• No results found

5 . 1 Motivation

The physical and chemical propert ies of isolated, small clusters of t he semiconduct ing group 14 elements

(

Sin , Gen , Snn

)

have been t he subject of intense research due to t he

fundamental interest and the possibility of diverse applications in nano-technologies [ 1 27] .

The electric dipole moment and dipole polarizability are especially interesting propert ies to probe, because t hey provide bot h a characterizat ion of t he geometrical and the elec­ t ronic structure of t hese small nano-sized part icles. Particularly tin clusters have become t he center of focus due to t heir occurrence in bot h covalent and metallic bulk phases [128]

and t heir abnormal higher melting temperat ures as compared to t he bulk value [1 29,130].

The higher rat io of surface to bulk atoms was believed to always cause size-dependent melt ing temperat ure depression of nanoclusters [1 3 1-133] 1. However, SnlO was the first reported example of a nano-particle melting at a higher t emperature t han its bulk phase

and contradicted t he above ment ioned standard paradigm based on t hermodynamic argu­ ments. Melting properties measured by nanocalorimet ry and ion mobi lity measurements and calculated by density funct ional met hods and molecular dynamics simulations are reported in [ 1 29, 1 30, 1 34-138] .

In conclusion, significantly higher melt ing temperatures t han t hat of bulk tin (505 K ) were found for most o f t he t in clusters and were attributed t o t he covalent nat ure of bonding and to geometrical factors2 .

1 For instance, Au clusters with a radius of 3 nm ( around 700 atoms) melt at about 800 1< , compared to a bulk melting point of 1 337 K [132] .

2The reported melting temperature of 8nlO, for instance, is around 2000 K [136] .

78

CHAPTER 5. TIN CL USTERS 79

The normal allotrope of tin under ambient conditions (jJ-Sn, white tin) is a metal with a body-centered tetragonal lattice, opposed to its lighter congeners, Si and Ge which are already semiconductors at room temperature. Below 286 K , there is a stable semiconducting phase (a-Sn, grey tin) with a small band gap of 0 . 1 eV and a cuhic, diamond structure similar to Si and Ge.

Due to t he high relevance of silicon clust ers in t he nano-electronic industry, a vast

range of t heoretical invest igations into the geometric stable structures of neutral and

charged, small and mid-sized silicon clusters have been published [139-153]. Jackson et

al. [154] and Maroulis et al. [ 155] calculated t heir response to an external stat ic electric

field. The former group found the per atom polarizability of Si2o-28 for compact struc­

t ures to decrease towards t he bulk limit , whereas pro late structures become increasingly p olarizable. Since jellium-based models for t he polarizability captured this trend nicely,

t he response to t he applied field is claimed to be metallic. The l atter group focused on

cluster sizes three to seven and found that , for t he different ial mean polarizability per

atom3, ab initio and density functional t heory based calculat ions yield distinctively dif­

ferent pictures. Earl ier, Schiifer et al. measured t he polarizabilit ies o f Sin clusters by a

laser vaporation beam deflect ion met hod at a nozzle t emperat ure of 300 K (9 :S

n

:S 1 20)

and found a very strong variation of t he polarizabilities per atom around t he bulk value of 3 . 7 1 A 3 [ 1 56 , 200] . These variat ions about t he bulk value were not found by J ackson et

at. [ 154] and have not been reported by t heoretical methods so far. A very revealing study

is t he spectroscopic evidence for the tricapped trigonal prism structure (TTP) , which is

a dominating structural motif in small, neutral Si, Ge and Sn clusters, for anionic silicon clusters [ 1 5 7] .

Germanium clusters have attracted similar theoretical and experimental interest com­ pared to silicon in the past two decades. They belong to t he most important m icro­ electronics materials and understanding t heir growth-habit and various electronic prop­ erties is of fundamental and substantial practical relevance. A vast range of t heoretical studies on the structures of neutral and charged germanium clusters up to 40 atoms have been reported [1 40, 1 58-1 64]. In general, the claimed ground-state structures were ob­ t ained using a genetic algorithm combined with t ight-binding methods or a basin-hopping algorithm coupled with plane-wave pseudopotential density functional calculations.

