The formation, reactions and structures of binary cobalt phosphide clusters [Co_xP_y]⁻ in the gas phase

Abstract

Laser ablation of CoP generated 152 binary anionic cobalt phosphide clusters [Co_xP_y]⁻, ranging up to [Co₂₅P₁₆]⁻. This is a more extensive set of [Co_xP_y]⁻ clusters than obtained in previous laser ablation of a mixture of Co metal and red phosphorus: the composition maps for the two experiments overlap, but with generally lesser y:x ratios for laser ablation of CoP. The reactivities, reactions, and collisional dissociation of selected ions have been investigated by Fourier transform ion cyclotron resonance mass spectrometry. The majority of the [Co_xP_y]⁻ ions react with H₂S by addition of (up to three) S atoms, substitution of P₂ by S, and with elimination of HP₂⁻. Reaction with NO₂ or N₂O causes addition of one O atom. Collisional activation causes dissociation of P₂, P₄, CoP₂ or CoP₄. The ions [CoP₈]⁻, [Co₄P₈]⁻, [Co₅P₉]⁻ and [Co₆P₁₀]⁻ are unreactive, and for these and the reactive [Co₄P₄]⁻ ion, density functional investigations of 30 postulated structures have been performed. The probable structures, calculated electron affinities, and electronic structures are reported. The structural principles evident so far are that P atoms, P₂ groups, and P₃ groups bridge the faces, and to a lesser extent the edges, of Co_x polyhedra. The lack of reactivity for relatively P-rich clusters correlates with the absence of co-ordinatively unsaturated metal sites on the clusters: it is postulated that the more reactive species undergo addition at under-co-ordinated Co atoms. The calculated electronic structures generally have close-lying electronic states with unpaired electrons, which explains the overall high reactivity of the [Co_xP_y]⁻ clusters. The structural differences between these [Co_xP_y]⁻ clusters and characterised molecules Co_xP_yL_z with ancillary ligands are discussed, as are possible applications of the new [Co_xP_y]⁻ clusters in synthesis of novel materials.

---

Introduction

This paper describes the formation in the gas phase of 152 binary anionic cobalt phosphide clusters [Co_xP_y]⁻, ranging up to [Co₂₅P₁₆]⁻, and investigations of the reactivities and structures of some of them in the gas phase. The clusters have been generated in the cell of a Fourier transform ion cyclotron resonance (FTICR) mass spectrometer, after laser ablation of solid cobalt phosphide, CoP. The atomic fragments in the resulting plasma associate during cooling to form clusters, in a process that can be regarded as the most fundamental synthesis of binary inorganic molecules.¹ We compare these products with those formed by association after laser ablation of mixtures of elemental cobalt and red phosphorus.² Key questions here concern the factors that influence the observed distributions of molecular binary clusters: preferred compositions for these [Co_xP_y] clusters could reflect fundamental thermodynamic stability or, since they are observed as anions, they could be revealing the species with the more competitive electron affinities.

In addition to gas phase metal–phosphide chemistry there is a related well-developed area of condensed-phase metal complexes with phosphide, polyphosphide, and polyphosphane ligands.^3–9 The preparative methods for these are very different from association in a cooling plasma, and these compounds are normally coated with ancillary ligands, and are often organometallic. The occurrence of small P_n fragments, acyclic, monocyclic and polycyclic, is a feature of these compounds.^3–5,8–26 Comparisons of the compositions of the cores of the condensed-phase compounds with the compositions of the gas phase clusters are intriguing and informative. For most condensed phase compounds there is definitive experimental knowledge of geometrical structure, but for the gas phase species theoretical calculations are required to approach questions of geometrical structure and to interpret the compositions revealed in the mass spectrum.

There is a small number of condensed-phase cobalt phosphorus clusters including [CoP₃(triphos)] and derivatives,^27,28 [(η⁵-C₅Me₅Co)₃(P₂)₃],²⁹ [Co(η⁵-Cp″)(P₄){(η⁵-Cp″Co)₂(μ-CO)}] (Cp″  = C₅H₃Bu₂^t),²⁰ [Co(η⁵-C₅Me₅)(CO)P₄] and [Co₂(η⁵-C₅Me₅)(CO)P₄],³⁰ but, as we report here, there is a much larger number of gas phase species. The demonstrated existence of stable gas phase species may inspire additional synthetic attempts to prepare and characterise analogues with ancillary terminating ligands. The accessibility of new compounds suggested by gas phase experiments depends on their stability and reactivity. We report some results on the reactivity and reactions of the [Co_xP_y]⁻ clusters in the gas phase. This allows identification of the more stable core compositions. Then there is the question of the geometrical and electronic structures of these stable core compositions, which we investigate using density functional calculations. The goal is to recognise structural features which confer stability and/or favourable electron affinity, and to understand bonding principles for this class of fundamental molecular cobalt phosphide clusters. Previous related theoretical work^29,31 includes density functional calculations (including molecular dynamics) of polyphosphanes and their anions.^32–35

In previous gas phase experiments involving phosphorus we have treated P₄(g) with metal ions, and with other binary clusters such as the copper sulfide cluster anions³⁶ and carbon anions.^36,37

Experimental

Samples of cobalt phosphide CoP were supplied by Dr Glaum and prepared at high temperature by treating cobalt metal and red phosphorus in the presence of iodine.³⁸ The samples were powdered and pressed into stainless steel probe tips.

Ion generation

Laser ablation–FTICR experiments were carried out using a Spectrospin CMS-47 Fourier Transform Ion Cyclotron Resonance (FTICR) mass spectrometer with a 4.7 T superconducting magnet, using techniques previously described.^1,37,39 The sample to be ablated was pressed into a satellite probe tip which was placed at one end of the cylindrical FTICR cell (r  = 30 mm, l  = 60 mm), in contact and flush with one of the trapping plates. The ions were generated by a 8 ns pulse from a Nd-YAG laser at 1064 nm, focussed to a spot size of 0.4 mm diameter. Laser power densities of 800 MW cm⁻² gave the best spectra. The mass spectra have a precision of m/z <0.1, and since both Co and P are monoisotopic, there is no ambiguity in the assignment of compositions to high mass peaks.

