# Easy Limit not working that should be 0?

I am trying to compute the following limit, which, tell me if I am wrong, should be 0 for beta->Infinity when z

 k[z_, beta_] = (Sinh[(1 - I)*Sqrt[beta/2]* z] -
2 z Sinh[((1 - I)/2)*Sqrt[beta/2]])/((1 - I)*
Sqrt[beta/2] Cosh[((1 - I)/2)* Sqrt[beta/2]] -
2 Sinh[((1 - I)/2) Sqrt[beta/2]]);
Limit[k[z, beta], beta -> Infinity]


How come I am not getting any output from Mathematica? With regard to the limit going to 0, please look at the following plot:

 Manipulate[Plot[Abs[k[z, beta]], {z, -1/2, 1/2}], {beta, 1, 10000000}]


Can we state the limit is 0 as the order of magnitude is decreasing?

• Try this and see what happens k[z, beta] /. beta -> Infinity – zhk May 1 '17 at 14:28
• If I specify a value for z say 1 the Limit is ComplexInfinity – george2079 May 1 '17 at 14:28

The limit is parameter dependent. Large z case:

Limit[k[z, beta], beta -> Infinity, Assumptions -> z > 1/2]

(* Out= ComplexInfinity *)


Small z case:

Limit[k[z, beta], beta -> Infinity,
Assumptions -> 0 < z < 1/2]

(* Out= 0 *)

• does it work also for -1/2<z<0? – Andrea G May 2 '17 at 10:48
• I actually ran your second limit and Mathematica didn't return any output and doesn't seem to work. Can it depend on the version? Im using 10.4.1.0 Student Edition – Andrea G May 2 '17 at 12:40
• Yes, it might depend on version. It also could depend on speed of hardware and software, since some subcomputations have to be time constrained (so Limit will not hang). As for -1/2<z<0, I just tried it: In:= Limit[k[z, beta], beta -> Infinity, Assumptions ->-1/2<z<0] Out= 0 – Daniel Lichtblau May 2 '17 at 13:56
• is there a way to remove the time constraint and try to get your same result for the one that I am not getting with my software/hardware? – Andrea G May 2 '17 at 15:23
• No way to remove time constraints, that's all in internal code. And if the result depends on version then removing them would not change things. – Daniel Lichtblau May 2 '17 at 15:25

No, the limit is in general not zero:

<< NumericalCalculus
Plot[ReIm@NLimit[k[z,beta], beta->Infinity, Terms->15], {z, -.8, .8},  PlotRange->All] • what about Manipulate[Plot[Abs[k[z, beta]], {z, -1/2, 1/2}], {beta, 1, 10000000}]? can we say the limit is 0 in z in the range of (-1/2,1/2)? – Andrea G May 1 '17 at 14:40

It actually appears to be zero for, Abs[z]<1/2 :

f[z_] := Limit[k[z, beta] , beta -> Infinity]
{#, f[#]} & /@ Range[-1, 1, 1/10]


{{-1, ComplexInfinity}, {-(9/10), ComplexInfinity}, {-(4/5), ComplexInfinity}, {-(7/10), ComplexInfinity}, {-(3/5), ComplexInfinity}, {-(1/2), 0}, {-(2/5), 0}, {-(3/10), 0}, {-(1/5), 0}, {-(1/10), 0}, {0, 0}, {1/10, 0}, {1/5, 0}, {3/10, 0}, {2/5, 0}, {1/2, 0}, {3/5, ComplexInfinity}, {7/10, ComplexInfinity}, {4/5, ComplexInfinity}, {9/10, ComplexInfinity}, {1, ComplexInfinity}}

It is zero for some complex z as well. Its not clear if the bound is Abs[z]<1/2 though.

This does not work BTW:

 Limit[k[z, beta] , beta -> Infinity,
Assumptions -> {Element[z, Reals], -1/2 < z < 1/2}]

• I agree with the fact that the limit is 0, which is the result that I expected. May I ask you why we cannot get a global result, rather than computing it discretely as you do? I just ran your last line but Mathematica didn't return an outoput but the input itself. Isn't Assumptions a function for Refine, Integrate and Simplify? Why are you using it here? – Andrea G May 1 '17 at 15:06
• I don't know. You may want to un-accept this. Maybe someone else will know a trick to get a general result. – george2079 May 1 '17 at 15:15

You can proof the result, -1/2<=z<=1/2, with a few transformations and then a series-expansion like that

 k[z_, ceta_] = (Sinh[(1 - I)*Sqrt[beta/2]*z] -
2 z Sinh[((1 - I)/2)*Sqrt[beta/2]])/((1 - I)*
Sqrt[beta/2] Cosh[((1 - I)/2)*Sqrt[beta/2]] -
2 Sinh[((1 - I)/2) Sqrt[beta/2]]) //. Sqrt[beta/2] -> ceta


Terms with Exp[-ceta/2] disapear,since ceta -> Infinity

 k1[z_, ceta_] = (k[z, ceta] // TrigExpand // Together //
TrigToExp) /. E^(-ceta/2) -> 0 // Simplify


Do ComplexExpand of the real part. (You get exactly the same for the imaginary part.)

 se = Simplify[
ComplexExpand[Re[k1[z, ceta]], TargetFunctions -> {Re, Im}],
Assumptions -> {z \[Element] Reals, ceta \[Element] Reals}] //
Expand


Do the series expansion and set Sin, Cos to Interval[{-1,1}]

 se2 = Series[se, {ceta, \[Infinity], 1}] // Normal // Expand

se3 = (se2) /. Cos[_] -> Interval[{-1, 1}] /.
Sin[_] -> Interval[{-1, 1}] // FullSimplify // Factor // Expand

(*    -(z/ceta) + (E^(-ceta (1/2 + z)) Interval[{-1, 1}])/ceta + (
E^(2 ceta z - ceta (1/2 + z)) Interval[{-1, 1}])/ceta     *)


Here you see, with Limit ceta->Infinity the exponetial terms either go to 0 or they explode, depending on z

Limit goes to 0, if

 Reduce[-1/2 - z <= 0, z]

(*   z >= -(1/2)   *)


and

 Reduce[2 z - (1/2 + z) <= 0, z]

(*     z <= 1/2    *)

• This is pretty much what Limit` does. – Daniel Lichtblau May 3 '17 at 20:43
• Not exactly. With Limit, you have to guess the convergence limits with the help of some other information. Here you get them directly and proofed to be the only one. – Akku14 May 4 '17 at 5:45