CHA PTER 5. CL USTERS 80

former group calculated t he polarizabilities of Gen

(

n = 2 -25) from B3LYP Lo -Alamos pseudopotential calculat ions using a valence double-zeta basis set

(LA L2DZ)

and con­ cluded that in general, the polarizabilities per atom ineTease wit h increasing cluster ize. but show various fluct uat ions around 5.4 A 3

In

for t he Clllst r siz 5 to 25. T h se fl uc­ t uations are claimed to cone pond t o clusters wit h large HO�IO-LUl\IO gap , resulting in smaller polarizabilitie . These are, however, ill cOllt rauict iouiJ wiL h t ho 'e of the latt er group. Chelikowsky et al. [165J invest igated t he polarizabilit ie of silicon, ger­ manium and germanium-ar enide clusters up to ten at oms and find t hese to deer'casc

wit h increasing size towards t he bulk limit . They find ident ical g omet rical for

Sb- 10 and Ge2

-

10

,

where the germanium cluster how an average increase of interatomic

d istances by about 4 o/c , and exhibit about 10 1C high r polarizabilit ies per atom t han Si clusters. Ion mobility measur ments of germanium cluster ion r vealed t hat clust rs \vith about 1 0 to 40 atoms follow a one-dimensional growth-pattern to give prolate geomct ries, those with

40

to 70 atoms retain roughly t he same aspect ratio and at about

70

atoms, the clusters reconstruct to a more spherical geometry [ 166]5 .

Early st udies on t in cluster date back to 1953. Honig ident ified ionic clusters of up

to five t in atoms upon vaporizat ion of t in

[167] .

The atomization energies of t in clu ter

with up to seven atoms and t he evaluation of t heir t ability under equilibrium condit ion

were invest igated by Gingerich et al. [16 J . Anderson calculated t he binding energies of

tin clu t ers wit h different geomet ric struct ures and compared them to the aforement ioned

t udy [ 169J . The energet ic separat ion and geometries of a variety of different neutral and positively charged tin cluster 6 (Sn2 > Sn3 , Sn.l) Snt , Sn5, Sn

t)

were reported by Balasubramanian et al. [1 70-1 75J.

The ground state vibrat ional frequency and equilibrium molecular const ants of Sn2 were measured by �Iiller et al. [ 1 76] . l\Ieasured photoionizat ion and -emission spectra are reported in [1 77-179]. No regions of unusual high stability were observ d in t he mass spectra of cationic ilicon 7 ) germanium and t in clusters by Schaber et al. [180J . St udies on the react ion of t in clu ter anions wit h oxygen revealed t hat no oxide products were observed for c lusters larger than Sn5 [181J . Calculated dissociation energies, bond lengt hs

4 Although otherwise tated by Zhao et al., who unfort unately do not compare their result with the

bulk limit polarizability.

5 An indication for metallic clusters.

61n most cases respective germanium and lead cluster are also reported. 7The produced ilicon clusters contained a significant amount of hydrogen.

CHAPTER

5. TIN

CL USTERS

8]

and harmonic frequencies of silicon, germanium and t in dimers are compared with ex­ perimental results in ref. [ 1 82] . J arrold et al. characterized t he struct ures of t in cluster

cations up to n = 68 by ion mobility measurements and summarized t hat up to n � 35,

tin cluster cations show a prolate growth-pattern [1 83]. They also probed structures of lead cluster cat ions up to n = 32 by t he same means and concluded that t hese clusters

adopt near-spherical geometries, indicating t he expected growth-pattern for metallic clus­ ters [ 1 84] . According to t hese findings, t he transition to » normal" metal cluster growth in t he group- 1 4 of singly positively charged clusters occurs between tin and lead . Lee

et al. calculated the ionization potent ials and binding energies of small germanium and t i n clusters with up to 13 atoms using a semi-empirical t ight binding met hod and found reasonable agreement with experimental results [185] . Much better agreement wit h ex­ perimental results for these properties were found by Maj umder et al. , who calculated the ground state geometries and energetics of neutral and singly positively charged t in clusters up to 20 atoms using the ultrasoft pseudopotential plane-wave method with gen­ eralized gradient approximations [186]. With respect to collision induced fragment ation processes of small tin cluster cations (Sn4-20) in t he energy range of 0-300 e V, Tai et al.

concluded t hat smaller clusters (n :s: 1 1 ) fragmented by t he atom loss process and t he l arger ones decayed by fission . Furt hermore, t hey found t hat t hese favored fragment ation paths resembled those for respective Si and Ge cluster ions, hence, confirming the struc­ t ur al similarities. They also backed their experimental results by t heoretical calculations utilizing various DFT methods [ 187, 1 88] .