Collision-induced dissociation (CID) experiments

After formation of the cobalt phosphide anions, the [Co_xP_y]⁻ ion to be investigated was selected and all other ions in the cell removed using a radio frequency pulse. The isolated anion was then excited, in the presence of argon at a static pressure of 1 × 10⁻⁵ Pa. using a radio frequency pulse with a frequency identical to the cyclotron frequency of the selected ion. A short delay (≈ 0.03–0.05 s) was used to allow one or two collisions of the anion with argon and then a spectrum of the products was observed. Similar experiments in the presence of carbon tetrachloride were undertaken in attempt to detect electron loss from [Co_xP_y]⁻ by appearance of Cl⁻.

Reactions

The anions were first prepared using an ionisation laser pulse and after a cooling delay the ions to be studied were selected using a radio frequency pulse which ejected all unwanted anions from the cell. The laser ablation process and the ion cooling occurred in the presence of argon and the reactant gas. Owing to the small mass differences between ions, e.g. [Co₃P₄]⁻ m/z 301, [Co₂P₆]⁻ 304, [CoP₈]⁻ 307, batches of ions with similar mass were selected for reaction. These ions were then allowed to react for various periods with the reactant gas (H₂S, NO₂ or N₂O) at an uncorrected pressure of (1–5) × 10⁻⁵ Pa. After the reaction period all the ions in the cell were excited and detected.

Density functional methodology

Density functional calculations of the geometrical and electronic structures of selected ions used the BLYP functional and double numerical basis sets, as implemented in the program DMol.^40,41 This DF methodology is accurate and efficient. The BLYP functional and the DMol methodology were used because they have consistently yielded good agreement with the observed structures of metal chalcogenide and related clusters.^42,43 All calculations were spin unrestricted. The generation of structures to be investigated and optimised is based on methods and principles that have been developed for various classes of binary metal clusters M_xE_y.^43,44 First, we use the cores and core types of crystallographically characterised clusters,^7,45–48 with E substituted by P, and stripped of peripheral ligands. This includes consideration of the Platonic and Archimedian polyhedra and their intersections. Second, substitution of the P₂ group for E is explored. Third, the merging of established small P_n moieties with established metal frameworks in clusters is considered. The calculational strategy is normally to optimise first with the highest possible symmetry, then test whether distortion allows improvement. Since the structures traversed during these calculations often have closely spaced orbitals at the Fermi level, and low-lying and variable electronic states, the calculations during geometry optimisation include an averaging over close-lying states by allowing partial occupancy of orbitals close to the Fermi level. In this way we attempt not to restrict the electronic state in exploration of the geometry–energy–electronic hypersurface. In the final stages of calculation for each structure this partial occupancy of higher orbitals is removed, to yield the electronic state for the local energy minimum. Geometry optimisation was effected by minimisation of the total energy. Second derivative calculations to confirm that stationary points are true wells have not been made in most cases, but experience shows that the behaviour of the first derivatives near the stationary point indicates the occurrence of a minimum or a saddle point. Electron affinities are reported as the energy of the optimised anion minus the energy of the optimised neutral. Details of the calculations, and the cartesian coordinates for the resulting structures, are provided in the electronic supplementary information.

Results

Ions generated by laser ablation

Laser ablation of CoP with a laser power of 800 MW cm⁻¹ produced over 150 cobalt phosphide cluster anions. The anions in the mass range 350–2000 are shown in Fig. 1, and the compositions of all of the anions observed are listed in Table 1. The [Co_xP_y]⁻ ions contain between 1 and 25 Co atoms, and between 3 and 17 P atoms. The smallest is [CoP₄]⁻, and the largest ions are [Co₂₄P₁₇]⁻ and [Co₂₅P₁₆]⁻. The most abundant is [Co₁₅P₁₁]⁻, but as is evident in Fig. 1, there are many ions extending over most of the composition range with abundances not much less than the maximum. Table 1 identifies the more abundant species. There are no ions or patterns of ions which have markedly enhanced relative abundance, and therefore no indications of enhanced stability. It can be seen from Table 1 that there is a region of the ion map around [Co₈P₈]⁻ where there are fewer abundant species: the ions [Co₈P₉]⁻, [Co₉P₆]⁻ and [Co₉P₇]⁻ are included in this group with low abundance. However, the ions [Co₈P₆]⁻ and [Co₈P₇]⁻ are amongst the most abundant in the spectrum, and we see no significance in these relative abundances in this region of the spectrum. The lower mass [Co_xP_y]⁻ ions are phosphorus-rich (y>x), while the more abundant ions with x⩾7 are phosphorus-deficient (y<x), with the high mass anions having Co:P of ≈ 5:3.

Fig. 1 The mass spectrum of ions [Co_xP_y]⁻ formed by laser ablation of CoP. The single broad-band spectrum in the range m/z 300–2000 is drawn in two sections with the same intensity scale. The identities of some of the peaks are marked, as x,y. The close-lying peaks are not isotopomers (Co and P are monoisotopic) but are separate [Co_xP_y]⁻ ions, as illustrated in the expansion. The laser power was 800 MW cm⁻¹.

Table 1 Compositions of the ions [Co_xP_y]⁻ observed by laser ablation of CoP. Bold italics signify major intensity

No. Co, x	No. P atoms, y
1		4	5	6		8
2		4	5	6	7	8
3		4	5	6	7	8	9
4	3	4	5	6	7	8
5		4	5	6	7	8	9
6		4	5	6	7	8	9	10	11
7			5	6	7	8	9	10	11
8			5	6	7	8	9	10	11
9				6	7	8	9	10	11	12
10				6	7	8	9	10	11	12
11			5	6	7	8	9	10	11	12	13
12			5	6	7	8	9	10	11	12	13	14
13					7	8	9	10	11	12	13	14	15
14					7	8	9	10	11	12	13	14	15
15					7	8	9	10	11	12	13	14	15	16
16					7	8	9	10	11	12	13	14	15	16
17						8	9	10	11	12	13	14	15	16
18								10	11	12	13	14	15	16
19									11	12	13		15	16	17
20										12	13	14	15
21										12	13	14
22										12	13	14
23											13	14
24												14	15	16	17
25												14	15	16

---

The positive ion spectrum of CoP, using similar laser powers to those which produce abundant anions, consists only of Co⁺, CoP₂⁺ and CoP₄⁺ with minor ions CoP⁺ and P₃⁺.

