We present results of a survey of Lyman $alpha$ emitters (LAEs) at $z=6.5$ which is thought to be the final epoch of the cosmic reionization. In a $approx530$ arcmin$^2$ deep image of the SSA22 field taken through a narrowband filter NB912 installed
in the Subaru/Suprime-Cam, we have found only 14 LAE candidates with $L_{rm Lyalpha}ga3times10^{42}$ erg s$^{-1}$. Even applying the same colour selection criteria, the number density of the LAE candidates is a factor of 3 smaller than that found at the same redshift in the Subaru Deep field (SDF). Assuming the number density in the SDF is a cosmic average, the probability to have a number density equal to or smaller than that found in the SSA22 field is only 7% if we consider fluctuation by the large-scale structure (i.e. cosmic variance) and Poisson error. Therefore, the SSA22 field may be a rare void at $z=6.5$. On the other hand, we have found that the number density of $i$-drop galaxies with $25.5<z<26.0$ in the SSA22 field agrees well with that in the SDF. If we consider a scenario that a larger neutral fraction of intergalactic hydrogen, $x_{rm HI}$, in the SSA22 field obscures a part of Ly$alpha$ emission, $x_{rm HI}$ in the SSA22 field should be about 2 times larger than that in the SDF. This can be translated into $x_{rm HI}<0.9$ at $z=6.5$ in the SSA22 field. A much larger survey area than previous ones is required to overcome a large fluctuation reported here and to obtain a robust constraint on $x_{rm HI}$ at the end of the reionization from LAEs.