By starting with h igh-symmetry structures and distorting them according to their

unstable modes , Pushpa et al. calculated low-energy structures of Mn (M = Sn, AI, As;

n = 4, 6,13) using DFT methods [189] . Alt hough t his is a conceptionally simple way

t o search for low-energy structures, and it also mimics t he way in which t he J ahn-Teller effect leads to particularly low-symmetry st ructures, it does not provide a proper route

to search for global minima. Maj umder et al. investigated t he structures, energetics and

fragmentation behavior of tin clusters up to 20 atoms by means of DFT methods and reported a different growth behavior to that of germanium and silicon clusters for tin clusters with more than 7 atoms [ 1 90] . A comparative t heoretical study with different exchange-correlation functionals of geometric structures and some electronic properties of

t i n clusters up to 20 atoms was published by Majumder et al. in 2005 [ 1 9 1] . A stable

CHAP TER 5. T IN CL USTERS

photoelectron spectl'oscoPY recently [1 92].

82

Considering the vast amount of theoret ical and experimental studies undertaken on small t in clusters, it become evident t hat neit her calculated , nor experimentally evalu­ ated static electric dipole polarizabilities have heen published. Furth rmore, all reported theoretical work on geometric structures was undertaken by using low-lying isomer struc­ t ures of silicon and germanium clusters as an illitial struct ure and then relaxing t hem into their local minima, rather than undertaking an unbiased search. Moreover, t he t heoretical approaches d here, invest igated only the lowest spin state of tin clusters. Th se three arguments are t he main motivat ion for t he present work.

5 . 2 Met hods

The predicted singlet and triplet global mll1lma ( G M ) struct ures o f t in clusters rang­ ing from seven up to twenty at oms were obt ained ut ilizing the genet ic a lgorithm code

B E L P H E G O R as described in det ail in section (3. 1 ) . The initial populations8 consisted

of randomly generated structures, predicted Lennard-Jones global minima and predicted low-lying minima st ructures from t in [ 190J and silicon clusters [147, 1 93] from t he litera­ t ure. The minimum energy difference 6i

V

was set to 0 . 0 1 eV, dmin and dmax parameters were fixed between 2.2 A and 3.8

A,

respectively. The termination criteria was 1 50 mat ing and local minimization steps for clusters up to 1 2 atoms and 100 steps for the remainder. The mutat ion probability was varied between 10 % and and 20 %.

In contr ast t o cesium and gold clusters, t he quest for the GM was performed in com­ binat ion with the ultrasoft pseudopotential plane-wave method, leaving 5s25p2 valence electrons, within t he local spin-density approximation (LDA) as implemented in t he VASP

program package [ 194]9 . The cut-off energy for t he plane-wave expansion was chosen to be 6 Ry. The clusters were placed in a cubic cell of side 16 A with periodic boundary conditions. D uring t he global opt imizat ion the cell was dynamically adapted, ensuring a dist ance of greater t han 8 A between t he clusters.

Typically, eight to ten of the t hus obtained energetically lowest-lying isomers were then fmt her relaxed from LANL20z basis set and pseudopot ntial calculat ions t o t heir local

minima as implemented in the G AUSSIAN03 program package [33 1 ] . Depending on the

8Typically ranging from 10 to 15 different structures.

9VASP makes a better compromise between a realistic treatment of the electronic structure and the

CHA P TER 5. TIN CL USTERS 83

energy distribution, two to four of the energet ically lowest- lying true minima structures

obtained by t he e means were t hen furt her optimi7.ed using t he extensive STUTTGART valence basis set together wit h t he energy-consistent relativistic pseudopotent ial for tin [195] . Finally, harmonic vibrat ional frequencies were computed to ensure t hat t he relaxed geometries are true minima on the potent ial energy surface. For t he exchange-correlation potential, the hybrid funct ional b3p86, according to the parameterizat ion suggested by Becke [196] and Perdew [ 1 12] , was applied in a self-con istent fashion as implemented in the GAUSS IAN03 program package. No symmetry constraints were applied during t he opt imizat ion proc dure.

The same approach was adopted to find t he best DFT functional reproducing t he mean static dipole polarizability as mentioned in the cases of gold and cesium clusters. At first , the polarizability of the ground st ate

e Po)

tin atom was calculated in the framework of CCSD(T) from the uncont racted and extensive STUTTGART valence basis set ( labelled Basis A [ 195] ) with a respective energy-consistent relat ivist ic pseudopotential. Then, this basis set was contracted and reduced (labelled Basis B [ 1 95] ) , and several DFT functionals, comprising different exchange and correlation funct ions such as b3p86, b3lyp, b3pw9 1 , blyp, bp86, pw9 1 pw9 1 , svwn, svwn5, pbepbe, mpw1 pw91 , b1 b95 and pbe1 pbe were tested to probe the best one reproducing the afor mentioned CCSD (T) value (8.04 A3) . As depicted in figure (5. 1 ) , t he b3p86 functional yields the smallest deviat ion and hence was applied for all optimizat ions and calculations of electronic properties of t he t in clusters presented in this work.