Collisional dissociation

Collision of the lower mass ions with argon gave the dissociation products listed in eqn. (1) to (6).

No ionic products were detected for similar CID experiments with the anions [Co₄P₅]⁻, [Co₅P₆]⁻, [Co₅P₇]⁻, [Co₈P₅]⁻ and [Co₈P₆]⁻, but the ion intensities decreased, indicative of dissociation processes. In order to investigate the possibility that electron detachment was occurring with these ions and was responsible for the reduced intensity, collisional studies of them with CCl₄ were made. These experiments did not produce Cl⁻, indicating that electron detachment did not occur, at least with CCl₄ which has higher mass than argon.

Reactions with H₂S

Selected [Co_xP_y]⁻ ions were treated with H₂S (g). Rapid reaction occurred in many cases, and a representative experiment is illustrated in Fig. 2. The complete results are listed in Table 2. In many cases the reaction is addition of one or more S atoms (see Fig. 2), and there are also reactions in which two P atoms are substituted by one S atom, as shown for [Co₃P₆]⁻ in Fig. 3.

Fig. 2 Reactions of [Co_xP_y]⁻ with H₂S (g). (Above) ions in the m/z range 460–700, as isolated: labels are x,y. (Below) ions observed after 10 s of reaction with H₂S (g) at an uncorrected pressure of 1 × 10⁻⁵ Pa: the products are [Co_xP_yS_z]⁻, labelled x,y,z.

Table 2 Products of reactions of [Co_xP_y]⁻ with H₂S, NO₂ and N₂O

Reactant	Products with H2S	Products with NO2	Products with N2O
a NR = no reaction products.b Dissociation occurred.c Experiment not done.
[CoP8]^−	NR^ab	*^c	*^c
[Co2P6]^−	[Co2P6S]^−	*	*
[Co2P7]^−	NR	*	*
[Co2P8]^−	NR	*	*
[Co3P4]^−	[Co3P4S]^−, [Co3P4S2]^−	*	*
[Co3P5]^−	[Co3P5S]^−, [Co3P5S2]^−	*	*
[Co3P6]^−	[Co3P6S]^−, [Co3P6S2]^−, [Co3P4S2]^−, [Co3P2S2]^−	*	*
[Co3P7]^−	[Co3P7S]^−, [Co3P5S]^−, [Co3P3S2]^−	[Co3P7O]^−	*
[Co3P8]^−	NR	[Co3P8O]^−	*
[Co4P3]^−	NR	*	*
[Co4P4]^−	[Co4P4S]^−, [Co4P4S2]^−, [Co4P4S3]^−	[Co4P4O]^−	*
[Co4P5]^−	[Co4P5S2]^−	[Co4P5O]^−	[Co4P5O]^−
[Co4P6]^−	[Co4P6S2]^−	[Co4P6O]^−	[Co4P6O]^−
[Co4P7]^−	[Co4P7S2]^−	[Co4P7O]^−	NR
[Co4P8]^−	NR	NR	NR
[Co5P4]^−	*	[Co5P4O]^−	*
[Co5P5]^−	[Co5P5S2]^−	[Co5P5O]^−	*
[Co5P6]^−	[Co5P6S]^−, [Co5P6S2]^−, [Co5P6S3]^−	[Co5P6O]^−	[Co5P6O]^−
[Co5P7]^−	[Co5P7S]^−	[Co5P7O]^−	[Co5P7O]^−
[Co5P8]^−	[Co5P8S]^−	*	NR
[Co5P9]^−	NR	NR	NR
[Co6P5]^−	[Co6P5S3]^−	[Co6P5O]^−	[Co6P5O]^−
[Co6P6]^−	[Co6P6S2]^−	[Co6P6O]^−	[Co6P6O]^−
[Co6P7]^−	[Co6P7S]^−, [Co6P7S2]^−	[Co6P7O]^−	[Co6P7O]^−
[Co6P8]^−	[Co6P8S]^−	[Co6P8O]^−	NR
[Co6P9]^−	[Co6P9S]^−	*	NR
[Co6P10]^−	NR	NR	NR
[Co7P5]^−	[Co7P5S2]^−, [Co7P5S3]^−	[Co7P5O]^−	[Co7P5O]^−
[Co7P6]^−	[Co7P6S]^−, [Co7P6S2]^−	[Co7P6O]^−	[Co7P6O]^−
[Co7P7]^−	[Co7P7S]^−, [Co7P7S2]^−, [Co7P7S3]^−	[Co7P7O]^−	[Co7P7O]^−
[Co7P8]^−	[Co7P8S]^−	*	NR
[Co8P6]^−	NR	[Co8P6O]^−	[Co8P6O]^−
[Co8P7]^−	*	[Co8P7O]^−	NR

---

Fig. 3 Substitution and addition reactions of H₂S (g) with [Co₃P₆]⁻. (Above) the ions [Co₄P₄]⁻ and [Co₃P₆]⁻ as isolated. (Below) product ions after reaction for 3 s with H₂S at an uncorrected pressure of 1 × 10⁻⁵ Pa: the products are [Co_xP_yS_z]⁻, labelled x,y,z.

In the following we demonstrate the types and patterns of reactions with H₂S. The ions [CoP₈]⁻, [Co₂P₇]⁻, [Co₂P₈]⁻, [Co₃P₈]⁻, [Co₄P₃]⁻, [Co₄P₈]⁻, [Co₅P₉]⁻, [Co₆P₁₀]⁻, and [Co₈P₆]⁻ were unreactive (although dissociation occurred in one case: see below), eqn. (7).

