Fullerene-like CPx: A first-principles study of the relative stability of precursors and defect energetics during synthetic growth

Thin Solid Films 515 (2006) 1028 – 1032 www.elsevier.com/locate/tsf

Fullerene-like CPx: A first-principles study of the relative stability of precursors and defect energetics during synthetic growth A. Furlan ⁎, G.K. Gueorguiev, H. Högberg, S. Stafström, L. Hultman Department of Physics, Chemistry, and Biology (IFM), Linköping University, S-581 83 Linköping, Sweden Available online 11 September 2006

Abstract Inherently nanostructured CPx compounds were studied by first-principles calculations. Geometry optimizations and cohesive energy comparisons show stability for C3P, C2P, C3P2, CP, and P4 (P2) species in isolated form as well as incorporated in graphene layers. The energy cost for structural defects, arising from the substitution of C for P and intercalation of P atoms in graphene, was also evaluated. We find a larger curvature of the graphene sheets and a higher density of cross-linkage sites in comparison to fullerene-like (FL) CNx, which is explained by differences in the bonding between P and N. Thus, the computational results extend the scope of fullerene-like thin film materials with FL-CPx and provide insights for its structural properties. © 2006 Elsevier B.V. All rights reserved. Keywords: Fullerene-like materials; Phosphorus carbide (CPx); First-principles calculations; Thin films; Precursors

1. Introduction The discoveries of fullerenes and carbon nanotubes have opened up new research perspectives in condensed matter physics and materials science. The molecular structure of these carbon allotropes gives rise to a unique set of properties, suggesting a multitude of highly diversified application areas in, for instance, electronics and armor, etc. A further advantage of these allotropes, however, less explored, is that they can also be alloyed with other elements, suggesting that their properties can be further tailored to yield a particular attribute of interest. For instance, recent studies show that substitutional nitrogen at C sites strongly influences the graphene layers by promoting bending and cross-linking [1,2]. This makes it possible to incorporate fullerene-like features in a solid matrix of C and N. The resulting so called fullerene-like (FL) compounds have been found to exhibit high compliance, low plasticity, and mechanical resiliency. Such a structure enables the material to extend the strength of a planar sp2-coordinated C network in three dimensions [3]. Typical for C and N atoms is their high electronegativity, low degree of polarizability, and similar distribution of the valence ⁎ Corresponding author. Tel.: +46 13 28 12 79; fax: +46 13 28 89 18. E-mail address: [email protected] (A. Furlan). 0040-6090/$ - see front matter © 2006 Elsevier B.V. All rights reserved. doi:10.1016/j.tsf.2006.07.176

electrons, which enables similar hybridization between their s and p electrons, i.e., sp, sp2, and sp3 bonding configurations. Within this context, P is an alternative dopant element, promising modified bonding characteristics. C, P, and N exhibit similarities in valency, but different electronegativity (C atom electronegativity is higher than that of P atom, yet lower than for N). Since P shows a great variety of bonding configurations, also the possibilities of stable P fullerenes have been theoretically explored. Seifert et al. investigated different P fullerenes, compared them to C fullerenes, and found some pure P cage structures metastable [4]. All these features, together with the preference for tetrahedral coordination of P, may offer prospects for synthesizing FL-CPx, however, with substantial structural and bonding differences between CPx and CNx. It has been shown earlier that the stable CxNy (x, y ≤ 2), and especially cyanogen (C2N2), are commonly present in the magnetron sputtered CNx deposition flux [5]. The presence of similar sized CxPy precursor clusters in a hypothetical FL-CPx deposition flux remains to be investigated. Phosphorus carbide has been synthesized as amorphous thin film material over a wide range of P:C compositions (up to a ratio 3:1), using capacitively coupled radio frequency plasma deposition from PH3/CH4 gas mixtures [6]. Such films, characterized by a high P content, show amorphous structure,

