Electron interaction with copper(II) carboxylate compounds

In the present study we have performed electron collision experiments with copper carboxylate complexes: [Cu2(t-BuNH2)2(µ-O2CC2F5)4], [Cu2(s-BuNH2)2(µ-O2CC2F5)4], [Cu2(EtNH2)2(µ-O2CC2F5)4], and [Cu2(µ-O2CC2F5)4]. Mass spectrometry was used to identify the fragmentation pattern of the coordination compounds produced in crossed electron – molecular beam experiments and to measure the dependence of ion yields of positive and negative ions on the electron energy. The dissociation pattern of positive ions contains a sequential loss of both the carboxylate ligands and/or the amine ligands from the complexes. Moreover, the fragmentation of the ligands themselves is visible in the mass spectrum below m/z 140. For the studied complexes the metallated ions containing both ligands, e.g., Cu2(O2CC2F5)(RNH2)+, Cu2(O2CC2F5)3(RNH2)2+ confirm the evaporation of whole complex molecules. A significant production of Cu+ ion was observed only for [Cu2(µ-O2CC2F5)4], a weak yield was detected for [Cu2(EtNH2)2(µ-O2CC2F5)4] as well. The dissociative electron attachment processes leading to formation of negative ions are similar for all investigated molecules as the highest unoccupied molecular orbital of the studied complexes has Cu–N and Cu–O antibonding character. For all complexes, formation of the Cu2(O2CC2F5)4−• anion is observed together with mononuclear DEA fragments Cu(O2CC2F5)3−, Cu(O2CC2F5)2− and Cu(O2CC2F5)−•. All dominant DEA fragments of these complexes are formed through single particle resonant processes close to 0 eV.