The following ions reacted by addition of one sulfur atom [Co₂P₆]⁻, [Co₃P₄]⁻, [Co₃P₅]⁻, [Co₃P₆]⁻, [Co₃P₇]⁻, [Co₄P₄]⁻, [Co₅P₆]⁻, [Co₅P₇]⁻, [Co₅P₈]⁻, [Co₆P₇]⁻, [Co₆P₈]⁻, [Co₆P₉]⁻, [Co₇P₆]⁻, [Co₇P₇]⁻, [Co₇P₈]⁻, eqn. (8). Several of these ions continued to add more sulfur atoms.

The following ions reacted by addition of two sulfur atoms: [Co₃P₄]⁻, [Co₃P₅]⁻, [Co₃P₆]⁻, [Co₄P₄]⁻, [Co₄P₅]⁻, [Co₄P₆]⁻, [Co₄P₇]⁻, [Co₅P₆]⁻, [Co₆P₆]⁻, [Co₆P₇]⁻, [Co₇P₅]⁻, [Co₇P₆]⁻, [Co₇P₇]⁻, eqn. (9).

While many [Co_xP_y]⁻ ions appear to add sequentially one then a second sulfur atom, for several ([Co₄P₅]⁻, [Co₄P₆]⁻, [Co₄P₇]⁻, [Co₅P₅]⁻, [Co₆P₅]⁻, [Co₆P₆]⁻, [Co₇P₅]⁻) there is addition of two or three S atoms but no evidence for addition of one. The ions which reacted by addition of three S atoms, are [Co₄P₄]⁻, [Co₅P₆]⁻, [Co₆P₅]⁻, [Co₇P₇]⁻, eqn. (10).

The ions [Co₃P₆]⁻ and [Co₃P₇]⁻, while able to undergo addition reactions with H₂S, also reacted by the replacement of a P₂ unit for each S atom added, eqn. (11) (z  = 1 or 2).

Two ions, [Co₄P₄]⁻ and [Co₄P₅]⁻, were selected for more detailed investigation of the reactivity with H₂S. The reaction of [Co₄P₄]⁻ was monitored for up to 15 s. Initially [Co₄P₄S]⁻ was formed and this ion reached a maximum intensity at 5 s and then rapidly decreased in intensity. [Co₄P₄S₂]⁻ was also formed and reached a maximum intensity at about 7 s, and was not present after 15 s of reaction. [Co₄P₄S₃]⁻ was observed at 2.5 s and was increasing in intensity at 15 s. These data demonstrate sequential addition of sulfur atoms to [Co₄P₄]⁻. However, when the reaction of [Co₄P₅]⁻ was studied for up to 20 s there were only two product ions [Co₄P₅S₂]⁻ and HP₂⁻, and no [Co₄P₅S]⁻ was observed.

HP₂⁻ was a product ion in most of the reactions, and presumably formed by the general reaction (12), although there is no definite information on associated neutral cobalt phosphide products.

Occasionally ions formed by dissociation were observed: in particular [CoP₄]⁻ was formed by exposure of [CoP₈]⁻ to H₂S.

Reactions with NO₂ or N₂O

Fig. 4 shows a representative experiment in which a range of [Co_xP_y]⁻ ions was exposed to NO₂. The results of all experiments with NO₂ and N₂O are contained in Table 2. There was no difference in the compositions of the products of reaction with NO₂ and N₂O, but some [Co_xP_y]⁻ ions which reacted with NO₂ did not react with N₂O. Some [Co_xP_y]⁻ ions did not react, and the remainder added one O atom [eqn. (13), Fig. 4]. There was no evidence of addition of more than one O atom.

Fig. 4 The reaction of selected [Co_xP_y]⁻ ions with NO₂ (g). Upper spectrum: the ions selected for reaction. Lower spectrum: the product spectrum after 2 s of reaction with NO₂ at an uncorrected pressure of 1 × 10⁻⁵ Pa: the products are [Co_xP_yO_z]⁻, labelled x,y,z.

The smaller [Co_xP_y]⁻ ions with x  = 1, 2, 3 could not be isolated due to their low abundance after the ion cooling process. The minor product ions PO₂⁻ (m/z 63) and PO₃⁻ (m/z 79) were observed in most of the reaction spectra. No reactions were observed when the [Co_xP_y]⁻ anions were exposed to NH₃, MeOH, EtOH and i-PrOH.

Structure, stability and reactivity

How can these observations be interpreted? At this point the principal questions about these [Co_xP_y]⁻ ions are: (1) what are their probable structures, and what are the relevant structural principles? (2) what processes or properties of the ions determine the map of observed compositions? does thermodynamic stability or electron affinity determine the species observed? (3) what is the relationship between structure and reactivity? what structural features confer reactivity, and how can the collisional dissociations and the reactions with other reagents be used to infer structure?

In approaching these questions we first consider the structural implications of the ion map. We note: (a) the swathe of entries on the ion map is broad, and much broader than the corresponding ion map of [Co_xS_y]⁻ generated by laser ablation of CoS.⁴⁶ In the cobalt dimension there are 17 ions with the composition [Co_xP₈]⁻, for x  = 1 to 17, 15 ions of type [Co_xP₇]⁻, 14 ions of type [Co_xP₁₁]⁻ and 14 ions of type [Co_xP₁₄]⁻. In the phosphorus dimension there are 8 ions [Co₆P_y]⁻, 10 ions [Co₁₂P_y]⁻, 10 ions for [Co₁₅P_y]⁻ and for [Co₁₆P_y]⁻. The implication of this is considerable structural diversity. It is unlikely that a small number of structural principles and moieties could construct this range of compositions. (b) There are two regions of higher abundance, a domain with smaller [Co_xP_y]⁻ centered about [Co₄P₆]⁻, up to Co₈, and a second domain with larger [Co_xP_y]⁻ centered around [Co₁₅P₁₁]⁻ in the cobalt range 10 to 20. (c) In the domain of smaller but abundant ions the compositions are P-rich, with only four ions (x/y  = 6/5, 7/6, 8/6, 8/7) having (just) less than one P per Co. This implies that structures with P–P bonded units are present. In this region of the ion map there are also few minor ions which are P-deficient. (d) In the domain of larger and abundant ions on the ion map, all of the major [Co_xP_y]⁻ ions are P-deficient, with y appreciably less than x. The three compositions [Co₁₀P₈]⁻, [Co₁₅P₁₁]⁻ and [Co₂₀P₁₃]⁻ indicate the general extent of phosphorus deficiency. There are few minor ions with y>x and x>10. The implication of observation (d) is that P–P connections are less likely in these structures. However, for these ions two general structural types can be conceived, one with dispersed Co and P atoms and maintained by Co–P bonds, and at the other extreme a cobalt metallic core coated with peripheral P–P nets. (e) In the smaller domain there is perhaps a slight preference for an even number of P atoms, but there is no evident preference for P_even in the larger domain.

