1:
2: \documentstyle[prb,%preprint,
3: aps,psfig]{revtex}
4: \begin{document}
5: \draft
6:
7:
8: %\onecolumn
9:
10: \title{Density functional study of Ni bulk, surfaces and the adsorbate
11: systems Ni(111)
12: $(\protect\sqrt{3} \times \protect\sqrt{3})$R30$^\circ$-Cl, and
13: Ni(111)${(2 \times 2})$-K}
14: \author{K. Doll}
15: \address{Institut f\"ur Mathematische Physik, TU Braunschweig,
16: Mendelssohnstra{\ss}e 3, D-38106 Braunschweig}
17: \maketitle
18:
19:
20: \begin{abstract}
21: Nickel bulk, the low index surfaces and the adsorbate systems
22: Ni(111)
23: $(\protect\sqrt{3} \times \protect\sqrt{3})$R30$^\circ$-Cl, and
24: Ni(111)${(2 \times 2)}$-K are studied with gradient corrected
25: density functional calculations. It is demonstrated that an approach based
26: on Gaussian type orbitals is capable of describing these systems.
27: The preferred adsorption sites and geometries are in good agreement
28: with the experiments. Compared to non-magnetic substrates, there
29: does not appear to be a huge difference concerning the structural
30: data and charge distribution. The magnetic moment of the nickel
31: atoms closest to the adsorbate is reduced, and oscillations of the
32: magnetic moments
33: within the first few layers are observed in the case
34: of chlorine as an adsorbate. The trends observed for the
35: Mulliken populations of the adsorbates are consistent
36: with changes in the core levels.
37: \end{abstract}
38:
39: \pacs{ }
40:
41: \narrowtext
42: \section{Introduction}
43:
44: Adsorption on surfaces is one of the main focuses in surface
45: science due to the enormous importance of topics
46: such as corrosion or catalysis. A huge number of experiments has
47: been performed to understand these processes.
48: In addition, theoretical studies have become capable of
49: modelling the adsorption process.
50:
51: The adsorption
52: of alkali metals or haloges on metallic surfaces may be considered
53: as two prototypes of adsorption processes. They are
54: fairly simple and experimentally well studied. However, there
55: have been quite a few surprises. For example, the occupancy
56: of top sites by alkali metals as adsorbates was unexpected (for reviews,
57: see references \onlinecite{DiehlMcGrath,Over,StampflScheffler} and more
58: general, reference \onlinecite{Zangwill}).
59: Halogens are another type of adsorbate which is experimentally
60: well studied. On close-packed
61: (111) surfaces, they usually occupy the face centered cubic (fcc)
62: hollow site, and in a few systems a partial occupancy (around
63: 15-25\%) of
64: hcp (hexagonal close-packed) sites has also been observed
65: \cite{Takata1992,Kadodwala}. In addition, there
66: is now a growing set of simulations for halogens on
67: metallic surfaces (e.g. chlorine
68: \cite{DollHarrison2000,DollHarrison2001,Jia2003},
69: bromine \cite{Wang2002,Blum2002} or iodine \cite{YunWang2001}).
70: Both for alkali metals and for halogens as adsorbates, there are
71: general questions such as the geometry, binding energies, diffusion
72: barriers, charge transfer and the binding mechanism.
73:
74: Magnetic surfaces pose additional difficulties and questions such
75: as how the magnetic moment changes after adsorption. It is
76: thus an interesting question which insights simulation can
77: give for these systems. Nickel surfaces and adsorption thereon
78: have been well studied by experimental groups and thus are a
79: prototype system for a computer simulation. For example,
80: hydrogen on Ni(111) \cite{Kresse2000}, benzene on Ni(111)
81: \cite{Mittendorfer2001},
82: CH$_3$ dehydrogenation on Ni(111) \cite{Michaelides2000}
83: or CO on Ni(110)\cite{Geetal2000} adsorption
84: have been simulated recently. The present article will deal with the
85: simulation of alkali and halogen adsorption, with Cl and K as
86: examples.
87:
88: The intention of this article is to give a broad overview over nickel
89: and the adsorbate systems. The computational technique
90: is based on a local basis
91: set of Gaussian type functions. As nowadays most calculations,
92: especially for metals, rely on plane-waves, the following section
93: \ref{Nibulksection} was devoted
94: to the fairly simple systems of Ni bulk and the low index surfaces.
95: This way, it will be demonstrated that results from a code based
96: on Gaussian type orbitals reproduce
97: the ones from plane-wave codes. In sections \ref{chlorineadsorptionsection}
98: and \ref{potassiumadsorptionsection},
99: the results for chlorine and potassium adsorption are presented,
100: respectively. Finally, the results are summarized and discussed.
101: Additional details on the calculations are given in the appendix.
102:
103: \section{Ni bulk and low index surfaces}
104: \label{Nibulksection}
105:
106:
107:
108: \subsection{Computational Parameters}
109:
110: A local basis set formalism was employed where the basis functions are
111: Gaussian type orbitals centered at the atoms as implemented in the code
112: CRYSTAL \cite{Manual}. The calculations
113: were done with spin-polarization, using the gradient corrected functional of
114: Perdew, Burke and Ernzerhof (PBE) \cite{PBE}, and in some additional
115: calculations the gradient corrected functional by Perdew and Wang (PW)
116: \cite{Perdewetal}. These functionals perform actually nearly identical
117: (see, e.g. references
118: \onlinecite{PBE,Hammeretal}, or with the CRYSTAL code for Ag
119: \cite{DollHarrison2001}), and also in the present article the results
120: with both functionals are very similar.
121: The PBE functional may be viewed as slightly favorable because
122: it was designed to have a simpler form and derivation, and
123: already revised versions exist (see, e.g., reference \onlinecite{Hammeretal}).
124:
125: The nickel basis set from reference \onlinecite{Towler} was employed, with the
126: innermost $[4s3p1d]$ exponents being unchanged. In addition, 2 $sp$
127: functions with exponent 0.63 and 0.13 and a $d$-function with exponent
128: 0.38 as optimized in calculations for the bulk were used, so that
129: the basis set as a whole was of the size $[6s5p2d]$.
130: The chlorine and potassium
131: basis sets were the ones used in previous studies on
132: Cl/Cu(111) and K/Cu(111), respectively\cite{DollHarrison2000,Doll2001KCu}.
133: For the calculations on the free atoms, enhanced basis sets with
134: additional diffuse functions were used \cite{basisdetail}.
135: For the numerical integration, the potential was fitted with
136: auxiliary basis sets (for Cl the same as in reference
137: \onlinecite{DollHarrison2000}, for K as in reference
138: \onlinecite{Doll2001KCu}, and for Ni the same as for Cu in reference
139: \onlinecite{Doll2001KCu}).
140: A $\vec k$-point net with 16 $\times$ 16 points in the surface
141: Brillouin zone and
142: 16 $\times$ 16 $\times$ 16 points for the bulk was used. A
143: smearing temperature of 0.001 $E_h$ was applied to the Fermi function
144: ($1 E_h$=27.2114 eV=315773 K).
145: This value for the temperature was chosen
146: relatively low
147: to ensure that the magnetic moment is not artificially modified by a too high
148: value. For more details and tests on the computational parameters and
149: basis sets, see also the appendix.
150:
151: The calculation of the charge and spin population is based on the Mulliken
152: population,
153: which is actually due to McWeeny\cite{McWeeny,Mulliken}
154: (see also reference \onlinecite{McWeenyIJQC}).
155: The Mulliken charge is defined as
156: $\sum_{\mu \in A}\sum_{\nu}P_{\mu\nu}S_{\nu\mu}$
157: (with density matrix $P$, overlap matrix $S$ and $A$ the set of basis
158: functions for which the population is computed). In the case of the
159: spin population
160: the density matrix for the corresponding spin polarization has to
161: be used. This type of population analysis thus
162: depends on the basis sets. It will be questionable if basis sets
163: of different quality are used for various atoms: if one basis set
164: is very small and the other very large, and especially if the larger
165: basis set also has more diffuse exponents (i.e. smaller exponents,
166: which results in a Gaussian with a relatively large spatial extension),
167: the larger
168: basis set may describe a part of
169: the charge of the atom with the smaller basis set as well,
170: and thus the Mulliken charge will be artificially large for the atom
171: with the larger basis set. It is thus important that balanced basis
172: sets are used.
173: Concerning the Mulliken populations computed in this article,
174: the population can be expected to be fairly
175: reliable in the case when identical basis sets are used for the
176: various atoms in a slab, so that the magnetic moments computed for
177: the various layers of
178: a nickel slab should not have a too large error due to the choice of the
179: basis set. Also, comparing results for various adsorption sites
180: should not be too unreasonable, because the basis sets for the systems
181: are identical. This means, that, for example,
182: a relative comparison of
183: the Mulliken charge for chlorine on the fcc site with
184: the charge on the top site should not be too unrealistic, although
185: the absolute value of the Mulliken charge may have a larger error.
186: A more difficult
187: issue is however, to compare the chlorine charge for Cl/Ni(111)
188: with the one for Cl/Ag(111) or Cl/Cu(111) because different basis
189: sets for the metals are used and because of the different geometries.
190:
191:
192:
193: The adsorbate systems Ni(111)
194: $(\protect\sqrt{3} \times \protect\sqrt{3})$R30$^\circ$-Cl
195: and Ni(111)${(2 \times 2)}$-K
196: were modeled with a slab consisting of 3 nickel
197: layers and the adsorbate layer, with the adsorbate layer on one side
198: of the slab. The model is truly two-dimensional and not periodically
199: repeated in the third dimension. This is different from the more common
200: approach where periodicity in three dimensions is used, and
201: a number of vacuum layers is required to separate the periodically repeated
202: slabs. In the latter case, this also introduces an artificial electrical
203: field due to the periodic boundary conditions in the third dimension,
204: which is usually compensated for by an additional dipole field
205: \cite{NeugebauerScheffler}. The approach used by the CRYSTAL code for
206: slabs avoids this artificial field as there are no periodic boundary
207: conditions in the third dimension, and thus there is no correction
208: required, and there is no parameter for the number of vacuum layers either.
209: The simulation was performed for the
210: fcc, hcp, bridge and top site
211: (see figures \ref{geometryfigure} and \ref{geometryfigure2}). To ensure
212: the stability of the results with respect to the number of layers,
213: additional calculations with thicker
214: slabs were performed for the systems Cl/Ni(111) (which is easier to
215: converge than K/Ni(111)).
216:
217: \subsection{Properties of bulk Ni and clean Ni surfaces}
218:
219: Results for cohesive properties of bulk Ni are displayed in Table
220: \ref{Nibulktable}. The computed values in this work reproduce
221: essentially the ones obtained with the gradient corrected functional
222: of Perdew and Wang, with the plane-wave
223: pseudopotential method \cite{Moroni}.
224: The binding energy is overestimated, compared to the experiment.
225: Also, the energy splitting between $d^8s^2$ and $d^9s^1$ is not
226: well reproduced (the $d^9s^1$ state should be lower by
227: -0.03 eV, after averaging the experimental energies over spin-orbit
228: components\cite{Moore,MartinHay1989}).
229: This is however, not a serious problem, as the
230: orbital occupancy in the solid is different from the atom and
231: thus the error cancellation does not work too well, and a density
232: functional description may be too crude to describe these effects.
233: The splitting between $d^8s^2$ and $d^9s^1$ state is a problem
234: well known in quantum chemistry. The relativistic effect to the
235: energy splitting between these states was
236: computed to be of the order of 0.3 eV
237: \cite{MartinHay1989,RaghavachariTrucks1989} (i.e. in a non-relativistic
238: calculation as the present one, the splitting should be
239: -0.33 eV instead of -0.03 eV) and thus is not negligible.
240: With highly accurate
241: configuration interaction schemes \cite{RaghavachariTrucks1989},
242: the calculated splitting is -0.18 eV; i.e.
243: there is still an error of 0.15 eV, compared to the experimental splitting,
244: after relativistic effects were subtracted from the experimental value.
245: The value computed here
246: (-0.035 $E_h$ $\hat=$ -0.95 eV); is in reasonable agreement with
247: other results from literature, e.g. in the range of -1.1 to
248: -1.2 eV, both at the level of the local density approximation
249: (LDA) and the level of the
250: generalized gradient approximation (GGA), with the PW functional
251: \cite{Moroni}. This difficulty
252: in obtaining the correct energy splitting between
253: $d^{n-1}s^1$ and $d^{n-2}s^2$
254: was also identified as a general problem for the transition metals
255: \cite{HarrisJones}, with the LDA, where a bias towards d$^{n-1}s^1$ of
256: $\sim$ 1 eV was found. With a different gradient corrected functional
257: (B-LYP), an energy splitting of -0.82 eV was computed\cite{Russoetal}.
258: There is, however, no reason why other properties should be affected
259: by this problem.
260:
261: The band structure is displayed in
262: figure \ref{nibulkbandstructure}. It
263: agrees well with the early tight binding
264: \cite{Langlinais1972,WangCallaway1974} or augmented plane
265: wave\cite{Connolly1967} results, and thus supports
266: the validity of the applied approach, relying
267: on Gaussian type orbitals.
268:
269: Surface formation energies (in $E_h$ per surface atom)
270: of the clean, unrelaxed surfaces are displayed in Table
271: \ref{Nisurfsummtable}. The surfaces were modeled with a finite number of
272: Ni layers (typically from 1 to 9).
273: The PBE energies obtained here are within literature values. For the
274: (111) surface, the surface energy
275: was additionally computed with the Perdew-Wang functional, and is
276: identical to the one with the PBE functional.
277:
278: Finally, in Table \ref{surfacemagneticmoment}, the magnetic moments of
279: the layers of the unrelaxed
280: slabs are given. In all cases, they were computed using
281: a 9 layer slab. They are in good agreement with literature values.
282: For the (111) surface
283: it is found that the second layer in the slab has the highest magnetic moment
284: and not the top layer, in agreement with two of the published articles.
285: For comparison, for the (111) surface, relaxation of the top layer was
286: additionally taken into account. This resulted in a small inwards
287: contraction by 0.03 \AA, which is consistent with experimental
288: \cite{Luetal1996} and theoretical \cite{Mittendorfer1999}
289: findings. The magnetic moments were virtually unchanged,
290: and thus relaxation does not seem to explain why in one publication
291: it was observed that the outermost layer should have the highest moment
292: \cite{Mittendorfer1999}. For comparison, also the Perdew-Wang functional
293: was used to compute the moments of the unrelaxed (111) surface, and the
294: results of the PBE functional were essentially reproduced.
295: As explained earlier, the calculation of the magnetic moments is based
296: on the Mulliken population analysis. The error should, however, not be
297: too large as identical basis sets are used for all the nickel atoms
298: of the slabs.
299:
300:
301:
302: \section{Cl on the Ni(111) surface}
303: \label{chlorineadsorptionsection}
304:
305:
306: As a first adsorbate system, the adsorption of Cl on the Ni(111)
307: surface was investigated. This system was studied experimentally
308: with LEED (low energy electron diffraction), Auger electron spectroscopy,
309: work function measurements and flash desorption \cite{Erley1977},
310: surface extended X-ray absorption fine
311: structure (SEXAFS) and soft X-ray standing wave (XSW)
312: \cite{Funabashi1989,Funabashi1991,Takata1992},
313: and angle-resolved photoemission fine-structure measurements
314: (ARPEFS)\cite{Wangetal}. At low coverages there were indications
315: of a $(2 \times 2)$ pattern, followed by a
316: $(\protect\sqrt{3} \times \protect\sqrt{3})$R30$^\circ$ structure, and
317: a complex $\left( 2 \ 1 \atop {4 \ 7} \right)$ pattern. The
318: $(\protect\sqrt{3} \times \protect\sqrt{3})$R30$^\circ$ structure is
319: the one which has been extensively studied experimentally and thus
320: also the calculations will focus on this pattern.
321: There was agreement that
322: the fcc hollow was the preferred site (with an fcc:hcp population
323: ratio of around 85:15\cite{Takata1992}, and the Cl-Ni bond length was
324: measured to be 2.40 $\pm$ 0.04 \AA \cite{Funabashi1989}, 2.33
325: $\pm$ 0.02 \AA \cite{Funabashi1991} or 2.332 $\pm$
326: 0.006 \AA \cite{Wangetal}. This corresponds to an interlayer distance of
327: 1.92 \AA \cite{Funabashi1989}, 1.83 \AA
328: \cite{Funabashi1991} or 1.837 \cite{Wangetal}.
329: The first nickel layer was found to be practically
330: unrelaxed in two experiments\cite{Funabashi1989,Funabashi1991},
331: whereas in the other article, a 5\% contraction with respect to the bulk
332: value was deduced \cite{Wangetal}.
333:
334: It is thus an interesting issue to compare with results from the
335: simulations. These results are presented in Table \ref{ClonNitable}.
336: A structural optimization was performed with 3 and 4 nickel layers.
337: An additional single point calculation with 5 nickel layers was
338: performed for fcc and hcp site, where the geometry from the thinner
339: slabs was used, with one additional nickel layer at the distance
340: corresponding to the bulk.
341: The vertical positions of the
342: chlorine adatoms and of the first nickel
343: layer were optimized. Lateral relaxations are not possible for fcc
344: and hcp site. As bridge and top site are much higher in energy,
345: lateral relaxations were not taken into account for these sites either.
346: Comparing the data for three and four layers, we see that the
347: geometry is well converged, and the difference in the
348: energy splitting between the
349: sites is within what is expected as the numerical noise (e.g.
350: fcc vs. hcp site, the energy splitting is 0.7 $mE_h$ with three nickel
351: layers,
352: 0.2 $mE_h$ with four layers, and 0.6 $mE_h$ with five layers).
353: Firstly, we note that the fcc hollow is indeed found to be lowest in
354: energy, with the hcp hollow being nearly degenerate at an energy
355: difference of only 0.2-0.7 $mE_h$ ($\sim$ 60-220 K).
356: Although the numerical noise is not
357: negligible when energy differences on this small scale are computed
358: (see appendix), this energy difference would be compatible with the
359: simple thermodynamic estimate given in reference \onlinecite{Kadodwala},
360: with copper as substrate.
361: The bridge site is higher in energy
362: by $\sim$ 3 $mE_h$ (or more than 900 K),
363: the top site by 18 $mE_h$, and thus thermodynamically fcc and hcp sites
364: are clearly favored. The energy splitting between fcc (or hcp site), and
365: the bridge site, may also be a crude estimate for the diffusion barrier
366: from one threefold-hollow site to another one.
367: These findings are
368: actually very similar to those
369: \cite{DollHarrison2000,DollHarrison2001,Jia2003} for
370: Cu(111)$(\protect\sqrt{3} \times \protect\sqrt{3})$R30$^\circ$-Cl and
371: Ag(111)$(\protect\sqrt{3} \times \protect\sqrt{3})$R30$^\circ$-Cl,
372: and also for the system Ag(111)$(1 \times 1)$-O \cite{Lietal2002}:
373: on the (111) surface,
374: the highly coordinated sites are clearly lowest in energy for adsorbates
375: with a small radius.
376: The bond lengths from chlorine to the nearest nickel
377: $d_{\rm {Cl-Ni \ nn}}$
378: also increase with increasing coordination number
379: and the binding energy increases with increasing coordination number,
380: as is to be expected from Pauling's argument \cite{Pauling}:
381: one strong Cl-Ni bond in the case of the top site will be shorter
382: than the three weaker bonds in the case of fcc and hcp, but the
383: three bonds as a whole are energetically more favorable.
384: The computed bond length agrees well with the values from the literature
385: \cite{Funabashi1989,Funabashi1991,Wangetal}.
386:
387: There are two different experimental results for the interlayer distance
388: between first and second nickel layer. The computed value from
389: this article agrees very well with the experimental from references
390: \onlinecite{Funabashi1989,Funabashi1991},
391: whereas the large contraction
392: of 0.10 \AA \ \cite{Wangetal}
393: with respect to the bulk appears to be unlikely.
394: A smaller contraction would also be in line with simulations
395: with Cu or Ag as substrate \cite{DollHarrison2000,DollHarrison2001,Jia2003}.
396:
397: An effective radius for Cl on Ni(111) could be computed by subtracting
398: an effective nickel radius of $\frac{3.53}{2\sqrt{2}}$\AA \ from
399: the bond length $d_{\rm {Cl-Ni \ nn}}$ of 2.34 \AA. This would result
400: in a radius of 1.09 \AA, in agreement with the computed values
401: \cite{DollHarrison2000,DollHarrison2001}
402: for Cl on Cu(111) (1.12 \AA) or Cl on Ag(111) (1.17 \AA). Thus the radius for
403: Cl is intermediate between the atomic (0.99 \AA) and ionic (Cl$^-$; 1.81 \AA)
404: \cite{Kittel}.
405: The bond length $d_{\rm {Cl-Ni \ nn}}$
406: is also in agreement with
407: experimental values for systems with similar atomic radii for
408: the substrate (e.g., bond lengths of
409: 2.39 \AA \ for the substrate Cu(111)\cite{Crapper1986},
410: 2.38 \AA \ for Rh(111)\cite{Shard1999}, 2.39 \AA \
411: for Pd(111)\cite{Shard2000}).
412:
413: In Table \ref{Clpopulationtable}, the Cl population
414: is displayed. The data is extracted from the calculations where
415: three nickel layers were used to model the substrate, but it
416: is virtually identical to the data obtained with four nickel layers.
417: The negative charge is largest for the top site, smaller
418: for the bridge site and smallest for hcp and fcc site. This is
419: consistent with the position of the 3$s$ core eigenvalue relative to
420: the Fermi level, which is lowest
421: for hcp and fcc, increasing for bridge and largest for the top site
422: (because the negative charge of the chlorine adsorbate is increasing and thus
423: the core levels are destabilized). Note that the core level for
424: majority spin is also slightly lower (by $\sim$ 0.002 $E_h$) than
425: for minority spin.
426: The density of states is visualized in figure \ref{ClNiDOS},
427: projected on the chlorine basis functions, and projected on all
428: basis functions (i.e. the total density of states). The different
429: occupancy of the two spin states becomes visible. Note that the
430: chlorine spin is mainly due to the different population of the second
431: and third $p$-shell, and not of the fourth (if we order the shells
432: according to the size of the exponents, starting with the highest).
433: This means that the outermost and thus most diffuse exponent does not
434: carry a huge spin and thus the chlorine spin is not an artifact
435: of the Mulliken population (if the spin was on the most diffuse
436: function only, one might argue that this basis function also describes
437: the spin on the nickel atoms and thus the Mulliken
438: population analysis would be questionable).
439:
440:
441: The computed Mulliken charge
442: is also in agreement with
443: the experimental estimates of 0.1 to 0.2 $|e|$ derived from work
444: function measurements\cite{Erley1977}. The overlap population
445: $\sum_{\mu \in A}\sum_{\nu \in B}P_{\mu\nu}S_{\nu\mu}$
446: (with density matrix $P$, overlap matrix $S$ and $A, \ B$ the set of basis
447: functions of atoms $A$ and $B$ for which the overlap population is computed)
448: is 0.09 for Cl and the nearest nickel atoms.
449: This may indicate that
450: there is no evidence for a strong covalent contribution to the binding
451: mechanism, similar to non-magnetic substrates
452: \cite{DollHarrison2000,DollHarrison2001,Doll2001KCu,Doll2002},
453: and now
454: also in the case of a
455: magnetic substrate.
456: For comparison, the overlap population between two
457: nearest nickel atoms in the top nickel layer is 0.10.
458: However, this data should not be over-interpreted, and the
459: overlap population should rather be viewed as a rough qualitative
460: measure and not a quantitative measure for covalency.
461: Another indication that the covalency is not strong might be that
462: in the case of the bridge site, where $p_x$ and $p_y$ orbitals do not have
463: the same occupancy, the orbital with the larger overlap with
464: the neighboring nickel atoms ($p_x$ in our choice of geometry) has
465: a slightly smaller population than $p_y$
466: (the total Cl $p$ population is
467: 3.73 for $p_x$, versus 3.82 for $p_y$).
468:
469:
470: Charge density, charge density difference
471: and spin density are visualized in figure
472: \ref{ClonNichargespindensity}. As usual,
473: the charge density difference is more
474: informative as the charge density.
475: The transfer from the top nickel layer to the chlorine atom
476: becomes obvious. In contrast to the charge density, the spin density
477: also offers more information:
478: it is very interesting that the spin density
479: varies relatively strong still
480: for the atoms in the third layer. This is
481: confirmed by the data from Mulliken population analysis for the
482: spin (table \ref{Clspintable}). The magnetic moments depend relatively
483: strong on the number of layers, and the analysis has been performed
484: by extracting the data from a single-point calculation for a
485: slab with 5 nickel layers, with an identical geometry as for
486: the 3-layer slab, i.e. the additional layers were added at a distance
487: as in the bulk.
488:
489: The data is presented in such a way that one number is given, if
490: all the atoms in a layer have the same distance to the chlorine atom, and
491: two numbers, if there are two different distances for the atoms in one
492: layer possible.
493: For example, in the case of the fcc site, looking at
494: the atoms in the first or second layer,
495: all three nickel atoms within a supercell
496: have the same distance to the adsorbed chlorine. However, in the third
497: layer there will be one atom which is closer than the other two. The
498: population of this closer atom (0.59) is displayed in the left column of
499: the results for the $3^{rd}$ layer, the data for the other two atoms (0.64)
500: in the right column.
501:
502: It is interesting to note that those nickel atoms in the first layer
503: which are closer to the chlorine atom have a reduced magnetic moment. This
504: is strongest in the case of the top site where the atom vertically under
505: the chlorine has its population reduced to 0.55, versus a moment of 0.76
506: for the other two atoms. It is also apparent for the bridge site
507: (0.66 vs 0.73), and in the case of fcc and hcp site the data is identical
508: because the nickel atoms in the first layer are identical by symmetry.
509:
510: In the second layer, the spin population is higher for those atoms
511: which are closer to the adsorbate (this is possible for the hcp and
512: bridge structure). In the third layer, again those atoms which have
513: a shorter distance to the adsorbate, have there moment reduced.
514: These oscillations are still visible in the fourth and fifth layer.
515:
516: Although this data depends relatively strong on the number of layers,
517: it becomes apparent that there are oscillations in the spin
518: density. It is also interesting to compare with the total charge:
519: the latter quantity is virtually constant within the layers,
520: from the second layer on, and has a value identical to that of
521: a clean slab from the third layer on (also table \ref{Clspintable}).
522: Only in the first layer, the atom(s) closer to the chlorine adsorbate
523: have a slightly higher charge (i.e. more electronic charge) than the
524: atom(s) which are further apart (visible for bridge and top site).
525:
526: Finally, the chlorine adsorbate has a magnetic moment of $\le$ 0.1 for
527: all sites (the smallest value for the top site). The moment is always
528: parallel to the nickel spin.
529:
530:
531: It is interesting to note that in the simulations
532: for CO on Ni(110) \cite{Geetal2000}, it was also found
533: that the moment of the atom(s) closest to the CO was reduced.
534:
535:
536: \section{K on the Ni(111) surface}
537: \label{potassiumadsorptionsection}
538:
539: In addition, the system Ni(111)${(2 \times 2)}$-K was studied with
540: this approach. Experimentally,
541: in a LEED study at various coverages, this was found
542: to be the only commensurate structure \cite{Chandavarkar1988}.
543: In a further LEED study\cite{Fisheretal,Kaukasoinaetal1993},
544: the top site was found to be occupied, with a K-Ni distance of
545: 2.82$\pm$ 0.04 \AA, a vertical rumpling of the first Ni layer of 0.12 \AA,
546: and a lateral displacement of 0.06 \AA \ of the three nickel atoms
547: in the top layer which are not vertically under the K adsorbate.
548: The adsorption site was later confirmed in SEXAFS
549: measurements\cite{Adler1993},
550: and a larger bond length of 2.92$\pm$ 0.02 \AA \ was obtained.
551: With the same technique, also the bond length for the system K/Cu(111)
552: was computed and deduced to be by 0.06 $\pm$ 0.04 \AA \ larger.
553: From
554: ARPEFS\cite{Huangetal1993},
555: a top site adsorption with K-Ni bond-length of 3.02 $\pm$ 0.01 \AA \
556: was obtained.
557: In addition, the first to second layer spacing was reduced by 0.13 \AA \
558: to 1.90 $\pm$ 0.04 \AA,
559: but no substrate rumpling in the top layer was observed. Lateral
560: displacements were not found either (0.00 $\pm$ 0.09 \AA) in this
561: experiment.
562: Finally, photoelectron diffraction (PhD) \cite{Davis1994} resulted
563: in a bond length of 2.87 $\pm$ 0.06 \AA, and a reduced first
564: to second layer distance of 1.86 $\pm$ 0.06 \AA \ was observed, with
565: virtually no rumpling in the first layer (0.01 $\pm$ 0.09 \AA).
566:
567: The experiments thus consistently support the top site as the
568: adsorption site, but the geometrical parameters are not identical.
569: Moreover, a comparison with the system K/Cu(111) is very interesting,
570: so that the influence of the partially occupied $d$-bands can be
571: investigated.
572:
573: The results of the simulations are summarized in Table \ref{KonNiEnergie},
574: with geometrical parameters as defined in figure \ref{geometricalparameters}.
575: Initially, the potassium position and the position of the top nickel layer was
576: optimized (i.e. the uniform relaxation of the top nickel layer). In
577: a second step, a vertical rumpling of the nickel atoms in the top layer
578: was allowed, as this is expected to be important for the top site
579: occupancy. A further refinement would have been the additional possibility
580: of lateral relaxations. These lateral relaxations were,
581: however, found to be of minor
582: importance for the related system K/Ag(111), where the energy
583: was lowered by the order of typically
584: 0.1 - 0.2 $mE_h$, when these relaxations were
585: allowed\cite{Doll2002}.
586: Thus, lateral relaxations were taken into account additionally
587: only for
588: the top site, which was found to be lowest in energy, and in order
589: to compare with the experimental data for the lateral displacement.
590:
591: Firstly, the top site is found to be the most favorable site. This is
592: in agreement with the experimental
593: findings \cite{Fisheretal,Kaukasoinaetal1993,Adler1993}.
594: By disabling
595: the possibility of substrate rumpling, it is also demonstrated that
596: substrate rumpling is crucial for top site occupancy, as it lowers
597: the energy by 1.6 $mE_h$, but only 0.4 $mE_h$ for the other
598: sites. These findings are actually similar to the
599: system K/Cu(111) \cite{Doll2001KCu}. The top site thus becomes the most
600: stable site because the nickel atom underneath is pushed into the substrate
601: relative to its neighbors and the potassium with its relatively large
602: radius thus also overlaps more with the second nearest nickel neighbors.
603:
604:
605: The computed bond length
606: (2.79 \AA) is also very similar to K/Cu(111)
607: (2.83 \AA) \cite{Doll2001KCu} so that these calculations do not
608: indicate a major difference compared to Cu.
609: Experimentally, using the same technique for both adsorbate systems,
610: it was measured to be slightly shorter
611: and it was speculated that this was due to greater adatom-substrate
612: interactions involving the partially filled Ni $d$ bands\cite{Adler1993}.
613: The possibility of a lateral displacement of the nickel
614: atoms as observed experimentally (0.06 \AA)\cite{Fisheretal}
615: was investigated. However, this could not be confirmed and
616: the energy was found to be lowest for the case where this
617: lateral displacement is exactly zero.
618:
619:
620: Bridge, fcc and hcp site are found to be energetically degenerate.
