1: \begin{abstract}

2: We investigate the critical endpoint of finite temperature phase transition of $N_f=3$ QCD at zero chemical potential.

3: We employ the renormalization-group improved Iwasaki gauge action and non-perturbatively $O(a)$-improved Wilson-clover  fermion action.

4: The critical endpoint is determined by using the intersection point 

5: of kurtosis for the temporal size $N_t$=4, 6, 8.  Spatial sizes of $N_l$=6--16 ($N_t$=4),  10--24 ($N_t$=6), and 12--24 ($N_t$=8) are employed.  We find that $N_t$=4 is out of the scaling region.   Using results for $N_t$=6 and 8, and making linear extrapolations in $1/N_t^2$, we obtain 

6: $\sqrt{t_0}T_\rmE=0.0975(14)(8)$, 

7: $\sqrt{t_0}m_\rmPSE=0.2254(52)(105)$ and 

8: $m_\rmPSE/T_\rmE=2.311(63)(13)$, 

9: where the first error is statistical error, the second error is systematic error, and $m_\rmPS$ is the pseudo scalar meson mass. 

10:  If one uses $1/\sqrt{t_0}=1.347(30)$~GeV reported by Borsanyi {\it et al.}, one finds

11: $T_\rmE=131(2)(1)(3)$~MeV, $m_\rmPSE=304(7)(14)(7)$~MeV and $m_\rmPSE/m_\rmPS^{\rm phys, sym}=0.739(17)(34)(17)$,

12: where the third error comes from error of $\sqrt{t_0}$ and $m_\rmPS^{\rm phys, sym}=\sqrt{(m_\pi^2+2m_K^2)/3}$.

13: Our current estimation of $\sqrt{t_0}m_\rmPSE$ in the continuum limit is about 25\% smaller than the SU(3) symmetric point.

14: \end{abstract}

15: