论文作者的导师是 Stewart Clark，CASTEP的开发者之一，是不是说明也没什么更好的办法了？关键是，就算重新优化QST maxima，得到一个新的构型，也不能确定这个到底是不是真正的过渡态。（At this point there was no way of differentiating between a true minimum and the minimum of a saddle point.）

我能想到的另一个办法也就是重新再找了。。。。。用过高斯的人表示CASTEP找过渡态太不灵活了

杜伦大学的一篇博士论文中提到了这个问题的解决办法，不知道诸位是不是也这么做？
Although the algorithm had failed to find a transition state, it had identified some structures different to both the reactant and product, new geometries with the potential to be transition states, two examples of which are shown in Figure Figure 4.4. If these geometries were potential transition states, relaxing them to their ground states would allow them to fulfil this potential, although this would require strict convergence criteria as the slightest noise in the computations would cause a transition state to relax to a reactant or product. As the convergence criteria used for the transition state algorithm involved strict convergences, the same geometry optimisation criteria described in Section 4.3.2 were used to relax the QST maxima. The relaxed geometries show that one fell into the product configuration and hence was not a potential transition state but more of an intermediate on the route to becoming a product -- this is purely geometrical and has no chemical significance. One of the QST maxima relaxed into a completely new configuration, implying that a new minimum on the potential energy surface had been reached. At this point there was no way of differentiating between a true minimum and the minimum of a saddle point.

附链接：http://cmt.dur.ac.uk/sjc/thesis_dlg/node68.html