The CID experiments could be made only on ions in the smaller, P-rich domain, with the general result of loss of P₂, P₄, CoP₂ and/or CoP₄. This also points to the occurrence of P₂ groups in the structures of these smaller ions. The collisional dissociation of Co from [Co₄P₄]⁻ to form [Co₃P₄]⁻ is the valuable exception, and is consistent with the expectation that the well known and ubiquitous cubanoid M₄X₄ structure with isolated P atoms could be expected for [Co₄P₄]⁻.

In considering the reactivities it is noted that there is no correlation of reactivity with the odd or even number of electrons in the [Co_xP_y]⁻ ion. The reactions are mainly addition, of one or more S atoms, or one O atom. Addition could occur at Co or at P, or, by insertion in Co–P or P–P bonds: the formation of P–S bonds was demonstrated in the gas phase reactions of copper sulfide anions with P₄.³⁶ The structural information implicit in the reactivities is more evident in the [Co_xP_y]⁻ ions that do not react, and presumably have structures without addition sites.

Density functional calculations of geometrical and electronic structure

We have made density functional calculations of possible structures for many of the observed compositions. Principles used in postulation of structures, and calculational methods for exploring the geometry–energy–electronic hypersurfaces, are outlined in the methodology section. The full results are too voluminous to include here, and will be published separately, but here we report on some of the key compositions which are relevant to the reactivity results and interpretations described in this paper. We report calculations on [CoP₈]⁻, [Co₄P₈]⁻, [Co₅P₉]⁻ and [Co₆P₁₀]⁻, which are P-rich and unreactive, and on [Co₄P₄]⁻ which reacts by addition. Our objective is to recognise the structural features and principles responsible for the non-reactivity and the concomitant stability. We have postulated structures, optimised their geometries using spin unrestricted calculations of the neutral species and negative ions, and calculated electron affinities. Structures described for [Co_xP_y] are denoted xyA, xyB, etc. in order of decreasing stability.

Fig. 5 shows the optimised structures for isomers of [CoP₈]⁻, and Table 3 contains the relative energies, electron affinities and electronic states. The structure 18A is most stable, and has the most favourable electron affinity. It contains two η⁴-P₄ units bound to Co in parallel staggered conformation, such that Co has square antiprismatic co-ordination. The corresponding eclipsed (D_4h) isomer, 18B, is slightly less stable. Other favourable structures have an opened P₄ tetrahedron functioning as a bidentate chelate ligand, with tetrahedral co-ordination (18C) better than square planar (18E). Structure 18H shows that η³-P₄ co-ordination is not favourable.

Fig. 5 The structures of isomers of [CoP₈]⁻, as optimised by density functional calculations: black is Co, grey is P.

Table 3 Properties of structures calculated for [CoP₈]⁻: structure labels refer to Fig. 5

Structure	Energy/kcal mol^−1	Relative energy/kcal mol^−1	Electron affinity/kcal mol^−1	Electronic structure spin state S (HOMO–LUMO gap/eV)
18A	−724.6	0	−83.5	0 (2.21)
18B	−720.1	+4.6	−78.5	0 (2.83)
18C	−703.1	+21.5	−51.6	0 (1.33)
18D	−697.7	+27.0	−59.7	0 (1.10)
18E	−684.1	+40.5	−44.2	0
18F	−673.4	+51.2	−43.7	0 (0.3)
18G	−672.0	+52.6	−48.9	0 (0.96)
18H	−670.0	+54.6	−42.0	0

---

All of the [CoP₈]⁻ structures have spin singlet ground states, and the more stable structures 18A, 18B, 18C and 18D have relatively large HOMO–LUMO gaps. All of the [CoP₈]⁻ structures have co-ordinatively saturated cobalt, consistent with the absence of addition reactions for this composition. The loss of P₄ in the CID experiments is consistent with all of these structures except 18D and 18G where dissociation also of P₂ would be expected.

For [Co₄P₄]⁻, Fig. 6 shows the optimised structures and Table 4 contains their relevant properties. The cubanoid structure 44A is substantially more stable than any other we have postulated. With symmetry T_d it has a t₂³ ground state, a competitive electron affinity and flattened tetrahedral stereochemistry at Co, as expected. We note that there is a local minimum for the more orthogonal cubanoid structure, 44B.

Fig. 6 The structures of isomers of [Co₄P₄]⁻, as optimised by density functional calculations: black is Co, grey is P.

Table 4 Properties of structures calculated for [Co₄P₄]⁻: structure labels refer to Fig. 6

Structure	Energy/kcal mol^−1	Relative energy/kcal mol^−1	Electron affinity/kcal mol^−1	Electronic structure of spin state S (HOMO–LUMO gap/eV)
44A	−737.4	0	−54.5	3/2, t2^3(0.59)
44B	−685.5	51.9	−43.7	3/2, t2^3(0.10)
44C	−661.7	75.7	−48.3	High spin (0.30)
44D	−654.3	83.1	−53.3	High spin (0.07)