A. Furlan et al. / Thin Solid Films 515 (2006) 1028–1032

in contrast of some expectations for segregated carbon and phosphorus phases. In addition, these films also exhibit a hydrogen content (up to 10%), originating from the hydrogen present in the gas mixture. Furthermore, they are prone to oxidation as revealed by secondary ion mass spectrometry (SIMS) and X-ray photoelectron spectroscopy (XPS) [7]. On the opposite end of the phosphorus content in the films, P-doping of diamond-like carbon (P-DLC) films [8] with P contents ≤ 11 at.% has been reported [9]. Pulsed laser ablation (PLA) offers an alternative route to producing H-free P-DLC thin films with controllable P/C ratios [10]. Overall, these P-DLC thin films show increased conductivity by nearly 5 orders of magnitude at room temperature and increased mechanical hardness. Unlike the hypothetical hexagonal β-C3N4 (isostructural with Si3N4) postulated to exhibit a large bulk moduli by virtue of short bond lengths and a high bond coordination [11], C3P4 was predicted to exhibit pseudocubic or defect zinc blende structure [12,13]. Although the laser ablation method promises to produce pure C–P compounds in a hydrogen free growth environment, potential problems are likely to arise due to gas-phase condensation when the super hot plume is expanded and cooled. This growth behavior was also encountered in PLA of elemental P where the formation of phosphorus clusters was frequently observed [14]. The occurrence of stable P clusters such as P4 will put restrictions in the synthesis of CPx since P4 will preferentially be incorporated between the graphene sheets rather than dissociate in contact with C. Being very stable, P4 cannot dissociate in contact with C but is instead incorporated in the graphitic structure by being trapped between graphene planes. Single P atom deposition can be achieved by means of chemical vapor deposition (CVD) and physical vapor deposition (PVD) [15]. However, in CVD, hydrogen is used as a reduction gas what inevitably leads to creation of strong P–H bonds, which can be difficult to dissociate in reaction with C. In order to achieve a viable deposition method for CPx films, an appropriate way to eliminate the considerable air sensitivity of the presently produced CPx films which rapidly oxidize/ hydrolyze and delaminate when in contact with air, needs to be found. One solution may be reactive magnetron sputtering of the material in high or ultrahigh vacuum chambers which has been successfully used for deposition of well structured FL-CNx followed by the application of a capping layer (diffusion barrier). However, no attempts to deposit CPx thin films by such technique have been reported to date, nor are theoretical investigations of the structure and the properties of potential FLCPx compounds available. In the present study, we report first-principles calculations for precursor species CnPm, (1 ≤ n, m ≤ 3) and Pn (n ≤ 4) that may be generated in the growth flux during vapor phase deposition of FL-CPx (x ≤ 0.3) thin films. The generic defects for FL-CPx are also evaluated by considering CPx finite model systems in which P is incorporated in an sp2-hybridized graphene sheet. The study involves both geometry optimizations and cohesive energy calculations performed within the framework of Density Functional Theory (DFT) in its Generalized Gradient Approximation (GGA). This theoretical

