[2]))*10^-(x + 1), {x, 1,
30}]
which gives
{-0.002158265971057576296292446726899637434277905244619602754711500291\
280, -0.00002326878830432348151050186189530613369186033070715442565971\
5184159, -2.\
344972498540413669051157226943914337661182682378121265593728211*10^-7,\
-2.34679530411087431851660597513831629183877508983564770824474081*10^\
-9, -2.3469777203157053575942148177302233368868838187419725158605417*\
10^-11, -2.\
346995963293941260610236486862825114739849987209310140146848*10^-13, \
-2.34699778760534365532625414499456151604960848632944848056957*10^-15,\
-2.3469979700366196841185707695593955974492155812208948712378*10^-17,\
-2.346997988279748644892286166044070017626312946190793075881*10^-19, \
-2.34699799010406155454860381961973372505412612579214679476*10^-21, \
-2.3469979902864928456500250473931591201743603058719161031*10^-23, \
-2.346997990304735974761525064795936837006618410461168616*10^-25, \
-2.34699799030656028767268864548247023705012450346433171*10^-27, \
-2.3469979903067427189638051393405861346146650120718647*10^-29, \
-2.346997990306760962092916790084292349947997733009464*10^-31, \
-2.34699799030676278640582795517224191773710290844785*10^-33, \
-2.3469979903067629688371190716811726639785895468696*10^-35, \
-2.346997990306762987080248183332067096497547977600*10^-37, \
-2.34699799030676298890456109449715655333023197512*10^-39, \
-2.3469979903067629890869923856136654991677088242*10^-41, \
-2.346997990306762989105235514725316393937004272*10^-43, \
-2.34699799030676298910705982763648148525584608*10^-45, \
-2.3469979903067629891072422589275980128067172*10^-47, \
-2.346997990306762989107260502056709849751672*10^-49, \
-2.34699799030676298910726232636962287534485*10^-51, \
-2.3469979903067629891072625088009325968910*10^-53, \
-2.346997990306762989107262527044247758914*10^-55, \
-2.34699799030676298910726252887042117380*10^-57, \
-2.3469979903067629891072625290714575021*10^-59, \
-2.346997990306762989107262529275751003*10^-61}
(Likewise I wonder what closed form starts with 2.34699799030676298910726?)
However here we have a different constant converged to by using the reciprocal
Table[NSum[(-1)^n (1/n^(1/(n + 10^x)) - 1), {n, 1, Infinity},
WorkingPrecision -> 70,
Method ->
"AlternatingSigns"] + (5 (Log[Pi] - Log[2]))*10^-(x + 1), {x, 1,
30}]
giving
{0.0027281794786312486595717428387460304770563641777457765720218456040\
40, 0.0000293550399240151127322722379565607620916546859019350100125647\
09573, 2.9551297193773754366885564592734651504693514196068487319926172\
96*10^-7,
2.95707365395153655826155461847169020429215068621347502129971599*10^-\
9, 2.9572678525605426063529814114458282642227732785054465491056943*10^\
-11, 2.957287270464422416597614464460409791592623556524877730647258*\
10^-13, 2.95728921223523164631517763738252605852113445125380329353929*\
10^-15, 2.9572894064121167732298398914267611418621574492035860745636*\
10^-17, 2.957289425829803327952191084785434676290469048265224666794*\
10^-19, 2.95728942777157196384472650970279641869142142361488168100*10^\
-21, 2.9572894279657488272381830467062488252286207798089293916*10^-23,
2.957289427985166513575570730343167129427371191629855892*10^-25,
2.95728942798710828220928991900621614638391196404033743*10^-27,
2.9572894279873024590726616420755146114008339317553828*10^-29,
2.957289427987321876758998812424474393527194892652000*10^-31,
2.95728942798732381852763252943979067109607097234091*10^-33,
2.9572894279873240127044959011411265018465393721695*10^-35,
2.957289427987324032122182238311258126951706025741*10^-37,
2.95728942798732403406395087202827126988436394602*10^-39,
2.9572894279873240342581277353999725840002517184*10^-41,
2.957289427987324034277545421737142715594072394*10^-43,
2.95728942798732403427948719037085973059533356*10^-45,
2.9572894279873240342796813672342314505144463*10^-47,
2.957289427987324034279700784920568806696226*10^-49,
2.95728942798732403427970272668920438421309*10^-51,
2.9572894279873240342797029208660863609516*10^-53,
2.957289427987324034279702940283958748494*10^-55,
2.95728942798732403427970294222758788593*10^-57,
2.9572894279873240342797029424403697865*10^-59,
2.957289427987324034279702942645837845*10^-61}
I could also ask about the constant 2.957289427987324034279702942.