---

Fig. 7 shows the wide range of structures investigated for [Co₄P₈]⁻, and relevant results are presented in Table 5. Three different structures, 48A, 48B and 48C, have very similar energies, but are differentiated by their electron affinities which favour the structures in the order 48C>48A>48B. Structure 48A (C_2v) is a tetrahedron of Co atoms, with bent P₃ straddling two faces, P₂ groups on the other two faces and a P atom bridging one edge: 48B is cubanoid Co₄P₄ with terminal η²-P₂ on two Co atoms; 48C is a rectangle of Co atoms with (μ₄-P)₂, (μ-P)₂ and (η²-μ-P₂)₂. Further, structures 48A and 48C have all Co atoms at least four-co-ordinate, while 48B has two exposed three-coordinate Co atoms. With the assumption that three-co-ordinate Co would be active to addition, which is not observed for [Co₄P₈]⁻, the conclusion is that 48B is an improbable structure, and that 48C and 48A are most probable. We note that 48C has a spin doublet ground state while 48A is S  = 5/2: all of the most favourable structures for this composition have very small energy gaps between HOMO and LUMO, and close-lying electronic states. From Table 5 it can be seen that isomer 48E has a competitive electron affinity and relative energy, and that 48G has the highest electron affinity, slightly greater than that of 48C. However, 48G is calculated with unusually short Co–Co bonds of 2.29 Å which also has the effect of increasing the electron affinity (EA), and it is possible that the higher EA of 48G is due to slight overbinding of Co–Co in our density functional methods. The Co–Co distances in 48C are “ normal” for relatively bare clusters such as these in the gas phase, and are very similar to the Co–Co distances in 48E. The conclusion from this analysis is that the most probable structure for [Co₄P₈]⁻ is 48C.

Fig. 7 The structures of isomers of [Co₄P₈]⁻, as optimised by density functional calculations: black is Co, grey is P. Co–Co distances are 2.47 Å × 4, and 2.57 Å × 1 in 48A, 2.43, 2.55 Å in 48C, 2.43, 2.57 Å in 48E, 2.29 Å in 48G, 2.82 Å in 48H.

Table 5 Properties of structures calculated for [Co₄P₈]⁻: structure labels refer to Fig. 7

Structure	Energy/kcal mol^−1	Relative energy/kcal mol^−1	Electron affinity/kcal mol^−1	Electronic structure of, spin state, S (HOMO–LUMO gap/eV)
48A	−1040.9	0	−63.0	5/2 (0.30)
48B	−1039.1	+1.7	−51.1	1/2 (0.05)
48C	−1038.6	+2.2	−72.2	1/2 (0.10)
48D	−1029.3	+11.6	−61.2	5/2 (0.25)
48E	−1027.0	+13.9	−64.2	1/2 (0.04)
48F	−1018.7	+22.2	−59.3	1/2 (0.29)
48G	−1017.9	+22.9	−74.7	3/2 (0.25)
48H	−1012.4	+28.5	−64.9
48I	−1011.6	+29.3	−54.8
48J	−1001.7	+39.2	−55.2
48K	−994.3	+46.5
48L	−991.4	+49.5
48M	−990.9	+50.1	−68.9	1/2 (0.18)

---

The three isomers investigated for [Co₅P₉]⁻, also unreactive to addition, are presented in Fig. 8 and Table 6. The most favourable structure, in terms of total energy and electron affinity, is 59A, which is a square pyramid of Co atoms, with P₂ bridging each Co₃ face, and μ₄-P on the basal face. There is square planar (P)₄ co-ordination of the apical Co atom, and five P atoms bonded to each of the basal Co atoms, and this co-ordination greater than threefold is consistent with the observed non-reactivity of [Co₅P₉]⁻. The two other isomers have three-co-ordinate Co atoms, and would be expected to be more reactive.

Fig. 8 The structures of isomers of [Co₅P₉]⁻, as optimised by density functional calculations. In 59A (C_4v) the Co–Co distances are 2.48 (to apical Co) and 2.50 Å (between basal Co).

Table 6 Properties of structures calculated for [Co₅P₉]⁻: structure labels refer to Fig. 8

Structure	Energy/kcal mol^−1	Relative energy/kcal mol^−1	Electron affinity/kcal mol^−1	Electron structure of spin state S (HOMO–LUMO gap/eV)
59A	−1243.7	0	−74.1	3/2 (0.13)
59B	−1212.0	+31.7	−59.7	5/2 (0.19)
59C	−1177.2	+66.5	−59.4	Close-lying spin states

---

For [Co₆P₁₀]⁻ two different structure types have been calculated as more stable than alternatives (see Fig. 9, Table 7). Structure 610A is a flattened Co₆ octahedron with P₂ and P₃ bridging the equatorial edges. Each Co is co-ordinated by at least 4 P atoms: the axial Co have approximately square planar co-ordination, and the equatorial Co have P₂ + P₃ planar five-co-ordination. Structure 610B is based on the well known M₆S₆ structure type with D_3d symmetry.⁴⁷ There is an additional capping P atom on the threefold axis, and three P₂ groups around the opposite edge. Again each Co atom is bonded to four P atoms, which is consistent with the absence of addition reactions. The energy of 610A is better, but the electron affinity of 610B is slightly more favourable. Both of these structures are possible.

Fig. 9 The structures of isomers of [Co₆P₁₀]⁻, as optimised by density functional calculations.

Table 7 Properties of structures calculated for [Co₆P₁₀]⁻: structure labels refer to Fig. 9

Structure	Energy/kcal mol^−1	Relative energy/kcal mol^−1	Electron affinity/kcal mol^−1	Electronic structure of spin state S (HOMO–LUMO gap/eV)
610A	−1433.4	0	−60.7	1/2 (0.19)
610B	−1400.5	+33.0	−63.3	Close-lying electronic states

---

Discussion

Laser ablation of CoP is clearly a rich source of [Co_xP_y]⁻ clusters, and many are unusually reactive. This abundance and reactivity have implications for applications of these clusters for further synthesis in the gas phase, and for their use in the generation of new condensed phases and films of cobalt phosphides, as we discuss further below. First we evaluate some of the fundamentals of these clusters.

We can now compare the compositions of [Co_xP_y]⁻ ions formed by two different experiments, one being laser ablation of a mixture of the elements Co and P,² the other being laser ablation of the compound CoP, reported here. The major ions observed in each experiment are compared on the ion map in Fig. 10. The fact that many ions are common to the two quite different generation processes indicates that the products are not excised fragments of CoP, but are formed during cooling of the high energy plasma generated in each case. This conclusion that synthesis occurs by associative processes in the gas phase and not by excision is the same as that from our other experiments in which binary clusters are generated by laser ablation.^1,49

Fig. 10 Comparison of the compositions of the major [Co_xP_y]⁻ ions observed by laser ablation of either the compound CoP or a mixture of cobalt metal and red phosphorus (2:1 w/w). Compositions which appear as major ions in both experiments are labelled B; those which are major ions only from CoP are labelled C, and those which are major ions only from the mixture are labelled M. For the compound the major ions were considered to be those with >20% intensity, while for the mixture the major ions considered are those with >10% intensity. In each experiment additional ions are formed.

The main difference between the two laser ablation experiments is that many of the major [Co_xP_y]⁻ ions formed from CoP are P-deficient, while the ions produced from the Co + P mixtures are P-rich (see Fig. 10). An exception here is [CoP₈]⁻ which was observed as a major ion from CoP but not from Co + P. The major ions that are formed in both experiments are generally P-rich, with the exception of [Co₈P₇]⁻, and two ([Co₄P₄]⁻ and [Co₇P₇]⁻) have equal numbers of Co and P atoms. As well as these differences in composition, the ablation of CoP produced larger [Co_xP_y]⁻ ions up to x  = 23, whereas ablation of Co + P was limited to x  = 13. A simple expectation for the laser plumes is that CoP would contain a 1:1 ratio of Co and P, and that Co + P would be P-rich because the red phosphorus precursor is more volatile than cobalt metal. This is consistent with the general richness of the [Co_xP_y]⁻ from Co + P, although [CoP₈]⁻ from CoP is the exception.

The [Co_xP_y]⁻ clusters are generally reactive to addition of S (from H₂S) and O from (NO₂ and N₂O), which is significant because anionic clusters are usually less reactive in the gas phase than are cationic counterparts. The few unreactive [Co_xP_y]⁻ clusters are P-rich, and this is interpreted in terms of the saturation of co-ordination at the Co atoms. Addition reactions are postulated to occur using exposed cobalt co-ordination sites, with minimal concomitant rearrangement of the cluster during addition. The unreactive P-rich clusters are believed to have sufficient P atoms in P_n moieties to achieve enclosing co-ordination at all Co atoms, while those with smaller P:Co ratios have some exposed Co atoms. This is consistent also with the density functional calculations of the probable structures of the unreactive clusters [CoP₈]⁻, [Co₄P₈]⁻, [Co₅P₉]⁻ and [Co₆P₁₀]⁻.

We note a subtle difference in reactivity of [Co₄P₄]⁻ (undergoes addition of S and O) and [Co₈P₆]⁻ which undergoes addition of O but not S. Calculations not reported here indicate that the probable structure of [Co₈P₆]⁻ is a highly symmetrical cube of Co with P atoms capping each face (hexahedro-Co₈-octahedro-P₆), 86A, in which each Co has trigonal pyramidal CoP₃ co-ordination. [Co₄P₄]⁻ as 44A also presents trigonal pyramidal CoP₃ co-ordination. However, there is a structural difference, in that the cobalt co-ordination is more acute in 44A (P–Co–P 107°) than in 86A (P–Co–P 110°), and therefore [Co₄P₄]⁻ is expected to be more reactive.

There are some P-rich clusters which undergo addition of S (see Table 2), and our hypothesis for these is that the structures are such as to allow some under-co-ordination of Co. This hypothesis is the subject of continuing density functional calculations. However, we recognise that electronic factors are important, and that the generally high reactivity of most of the [Co_xP_y]⁻ clusters is probably also due to their close-lying electronic states with S>0.

The structural principles evident so far in the most stable calculated structures are that P atoms, P₂ groups and P₃ groups occur, and that they bridge the faces, to a lesser extent the edges, of Co_x polyhedra. The map of compositions of the observed [Co_xP_y]⁻ clusters (Table 1) is characterised by extensive ranges of y for most values of x. Our preliminary interpretation of this is that structures with the same or similar Co_x polyhedra can be bridged by P, P₂ or P₃ at the various faces and edges. This is being tested with density functional calculations in progress. In this context we note that many of the organometallic complexes containing P_n moieties demonstrate variations in size (n) but contain a small number (often one) of highly catenated cycles and polycycles of phosphorus,⁹ and have little similarity to the structures demonstrated and postulated in this paper. As an illustration of the structural difference for related polyarsenides, [(Et₂PhPCo)₆As₁₂] contains As₆ and As₃ cycles.²⁹

Finally, we consider the implications of our results for applications, especially to the syntheses of new compounds and materials. First, generation in high vacuum is not difficult, and many of the products of laser ablation are quite reactive, which means that gas phase syntheses at very low pressures are feasible. However, these high reactivities mean that condensation, as in the deposition of films, is likely to be accompanied by substantial rearrangement to more stable phases. However, the addition reactions which are reported here tend to reduce reactivity by completion of co-ordination, and so condensation of [Co_xP_y] in the presence of terminating ancillary ligands L is expected to form stabilised new compounds Co_xP_yL_z. Such products with multiple smaller P_n moieties would be quite different from most of the known compounds with fewer larger P_n moieties, and we predict the existence of such compounds while recognising the difficulties of synthetic access.

Acknowledgements

This research is supported by the Australian Research Council and the University of New South Wales. We thank Dr Robert Glaum for the sample of CoP.

