Project questions for CAMD-group 12

Computational Materials Design Project (group10)
New Materials for Hydrogen Storage
Project description for CAMD group10 (update 3)
Calculations and setup
The future development of new energy materials, e.g. for energy storage, will depend extensively
on computational techniques. As part of a joined CAMD summer school project on combining DFT
calculations, database methods and screening techniques, you will be investigating the structure and
stability of possible alloy structure of metal borohydrides - a promising class of materials for
reversible hydrogen storage. Finding new alloys with high hydrogen content and optimal
thermodynamic stability is essential if hydrogen is going to be used as a commercial fuel in the
transport sector. The alloys we will investigate contain 1 alkali metal (Li,Na,K) and 1 earth alkali or
3d/4d transition metal atom plus either 3 or 4 (BH 4 ) - groups.
Your group (group 10) will investigate a number of different structures, where the local
coordination around the metal atoms is varied to obtain the most stable structure. Each group will
be responsible for their own part of the “alloy phase space”; your group will be investigating the
following borohydride alloys:
• LiZn(BH 4 ) 3,4 and LiCd(BH 4 ) 3 , 4
To help you get started, we have made a number of scripts which sets up the practical parameters
(k-points, cut-offs, etc.) and initial guesses for the positions of the atoms in the given local
coordination you have specified, based on the Pauling radius of the ions (radii.py).
(BH 4 ) - groups are blue.
For metal borohydrides, the metal atoms are predominantly
coordinated either tetrahedrally (tetra) or octahedrally (octa) to the
(BH 4 ) - groups.
Recent work (Voss et al. 2008) has shown that model structures with
correct local coordination provide a very good estimate of the ground
state energy of the true ground state.
Your group should investigate the following 8 structures and minimize the energy of the alloys in
the different configurations using ASE and Dacapo (https://wiki.fysik.dtu.dk/dacapo):
• Structure I: LiZn(BH 4 ) 3 – tetra
• Structure II: LiZn(BH 4 ) 3 - octa
• Structure III: LiZn(BH 4 ) 3 – tetra/octa (Li in tetra and Zn in octahedral coordination)
• Structure IV: LiZn(BH 4 ) 3 – octa/tetra
• Structure V: LiZn(BH 4 ) 4 – tetra
• Structure VI: LiZn(BH 4 ) 4 – octa
• Structure VII: LiZn(BH 4 ) 4 – tetra/octa
• Structure VIII: LiZn(BH 4 ) 4 – octa/tetra
- and similarly for the 4 equivalent LiCd(BH 4 ) 3 and the 4 LiCd(BH 4 ) 4 structures
CAMD Summer School in Electronic Structure Theory 21-08-2008

Computational Materials Design Project (group10)
New Materials for Hydrogen Storage
Setting up the environment
All tasks for this project must be performed on the login nodes school1 or school2. To login do
- choose any window manager
- right-click mouse -> Terminal ->
$>ssh school1.fysik.dtu.dk or $>ssh school2.fysik.dtu.dk
CAMd users: please use your niflheim username: $>ssh username@school1.fysik.dtu.dk
To setup the paths in your .tcshrc or .bashrc run the script (don't type the $>) do
$> /home/niflheim/dlandis/ss08scripts/install.py
- your password is derived from you name: ‘First initial’userid’Last initial’: Xcamd0xxY
Setting up the calculations
Setup.py (one external person per group, i.e. not a CAMD member)
Step 1: Run the python script setup.py
$> cd
$> setup.py
to generate the necessary python files and directories, and follow the instructions on the screen. The
script should be run twice to setup both your LiZn and LiCd system.
For each of the 4 alloys (LiZn(BH 4 ) 3,4 and LiCd(BH 4 ) 3 , 4 ) a directory is created with subdirectories
for the different coordinations. An initial guess for the alloy in the chosen coordination is
automatically generated and placed it in the corresponding subdirectory.
Setup.py (all other persons in the group)
Step 1: Run setup.py
$> setup.py
Select option ‘1’: Update your local files…”. This will load the group scripts onto your
~/project08 folder. You can re-run setup.py in order to obtain the submitted files from other group
members as well.
For everyone again
Step 2: Go to a structure subdirectory, i.e. t, o, ot or to (not free…) and run $> python
‘name’.py (please do not rename the scripts). Analyze the structure ‘name’_init.xyz using
Firefox http://dcwww.fys.dtu.dk/~strabo/school/group10 by pressing the loadlast button
and selecting the filename in the dialog box to check that the structure is okay, i.e. tetra/octa
coordination and check atomic distances by double clicking on the atoms. If so, proceed with the
Dacapo optimization; this is done by uncommenting the line ‘#bh.optimizeVolume()’ in the script
(‘name’.py) and using the command:
$>qsub -l nodes=2:ppn=4:switch5 -q long ‘name’.py
(external)
$>qsub –W group_list=gcourse -l nodes=2:ppn=4:switch5 -q long ‘name’.py (CAMD)
CAMD Summer School in Electronic Structure Theory 21-08-2008