621: The fact that all the sites are energetically very close
622: (e.g. much closer than in the case of chlorine)
623: is consistent with earlier findings
624: \cite{NeugebauerScheffler,Doll2001KCu,Doll2002}
625: and was
626: explained due to the large radius of the adsorbate which makes
627: it experience only a small substrate electron density corrugation
628: \cite{NeugebauerScheffler}. This small energy splitting between the
629: various sites is also consistent with the experiment
630: because of various observations\cite{Adler1993}: firstly,
631: the SEXAFS data indicated enhanced
632: in-plane motion resulting from a wide and shallow potential well.
633: Secondly, no SEXAFS signal was observed for low coverages below 0.13
634: monolayers. It was thus argued that lateral motion was essentially hindered
635: by the presence of other adatoms. Finally, with increasing temperature
636: (already in the range of 120-145 K)
637: the SEXAFS amplitude was found to decrease. This was explained due
638: to the occupancy of other sites, which leads to a destructive interference.
639:
640: The bond length $d_{\rm {K-Ni \ nn}}$
641: increases with the number of nearest nickel neighbors in
642: the order top, bridge, fcc and hcp. The effective radius of the
643: potassium adsorbate is 2.79-$\frac{3.53}{2\sqrt{2}}$ \AA=1.54 \AA \ for
644: the top site. This
645: number is in good agreement with experimental values for potassium
646: when occupying the top site \cite{DiehlMcGrath}, or the computed
647: value for K/Cu(111) (1.55 \AA\cite{Doll2001KCu}).
648:
649: Looking at the populations in Table
650: \ref{Clpopulationtable},
651: we see that they ($\sim$ 0.3 $|e|$)
652: are very similar for all considered
653: sites, only for the top site the charge is slightly less positive.
654: In agreement with this, the core eigenvalues of the $3s$ and $3p$
655: orbital are virtually independent of the site. The core levels
656: are identical for both spin polarizations which agrees with the
657: finding that the magnetic moment of the potassium adatoms is negligible
658: (see below). In the calculations for K/Ag(111) \cite{Doll2002}, it
659: was recently shown that the adsorbate charge and thus also the
660: core eigenvalue (and other properties such as the bond
661: length) change with the coverage.
662: Such a change in the core eigenvalues was observed experimentally for
663: alkali metals on metallic surfaces, e.g.
664: for the systems Na/Cu(111) and Na/Ni(111) \cite{Shietal1993},
665: or for alkali metals
666: on Ru(100) \cite{Sheketal1990} or W(110) \cite{Riffeetal1990}.
667:
668: The overlap population
669: is $\sim$ 0.03 for
670: K and the nearest nickel atom, so that there is also in this case
671: no evidence for a strong covalent contribution to the binding mechanism.
672:
673: The density of states is displayed in figure \ref{KNiDOS}.
674: The integrated density of states, when projected on the potassium
675: basis functions, differs so little for majority and
676: minority spin that there is nearly no magnetic
677: moment. For the top site, the spin is even antiparallel to the nickel sites,
678: in contrast to the other sites (and in contrast to chlorine).
679:
680:
681:
682: Again, the charge density is not very informative, and
683: it is more interesting to look at
684: the charge density difference (figure \ref{KonNispin}). Electrons
685: have flown from the adsorbate mainly to the top nickel layer
686: (and a relatively large part
687: to the atom vertically under the potassium adatom). Thus, the
688: charge of the potassium overlayer has decreased.
689: It is apparent that the spin density of the nickel atom vertically
690: under the top atom is slightly lower than the spin density of its neighbors
691: in the same layer, as the contours are closer to the nucleus. This is
692: confirmed in the Mulliken spin population (table \ref{Kspintable}).
693: Again, those atoms in the top layer which are closest to the potassium
694: adsorbate, have their moment reduced.
695: No noteworthy variation is found in the second and third nickel layer,
696: and the values for the spin population are very close to
697: the ones for the clean Ni slab.
698: The moment of the potassium adsorbate is negligible, as its magnitude
699: is $\le$ 0.003.
700: Also, the charge only varies in the first layer, and already the second
701: and third layer have charges which are identical to that of a clean slab.
702:
703: There are thus no oscillations of the spin population beyond the first
704: nickel layer, probably because the potassium magnetic moment is much
705: smaller than that of chlorine.
706:
707:
708: \section{Summary}
709:
710: We have studied Ni bulk, the clean low index surfaces and
711: the adsorption of chlorine and potassium on the Ni(111) surface.
712:
713: The results for the adsorption geometry confirm the
714: experimental findings. It turns out that there is no huge
715: difference to copper as substrate; the preferred adsorption site and
716: energies are very similar.
717:
718: The reason for the preference
719: of the top site for K is substrate rumpling which helps to increase the
720: overlap of the K adsorbate with the second nearest nickel neighbors.
721: This rumpling is thus crucial
722: for the site preference. In the case of chlorine, the
723: threefold hollow sites are very close in energy. In general, the adsorbate
724: atoms
725: prefer sites with the highest possible coordination number, or they try
726: to increase the overlap to neighboring atoms by substrate rumpling.
727:
728: There appear to be no strong covalent contributions to the binding so
729: that the mechanism appears to be mainly an ionic bond (although
730: the charge transfer is rather small), and in the case of
731: potassium, additionally a metallic binding mechanism. However,
732: the quantitative analysis of the binding mechanism is certainly difficult.
733: The computed charges seem to be fairly reliable, but the overlap
734: population is probably only a qualitative tool to estimate the degree
735: of covalency.
736:
737: Charges and core levels were found to be consistent, i.e. with
738: increasing negative charge of the adsorbate, the core levels of the
739: adsorbate are slightly destabilized.
740: It is also
741: demonstrated that the magnetic moments computed with the help
742: of the Mulliken population analysis agree well with data from other
743: schemes, e.g. pseudopotential plane wave. The calculation of work
744: functions, however, seems to be difficult with a local basis set
745: and the results are strongly basis-set dependent.
746:
747: Finally, the magnetic moment of those nickel atoms closest to the
748: adsorbate are always reduced. Oscillations of the spin population
749: within one nickel layer are notable for the first three layers for Cl/Ni(111),
750: but only for the first nickel layer for K/Ni(111), probably because
751: of the very small magnetic moment of the potassium adsorbate,
752: compared to chlorine.
753:
754: \section{acknowledgments}
755:
756: All the calculations were performed at the computer centre of
757: the TU Braunschweig (Compaq ES 45).
758:
759: \appendix
760: \section{Test of computational parameters}
761: The calculations require various sets of parameters, with the
762: most important being: a finite number
763: of $\vec k$-points, a finite temperature to smoothen the integrand
764: and thus to facilitate the numerical integration of the exchange-correlation
765: functional, basis set parameters, and truncation schemes to reduce the
766: number of matrix elements (i.e. integrals of operators such as
767: kinetic energy, nuclear attraction, coulomb and exchange, with the
768: basis functions). In this first part of the
769: appendix, the stability of the results
770: with respect to the variation of parameters concerning the grid
771: and the parameters concerning the selection of the integrals is tested.
772: For further tests, see also the extensive tests performed for lithium
773: \cite{KlausNicVic}, or tests for copper\cite{DollHarrison2000}
774: and silver\cite{Doll2002} ($\vec k$-points, smearing temperature).
775: In tables \ref{ClNiparametertest} and \ref{KNiparametertest},
776: the binding energies are computed with the default
777: grid and truncation parameters, and with an improved grid and
778: two sets of stricter and thus more accurate truncation parameters.
779:
780: The improved grid has roughly six times more sampling points than the default
781: grid for the various structures.
782: As is displayed in the tables, the grid has virtually no impact on the
783: energy splitting. For Cl/Ni(111), the binding energy changes by
784: at most 0.2 $mE_h$ for fcc, hcp and bridge site, and only for
785: the top site an energy change of 0.9 $mE_h$ is observed, with respect
786: to the default grid. This has, however,
787: no impact as the top site is much higher in energy anyway.
788: For K/Ni(111) the absolute value of the
789: binding energy is shifted by
790: 0.7 $mE_h$, and the relative energies of the various sites
791: are identical.
792:
793: There are two truncation parameters in calculations when
794: no Fock exchange is involved\cite{Manual}:
795: a general one ("ITOL 1") which leads to
796: a neglect of all integrals
797: with an overlap below a certain threshold (default: 10$^{-6}$), and
798: a second parameter ("ITOL 2")
799: which leads to the evaluation of the coulomb integrals at a
800: lower level of accuracy by means of a multipolar expansion,
801: if the overlap is below another threshold (default: 10$^{-6}$). As
802: the results turned out to be more sensitive to these parameters, two sets of
803: higher thresholds were used (10$^{-7}$ for ITOL 1 and 2, and 10$^{-8}$
804: for ITOL 1 and 2).
805: In the case of chlorine, the largest error is found for the
806: splitting between fcc and hcp site: it ranges from 0.7 to 2.1 $mE_h$,
807: for the various parameters, with the fcc site always being more
808: favorable. The bridge site is always clearly
809: higher than the fcc site by 3.3 to 4.2 $mE_h$, and the top site
810: is much higher in energy.
811: For K/Ni(111), the impact of these parameters is smaller, and
812: for example, the splitting from the top site to the site closest
813: in energy ranges from 0.5 $mE_h$ to 0.8 $mE_h$.
814:
815: As a whole, the impact of these computational parameters is certainly
816: not negligible, but the conclusions are not affected. The fcc site
817: is favored for Cl/Ni(111), with the hcp site being nearly degenerate.
818: In the case of K/Ni(111), the top site is favored, with the other
819: sites being very close in energy.
820:
821:
822:
823:
824:
825: \section{Basis set dependence of the results}
826: \label{Basissetappendix}
827:
828: The choice of the basis set is of high importance for the quality
829: of the results. In this part of the appendix, the
830: results of various tests are displayed, to investigate the dependence
831: of the results on the basis set. These tests focus on nickel bulk,
832: the clean surfaces, and Cl/Ni(111).
833:
834: Firstly, one might consider reoptimizing the exponents for the
835: clean surface, for example keeping the exponents of the inner layer
836: like in the bulk
837: and reoptimizing the exponents of the outermost layers.
838: This would result (with a slab with three layers)
839: in exponents of 0.12 ($sp$, instead of 0.13 for the
840: bulk) and 0.38 ($d$, as in bulk nickel) for the outer layers.
841: This is thus a tiny change only
842: and the results are virtually identical, for example the surface energy
843: per atom changes by 0.6 $mE_h$ which is negligible. If an outermost
844: $sp$-exponent of 0.12 instead of 0.13 is used for the bulk,
845: exactly the same results as in table \ref{Nibulktable} are obtained.