References

1. K. J. Fisher, I. G. Dance, G. D. Willett and M. Yi, J. Chem. Soc., Dalton Trans., 1996, 709.
2. P. F. Greenwood, I. G. Dance, K. J. Fisher and G. D. Willett, Inorg. Chem., 1998, 37, 6288.
3. H. G. von Schnering and W. Hönle, Chem. Re[italic v]., 1988, 88, 243.
4. O. J. Scherer, Angew. Chem., Int. Ed. Engl., 1990, 29, 1104.
5. O. J. Scherer, B. Rink, G. Berg, J. Braun, B. Hobel, P. Walther, R. Winter, G. Heckmann and G. Wolmershauser, Phosphorus Sulfur Silicon Relat. Elem., 1993, 77, 25.
6. D. Fenske and W. Holstein, Angew. Chem., Int. Ed. Engl., 1994, 33, 1290.
7. K. H. Whitmire, Ad[italic v]. Organomet. Chem., 1998, 42, 1.
8. O. J. Scherer, C. Hofmann, M. Ehses, S. Weigel and G. Wolmershauser, Phosphorus Sulfur Silicon Relat. Elem., 1999, 144–146, 141.
9. O. J. Scherer, Acc. Chem. Res., 1999, 32, 751.
10. M. Di Vaira and L. Sacconi, Angew. Chem., Int. Ed. Engl., 1982, 21, 330.
11. O. J. Scherer, H. Sitzmann and G. Wolmershauser, Angew. Chem., Int. Ed. Engl., 1985, 24, 351.
12. M. Baudler, S. Akpapoglou, D. Ouzounis, F. Wasgestian, B. Meinigke, H. Budzikiewicz and H. Münster, Angew. Chem., Int. Ed. Engl., 1988, 27, 280.
13. W. Honle, H. G. von Schnering, G. Fritz and H. W. Schneider, Z. Anorg. Allg. Chem., 1990, 584, 51.
14. L. Y. Goh, R. C. S. Wong and E. Sinn, J. Chem. Soc., Chem. Commun., 1990, 1484.
15. L. Y. Goh, R. C. S. Wong, C. K. Chu and T. W. Hambley, J. Chem. Soc., Dalton Trans., 1990, 977.
16. O. J. Scherer, B. Werner, G. Heckmann and G. Wolmershauser, Angew. Chem., Int. Ed. Engl., 1991, 30, 553.
17. O. J. Scherer, B. Hobel and G. Wolmershauser, Angew. Chem., Int. Ed. Engl., 1992, 31, 1027.
18. M. Scheer, U. Becker, J. C. Huffman and M. H. Chisholm, J. Organomet. Chem., 1993, 461, C1.
19. L. Y. Goh, R. C. S. Wong and E. Sinn, Organometallics, 1993, 12, 888.
20. M. Scheer, U. Becker, M. H. Chisholm, J. C. Huffman, F. Lemoigno and O. Eisenstein, Inorg. Chem., 1995, 34, 3117.
21. M. Scheer, U. Becker and E. Matern, Chem. Ber., 1996, 129, 721.
22. C. Hofmann, O. J. Scherer and G. Wolmershauser, J. Organomet. Chem., 1998, 559, 219.
23. O. J. Scherer, S. Weigel and G. Wolmershauser, Chem. Eur. J., 1998, 4, 1910.
24. O. J. Scherer, T. Hilt and G. Wolmershaeuser, Organometallics, 1998, 17, 4110.
25. O. J. Scherer, M. Ehses and G. Wolmershauser, Angew. Chem., Int. Ed., 1998, 37, 507.
26. O. J. Scherer, T. Volmecke and G. Wolmershauser, Eur. J. Inorg. Chem., 1999, 945.
27. M. Di Vaira, P. Stoppioni, S. Midollini, F. Laschi and P. Zanello, Polyhedron, 1991, 10, 2123.
28. M. Di Vaira, D. Rovai and P. Stoppioni, Polyhedron, 1990, 9, 2477.
29. R. Ahlrichs, D. Fenske, K. Fromm, H. Krautscheid, U. Krautscheid and O. Treutler, Chem. Eur. J., 1996, 2, 238.
30. O. J. Scherer, M. Swarowsky and G. Wolmershauser, Organometallics, 1989, 8, 841.
31. M. Haser, U. Schneider and R. Ahlrichs, J. Am. Chem. Soc., 1992, 114, 9551.
32. R. O. Jones and G. Seifert, J. Chem. Phys., 1992, 96, 7564.
33. R. O. Jones, New J. Chem., 1992, 16, 1099.
34. R. O. Jones and D. Hohl, J. Chem. Phys., 1990, 92, 6710.
35. P. Ballone and R. O. Jones, J. Chem. Phys., 1994, 100, 4941.
36. K. Fisher and I. G. Dance, J. Chem. Soc., Dalton Trans., 1997, 2381.
37. K. J. Fisher, I. G. Dance and G. D. Willett, Eur. J. Mass Spectrom., 1996, 2, 369.
38. A. Schmidt and R. Glaum, Z. Anorg. Allg. Chem., 1995, 621, 1693.
39. I. G. Dance, K. J. Fisher and G. D. Willett, Inorg. Chem., 1996, 35, 4177.
40. B. Delley, in Modern density functional theory: a tool for chemistry, eds. J. M. Seminario and P. Politzer, Elsevier, Amsterdam, 1995, pp. 221–254..
41. MSI, http://www.msi.com, 1998..
42. I. G. Dance, in Transition Metal Sulfur Chemistry: Biological and Industrial Significance, eds. E. I. Stiefel and K. Matsumoto, American Chemical Society, Washington, DC, 1996, pp. 135–152..
43. I. Dance, Chem. Commun., 1998, 523.
44. I. G. Dance and K. J. Fisher, J. Chem. Soc., Dalton Trans., 1997, 2563.
45. I. G. Dance, in Comprehensi[italic v]e Coordination Chemistry, ed. G. Wilkinson, Pergamon, Oxford, 1988, vol. 1, ch. 4, pp. 135–177..
46. J. H. El Nakat, K. J. Fisher, I. G. Dance and G. D. Willett, Inorg. Chem., 1993, 32, 1931.
47. I. G. Dance and K. J. Fisher, Prog. Inorg. Chem., 1994, 41, 637.
48. H. Ogino, S. Inomata and H. Tobita, Chem. Re[italic v]., 1998, 98, 2093.
49. J. H. El Nakat, I. G. Dance, K. J. Fisher and G. D. Willett, Inorg. Chem., 1991, 30, 2957.

---

Footnote

† Electronic supplementary information (ESI) available: details of density function calculations and Cartesian coordinates for optimised structures as monoanions. See http://www.rsc.org/suppdata/nj/b0/b006821g/

---