1029

approach was successfully applied to FL-CNx [16,17] leading to important findings regarding the role of the CN precursors for the pentagon formation and the evolution of curvature of the graphene sheets [16] and the cross-linkages between them [17]. We find that the preformed species C3P, C2P, C3P2, and CP, as well as pure P clusters such as P2 and P4 are the most important candidates for the incorporation in FL-CPx. In addition, the tetragon defects, together with pentagons, emerge as a likely feature in FL-CPx. These are shown to cause more strongly curved and shorter graphene layers in FL-CPx compared to FL-CNx. 2. Computational details The energy cost for substitutional P at C sites in graphene layers was investigated by geometry optimizations covering a wide diversity of defects resulting from the P incorporation. The defects were chosen by analogy with defects typical for FL-CNx and consideration of defects and bonding configurations known from the organic chemistry of P. P atoms intercalated in the graphene network were also considered since such may be trapped in vapor phase deposition at low temperature. A systematic classification of film-forming CnPm (1≤n, m ≤ 3) and Pn (n ≤ 4) species was built by geometry optimizations of different geometries of these small clusters. All comparisons of different possible structures are based on differences in cohesive energies |ΔEcoh| normalized by the total number of C and P atoms. The details of the optimization strategies are presented elsewhere [16,17]. The calculations were carried out using the Gaussian 03 program [18] within the framework of DFT-GGA. For all results reported here, 6-31G* basis set (which is augmented with polarization functions) was employed. All numerical data presented and discussed in this work were obtained by making use of the B3LYP hybrid functional [19] which is known to provide an accurate description of the structural and electronic properties of fullerene-like thin films [16,17] and similar covalent systems [20–22]. In order to ensure that the results reported here do not depend on the level of theory chosen, test calculations employing also different basis sets (e.g., 6-311G(d)) and exchange correlation functionals (e.g., Perdew-Wang 91) have been performed. No significant differences with respect to the B3LYP/6-31G⁎ results were found. 3. Results and discussion 3.1. Precursors for the formation of CPx Table 1 summarizes the cohesive energy data obtained for CnPm (1 ≤ n, m ≤ 3) and pure Pn species. The most stable pure Table 1 B3LYP cohesive energies per atom, corresponding to different volatile species of interest as incorporation units during synthetic growth of FL-CPx Species

CP

C2P

C3P

CP2

C2P2 C3P2 CP3

Ecoh/at (eV) 4.40 5.50 5.85 4.67 5.07

5.77

C2P3 P2

4.71 5.16

P4

4.09 4.73

1030

A. Furlan et al. / Thin Solid Films 515 (2006) 1028–1032

Fig. 1. Most stable CnPm and Pn precursor species: a) C2P, b) C3P, c) C3P2 and d) P4.

phosphorus precursors Pn (n ≤ 4), namely the dimer P2 and the tetramer P4, were studied. Both of these small evennumbered clusters are known to be locally (as compared to theirs neighbours in size) stable [23,24], and as such are important candidates for incorporation in FL-CPx. Particularly interesting as a precursor is the tetramer P4 belonging to the Td group of symmetry. Its shape and relative stability result from P's tendency for tetrahedral coordination. In fact, P–P bonds are energetically conceivable in CPx compounds, being only 0.2–0.3 eV energetically more costly than a C–P bond. However, this higher energy cost, along with the possibility for some local (with respect to the P incorporation site) rearrangements of the atomic positions at the incorporation site, the P concentration (≤ 30%) considered and the relatively low stability of the P2 dimer makes the role of the P–P bond secondary. This condition is in contrast to the case of FL-CNx thin film growth, where the N2 molecule was excluded from the simulations due to its high stability and the high energy barrier and energy cost for the incorporation of N–N bonds in CNx, what caused it to appear mostly as a desorption agent [16]. Fig. 1 shows the most stable CnPm and Pn precursor species. The particular stability of C3P, C2P, C3P2 clusters, and to less extent CP, is due to their higher cohesive energies per atom (Ecoh/at), and smaller number of dangling and P–P bonds with respect to the other CnPm (1 ≤ n, m ≤ 3) clusters. These may be formed in film-forming flux by C and P atoms interaction, and defined as precursors. Due to the stability of larger CnPm precursors, they can be seen as defect introducing agents (crosslinkages, P-segregation) during incorporation in FL-CPx. In contrast to the case of pure P clusters, to our best knowledge, there are no previous works on mixed CnPm. 3.2. Phospho-fullerenes In [17] the idea of a structural similarity between the fullerene adducts and heterofullerenes such as the aza-fullerenes, and the CNx compounds, was explored. The cohesive energy of the most stable isomers of the dodeca-aza-fullerene (C48N12, with one N atom per each pentagon) [20] was found to be comparable to the energy values for both CNx graphene network and CNx model systems incorporating pentagon defects. Qualitatively, this result can be viewed as a test for our first-principles approach, since the dodeca-aza-fullerene contains both pentagons and hexagons, but in contrast with model clusters considered it is a closed cage molecule without hydrogen terminations and, thus, not affected by boundary