846:
847: There is however a property which depends extremely strong on the
848: basis set: the computed work function changes drastically
849: when the outermost exponent is changed by a tiny value. For example,
850: with the $sp$-exponent used for all the calculations (0.13),
851: the work function of the Ni(100) surface is 0.142 $E_h$
852: (exp.\cite{Baker,Michaelson}: 0.192 $E_h$),
853: of the Ni(110) surface 0.129 $E_h$ (exp.\cite{Baker,Michaelson}:
854: 0.185 $E_h$), and of the
855: Ni(111) surface 0.155 $E_h$ (exp.\cite{Baker,Michaelson}: 0.197 $E_h$).
856: This data
857: is practically stable with respect to the number of layers (i.e.
858: 3 layers are sufficient to obtain a stable number).
859: When,
860: however, for example, the outermost $sp$-exponent
861: of the outer layers is changed from 0.13 to 0.12, the
862: work function of the Ni(111) surface changes by 0.01 $E_h$ to 0.165 $E_h$,
863: and it further increases with smaller exponents (e.g. 0.183 $E_h$, if
864: the outermost $sp$-exponent of the outer layers is 0.10 instead of 0.13).
865: Also, increasing this exponent leads to a lower work function.
866: The same problem shows also up with other metals, such as for
867: example copper or silver (and also when the exponents of all the
868: layers and not just the outermost layers are changed).
869: It appears thus that the work function is a quantity which is not
870: well described with a local basis set.
871: The work functions may be qualitatively reasonable, but a quantitative
872: comparison seems unadvised. Thus, also the work function data computed
873: for Ag and the adsorption thereon
874: \cite{Doll2002} has to be considered with caution. A problem with
875: computing work functions with a local basis set was already discussed
876: earlier \cite{Boettgeretal1995}.
877:
878: The results thus indicate that
879: the only property which appears to be strongly basis set dependent is
880: the work function
881: (total energy, surface energy,
882: bulk modulus and lattice constant are essentially
883: stable with respect to small variations in the basis set).
884: Mulliken charge and magnetic moment are stable if an identical basis
885: is used for all the nickel atoms in a slab.
886: Of course, if
887: basis sets of different type are used for the inner layer of
888: a slab and the outermost layers, the Mulliken population also gets
889: less reliable.
890:
891: Further tests were performed for the adsorbate system Cl/Ni(111).
892: One test was done by choosing a more tight outermost
893: $sp$-exponent of 0.15 instead of 0.09 for chlorine. When a structural
894: optimization is performed, virtually the same structural data as in table
895: \ref{ClonNitable} is obtained (the maximum deviation is 0.01 \AA).
896: The binding energy is lower by $3 mE_h$
897: because of the poorer chlorine basis set.
898: However, what is important is that the energy splitting for the various
899: sites remains practically identical (the binding energy is
900: -0.1326 $E_h$ for the fcc site,
901: -0.1320 $E_h$ for the hcp site, -0.1292 $E_h$ for the bridge site,
902: and -0.1151 $E_h$ for the top site).
903: Also, properties such as the population, magnetic moment
904: and the relative position of
905: the core levels with respect to the Fermi energy are essentially stable
906: with respect to this change of the basis set
907: (but again the work function changes strongly).
908:
909: As a whole, the basis set needs of course to be carefully chosen.
910: When this is done, the results for the geometry, energies, core
911: eigenvalues, populations, and magnetic moments are reliable and
912: not too strongly basis set dependent. The accurate determination of the
913: work function, however, appears to be difficult with a local basis set.
914:
915:
916: \begin{references}
917: \bibitem{DiehlMcGrath}
918: R. D. Diehl and R. McGrath, Surf. Sci. Rep. {\bf 23}, 43 (1996);
919: R. D. Diehl and R. McGrath, J. Phys.: Condensed Matter
920: {\bf 9}, 951 (1997).
921: \bibitem{Over} H. Over, Progr. in Surf. Sci. {\bf 58}, 249 (1998).
922: \bibitem{StampflScheffler} C. Stampfl and M. Scheffler, Surf. Rev. Lett.
923: {\bf 2}, 317 (1995); M. Scheffler and C. Stampfl, in:
924: Handbook of Surface Science, Vol. 2: Electronic Structure,
925: Editors: K. Horn and M. Scheffler, Amsterdam (1999).
926: \bibitem{Zangwill} A. Zangwill, Physics at Surfaces, University
927: Press, Cambridge (1988).
928: \bibitem{Takata1992} Y. Takata, H. Sato, S. Yagi, T. Yokoyama, T. Ohta,
929: and Y. Kitajima, Surf. Sci. {\bf 265}, 111 (1992).
930: \bibitem{Kadodwala}
931: M. F. Kadodwala, A. A. Davis, G. Scragg, B. C. C. Cowie, M. Kerkar,
932: D. P. Woodruff, R. G. Jones, Surf. Science {\bf 324}, 122 (1995).
933: \bibitem{DollHarrison2000} K. Doll and N. M. Harrison, Chem. Phys. Lett.
934: {\bf 317}, 282 (2000).
935: \bibitem{DollHarrison2001} K. Doll and N. M. Harrison, Phys. Rev. B
936: {\bf 63}, 165410 (2001).
937: \bibitem{Jia2003} L. Jia, Y. Wang and K. Fan, J. Phys. Chem. B {\bf 107},
938: 3813 (2003).
939: \bibitem{Wang2002}
940: S. Wang and P. A. Rikvold, Phys. Rev. B {\bf 65}, 155406 (2002).
941: \bibitem{Blum2002} V. Blum, L. Hammer, K. Heinz, C. Franchini, J. Redinger,
942: K. Swamy, C. Deisl and E. Bertel, Phys. Rev. B {\bf 65}, 165408 (2002).
943: \bibitem{YunWang2001} Y. Wang, W. Wang, K. Fan and J. Deng,
944: Surf. Sci. {\bf 487}, 77 (2001).
945: \bibitem{Kresse2000} G. Kresse and J. Hafner, Surf. Sci. {\bf 459},
946: 287 (2000).
947: \bibitem{Mittendorfer2001} F. Mittendorfer and J. Hafner, Surf. Sci.
948: {\bf 472}, 133 (2001).
949: \bibitem{Michaelides2000} A. Michaelides and P. Hu, J. Chem. Phys.
950: {\bf 112}, 8120 (2000).
951: \bibitem{Geetal2000} Q. Ge, S. J. Jenkins and D. A. King, Chem. Phys. Lett.
952: {\bf 327}, 125 (2000).
953: \bibitem{Manual} V. R. Saunders, R. Dovesi, C. Roetti, M. Caus\`a,
954: N. M. Harrison, R. Orlando, C. M. Zicovich-Wilson {\sc crystal 98} User's
955: Manual, Theoretical Chemistry Group, University of Torino (1998).
956: \bibitem{PBE} J. P. Perdew, K. Burke and M . Ernzerhof, Phys. Rev. Lett.
957: {\bf 77}, 3865 (1996).
958: \bibitem{Perdewetal} J. P. Perdew, J. A. Chevary, S. H. Vosko,
959: K. A. Jackson, M. R. Pederson, D. J. Singh, and C. Fiolhais,
960: Phys. Rev. B {\bf 46}, 6671 (1992).
961: \bibitem{Hammeretal} B. Hammer, L. B. Hansen, and J. K. N{\o}rskov,
962: Phys. Rev. B {\bf 59}, 7413 (1999).
963: \bibitem{Towler}M. D. Towler, N. L. Allan, N. M. Harrison, V. R. Saunders,
964: W. C. Mackrodt, E. Apr\`a, Phys. Rev. B {\bf 50}, 5041 (1994).
965: \bibitem{Doll2001KCu} K. Doll, Euro. Phys. J. B {\bf 22}, 389 (2001).
966: \bibitem{basisdetail} The energies of the neutral atoms were computed as
967: follows: the outermost Cl $sp$-shells were 0.315 and 0.119
968: instead of 0.294 and 0.090, yielding an energy of -459.9385 $E_h$. For K,
969: three diffuse $sp$ shells (0.39, 0.21, 0.03) replaced the two used
970: for the adsorption studies (0.29, 0.08), yielding an energy of -599.6672
971: $E_h$. For the nickel atom, 7 additional even tempered diffuse functions
972: in the range from 0.01 to 0.25 ($sp$), and 5 from 0.016 to 0.6 ($d$),
973: were employed,
974: the $[5s4p1d]$ tight inner exponents were kept as in the bulk.
975: Thus, as a whole a $[12s11p6d]$ basis set was used. The PBE-energies
976: were -1507.933 $E_h$ $(d^8s^2)$; (PW:
977: -1508.311 $E_h$), and -1507.967 $E_h$ ($d^9s^1$); (PW: -1508.346 $E_h$).
978: The energies of the nickel atom were calculated with integer occupation
979: numbers, the $d_{z^2}$ and $d_{x^2-y^2}$ orbital were singly occupied
980: in the case of the $d^8s^2$ state, and the $d_{x^2-y^2}$ orbital in
981: the case of the $d^9s^1$ state.
982: \bibitem{McWeeny} R. McWeeny, J. Chem. Phys. {\bf 19}, 1614 (1951).
983: \bibitem{Mulliken} R. S. Mulliken, J. Chem. Phys. {\bf 23}, 1833 (1955).
984: \bibitem{McWeenyIJQC} D. Cook and B. Sutcliffe,
985: Int. J. Quantum Chem. {\bf 60}, 1 (1996);
986: R. McWeeny, Int. J. Quantum Chem. {\bf 60}, 3 (1996).
987: \bibitem{NeugebauerScheffler} J. Neugebauer and M. Scheffler,
988: Phys. Rev. B {\bf 46}, 16067 (1992).
989: \bibitem{Moroni} E. G. Moroni, G. Kresse, J. Hafner and J. Furthm\"uller,
990: Phys. Rev. B {\bf 56}, 15629 (1997).
991: \bibitem {Moore} C. E. Moore, Atomic Energy Levels, NSRDS-NBS 35 / Vol. I-III,
992: Nat. Bur. Standards (Washington, DC, 1949, 1952, 1958).
993: \bibitem{MartinHay1989}R. L. Martin, P. J. Hay, J. Chem. Phys. {\bf 75}, 4539
994: (1981)
995: \bibitem{RaghavachariTrucks1989} K. Raghavachari and G. W. Trucks,
996: J. Chem. Phys. {\bf 91}, 1062 (1989).
997: \bibitem{HarrisJones} J. Harris
998: and R. O. Jones, J. Chem. Phys. {\bf 70}, 830 (1979).
999: \bibitem{Russoetal} T. V. Russo, R. L. Martin, and J. P. Hay,
1000: J. Chem. Phys. {\bf 101}, 7729 (1994).
1001: \bibitem{Langlinais1972} J. Langlinais and J. Callaway,
1002: Phys. Rev. B {\bf 5}, 124 (1972).
1003: \bibitem{WangCallaway1974} C. S. Wang and J. Callaway, Phys. Rev. B
1004: {\bf 9}, 4897 (1974).
