Need help on why making fine mesh throwing off the result from intended one

[ashokgrg19]

Dear all,
I am trying to simulate a problem where we apply a some harmonic perodic function [E= E0(sin(wt)) ] (represented by a black solid line on attached figure) and response funcition is a perodic function in nature dP= dP0*sin(wt+theta) which has same frequency as that of applied frequency but with some phase difference (theta).
I choose simulation box of size 18X1X1 nm, but if i choose element size 36X3X3, the profile of reponse function looks what is desired for(Please refer figure attached represent by red solid line). However, increasing mesh resolution with element size 72X3X3, the profile of response completely deviates from the sin profile (Please refer figure attached represent by red dash line).
I used HEX8 for doing this solution.
Now, I have question why increasing element size throwing off my solution from intented solution.
Is it PETSC option or My executioner? I tried to play with Executioner and PETSC options but haven't had a sucess.

I would really appreciate if you guys share your thoughts and insights on this matter.
Thx,
Ashok
This are my precondition and executioner options.
[Preconditioning]

###############################################

Numerical preconditioning/solver options

###############################################

[./smp]
type = SMP
full = true
petsc_options = '-snes_ksp_ew'
petsc_options_iname = '-ksp_gmres_restart -snes_atol -snes_rtol -ksp_rtol -pc_type -build_twosided'
petsc_options_value = ' 160 1e-10 1e-8 1e-6 bjacobi allreduce'
[../]
[]
[Executioner]

##########################################

Time integration/solver options

##########################################

type = Transient
solve_type = 'PJFNK'
#scheme = 'crank-nicolson'

dtmin = 1e-16

dt = 0.0001

dtmax = 1e-2

verbose = true

num_steps = 1200
[TimeIntegrator]
type = NewmarkBeta
inactive_tsteps = 1
[]
[]