effects. In order to make a theoretical study of the isomers of phospho-fullerenes, and identify the energetically most favorable combinations of defects, which can exist in nanostructured CPx, we calculated the cohesive energy of the dodeca-phosphofullerene C48P12, which is a direct structural analogue to the dodeca-aza-[60-S6]-fullerene [20]. Optimized in the similar computational framework as in [20], this particular isomer of C48P12 is used as an indicator of the relative stability and possible prevalence of mixed pentagon/hexagon CPx systems. The geometry of C48P12 exhibits characteristic pentagon distortions at the P sites (Fig. 2a). For comparison in the inset of Fig. 2a, the emblematic, not distorted and symmetric azafullerene is displayed. The pentagon distortions are due to the large C–P bond lengths (∼ 0.185 nm). The optimized C48P12 is considerably less stable than dodeca-aza-fullerene (by 0.92 eV per atom), although converged to a local minimum of the potential energy hypersurface (PES). In addition, C48P12 is unstable with respect to the reaction of dissociation into a C48 fullerene1 and three P4 clusters (Fig. 2b). A number of additional C48P12 isomers were also studied but most of them are not minima of PES at all. Due to its particular tendency to tetrahedral coordination, P introduces different types of defects in the graphene matrix than for N. The nature of these defects will be discussed in the next section. 3.3. Energy cost for substitutional P at C sites in graphene layers and related defects By optimizing the model clusters (Fig. 3) containing substitutional P atoms at C sites (both internal and peripheral for the cluster), we calculate the energy cost for substitutional P at C sites in purely hexagonal (graphene) structure of CPx (Fig. 3a), cluster with pentagon defects (a double pentagon defect, Fig. 3b), cluster containing one tetragon (Fig. 3c), and a formation corresponding to the Stone–Wales (SW) defect. The specific double pentagon defect shown in Fig. 3b was found to be energetically favorable in FL-CNx [17] and in the present calculations we confirmed also in CPx that it has slight energetic advantages compared to the single pentagon defect. Table 2 contains the cohesive energy results for the model clusters in Fig. 3, together with the data pertaining to the FL1 Within GGA-DFT we optimized some of the most symmetric C48 isomers; the one considered here is the most stable and belongs to the D6d group of symmetry.

A. Furlan et al. / Thin Solid Films 515 (2006) 1028–1032

1031

Fig. 2. a) The optimized dodeca-phospho-fullerene C48P12 and its emblematic prototype — the dodeca-aza-fullerene C48N12 (inset), and b) dissociation of C48P12 into C48 fullerene and P4 clusters.

CNx obtained in [17] for the structural analogues of those clusters. The incorporation of P atoms in a graphene network is energetically more expensive (by 0.2–0.3 eV, depending if the substitution site is internal or peripheric for the model cluster, and also on the presence of previously incorporated P atoms) than incorporation of N atom to form CNx. Overall, the same is valid also for the energy cost of introduction of single and double pentagons, and the SW defect. The higher cost of the substitutional P at C sites (Table 2) is related to the larger covalent radius and lower electronegativity of the P atoms compared to those of the N atom. This makes the P atom more difficult to place substitutionaly in a graphene sheet, however, it enhances the site reactivity. The double C– P bond works also as a curvature-enhancing factor. The pentagon defects (single pentagons, or combinations of them) are less favorable in CPx than in CNx — i.e., they are by 0.56 eV per atom higher in energy than the graphene structure (against only 0.26 eV per atom for CNx). The same to a lesser extent is valid for the SW defects, which is not surprising since they also contain pentagons. It is remarkable that the energy cost of the tetragons in CPx appears to be in average by 0.2 eV per atom lower than in CNx (Table 2, Ecoh/at for structures containing defects is compared to the Ecoh/at for graphene systems). In the case of CPx, the results in Table 2 show that the tetragons are approximately as stable as pentagon defects and/or SW defects. The predicted stability of the tetragons is an important indication for stronger curved and shorter graphene layers in FL-CPx than