1005: \bibitem{Connolly1967} J. W. D. Connolly, Phys. Rev. {\bf 159}, 415 (1967).
1006: \bibitem{Luetal1996} H. C. Lu, E. P. Gusev, E. Garfunkel, and
1007: T. Gustafsson, Surf. Sci. {\bf 352-354}, 21 (1996).
1008: \bibitem{Mittendorfer1999}F. Mittendorfer, A. Eichler and J. Hafner,
1009: Surf. Sci. {\bf 423}, 1 (1999).
1010: \bibitem{Erley1977} W. Erley and H. Wagner, Surf. Sci. {\bf 66}, 371 (1977).
1011: \bibitem{Funabashi1989} M. Funabashi, Y. Kitajima, T. Yokoyama, T. Ohta
1012: and H. Kuroda, Physica B {\bf 158}, 664 (1989).
1013: \bibitem{Funabashi1991} M. Funabashi, T. Yokoyama, Y. Takata, T. Ohta,
1014: Y. Kitajima and H. Kuroda, Surf. Sci. {\bf 242}, 59 (1991).
1015: \bibitem{Wangetal} L.-Q. Wang, Z. Hussain, Z. Q. Huang, A. E. Schach
1016: v. Wittenau, D. W. Lindle and D. A. Shirley, Phys. Rev. B {\bf 44}, 13711
1017: (1991).
1018: \bibitem{Lietal2002} W.-X. Li, C. Stampfl and M. Scheffler, Phys. Rev. B
1019: {\bf 65}, 075407 (2002).
1020: \bibitem{Pauling}L. Pauling, The nature of the chemical bond and the
1021: structure of molecules and crystals, Cornell University Press (1960).
1022: \bibitem{Kittel} C. Kittel, Introduction to solid state physics,
1023: 7th edition, Wiley, New York, Chichester, Brisbane, Toronto, Singapore
1024: (1996).
1025: \bibitem{Crapper1986} M. D. Crapper, C. E. Riley, P. J. J.
1026: Sweeney, C. F. McConville, and D. P. Woodruff, Europhys. Lett. {\bf 2},
1027: 857 (1986).
1028: \bibitem{Shard1999} A. G. Shard, V. R. Dhanak and A. Santoni,
1029: Surf. Sci. {\bf 429}, 279 (1999).
1030: \bibitem{Shard2000} A. G. Shard, V. R. Dhanak and A. Santoni,
1031: Surf. Sci. {\bf 445}, 309 (2000).
1032: \bibitem{Doll2002} K. Doll, Phys. Rev. B {\bf 66}, 155421 (2002).
1033: \bibitem{Chandavarkar1988} S. Chandavarkar and R. D. Diehl, Phys. Rev. B
1034: {\bf 38}, 12112 (1988).
1035: \bibitem{Fisheretal} D. Fisher, S. Chandavarkar, I. R. Collins, R. D.
1036: Diehl, P. Kaukasoina and M. Lindroos, Phys. Rev. Lett. {\bf 68},
1037: 2786 (1992).
1038: \bibitem{Kaukasoinaetal1993} P. Kaukasoina, M. Lindroos, R. D. Diehl,
1039: D. Fisher, S. Chandavarkar, and I. R. Collins, J. Phys.: Cond. Matter
1040: {\bf 5}, 2875 (1993).
1041: \bibitem{Adler1993} D. L. Adler, I. R. Collins, X. Liang, S. J. Murray,
1042: G. S. Leatherman, K.-D. Tsuei, E. E. Chaban, S. Chandavarkar,
1043: R. McGrath, R. D. Diehl and P. H. Citrin, Phys. Rev. B {\bf 48}, 17445
1044: (1993).
1045: \bibitem{Huangetal1993} Z. Huang, L. Q. Wang, A. E. Schach von Wittenau,
1046: Z. Hussain, and D. A. Shirley, Phys. Rev. B {\bf 47}, 13626 (1993).
1047: \bibitem{Davis1994} R. Davis, X.-M. Hu, D. P. Woodruff, K.-U. Weiss,
1048: R. Dippel, K.-M. Schindler, Ph. Hofmann, V. Fritzsche and A. M. Bradshaw,
1049: Surf. Sci. {\bf 307}, 632 (1994).
1050: \bibitem{Shietal1993} X. Shi, D. Tang, D. Heskett, K.-D. Tsuei,
1051: H. Ishida, Y. Morikawa, and K. Terakura, Phys. Rev. B {\bf 47}, 4014
1052: (1993).
1053: \bibitem{Sheketal1990} M.-L. Shek, J. Hrbek, T. K. Sham, and G.-Q. Xu,
1054: Phys. Rev. B {\bf 41}, 3447 (1990).
1055: \bibitem{Riffeetal1990} D. M. Riffe, G. K. Wertheim, and P. H. Citrin,
1056: Phys. Rev. Lett. {\bf 64}, 571 (1990).
1057: \bibitem{KlausNicVic} K. Doll, N. M. Harrison, and V. R. Saunders,
1058: J. Phys.: Condensed Matter {\bf 11}, 5007 (1999).
1059: \bibitem{Baker} B. G. Baker, B. B. Johnson and G. L. C. Maire,
1060: Surf. Sci. {\bf 24}, 572 (1971).
1061: \bibitem{Michaelson} H. B. Michaelson, J. Appl. Phys. {\bf 48}, 4729
1062: (1977).
1063: \bibitem{Boettgeretal1995} J. C. Boettger, U. Birkenheuer, S. Kr\"uger,
1064: N. R\"osch, and S. B. Trickey, Phys. Rev. B {\bf 52}, 2025 (1995).
1065: \bibitem{Gschneider} K. A. Gschneidner, Jr., Solid State Phys. {\bf 16},
1066: 276 (1964).
1067: \bibitem{Alden} M. Ald\'en, H. L. Skriver, S. Mirbt and B. Johansson,
1068: Surf. Sci. {\bf 315}, 157 (1994).
1069: \bibitem{Vitosetal} L. Vitos, A. V. Ruban, H. L. Skriver and J. Koll{\'a}r,
1070: Surf. Science {\bf 411}, 186 (1998).
1071: \bibitem{FuFreeman} C. L. Fu and A. J. Freeman, Journal de Physique C49,
1072: Colloque C8, 1625 (1988).
1073: \bibitem{Wimmer1984} E. Wimmer, A. J. Freeman, and H. Krakauer,
1074: Phys. Rev. B {\bf 30}, 3113 (1984).
1075: \end{references}
1076: \onecolumn
1077:
1078: \newpage
1079: \begin{table}
1080: \begin{center}
1081: \caption{The ground state properties of bulk Ni. }
1082: \label{Nibulktable}
1083: \vspace{5mm}
1084: \begin{tabular}{ccccccc}
1085: & & & & \\
1086: & lattice constant $a_0 \ [{\rm \AA}]$ & $E_{coh} \ [E_h] $ & $B$ [GPa] & magnetic moment
1087: [$\mu_B$] \\
1088: PBE, this work & 3.53 & 0.218$^a$ 0.184$^b$ & 203 & 0.62 \\
1089: PW, this work & 3.53 & 0.220$^a$ 0.185$^b$ & 203 & 0.61\\
1090: Ref. \onlinecite{Moroni}, LDA & 3.43 & & 255 & 0.59 \\
1091: Ref. \onlinecite{Moroni}, GGA & 3.53 & & 195 & 0.61 \\
1092: exp. & 3.52 \cite{Gschneider} & 0.163$^a$, 0.164$^b$
1093: \cite{Gschneider} & 190 \cite{Gschneider} & 0.61 \cite{Kittel} \\
1094: \end{tabular}
1095: $^a$ energy with respect to a free nickel atom in its $d^8s^2$ state \hfill \\
1096: $^b$ energy with respect to a free nickel atom in its $d^9s^1$ state \hfill \\
1097: \end{center}
1098: \end{table}
1099:
1100:
1101:
1102: %\newpage
1103: \begin{table}
1104: \begin{center}
1105: \caption{\label{Nisurfsummtable}The surface energy $[\frac{E_h}
1106: {\rm surface \ atom}]$ of
1107: the low index nickel surfaces.}
1108: \begin{tabular}{ccccccc}
1109: surface & PBE, this work & Ref. \onlinecite{Mittendorfer1999} & Ref.
1110: \onlinecite{Alden} & Ref. \onlinecite{Vitosetal} \\
1111: (100) & 0.038 & 0.031 & 0.039 & 0.036 \\
1112: (110) & 0.056 & 0.046 & - & 0.049 \\
1113: (111) & \hspace{1.7cm} 0.028 (PW: 0.028) & 0.024 & 0.033 & 0.026 \\
1114: \end{tabular}
1115: \end{center}
1116: \end{table}
1117:
1118: %\newpage
1119: \begin{table}
1120: \begin{center}
1121: \caption{{Magnetic moments [$\mu_B$] of
1122: the low index nickel surfaces, computed using a slab with 9 Ni layers.}}
1123: \label{surfacemagneticmoment}
1124: \begin{tabular}{cccccccc}
1125: layer & PBE, this work & PW, this work &
1126: Ref. \onlinecite{Mittendorfer1999} (9 layers, \\
1127: & & & 3 surface layers on both & Ref. \onlinecite{Alden} & Ref. \onlinecite{FuFreeman} & Ref. \onlinecite{Wimmer1984} \\
1128: & & & sides are allowed to relax) & & (7 layers) & (7 layers)\\
1129: (100) surface \\
1130: S & 0.729 & & 0.76 & 0.69 & & 0.68\\
1131: S-1 & 0.634 & & 0.68 & 0.64 & & 0.60\\
1132: S-2 & 0.632 & & 0.66 & 0.66 & & 0.59\\
1133: S-3 & 0.611 & & & 0.64 & & 0.56\\
1134: S-4 & 0.619 & &\\
1135: (110) surface \\
1136: S & 0.757 & & 0.76 \\
1137: S-1 & 0.637 & & 0.66\\
1138: S-2 & 0.618 & & 0.64\\
1139: S-3 & 0.620 & \\
1140: S-4 & 0.607 & \\
1141: (111) surface \\
1142: S & 0.648 (relaxed: 0.649) & 0.638 & 0.68 & 0.62 & 0.63\\
1143: S-1 & 0.658 (relaxed: 0.657) & 0.649 & 0.65 & 0.67 & 0.64\\
1144: S-2 & 0.622 (relaxed: 0.622) & 0.613 & 0.62 & 0.65 & 0.58\\
1145: S-3 & 0.617 (relaxed: 0.617) & 0.604 & & 0.63 & 0.58\\
1146: S-4 & 0.613 (relaxed: 0.613) & 0.605 \\
1147: \end{tabular}
1148: \end{center}
1149: \end{table}
1150:
1151:
1152:
1153: %\newpage
1154: \begin{table}
1155: \begin{center}
1156: \caption{Adsorption of Cl on the Ni(111) surface.