Computational Materials Design Project (group10)
New Materials for Hydrogen Storage
The setup structure is based on the Pauling radius it is not optimized yet, so the following steps are
therefore performed in the script:
• Relaxing the internal degrees of freedom in the initial structure (keeping the position of the
metal atoms fixed).
• Search for a better “effective lattice parameter” (for the non-cubic template structures, the
ratio between the lattice constants will be kept fixed based on the initial Pauling-guess
structure). The script uses a function which automatically maintains the optimal B-H
distances, when the lattice constant is changed.
• As starting parameters a 10% volume contraction to 10% expansion screening is performed.
An equation of state fit is then used to determine the optimal effective lattice constant.
• Perform an internal optimization of the structure with the optimal effective lattice constants
(metal atoms are still fixed in scaled coordinates).
Step 3: You should analyze the results of the equation of state fit and assess the plot (‘name’-
eos.png) using gthumb, to ensure that the optimal lattice constant is accurately determined (points
above and below the minimum). If not, set ‘reoptimize=True’ and resubmit the calculation.
Step 4: You must verify that your final minimization has been completed, if not, try to resubmit the
script – if it fails, contact a student assistant. Inspect the final xyz file to see whether the structure is
maintained or distorted compared to the original setup.
• This process should be completed for all remaining structures once you have completed the
check in described below.
Check files and check-in functions
If you want to get the results of the simulations other members of your group have submitted, run
$> setup.py
and choose 1 (Update your local files....).
If you have some new converged results that you would like to check run check.py
$> check.py
A list with the invalid and valid calculations is now shown. From the valid ones you can choose the
one you want to check. Then a list with some results is shown; make sure to keep the information,
e.g. the total energy and volume, since it will be used later.
Finally, you are prompted if you want to submit your results. If you are confident in the relaxed
structure; press 'y'. This will upload the generated output to the database together with information
from the nc-files. Now other group members can use setup.py to get the result files into their own
home directory.
• Complete the steps above for all 2x8 structures
CAMD Summer School in Electronic Structure Theory 21-08-2008

Computational Materials Design Project (group10)
New Materials for Hydrogen Storage
Data analysis and hand-in
You have now completed the initial simulations needed for you sub-project and you should be able
to see your data as points in the different shared plots which can be found at:
http://dcwww.camd.dtu.dk/school08/
By comparing the results from the optimized structures with reference values
(http://dcwww.camd.dtu.dk/school08/)for, e.g. LiBH 4 and Zn(BH 4 ) 3 and assuming a general
decomposition pathway to determine the decomposition energy, e.g. for LiZnB4_t: LiZn(BH 4 ) 4 →
LiH + Zn + 4 B + 7.5 H 2 , a number of different values have been extracted from your data, e.g.:
• ‘Decomposition energy’ [eV/H 2 ]: ΔE decomp = E LiZnB4_t - (E LiH + E Zn + 4 E B + 7.5 E H2 )
• Alloy stability [eV/H 2 ]: ΔE alloy = E LiZnB4_t - (E LiBH4 + E Zn(BH4)3 )
The following graphs are available, showing the results from all groups:
• Weight percent (wt.%) hydrogen as a function of ΔE decomp
• Hydrogen density (ρ hydrogen ) as function of ΔE decomp
• ΔE decomp as a function of ΔE alloy
• ΔE alloy for Li/Na/K for all 3d-transition metals (+Mg and Al) using only the stability
of the preferred M1M2B3 structure (tetra:□, octa: ○, tetra/octa:+, octa/tetra:Δ)
• ΔE alloy for Li/Na/K for all 4d-transition metals (+Ca and Ga) using only the stability
of the preferred M1M2B3 structure (tetra:□, octa:○, tetra/octa:+, octa/tetra:Δ)
• ΔE alloy for Li/Na/K for all 3d-transition metals (+Mg and Al) using only the stability
of the preferred M1M2B4 structure (tetra:□, octa:○, tetra/octa:+, octa/tetra:Δ)
• ΔE alloy for Li/Na/K for all 4d-transition metals (+Ca and Ga) using only the stability
of the preferred M1M2B4 structure (tetra:□, octa:○, tetra/octa:+, octa/tetra:Δ)
Assignment – part I
Please provide short answer the following 7 (5+2) questions and hand them in (see below).
1. Which alloy structures were found to be most stable for your systems LiZn(BH 4 ) 3,4 and
LiCd(BH 4 ) 3 , 4 ?
2. What storage capacities did you obtain (wt.% and ρ hydrogen )?
3. Check the ΔE alloy and ΔE decomp obtained when you ran check.py (reference values at
http://dcwww.camd.dtu.dk/school08/reference_energies)– do they match the plots?
4. Compare the alloy stabilities ΔE alloy to its “binary” constituents (e.g. LiBH 4 and Zn(BH 4 ) 3 ).
Did you obtain a stable alloy? Evaluate and discuss your results.
5. How did your structures compare to those from the other groups and the optimal value
ΔE decomp ≈ -0.3 eV/H 2 (see http://dcwww.camd.dtu.dk/school08/plots)?
Final part of the exercise
Based on the results you have obtained regarding the preferred coordination and structural stability
of your systems, you are now able improve the best structures even further. The template structures
have a number of constraints, which should now be lifted.
CAMD Summer School in Electronic Structure Theory 21-08-2008