in FL-CNx, which implies also higher density of interlayer cross-linkages [25]. Viewed in the context of the P chemistry, the appearance of tetragons is not surprising, since P is a third row element with low energy d-orbitals which allow it to expand its octet and to form four-membered ring transition states and intermediate structures. This electronic property of P is well known as defining its bonding nature, especially in the Pcontaining ring systems and cyclic compounds which represent an important part of the organic chemistry of phosphorus [26]. Finally, the deffects of P intercalated between the hexagonal or curved graphene layers – a type of defect which cannot be ruled out in ion-assisted vapor deposition – were investigated. We found such defects not likely to prevail. In fact, during optimization, the P atom migrates to one of the graphene layers changing locally the network by rearrangement and sometimes generates a new pentagon or tetragon defect. Such migration, incorporation in the graphene sheet and local rearrangement of the atomic positions lead to a total gain in Ecoh of 1.2–1.8 eV (depending on the number of P atoms close to the incorporation site). 4. Conclusions Using first-principles calculations in the DFT-GGA framework, together with an optimization strategy based on random changes in the atomic positions and subsequent optimizations,

Fig. 3. Optimized CPx model systems representing substitutional P at C sites in: a) hexagonal (graphene) network, b) structure containing pentagons, c) structures with a four-membered ring (the P atom at an internal site — upper panel, and at a peripheral position — lower panel), and d) a system with a Stone–Wales (SW) defect.

1032

A. Furlan et al. / Thin Solid Films 515 (2006) 1028–1032

Table 2 Comparison of B3LYP cohesive energies per atom (Ecoh/at), corresponding to the fully relaxed finite model systems displayed in Fig. 3 Structure

Ecoh/at (eV) ΔEcoh/at (eV)

C23P1H12 (C23N1H12) hexagonal

C13P1H8 (C13N1H8) two pentagons

6.30 (6.53) 0 (0)

5.74 (6.27) 0.56 (0.26)

C15P1H8 (C15N1H8) tetragonal defect P(N) at internal site

P(N) at external site

5.65 (5.71) 0.65 (0.82)

5.57 (5.59) 0.73 (0.94)

C27P1H14 (C27N1H14) SW defect 5.67 (6.12) 0.63 (0.41)

B3LYP cohesive energies for the FL-CNx analogous systems (i.e., in which P is substituted for N) are shown in parentheses. The bottom row represents the energy cost for the corresponding defects relative to the hexagonal structures C23P1H12 (and C23N1H12, respectively) (ΔEcoh/at). In all these results the lowest energy positions of the P or the N atom were used.

CnPm (1 ≤ n, m ≤ 3) species of interest as precursors for vapor phase synthesis of FL-CPx thin solid films, as well as the stability of pentagon and tetragon defects due to P incorporation in a corresponding graphene network have been studied. We find a larger diversity of potential precursors in the C–P system compared to the C–N system. The enhanced role of Pcontaining species implies more energetically demanding incorporation mechanism, higher concentration of defects, and higher probability for disruption of the regular graphene network than in FL-CNx. In sharp contrast to the FL-CNx structures, the tetragon defects are likely in FL-CPx, coexisting with combinations of pentagons and SW defects. Due to the specific bonding nature of P, the tetragon defects emerge as a characteristic structural feature of FL-CPx, defining stronger curvature of its graphene layers and energetically favoring higher density of crosslinkages between them. In addition, P atoms intercalated between the graphene sheets can be ruled out as energetically unfavorable and structurally unstable, with respect to substitutional incorporation of P in graphene layers. Since the dodeca-aza-fullerene is considered as a prototypical structure for FL-CNx, dodeca-phospho-fullerene analogues were investigated in the context of FL-CPx. However, even the most stable C48P12 – which is the structural analogue to the dodeca-aza-[60-S6]-fullerene – was found to be not particularly stable and cannot be considered as a direct bonding example for FL-CPx. The present results provide theoretical insights useful for the realization of a synthesis process for FL-CPx thin films. Acknowledgements The research was supported by the Swedish Foundation for Strategic Research (SSF) Strategic Research Center in Materials Science for Nanoscale Surface Engineering and the European Commission. The National Supercomputer Center in Linköping is gratefully acknowledged for providing high performance computing resources.

References [1] H. Sjöström, S. Stafström, M. Boman, J.-E. Sundgren, Phys. Rev. Lett. 75 (1995) 1336. [2] L. Hultman, J. Neidhardt, N. Hellgren, H. Sjöström, J.-E. Sundgren, MRS Bull. 28 (2003) 19. [3] J. Neidhardt, Z.S. Czigány, I.F. Brunell, L. Hultman, J. Appl. Phys. 93 (2003) 3002. [4] G. Seifert, T. Heine, P.W. Fowler, Eur. Phys. J., D 16 (2001) 341. [5] J. Neidhardt, B. Abendroth, R. Gago, W. Möller, L. Hultman, J. Appl. Phys. 94 (2003) 7059. [6] S.R.J. Pearce, P.W. May, R.K. Wild, K.R. Hallam, P.J. Heard, Diamond Relat. Mater. 11 (2002) 1041. [7] F. Claeyssens, G.M. Fuge, N.L. Allan, P.W. May, S.R.J. Pearce, M.N.R. Ashfold, Appl. Phys., A 79 (2004) 1237. [8] V.S. Veerasamy, G.A.J. Amaratunga, C.A. Davis, A.E. Timbs, W.I. Milne, D.R. McKenzie, J. Phys., Condens. Matter 5 (1993) L169. [9] A. Helmbold, P. Hammer, J.U. Thiele, K. Rohwer, D. Meissner, Philos. Mag., B 72 (1995) 335. [10] G.M. Fuge, P.W. May, K.N. Rosser, S.R.J. Pearce, M.N.R. Ashfold, Diamond Relat. Mater. 13 (2004) 1442. [11] E.G. Wang, Prog. Mater. Sci. 41 (1997) 241. [12] A.T.-L. Lim, J.-C. Zheng, Y.P. Feng, Int. J. Mod. Phys. B 16 (2002) 1101. [13] F. Claeyssens, J.M. Oliva, P.W. May, N.L. Allan, Int. J. Quant. Chem. 95 (2003) 546. [14] A.V. Bulgakov, O.F. Bobrenok, I. Ozerov, W. Marine, S. Giorgio, A. Lassesson, E.E.B. Campbell, Appl. Phys., A 79 (2004) 1369. [15] M. Ohring, Materials Science of Thin Films, Academic Press, 2002. [16] G.K. Gueorguiev, J. Neidhardt, S. Stafström, L. Hultman, Chem. Phys. Lett. 401 (2005) 288. [17] G.K. Gueorguiev, J. Neidhardt, S. Stafström, L. Hultman, Chem. Phys. Lett. 410 (2005) 228. [18] M.J. Frisch, Gaussian, Inc., Wallingford, CT, 2003. [19] A.D. Becke, J. Chem. Phys. 98 (1993) 5648. [20] S. Stafström, L. Hultman, N. Hellgren, Chem. Phys. Lett. 340 (2001) 227. [21] R.-H. Xie, G.W. Bryant, V.H. Smith Jr., Chem. Phys. Lett. 368 (2003) 486. [22] R.-H. Xie, G.W. Bryant, L. Jensen, J. Zhao, V.H. Smith Jr., J. Chem. Phys. 118 (2003) 8621. [23] G. Seifert, R.O. Jones, Z. Phys., D 26 (1993) 349. [24] L. Guo, H. Wu, Z. Jin, J. Mol. Struct., Theochem 677 (2004) 59. [25] G.K. Gueorguiev, A. Furlan, H. Högberg, S. Stafström, L. Hultman, Chem. Phys. Lett. 426 (2006) 374. [26] L.D. Quin, The Heterocyclic Chemistry of Phosphorus: Systems Based on the Phosphorus–Carbon Bond, Wiley-Interscience, New York, 1981.