1157: $d_{\rm {Cl-Ni \ top \ layer}}$ is the interlayer distance between the Cl
1158: layer and the top Ni layer, $d_{\rm {Cl-Ni \ nn}}$ is the bond length between
1159: Cl and nearest neighbor Ni. $d_{Ni1-Ni2}$ is the distance between
1160: first and second nickel layer. The distance between second and
1161: third nickel layer $d_{\rm Ni2-Ni3}$
1162: is held fixed at the bulk value (and for the thicker slabs also
1163: $d_{Ni3-Ni4}$ and $d_{Ni4-Ni5}$ are fixed at the bulk value).
1164: The adsorption energy is the difference
1165: $E_{\rm {Cl \ at \ Ni(111)}}-{E_{\rm Ni(111)}-E_{\rm Cl}}$.}
1166: \label{ClonNitable}
1167: \begin{tabular}{ccccc}
1168: Site & $d_{\rm {Cl-Ni \mbox{ }top\mbox{ } layer}}$ &
1169: $d_{Ni 1-Ni 2}$ &
1170: $d_{\rm {Cl-Ni \ nn}}$ & $E_{adsorption}$\\
1171: & [\AA] & [\AA] & [\AA] & $\left[\frac{E_h}{Cl \ atom}\right]$\\
1172: \multicolumn{5}{c}{3 nickel layers} \\
1173: fcc & 1.84 & 2.02 & 2.34 & -0.1358 \\
1174: hcp & 1.85 & 2.02 & 2.35 & -0.1351 \\
1175: bridge& 1.90 & 2.02 & 2.27 & -0.1325 \\
1176: top & 2.14 & 2.02 & 2.14 & -0.1177 \\
1177: \multicolumn{5}{c}{4 nickel layers} \\
1178: fcc & 1.83 & 2.02 & 2.33 & -0.1338 \\
1179: hcp & 1.84 & 2.02 & 2.34 & -0.1336 \\
1180: bridge & 1.89 & 2.02 & 2.27 & -0.1307 \\
1181: top & 2.14 & 2.02 & 2.14 & -0.1159 \\
1182: \multicolumn{5}{c}{5 nickel layers} \\
1183: fcc (single point calculation) & & & & -0.1337 \\
1184: hcp (single point calculation) & & & & -0.1331 \\
1185: \\
1186: exp. (ARPEFS) \cite{Wangetal} & 1.837 & 1.926 & 2.332 & \\
1187: exp. (SEXAFS, XSW)
1188: \cite{Funabashi1989,Funabashi1991} & 1.92, 1.83 & 2.03, 2.02
1189: & 2.40, 2.33 & \\
1190: \end{tabular}
1191: \end{center}
1192: \end{table}
1193:
1194:
1195:
1196:
1197: %\newpage
1198: \begin{table}
1199: \begin{center}
1200: \caption{Charge and position of the $3s$ eigenvalue
1201: for Cl and the $3s$ and $3p$ eigenvalues for K on different adsorption sites,
1202: relative to the Fermi energy. For chlorine, the peaks are at slightly
1203: different positions for majority and minority bands, for potassium they
1204: are at the same position.}
1205: \label{Clpopulationtable}
1206: \begin{tabular}{cccc}
1207: site & charge, in $|e|$ & $3s$ level, relative to $E_F$ $[E_h]$ & $3p$ level,
1208: relative to $E_F$ $[E_h]$ \\
1209: \multicolumn{3}{c}{Cl on Ni:}\\
1210: fcc & -0.050 & -0.589 (majority spin); -0.586 (minority spin) \\
1211: hcp & -0.046 & -0.586 ; -0.584 \\
1212: bridge & -0.068 & -0.580 ; -0.578\\
1213: top & -0.152 & -0.537 ; -0.535 \\
1214: \multicolumn{3}{c}{K on Ni:}\\
1215: fcc & +0.294 & -1.178 (both spins) & -0.583 (both spins) \\
1216: hcp & +0.295 & -1.178 & -0.582 \\
1217: bridge & +0.291 & -1.178 & -0.583 \\
1218: top & +0.269 & -1.179 & -0.583 \\
1219: \end{tabular}
1220: \end{center}
1221: \end{table}
1222:
1223:
1224:
1225: \begin{table}
1226: \begin{center}
1227: \caption{Mulliken spin and charge population for Cl/Ni(111),
1228: on different adsorption sites, extracted from calculations with 5 nickel
1229: layers. The population for the nickel atoms in
1230: the individual layers is given. The number in parenthesis gives the
1231: number of atoms which are identical because of symmetry and thus have an
1232: identical population. The left column corresponds to that atom(s)
1233: in the layer which is closer to the chlorine adsorbate. For example,
1234: in the fcc case, all three nickel atoms of a supercell
1235: in the first and second layer have the same distance
1236: and thus the population is identical,
1237: but one atom in the third layer is closer than the other two atoms
1238: resulting in two different values for the spin population. In the case
1239: of the clean slab, all atoms within one layer are degenerate and thus
1240: only one number is displayed.}
1241: \label{Clspintable}
1242: \begin{tabular}{ccccccc}
1243: site & 1$^{st}$ layer & 2$^{nd}$ layer & 3$^{rd}$ layer & 4$^{th}$ layer
1244: & 5$^{th}$ layer & Cl \\ \\
1245: \multicolumn{7}{c}{spin}\\
1246: fcc & 0.67 (3 atoms) & 0.64 (3 atoms) & 0.59 (1 atom) 0.64 (2 atoms)
1247: & 0.66 (3 atoms) & 0.65 (3 atoms) & 0.08 \\
1248: hcp & 0.68 (3 atoms) & 0.68 (1) 0.63 (2) & 0.62 (3) & 0.66 (3) & 0.63 (1) 0.66 (2) & 0.08 \\
1249: bridge & 0.66 (2) 0.73 (1) & 0.67 (1) 0.63 (2) & 0.60 (1) 0.63 (2) & 0.66 (2) 0.65 (1) & 0.64 (1) 0.66 (2) & 0.08\\
1250: top & 0.55 (1) 0.76 (2) & 0.64 (3) & 0.62 (3) & 0.67 (1) 0.64 (2) & 0.65 (3) & 0.04 \\
1251: clean slab & 0.66 & 0.66 & 0.63 & 0.66 & 0.66\\ \\
1252: \multicolumn{7}{c}{charge}\\
1253: fcc & 27.97 (3) & 28.02 (3) & 28.00 (1) 28.00 (2) & 28.05 (3)& 27.95 (3) &
1254: 17.05 \\
1255: hcp & 27.97 (3) & 28.02 (1) 28.00 (2) & 28.00 (3) & 28.05 (3) & 27.95 (1) 27.95 (2) &
1256: 17.05 \\
1257: bridge & 28.00 (2) 27.90 (1) & 28.02 (1) 28.01 (2) & 28.00 (1) 28.00 (2) & 28.05 (2) 28.05 (1) & 27.95 (1) 27.95 (2) & 17.07 \\
1258: top & 28.12 (1) 27.84 (2) & 28.02 (3) & 28.00 (3) & 28.05 (1) 28.05 (2) & 27.95 (3) &
1259: 17.15 \\
1260: clean slab & 27.95 & 28.05 & 28.00 & 28.05 & 27.95 \\
1261: \end{tabular}
1262: \end{center}
1263: \end{table}
1264:
1265: %\newpage
1266: \begin{table}
1267: \begin{center}
1268: \caption
1269: {Adsorption of K on the Ni(111) surface.
1270: $d_{\rm {K-Ni1a}}$ is the interlayer
1271: distance between the K layer and the layer made of those
1272: nickel atoms in the first layer closest
1273: to the K layer. $d_{\rm {Ni1b-Ni2}}$ is the distance between the layer made
1274: of those nickel atoms of the first layer which have moved away from
1275: the adsorbate and the second Ni layer.
1276: $\delta=d_{Ni1a-Ni1b}$ is the rumpling within the first layer.
1277: $d_{\rm {K-Ni \ nn}}$ is the bond length between
1278: K and nearest neighbor Ni. Again, the distance between second and
1279: third nickel layer $d_{\rm {Ni2-Ni3}}$ is held fixed at the bulk value.
