A possible mechanism of copper corrosion in anoxic waterAnatoly B Belonoshko and Anders Rosengren

Theoretical physics, KTH

Background

• Common belief• Thermodynamic databases• Electronic structure theory• Other theoretical studies, other surfaces

(Ren and Meng, Taylor, Feibelman)

Our calculations

• We study (100) surface• A supercell, six layers of Cu in (001) direction and

a vacuum layer , periodic boundary conditions. The size 10.905x10.905x21.810 Å3

• Surface energy 1.388 J/m2 , exp 1.83 for (111)• Adsorption energy of a water molecule 0.22 eV,

same as obtained by Tang and Chen 2007• OH adsorption energy in excellent agreement

with Nørskov et al 2007

• Then inbetween slabs place OH and H separated laterally

• Calculate energy of adsorbed OH and H, i.e. of the dissociated water molecule. This energy is lower than the energy of H2O adsorbed intact.

• Thus we find dissociative adsorption of water on the surface in agreement with Taylor . Recently confirmed by another calculation.

Computational cell

Continuous supply of free surface?

• A mechanism that continously provides free copper surface for water dissociation

• We have earlier suggested one mechanism, nanoparticles, that would provide this surface

• Another way to increase this surface is to take grain boundary corrosion into account. If grain boundaries facilitate the removal of OH from the surface, the available surface for OH adsorption is essentially the surface of all grains in the sample

Clusters

• Magic number clusters N=13, 38, 55, 75, … unusually stable

• Cu clusters have been studied by EAM for 2 to 150 atoms. First principles, up to 13 atoms

• We apply first principles methods from 2 to 55. Put them in cubic box with edge 15 Å.

• Up-method and Down-method

The Cu cluster of 55 atoms

• OH binding to cluster, cluster size + # hydroxyls• Binding energy of OH to Cu(100) surface is -

2.61 eV. This is higher than the OH binding energy to a reasonably large cluster.

• Question: Can this gain in binding energy compensate the cost in energy for transferring Cu atoms from the bulk to the cluster?

• We calculated Cu55(OH)42.

The cluster of 55 Cu and 42 OH

Result

• The energy of Cu55 is -166.63 eV• The energy of Cu55(OH)42 is -620.07 eV• The energy of isolated hydroxyls is -378.78 eV• This gives OH binding energy to cluster -3.21 eV• But transfer of 55 Cu atoms from the bulk and 42

OH from the surface is larger by 9.89 eV• Conclusion: Formation of nanoparticulates

requires considerable energy and is not relevant.

Diffusion in grain boundaries

• Diffusion of O in bulk Cu is negligible• Removal of OH adsorbed on the Cu surface is

possible via grain boundaries only• Grain boundary penetration or intergranular

attack• At high temperature a grain boundary might

be approximated by a liquid structure due to premelting

Modeling the grain boundary

• Heat solid Cu to 4000 K• Anneal the liquid to 300K, 1200 K and 2200K• At 300 K and 1200 K Cu is solid (no self-

diffusion), however the radial distribution function remained non-solid. Formation of quasi-crystalline planes is seen

• At 2200 K the structure is liquid and quasi-crystalline planes vanish

Embedding OH in the grain boundary

• Two adjacent Cu atoms were removed from the center of the computational cell

• One position filled with O the other with H• O and H were shifted towards each other to

form the OH bond. Initial configuration.• Run molecular dynamics • D=2.25x10-8 (2200 K), 1.04x10-8 (1200 K) and

2.08x10-9 (300 K) m2/s

Discussion

• The quantity of emitted hydrogen in the ongoing experiment was 3x10-6 g/cm2

• A typical grain size in the Cu foil was 10-5 m. Approximate grains with fcc cubes with edge 10-

5 m. • Assume all surfaces of grains have adsorbed OH

to the same extent as the Cu surface• Grain boundary thickness 2-10 atomic distances

Order-of-magnitude estimate

• We obtain 10-6 g/cm2.• The release of hydrogen will continue• Some hydrogen will stay in the copper• Calculations show that OH dissociates

immediately and O and H diffuse independently

• Strong bond forms between O and Cu, and H is carried away

• Even more hydrogen is produced

• Copper oxide will be formed inside the crystal, probably as nanocrystals

• Hydrogen saturation leads to de-cohesion – as observed in experiments

• Oxidation will lead to a lattice expansion process, which might give rise to cracks and even more copper surface will be available

Conclusions

• We have investigated 2 possible mechanisms for OH removal from Cu surface

• Formation of Cu clusters with OH adsorbed• Diffusion of OH in grain boundaries• Possible formation of nanocrystals of copper

oxide. Cracks.