Introduction
Present technological changes require the development of new methods and new materials for preparation of thin layers or 3D nanostructures. Complexes and metalorganic compounds are used as precursors in modern nano scale layer techniques. After activation, molecules undergo dissociation on the surface. Volatile parts of molecules are removed from the surface while a metal component remains and forms the layer. Activation of the precursor molecules can be induced by several processes. For instance, a catalytic or a thermal dissociation can occur. Plasma activated processes such as plasma enhanced chemical vapor deposition (PECVD) can be used for coating of the surface [1]. In the latter, reactive chemical species (radicals) and electrons lead to activation of molecules and this process can be controlled well on large scales.
One of the most innovative techniques, known as EBID or FEBID (Focused Electron Beam Induced Deposition) [2,3], uses a high energy electron beam that can be focused into a spot of diameter in the nanometer range for a spatially confined activation of precursor molecules on a very narrow range of the surface. The presence of high energy electrons from the primary beam (usually around 10 keV) causes ionization inside the wafer, with a high yield of secondary low energy electrons (below 100 eV). These electrons can diffuse to the surface and initiate reactions in the precursor molecules. As a result, a deposit is formed. This technique enables the production of free standing 3D nanostructures and is already used commercially for the repair of photolithographic masks [4,5].
However, the underlying chemical reactions on the surface are still not well known. Moreover, the main problems of FEBID are co-deposited impurities resulting from incomplete dissociation of the precursor molecules. The level of purity strongly depends on the type of the precursor molecule. Only a few types of precursors are known to produce a layer with purity over 80% [3]. Moreover, there is no clear connection between the layer purity and the type of ligand in the precursor; an iron deposit from Fe(CO) 5 leads to purity over 95% of Fe, while tungsten layers from W(CO) 6 can reach purities of W from 55 to 70% [6,7]. For comparison, the deposition of cobalt from Co(CO) 3 (NO) leads to around 50% purity [8] or satisfying purity over 95% using the dimer Co 2 (CO) 8 [9]. Only few types of precursor molecules can be converted into a layer with satisfying level of purity, for other elements there is still demand for new precursors with other satisfactory parameters like vapor pressure, toxicity, thermal stability, and stability over time.
Electrons present on the surface are combination of high energy electrons from the primary electron beam and secondary electrons emitted from the surface. High energy electrons have low interaction cross section with the target molecule; their interaction efficiency is therefore very low. Secondary electrons, on the other hand, play important role through electron attachment (EA), dissociative electron attachment (DEA) [10][11][12][13][14][15][16] and electron ionization (EI), dissociative ionization (DI) [17][18][19] processes. Their kinetic energy is only a few eV, with energy distribution determined by the type of wafer and energy of primary beam [20,21]. Thorman et al. have compared gas phase and surface data on low energy electron interactions with some commonly used FEBID precursors [22] and have shown that in some cases a single ligand loss dominates the initial fragmentation following electron induced ionization or attachment. This may then induce other surface interactions. They also conclude that dissociation through neutral dissociations induced via electron impact excitation [23] can be as important as DEA.
Copper is an important material commonly used in the advanced metallization of microelectronic and optoelectronic devices and ultralarge-scale integrated (ULSI) circuits due to its low electrical resistivity, high stress-induced deformation, and electromigration resistance which is higher than for aluminum [24][25][26][27]. Copper nanostructures, especially nanowires, are applied in opto-electronic devices, solar cells, field-emission displays, catalysis, electronic skins, and sensor devices. Moreover, they can find medical applications because copper exhibits antibacterial and antifungal properties [28].
In FEBID experiments, the deposited Cu-C lines and squares, obtained from the fluorinated copper(II) β-diketonate  [29,30]. Therefore, new copper FEBID precursors are still necessary. When designing such new precursors, it should be considered that copper(II) derivatives are more user-friendly than copper(I) compounds, which are usually air and moisture sensitive, which results in decomposition of the precursor itself. Also, introduction of amine ligands was expected to be advantageous. In fact, the reducing action of ammine ligands was discussed previously with respect to FEBID experiments using cisplatin [Pt(NH 3 ) 2 Cl 2 ] as precursor [31]. These latter experiments were motivated by gas phase DEA studies on this compound showing also that it can be evaporated intact [32].
Gas phase electron-precursor collision experiments should allow to determine the potential usefulness of new compounds in FEBID. Moreover, these results can be interesting for the development of new metalorganic or coordination compounds suitable for FEBID.
Previous gas phase studies performed in Bratislava with coordination compounds have shown the importance of DI and DEA processes in FEBID [10][11][12][17][18][19]35]. The partial decomposition of the metal complex via DI and DEA together with the role of secondary electrons in FEBID acting on much wider range than is the focus of the primary beam leads to broadening of the deposited structure [36]. The importance of the DI and/or DEA processes in FEBID was also discussed by Warneke et al. [37] with the focus on the appropriate choice of the ligands on the metal complex. In this work, acetylacetone was studied under electron impact and it was concluded from gas phase experiments that radicals released by electron-induced fragmentation react with intact molecules to produce a non-volatile residue that was detected using XPS. These results emphasize that both the proper choice of ligands and the knowledge of elementary processes under DI and DEA in gas phase are important for the understanding and development of FEBID precursors.
In the present experiments electron impact ionization, electron attachment, and subsequent dissociation processes have been studied for the first time on the following copper(II) pentafluoropropionate derivatives: [Cu 2 (t-BuNH 2 ) 2 (µ-O 2 CC 2 F 5 ) 4  The carboxylate copper(II) complexes with tert-butylamine ([Cu 2 (t-BuNH 2 ) 2 (µ-O 2 CR) 4 ], where R = C n F 2n+1 , n = 1-6) were previously applied as Cu CVD precursors [34,38] for the formation of copper nanomaterials. This fact confirms that copper(II) carboxylate compounds can be considered as copper sources in vapor deposition processes. The influence of secondary ligands on the physicochemical properties and, with regards to FEBID, the electron-induced fragmentation behaviour of pentafluoropropionate Cu(II) complexes as studied here is thus of particular interest.