1280: The adsorption energy is the difference
1281: $E_{\rm {K \mbox{ }at \mbox{ } Ni(111)}}-{E_{\rm Ni(111)}-E_{\rm K}}$.}
1282:
1283: \label{KonNiEnergie}
1284: \begin{tabular}{cccccccc}
1285: Site & $d_{K-Ni1a}$ & $d_{Ni1b-Ni2}$ & $\delta$ & $d_{K-Ni \ nn}$ &
1286: $E_{adsorption}$ \\
1287: & [\AA] & [\AA] & [\AA] & [\AA] & $\left[\frac{E_h}{K \ atom}\right]$ \\
1288: \\
1289: \multicolumn{6}{c} {without rumpling}\\
1290: fcc & 2.71 & 2.01 & 0 & 3.07 & -0.0532 \\
1291: hcp & 2.70 & 2.01 & 0 & 3.06 & -0.0531 \\
1292: bridge & 2.70 & 2.01 & 0 & 2.98 & -0.0531 \\
1293: top & 2.76 & 2.01 & 0 & 2.76 & -0.0528\\
1294: \\
1295: \multicolumn{6}{c} {with rumpling}\\
1296: fcc & 2.63 & 1.99 & +0.07 & 3.06 & -0.0536\\
1297: hcp & 2.63 & 1.99 & +0.07 & 3.06 & -0.0535 \\
1298: bridge & 2.65 & 1.97 & +0.07 & 2.99 & -0.0535 \\
1299: top & 2.68 & 1.93 & +0.11 & 2.79 & -0.0544 \\
1300: \\
1301: \multicolumn{6}{c} {experiment}\\
1302: exp. (LEED) \cite{Fisheretal,Kaukasoinaetal1993} & 2.70$\pm$0.04 &
1303: 1.90$\pm$0.03 & 0.12$\pm$0.02 & 2.82$\pm$0.04 & \\
1304: exp. (SEXAFS) \cite{Adler1993} & & & & 2.92$\pm$0.02 \\
1305: exp. (ARPEFS) \cite{Huangetal1993} & & 1.90$\pm$0.04 & 0.00$\pm$0.03
1306: & 3.02 $\pm$ 0.01 \\
1307: exp. (PhD) \cite{Davis1994} & & $1.86 \pm 0.06$
1308: & 0.01$\pm$0.09 & 2.87$\pm$0.06 \\
1309: \end{tabular}
1310: \end{center}
1311: \end{table}
1312:
1313:
1314:
1315: \begin{table}
1316: \begin{center}
1317: \caption{Mulliken spin and charge population for K/Ni(111),
1318: on different adsorption sites,
1319: extracted from calculations with 3 nickel
1320: layers. The population for the nickel atoms in
1321: the individual layers is given. The left column corresponds to that atom
1322: in the layer which is closest to the potassium adsorbate. For example,
1323: in the top case,
1324: one atom in the first layer is closer than the other three atoms
1325: resulting in two different values for the population. In the case
1326: of the clean slab, all atoms within one layer are degenerate and thus
1327: only one number is displayed. Also, when the populations were found to be
1328: identical, only one number was displayed (usually from the second Ni layer
1329: on, marked with 'all').}
1330: \label{Kspintable}
1331: \begin{tabular}{ccccccc}
1332: site & 1$^{st}$ layer & 2$^{nd}$ layer & 3$^{rd}$ layer & K \\ \\
1333: \multicolumn{5}{c}{spin}\\
1334: fcc & 0.58 (3 atoms) 0.62 (1 atom) & 0.68 (all) & 0.66 (all) & 0.001 \\
1335: hcp & 0.59 (3 atoms) 0.61 (1 atom) & 0.68 (all) & 0.66 (all) & 0.001 \\
1336: bridge & 0.58 (2 atoms) 0.61 (2 atoms) & 0.68 (all) & 0.66 (all) & 0.000 \\
1337: top & 0.55 (1 atom)0.61 (3 atoms)& 0.68 (3 atoms) 0.69 (1 atom) & 0.66 (all) & -0.003 \\
1338: clean slab & 0.66 & 0.69 & 0.66 \\ \\
1339: \multicolumn{5}{c}{charge}\\
1340: fcc & 28.04 (3 atoms)28.00 (1 atom)& 28.10 (all) & 27.95 (all) & 18.71 \\
1341: hcp & 28.04 (3 atoms)28.00 (1 atom)& 28.10 (all) & 27.95 (all) & 18.71 \\
1342: bridge & 28.05 (2 atoms) 28.00 (2 atoms) & 28.10 (all) & 27.95(all) & 18.71 \\
1343: top & 28.08 (1 atom) 28.00 (3 atoms) & 28.10 (all) & 27.95 (all) & 18.73 \\
1344: clean slab & 27.95 & 28.09 & 27.95 & \\
1345: \end{tabular}
1346: \end{center}
1347: \end{table}
1348:
1349:
1350: \begin{table}
1351: \begin{center}
1352: \caption{Binding energies for Cl/Ni(111),
1353: extracted from calculations with 3 nickel
1354: layers, as a function of the computational parameters. The geometry was
1355: held fixed at the one optimized with the default parameters (table
1356: \ref{ClonNitable}).}
1357: \label{ClNiparametertest}
1358: \begin{tabular}{ccccc}
1359: site & default grid, &
1360: better grid, &
1361: default grid, &
1362: default grid, \\
1363: & default ITOL ($10^{-6}$, $10^{-6}$) & default ITOL ($10^{-6}$, $10^{-6}$)
1364: & higher ITOL ($10^{-7}$, $10^{-7}$) &
1365: even higher ITOL ($10^{-8}$, $10^{-8}$)\\
1366: fcc & -0.1358 & -0.1357 & -0.1359 & -0.1357 \\
1367: hcp & -0.1351 & -0.1349 & -0.1338 & -0.1339 \\
1368: bridge & -0.1325 & -0.1325 & -0.1317 & -0.1317 \\
1369: top & -0.1177 & -0.1186 & -0.1180 & -0.1179 \\
1370: \end{tabular}
1371: \end{center}
1372: \end{table}
1373:
1374:
1375:
1376: \begin{table}
1377: \begin{center}
1378: \caption{Binding energies for K/Ni(111),
1379: extracted from calculations with 3 nickel
1380: layers, as a function of the computational parameters. The geometry was
1381: held fixed at the one optimized with the default parameters (table
1382: \ref{KonNiEnergie}).}
1383: \label{KNiparametertest}
1384: \begin{tabular}{ccccc}
1385: site & default grid, &
1386: better grid, &
1387: default grid, &
1388: default grid, \\
1389: & default ITOL ($10^{-6}$, $10^{-6}$) & default ITOL ($10^{-6}$, $10^{-6}$)
1390: & higher ITOL ($10^{-7}$, $10^{-7}$) &
1391: even higher ITOL ($10^{-8}$, $10^{-8}$)\\
1392: fcc & -0.0536 & -0.0529 & -0.0534 & -0.0533 \\
1393: hcp & -0.0535 & -0.0528 & -0.0532 & -0.0531 \\
1394: bridge & -0.0535 & -0.0528 & -0.0532 & -0.0532 \\
1395: top & -0.0544 & -0.0537 & -0.0539 & -0.0541 \\
1396: \end{tabular}
1397: \end{center}
1398: \end{table}
1399:
1400:
1401: \newpage
1402: \begin{figure}
1403: \caption{The threefold hollow structures considered for Cl,
1404: adsorbed on the Ni(111) surface, at a coverage of one third of a monolayer,
1405: $(\protect\sqrt 3 \times \protect\sqrt 3)$R30$^\circ$ unit cell.
1406: The nickel atoms in the top layer
1407: are displayed by open circles. The displayed chlorine adsorption sites are
1408: the threefold hollow sites (fcc or hcp hollow,
1409: filled circles). Note that these threefold hollow
1410: sites can not be distinguished in this figure. The ratio of the
1411: radii is chosen
1412: according to the computed values, i.e. using a nickel radius of 1.25 \AA
1413: \ and a chlorine radius of 1.09 \AA. The top site would be vertically
1414: above a nickel atom, the bridge site vertically above a point which is in
1415: the middle between two neighboring nickel atoms.}
1416: \label{geometryfigure}
1417: \centerline
1418: {\psfig
1419: {figure=figure1.eps,width=15cm,angle=270}}
1420: \end{figure}
1421:
1422: \newpage
1423: \begin{figure}
1424: \caption{Potassium,
1425: adsorbed on
1426: the Ni(111) surface, at a coverage of one fourth of a monolayer,
1427: $(2\times 2)$ unit cell. The nickel atoms in the top layer
1428: are displayed by open circles. The displayed potassium adsorption site is
1429: the top site vertically above the nickel atoms (filled circles).
1430: The ratio of the radii is chosen
1431: according to the computed values, i.e. using a nickel radius of 1.25 \AA
1432: \ and a potassium radius of 1.54 \AA.}
1433: \label{geometryfigure2}
1434: \centerline
1435: {\psfig
1436: {figure=figure2.eps,width=15cm,angle=270}}
1437: \end{figure}
1438:
1439:
1440:
1441:
1442: \newpage
1443: \begin{figure}
1444: \caption{Band structure of ferromagnetic nickel, with the PBE functional.
1445: The majority bands are drawn with full lines, the minority bands with dashed
1446: lines.}
1447: \label{nibulkbandstructure}
1448: \end{figure}
1449: \centerline{\psfig{figure=figure3.eps,width=15cm,angle=270}}
1450: \vfill
1451:
1452:
1453:
1454:
1455: \newpage
1456: \centerline{\psfig{figure=figure4.eps,width=18cm,angle=270}}
1457: \begin{figure}
1458: \caption{The density of states, projected on the chlorine basis functions,
1459: (upper two graphs); and projected
1460: on all basis functions (lower two graphs), for majority (first and
1461: third graph) and minority band (second and fourth graph).}
1462: \label{ClNiDOS}
1463: \end{figure}
1464: \vfill
1465:
1466: \newpage
1467: \centerline{\psfig{figure=figure5.ps,width=20cm,angle=90}}
1468:
1469: \vspace{-6cm}
1470: \begin{figure}
1471: \caption{Charge density (left), charge density difference (middle)
1472: and spin density (right) for Cl on Ni(111),
1473: from a three-layer slab, fcc site.
1474: The charge density is displayed with full lines, increasing in steps
1475: of 0.01, from 0 onwards. The charge density difference
1476: (charge of adsorbate system minus charge of a layer of chlorine atoms minus
1477: charge of a clean nickel surface)
1478: is in steps of 0.001, excess negative
1479: charge is displayed with full lines (from 0 onwards),
1480: electron depletion with dashed lines (from -0.001 onwards).
1481: Positive spin density is displayed
1482: with full lines, increasing from 0 in steps of 0.001. Negative spin density
1483: is displayed with dashed lines, decreasing from -0.0002
1484: in steps of -0.0002. The
1485: plane goes through the center of the Cl adsorbate, and through the
1486: nickel atom in the third layer vertically under the chlorine, and through
1487: the nearest neighbors of this nickel atom in the third layer.}
1488: \label{ClonNichargespindensity}
1489: \end{figure}
1490: \vfill
1491:
1492:
1493: \newpage
1494: \centerline{\psfig{figure=figure6.eps,width=15cm,angle=270}}
1495: \begin{figure}
1496: \caption{Geometrical parameters for the adsorption studies. }
1497: \label{geometricalparameters}
1498: \end{figure}
1499: \vfill
1500:
1501:
1502:
1503: \newpage
1504: \centerline{\psfig{figure=figure7.eps,width=18cm,angle=270}}
1505: \begin{figure}
1506: \caption{The density of states, projected on the potassium basis functions,
1507: (upper two graphs); and projected
1508: on all basis functions (lower two graphs), for majority (first and
1509: third graph) and minority band (second and fourth graph).}
1510: \label{KNiDOS}
1511: \end{figure}
1512: \vfill
1513:
1514:
1515: \newpage
1516: \centerline{\hspace{2cm}\psfig{figure=figure8.ps,width=25cm,angle=90}}
1517: \begin{figure}
1518: \caption{
1519: Charge density (left), charge density difference (middle)
1520: and spin density (right) for K on Ni(111),
1521: from a three-layer slab, top site.
1522: The charge density is displayed with full lines, increasing in steps
1523: of 0.01, from 0 onwards. The charge density difference
1524: (charge of adsorbate system minus charge of a layer of potassium atoms minus
1525: charge of a clean nickel surface)
1526: is in steps of 0.001, excess negative
1527: charge is displayed with full lines (from 0 onwards),
1528: electron depletion with dashed lines (from -0.001 onwards).
1529: Positive spin density is displayed
1530: with full lines, increasing from 0 in steps of 0.001. Negative spin density
1531: is displayed with dashed lines, decreasing from -0.0002
1532: in steps of -0.0002. The
1533: plane goes through the center of the K adsorbate, and through the
1534: nickel atom vertically below, and through
1535: the nearest neighbors of this nickel atom in the first layer.
1536: }
1537: \label{KonNispin}
1538: \end{figure}
1539: \vfill
1540:
1541:
1542: \end{document}
1543:
1544:
1545:
1546:
1547:
1548:
1549:
1550: