ONIOM: Microiterations cycle 26 out of a maximum of 25 Maximum number of ...

xylz6188

各位计算化学公社的朋友们，大家好。近来我采用Gaussian做反应机理，方法是QM/MM（High采用QM，Low采用MM），考虑溶剂化效应（显性模型）的时候遇到一个问题，经查询处理无果，希望能在计算化学公社得到帮助，谢谢大家！
错误提示：ONIOM: Microiterations cycle   26 out of a maximum of  25
Maximum number of microiterations cycles exceeded!
我自己处理的办法：1、加大循环次数，但还是失败了；2、Opt=(QuadMac,CalcFC)换成Opt=(QuadMac,CalcAll)也失败了。错误提示都是同样的结果。
希望遇到此问题已有解决办法的网友给以提示为谢！

附件：输入文件、输出文件（太大，没法上传，我粘贴出最后错误提示）
错误提示：
NAtoms=  924 NActive=   24 NUniq=   24 SFac= 1.00D+00 NAtFMM=   60 NAOKFM=F Big=F
Integral buffers will be    131072 words long.
Raffenetti 1 integral format.
Two-electron integral symmetry is turned on.
UFF calculation of energy and first derivatives.
Energy=    -83.2201487     NIter=   0.
Dipole moment=       2.924218      -0.500967       0.467056
***** Axes restored to original set *****
Cartesian Forces:  Max     0.076101147 RMS     0.003887689
ONIOM: saving gridpoint  1
ONIOM: restoring gridpoint  3
ChrgS2: IScale=     5     5     5     5     0     0 (distance 6..1 bonds)
ChrgS2:  Scale=  1.00  1.00  1.00  1.00  0.00  0.00 (distance 6..1 bonds)
ChrgS2: IScale=     5     5     5     5     0     0 (distance 6..1 bonds)
ChrgS2:  Scale=  1.00  1.00  1.00  1.00  0.00  0.00 (distance 6..1 bonds)
Using ESP charges for embedding optimization.
Test: gridpoint  1 method:  low   system:  model energy:     0.122755489577
Test: gridpoint  2 method:  high  system:  model energy:  -576.662671505953
Test: gridpoint  3 method:  low   system:  real  energy:    -3.093674108103
Test: extrapolated energy =    -579.879101103633
ONIOM: Microiterations cycle   26 out of a maximum of  25
Maximum number of microiterations cycles exceeded!
Terminating job.
Error termination via Lnk1e in /public/software/g09/l120.exe at Thu Sep 21 12:45:46 2017.
Job cpu time:       1 days 21 hours 37 minutes 37.2 seconds.
File lengths (MBytes):  RWF=   2423 Int=      0 D2E=      0 Chk=    180 Scr=      1

xylz6188

自己顶一下，这是Tao, P., et al. (2009). "Matrix Metalloproteinase 2 Inhibition: Combined Quantum Mechanics and Molecular Mechanics Studies of the Inhibition Mechanism of (4-Phenoxyphenylsulfonyl)methylthiirane and Its Oxirane Analogue." Biochemistry 48(41): 9839-9847.发的文章。
里面是采用计算单点能，没有计算力常数

sobereva

真有必要包那么多水？可以去一部分再试
另外整体结构可以先在UFF下优化，然后再尝试改用ONIOM；或者先尝试不用电子嵌入，优化成功了再开启

论坛空间非常有限，较大文件可以压缩后传网盘

xylz6188

sobereva 发表于 2017-9-21 19:03
真有必要包那么多水？可以去一部分再试
另外整体结构可以先在UFF下优化，然后再尝试改用ONIOM；或者先尝试 ...

谢谢sobereva老师，我尝试一下你所说的方法。或者我先考虑算个单点能试一试。考虑在旁边加几个水分子我也想过，其实我也是有意向朝这个方向发展，以后我会用到混合溶剂，显性溶剂说服力要强一点。

xylz6188

https://www.researchgate.net/pos ... tronic_embedding_EE
a year ago
Gergely Bánóczi
Added an answer
Hello Everyone!
Actually I have a solution for this problem. Strangely, it seems there is no keyword to change the maximum number of microiteration cycles. But there is a way to get around it. Use the keywords: " opt=quadmacro iop(1/98=66,1/19=7)" Useful options might be iop(1/8=3,1/13=6). The deal is the following: "quadmacro" stands for quadratic macrostep. This is an awesome feature in g09, I heavily encourage everybody to use it in all QM/MM calculations, especially for TS searches. Iop(1/98=66) makes a twist in the optimization workflow, it disables microiterations, only second order steps are made everytime in both the QM and MM regions. Keep in mind that the MM region needs much more steps to converge than the QM region because of its sheer size. If you run an optimization like this, you will notice that it will take steps faster than the standard protocol, and eventually the QM region will reach convergence, but the MM region will most probably be far from convergence. So after like a few tens of steps (for example 50-100) stop, get the latest structure, and start an optimization with microiterations. Mine looks like: "opt=(quadmacro,readfc) iop(1/8=3,1/128=1,1/110=2,1/19=7,1/13=6)". 1/128 and 1/110 are microiteration realted options I think they are useful too in a lot of cases. 1/19=7 is a must (at least for my calculations) with quadmacro because of a bug. If you get still the same error, than get another round of pire quadmacro optimization. And try again microiterations afterwards. You can reach full convergence without microiterations, I suppose, but it would be quite inefficient, I think.
Best of luck, cheers!