Results
Mass spectrum of [Cu 2 (µ-O 2 CC 2 F 5 ) 4 ] [Cu 2 (µ-O 2 CC 2 F 5 ) 4 ] represents the basic chemical structure for all discussed carboxylate compounds in this work. In Figure 2 the fragmentation pattern of [Cu 2 (µ-O 2 CC 2 F 5 ) 4 ] is presented. The spectrum was obtained with higher mass resolution at which it is easy to resolve to atomic masses for m/z from 10 to 150. For m/z from 150 up to 800 lower mass resolution was used to increase the signal intensity. Some of the peaks obtained at low resolution but with satisfying intensity were re-measured with higher resolution again.
The parent ion Cu 2 (µ-O 2 CC 2 F 5 ) 4 +• has been detected at m/z 788. Its electron induced dissociation can be characterized by several fragmentation patterns. The first is the sequential loss of the ligand radical O 2 C-C 2 F 5 (L), which is characterized by formation of Cu 2 L 3 + , Cu 2 L 2 +• , and Cu 2 L + ions up to a final decomposition to the copper dinuclear Cu 2 +• and the Cu + ion as well. Another pattern results from the fragmentation of the O 2 C-C 2 F 5 ligand itself, as can be seen for the C-C bond dissociation between the CO 2 and C 2 F 5 units. The rupture of this bond in the coordinated carboxylate gives rise to the Cu 2 (L)(O 2 C) + ion. The remaining products are formed via additional fragmentation of the ligand or rearrangement of some of its atoms, e.g., Cu 2 CF 3 + . Moreover, we have identified the simultaneous C-F and C-C bond dissociation via fragmentation of neutral tetrafluoroethene (C 2 F 4 ) from the complex, while the remaining fluorine atom is bound in Cu 2 F + . Finally, the formation of the Cu + ion has been observed only in the mass spectrum at higher temperature (150 °C) of the molecular beam source. This copper fragment abundance achieved the highest value among signals registered below m/z 300. Over this m/z range the signal intensity is not reduced by the QMS (Experimental).
An unusual product in the mass spectrum of this molecule can be seen at m/z 45, which is according to our conclusions the COOH + ion. The only explanation for the formation of this product from a sample without any hydrogen atom is the interaction between [Cu 2 (µ-O 2 CC 2 F 5 ) 4 ] and adsorbed water traces.
The possibility that the [Cu 2 (µ-O 2 CC 2 F 5 ) 4 ] sample was contaminated by some free acid was excluded.
Appearance energies have been measured for several fragments ( Figure 3). The values presented in Table 1 are estimated with uncertainty ±0.5 eV, which results from the resolution of the electron beam that varied from 100 meV up to 500 meV. The largest value was taken to estimate the margin of error. The resolution of the electron beam was lowered to increase the electron current and ion yields when the sample has adsorbed on the electrodes of the monochromator and thus affected their electric fields. The appearance energy for the ion with m/z 18 is AE 18 = 12.6 eV and is in a good agreement with the ionization energy of the water molecule, IE(H 2 O) = 12.62 eV [42]. It cannot be produced from [Cu 2 (µ-O 2 CC 2 F 5 ) 4 ] itself due to its lack of hydrogen atoms; obviously must have desorbed from the sample. This is in agreement with the presence of the previously discussed COOH + ion that is assumed to be formed via interaction of residual water molecules with the sample.  .9 eV without any other distinguishable threshold. The origin of the second threshold can be a) an energetically higher excited state of the same fragment, b) a different stoichiometric fragment or c) a different energetically higher process, or doubly charged product. We did not find any other stoichiometric product for the given masses so the existence of an excited ionic state is the reasonable explanation. An interesting fact is that the loss of two ligands requires less energy than the dissociation of first and third ligand.
As mentioned before we did not observe any signal for a ligand cation itself, therefore we did not obtain any AE value for its production. A decreasing trend of AE for C x F y fragments was found: AE 119 = 15.1 eV, AE 100 = 12.2 eV, AE 69 = 11.7 eV, AE 31 = 11.0 eV except of CF 2 + ion (m/z 50) that clearly requires additional energy for dissociation of F atom from CF 3 + ion. Ion COOH + with m/z 45 proposed as an impurity has AE 45 = 13.5 eV, which is however much higher than the ionization energy of hydrocarboxyl radical IE(COOH) = 8.2 eV [43]. According to our conclusions its formation is accompanied by dissociation of two Cu-O bonds and a C-C bond on the complex as well as O-H bond dissociation on the water molecule.  Figure 5) is characterized by a lack of parent cation. The highest visible mass (for details Experimental) for both molecules is produced via detachment of one carboxylate ligand (Cu 2 L 3 A 2 + ). As discussed before for the previous molecules, the range of the ions formed by whole carboxylate or amine ligand loss was registered. They appear for both tertbutyl and sec-butyl substituents above mass m/z 130 ( Figure 5). We can recognize dicopper fragments, e.g., Cu 2 LA 2 + , Cu 2 LA + ,  (Table 3). However, the appearance energies were measured only for the fragments with highest intensity due to the charging effects of the monochromator electrodes relating to the measured sample. To avoid the total loss of the signal the electron current was increased which lead to decrease of the electron beam resolution and thus larger uncertainties in the determination of the appearance energies.
The present results are moreover supported by independent measurements of photoelectron spectra (PES) [45,46] for two of

Negative ions
Mass spectra resulting from electron attachment to all four measured compounds are shown in Figure 7 and reveal many common features. We have observed the production of the transient negative ion (TNI)   where the signal was insufficient. A strong feature of a single particle resonance close to 0 eV is seen for all DEA products. This is in fact possible only when the electron affinities (EA) of the individual products are large enough to compensate the corresponding bond dissociation energies (BDE) [47]. However, we have no quantitative information about the EA and BDE of the individual products shown in Figure 8.

Positive ions
The dissociation pattern of dissociative ionization of the investigated molecules showed loss of entire ligands as well as fragments containing only copper atom ( Figure 9). Limited numbers of fragments containing more than two carboxylate ligands are observed (Cu 2 L 4 +• , Cu 2 L 3 + , Cu 2 L 3 A 2 + ). However, it is important to note that intensity of fragments with higher masses is discriminated during their transition through the ion optics and quadrupole system of the experiment. As shown in Figure 3 and listed in  [49,50]. An additional comparison may be provided with hexafluoroethane (C 2 F 6 ) [51], where the appearance energy difference between CF 3 + and CF + ion is ≈3 eV in relation to almost 7 eV difference for the [Cu 2 (µ-O 2 CC 2 F 5 ) 4 ] compound. Moreover, the formation of the C 2 F 4 +• ion requires 5.7 eV more than formation of C 2 F 5 + ion from hexafluoroethane. We detected an opposite trend with the difference of slightly over 3 eV. This significant effect is provided by the presence of the carboxyl group.
For all three studied complexes, the fragmentation of the alkylamine ligand cation is observed with higher intensity than that of the carboxylate ligand. For [Cu 2 (EtNH 2 ) 2 (µ-O 2 CC 2 F 5 ) 4 ] and [Cu 2 (s-BuNH 2 ) 2 (µ-O 2 CC 2 F 5 ) 4 ], it is produced as a deprotonated (A-H) + ion instead of A + , visible in the spectrum of [Cu 2 (t-BuNH 2 ) 2 (µ-O 2 CC 2 F 5 ) 4 ]. In the case of [Cu 2 (EtNH 2 ) 2 (µ-O 2 CC 2 F 5 ) 4 ], the fragmentation leads only to the formation of a fragment with m/z 30 as assigned to C 2 H 6 +• or more probable to H 2 NCH 2 + . The registered spectrum is the superposition of the thermal loss EtNH 2 and the complex spectra what influenced the observed signals intensities. Mass spectra for sec-butylamine and tert-butylamine show one dominant dissociation product: H 2 NC(H)CH 3 + with m/z 44 and H 2 NC(CH 3 ) 2 + with m/z 58, respectively [44]. Both of these processes are also observed in corresponding dissociation patterns of the molecules investigated here. The ion with m/z 44 may also be associated with CO 2 +• . However, the appearance energy of m/z 44 from [Cu 2 (EtNH 2 ) 2 (µ-O 2 CC 2 F 5 ) 4 ] is lower than that of CO 2 +• from CO 2 [52]. This points to an assignment to (A-H) + as may also be expected for the s-BuNH 2 and t-BuNH 2 complexes. Moreover, the AE detected for H 2 NC(CH 3 ) 2 + (m/z 58) from [Cu 2 (t-BuNH 2 ) 2 (µ-O 2 CC 2 F 5 ) 4 ] of AE 58 = 9.1 eV is similar to the PES value of its IE = 9.3 eV. We can thus conclude that the dissociation of the amine ligand from the complex and the consequential dissociation of one methyl group as well as the second methyl group are extremely effective processes. For [Cu 2 (s-BuNH 2 ) 2 (µ-O 2 CC 2 F 5 ) 4 ], the loss of an ethyl group is a dominant product of alkylamine ligand fragmentation. Additionally, the single methyl group dissociation is observed. Other remaining fragments of the studied molecules relate to additional hydrogen or carbon dissociations.

Negative ions
In the fragmentation pattern obtained via DEA, we observed with particularly high intensity the symmetrical splitting of the Cu 2 (O 2 CC 2 F 5 ) 4 −• ion into the Cu(O 2 CC 2 F 5 ) 2 − and Cu(O 2 CC 2 F 5 ) 2 fragments. Therefore, we can suppose that the lowest unoccupied molecular orbital (LUMO) of the measured complexes has an antibonding character and consists from d-orbitals on Cu atoms, and p-orbitals from the corresponding O atoms. However, Figure 7 shows a difference in the relative intensity of Cu(O 2 CC 2 F 5 ) 2 − ion formed via DEA for all four compounds. While it is the most intensive product for the [Cu 2 (O 2 CC 2 F 5 ) 4 ] measured at electron energy close to 0 eV, its intensity is significantly lower in the spectra of the other three complexes, measured at the same electron energy. The maximum of this DEA channel in the [Cu 2 (RNH 2 ) 2 (µ-O 2 CC 2 F 5 ) 4 ] compounds is shifted towards higher energies (Figure 8), which in fact is a consequence of more bonds to be dissociated to produce the same product Cu ( ion directly from TNI. The position of the resonance agrees with the formation of C 2 F 5 − anion from C 2 F 6 molecule, its maximum cross section is here located at a similar energy of 4.8 eV [53]. The presence of EtNH 2 or t-BuNH 2 in the com-plex closes this channel and the C 2 F 5 − anion is formed only via the single particle resonance at almost 0 eV.

