One significant difference between charge doping and vacancy doping appeared as a spin-glass like behavior for the charge doped systems (x=9.5, 14.9% and 30.5%). From the cusp of the zero-field-cooling (ZFC) susceptibilities, the glass temperatures () were determined as 2.5, 2.9 and 3.0 K for the x=9.5, 14.9 and 30.5 % systems, respectively. The spin-glass like behavior indicates that the doped hole destroy the non-magnetic ground state.
The vacancy doped systems (y=1.7 and 4.1%) remain paramagnetic down to 2 K. As presented in the next section, the absence of a magnetic order was confirmed to the milli-Kelvin regime.
One possible reason for the different magnetic behaviors of the charge/vacancy doped systems may be the existence/absence of an interaction across the doping site; although we do not know the local structure of the doping-induced paramagnetic moments, the interaction across a non-magnetic Mg2+ ion is probably negligible. Therefore, the broken chains of the vacancy doped system are isolated from each other, leaving no chance for magnetic order. In the charge doped system, on the other hand, there should be a super-exchange coupling (J') across the doping site, Ni2+-O--Ni2+ (or Ni2+-Ni3+-Ni2+), as was the case in an analogous system (Cu doped NENP) . The existence of the interaction between the chain segments may allow the unpaired spins to freeze at kTJ', if there is a strong coupling within the broken chain segments. The failure of the simple VBS picture to describe the number of unpaired spins in the Zn-doped systems  seems to support the existence of strong coupling within the broken chains. To estimate the magnitude of J' is difficult; if one uses J' in Cu doped NENP (0.70.9 K), and assumes a scaling with the magnitude of Haldane gap (), J' in our system falls on the order of a few Kelvin, which is consistent with the cusp temperatures of the Ca doped systems.