[CP2K:8433] Re: How to obtain accurate frequencies from CP2K? (inside the post the frequencies are compared between CP2K and Gaussian)

[Satish Kumar]

You can some tests with your preferred values of DX and see how your 
vibrational frequencies compare with available experimental data for 
example. 

You see multiple *.out files based on the value of NPROC_REP. If you run 
your calculation on 48 processors for example with NPROC_REP of 24, you 
will see 48/24 =2 output files.

On Saturday, November 26, 2016 at 4:38:10 PM UTC-5, J. Ye wrote:
>
> Hi S Ling, 
>
> Thanks very much! 
>
> How to choose  a correct or reasonable step size (DX in CP2K)? I think DX 
> value should be as close as the displacement of atoms. However, the 
> displacement of atoms in real condition should be different. In the 
> calculation, we are forced to use the same displacement (DX) to calculate 
> the force constant. How does this affect the frequencies. Just curious? 
> Maybe it is very minor. 
>
> Why there's many other output generate other than f.out? 
> For example, I have f.inp, f.out and other output files (xxr-1.out, 
> xx-r-2.out, xx-r-3.out, .......xx-r-9.out). Each xx-r-x.out file looks like 
> a single point calculations. 
>
> Best,   
>
> On Thursday, November 24, 2016 at 4:59:44 AM UTC-6, S Ling wrote:
>>
>> Hi
>>
>> In addition to what Matt said, if you really want to compare results from 
>> CP2K and Gaussian, you should: (1) use the all-electron method, i.e. GAPW, 
>> in CP2K; (2) use the same basis sets; (3) use the same method to calculate 
>> frequencies. For (3), Gaussian calculates frequencies analytically by 
>> default, while CP2K calculates frequencies numerically. To calculate 
>> frequencies numerically in Gaussian, you should specify Freq=Numer. And 
>> also the calculated frequencies may be sensitive to the the step size for 
>> numerical differentiation, therefore, you should use the same step size in 
>> both calculations. In Gaussian, this is controlled by the Step keyword, see 
>> http://www.gaussian.com/g_tech/g_ur/k_freq.htm; in CP2K, this is 
>> controlled by the DX keyword.
>>
>> SL
>>
>>
>>
>> On 24 November 2016 at 10:05, Matt W <Matt... at gmail.com> wrote:
>>
>>> Hi,
>>>
>>> you are not consistently running a non-periodic system - you need a 
>>> nonperiodic poisson section too. And 
>>>
>>> 1. You have 
>>>
>>> &VIBRATIONAL_ANALYSIS
>>>     DX 1.0E-03
>>>     NPROC_REP  24
>>>     FULLY_PERIODIC TRUE
>>>
>>> the fully periodic keyword means that CP2K won't remove rotations 
>>> (because they aren't there for a crystal with pbc) - so you get an extra 3 
>>> modes.
>>>
>>> 2. You are running a totally different 'model chemistry' in Gaussian 
>>> terminology. You can't compare M062X with PBE directly. PBE (GGAs) 
>>> typically underbind, bonds too long, frequencies too small. M06 maybe is 
>>> better on average.
>>>
>>> 3 So you are mixing two things - accurate, i.e. properly performed 
>>> calculation at a given level of theory - and agreement with experiment. 
>>> Changing the model will change the frequencies. Changing settings (CUTOFF 
>>> mainly, probably) will just ensure you get the correct numbers for the 
>>> model. You should run PBE in Gaussian to compare to, if this is what you 
>>> want.
>>>
>>> Matt
>>>
>>> On Wednesday, November 23, 2016 at 9:40:31 PM UTC, J. Ye wrote:
>>>>
>>>> Hi all, 
>>>>
>>>> I calculate frequency for small molecules and try to compare with 
>>>> Gaussian 09 and experimental results. I show the results from *CP2K *(PBE 
>>>> functional-BASIS_MOPLOT, GTH potental, nonperiodic) and * Gaussian09 *(M062X/6-311g(d,p)) 
>>>> . 
>>>>
>>>> I have several questions:
>>>>
>>>> 1. You will see the total number of modes obtained from CP2K is 15, 
>>>> however 12 modes from Gaussian. Are the first three modes are translation 
>>>> modes or rotation modes for C2H4?
>>>> 2. If we only compare the 12 vibration modes, the value for most of 
>>>>  modes from Gaussian is larger than CP2K. What's the reason for getting 
>>>> smaller frequencies from CP2K?  
>>>> 3. What is the keyword can strongly affect the frequencies?  Affect the 
>>>> accuracy of frequencies? 
>>>>
>>>> Thanks very much! 
>>>>
>>>> CP2K Gaussian 
>>>> c2h4 
>>>> 15 modes  
>>>> -19.0703 
>>>> -5.64472 
>>>> 19.50384 
>>>> 807.9613 831.8134 
>>>> 942.954 988.1607 
>>>> 995.6397 996.8044 
>>>> 1045.988 1071.741 
>>>> 1206.11 1239.784 
>>>> 1373.488 1386.642 
>>>> 1446.925 1470.955 
>>>> 1646.778 1717.766 
>>>> 3158.382 3156.212 
>>>> 3171.654 3172.992 
>>>> 3269.88 3235.274 
>>>> 3270.754 3260.871 
>>>>
>>> -- 
>>> You received this message because you are subscribed to the Google 
>>> Groups "cp2k" group.
>>> To unsubscribe from this group and stop receiving emails from it, send 
>>> an email to cp2k+... at googlegroups.com.
>>> To post to this group, send email to cp... at googlegroups.com.
>>> Visit this group at https://groups.google.com/group/cp2k.
>>> For more options, visit https://groups.google.com/d/optout.
>>>
>>
>>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20161129/317fac11/attachment.htm>