Computational Materials Design Project (group10)
New Materials for Hydrogen Storage
• When you ran setup for your structures, a few extra directories called free1, free2, ... were
created. Every group member should use a different directory for individual optimizations.
Go to http://dcwww.camd.dtu.dk/school08/best.html and download the nc-files of
the best LiZn(BH 4 ) x and LiCd(BH 4 ) x structures (2 structures in total) to two of the relevant
…/free* directories. The nc file will only be available once all fixed structures have been
calculated and checked in.
• In …/free* you will find a template script, which reads in the nc-file, releases all constraints
and re-optimizes the structure (internal and external degrees of freedom). Please delete or
out-comment the line ’#sys.path.insert(0,'/usr/local/ase3-3.0.0.507-
1/lib64/python2.3/site-packages')before submitting the jobs. You should use 2 nodes
and the verylong queue for the free optimizations (and resubmit using the last nc-file as
input if the job completed or converged)
>qsub -l nodes=2:ppn=4:switch5 -q verylong ‘name’.py
>qsub –W group_list=gcourse -l nodes=2:ppn=4:switch5 -q verylong ‘name’.py
• Analyze if the energy of LiZn(BH 4 ) x -free was lowered and whether the system retained its
coordination (analyze the xyz file). Repeat the procedure for LiCd(BH 4 ) x -free.
• If you are very ambitious – try the following (else go to next bullet)! Use the insight you
have gained from your calculations to try to setup and generate a new structure with the
preferred local coordination. You can also use the reference structures for inspiration
http://dcwww.fysik.dtu.dk/~strabo/school/reference/. Run
/home/camp/dlandis/ss08scripts/setup2.py twice to generate the LiZnB and LiCdB
directories. From the script /home/camp/dlandis/ss08scripts/structures.py you can
copy a suitable template function, e.g. m1m2b2_t, into the LiZnB.py or LiCdB.py scripts
found here to setup a new init structure (modify the template to change, distances or
angles,…). Use the web interface
(http://dcwww.fys.dtu.dk/~strabo/school/group10) to visualize your structures. Use
your new optimization script to minimize the energy of the new structure; your can try to
analyze many different structures. If you get a lower energy, find an assistant – quickly!
• Finally, when you have completed this part of the exercise, use the check.py function to
upload the most stable – free - structures you obtained for LiZn(BH 4 ) x and LiCd(BH 4 ) x to
the database. Please check the nc and python files in.
Assignment – part II
Finally, you should evaluate your new structure(s)
6. Describe the preferred coordination of “Li” and “Zn/Cd”in the investigated structures. Did
they remain in the tetra/octahedral setup or did they form a new structure?
7. Did the free structures become more stable, and if so provide a possible explanation?
Editing the paper
If you made some interesting observations and want to contribute more to the paper, please check
out the tex-file and update it. If you experience problems please ask an assistant!
A checkout can be done:
CAMD Summer School in Electronic Structure Theory 21-08-2008

Computational Materials Design Project (group10)
New Materials for Hydrogen Storage
$> svn checkout https://svn.fysik.dtu.dk/projects/project08/paper
~/project08/paper
After this you can edit the file using gedit or your favourite editor.
$> gedit ~/project08/paper/project08.tex &
When you are done updating the text you can do a checkin – but other people might have changed
the document, so run a first
$> svn update ~/project08/paper
It might be that you have to merge – this task is not that easy – so please ask an assistant. After
merging you can commit:
$> svn commit ~/project08/paper -m “Updated tex file”
After the summer school we will complete the data analysis and text editing; the goal is to submit
our results as a paper to J. Chem. Phys. – with all active participants as co-authors!
Completion of the project
Your group should hand in the answers to the 7 questions, attached to and send via email to
project08@fysik.dtu.dk, in order to pass the summer school.
CAMD Summer School in Electronic Structure Theory 21-08-2008