Conclusion
This article presents an investigation of the fragmentation following electron impact ionization of and electron attachment to four copper(II) carboxylate complexes.
Regarding electron impact ionization, the cross sections for formation of the parent molecular ions were very weak.  4 ] can efficiently decompose to Cu + ion via electron impact in FEBID. The amine ligand introduction decreased the evaporation temperature but unfortunately suppressed the copper ion formation. On the other hand, this phenomenon can be useful for the "halo" effect limitation in FEBID processes.
Regarding electron attachment, we have registered the first spectra of negative ions for copper carboxylates compounds. Comparable negative ions are formed for all investigated molecules. The electron attachment processes occur mainly at incident electron energy close to 0 eV, through single particle resonances but specific fragments are also formed with smaller intensity through higher-lying resonances.

Experimental
Investigation of electron induced processes was carried out by crossed electron and molecular beam experiments [54]. The electron beam was created by a trochoidal electron monochromator operating with energy resolution down to 100 meV in the range 0-120 eV. In the case of low signals, which were either an inherent property of the sample or resulted from the effect of deposition of the sample on the monochromators electrodes, the electron resolution was reduced up to 300-500 meV. A molecular beam was created by sublimation/evaporation of solid/gel samples into a small chamber. The chamber is connected with a main reaction chamber by small capillary, which creates a molecular beam that perpendicularly collides with the electron beam. Ionic products are then forced by a weak electric field into the ion optics of the quadrupole mass spectrometer. After separation of the products with different mass-to-charge ratios (m/z) the ions are detected by an electron multiplier. A constant electron energy of 70 eV was applied to register the mass spectra, i.e., the ion intensity as function of the m/z ratio of the measured ions. For a selected product (selected m/z ratio) the ion yield dependences were then measured by varying the electron energy. In the case of the negative ions the recorded mass spectrum strongly depends on the electron energy due to a resonant character of attachment reaction.
For the measured cross section of electron ionization and dissociative ionization we can evaluate the threshold value of the corresponding ion formation by a fitting procedure using a modified Wannier law [55]. This value then represents an ionization potential or appearance energy of electron ionization or dissociative ionization respectively.
where b represent background, AE represent appearance (or ionization) energy, ε is electron energy and a, d are independent fitting parameters.
Calibration of the electron energy has been carried out by measurement of the ionization potential of Ar atoms and calibration to its known value 15.76 eV [56] and with reference to the maximum of the electron attachment resonance on SF 6 molecule at energy ≈0 eV [57].  4 ]. In the present experiment the intensity of ions detected with masses above approximately m/z 300 is reduced by the QMS. This can be avoided by decreasing the mass resolution (usually defined as the ratio of mass m and full width at half maximum of the peak Δm) through changing the software parameter, which defines the resolution of the ion peak in the mass spectrum (see, for instance, Figure 2). Heavy ions can thus be detected. However, the position of the peak is then not measured precisely as the signal is broadened over several masses. High resolution measurements presented in the paper represent a peak FWHM ≈0.6 amu and for medium resolution ≈1.6 amu. Measurements with low and very low resolution yield FWHM of peaks ≈4.5 both, however with higher transmittance for the second one. (Regular m/Δm ratio can be hardly evaluated due to dynamic resolution behavior.) Photoelectron spectra (PES) were registered with a Perkin Elmer He I photoelectron spectrometer [45,46]. Photons with energy 22.21 eV ionize the studied molecules in the gas phase. The photoelectrons depart from the chamber through a narrow slit and are analyzed with a cylindrical electrostatic analyzer. In the present measurements, electrons pass through an analyzer at the fixed predefined energy, while the potential of the analyzer is varied with respect to the target chamber. An energy calibration was carried out by measuring known argon and xenon ionization potentials.

Instrumentation for complexes characteristics
The first mass spectra were detected with a Finnigan MAT 95 mass spectrometer, using electron ionization (EI) method over the temperature range 30