Chemical vapor deposition of ruthenium-based layers by a single-source approach †

A series of ruthenium complexes of the general type Ru(CO)2(P(n-Bu)3)2(O2CR)2 (4a, R = Me; 4b, R = Et; 4c, R = i-Pr; 4d, R = t-Bu; 4e, R = CH2OCH3; 4f, R = CF3; 4g, R = CF2CF3) was synthesized by a singlestep reaction of Ru3(CO)12 with P(n-Bu)3 and the respective carboxylic acid. The molecular structures of 4b, 4c and 4e–g in the solid state are discussed. All ruthenium complexes are stable against air and moisture and possess low melting points. The physical properties including the vapor pressure can be adjusted by modification of the carboxylate ligands. The chemical vapor deposition of ruthenium precursors 4a–f was carried out in a vertical cold-wall CVD reactor at substrate temperatures between 350 and 400 1C in a nitrogen atmosphere. These experiments show that all precursors are well suited for the deposition of phosphorus-doped ruthenium layers without addition of any reactive gas or an additional phosphorus source. In the films, phosphorus contents between 11 and 16 mol% were determined by XPS analysis. The obtained layers possess thicknesses between 25 and 65 nm and are highly conformal and dense as proven by SEM and AFM studies.


Introduction
The ongoing miniaturization of devices in semiconductor industry has introduced new manufacturing and materials challenges. 1 A promising candidate which can replace current materials in many microelectronic applications in future technology nodes is ruthenium, as it possesses a high thermal and chemical stability, a low electrical resistance and a negligible solid solubility with copper. 1,24][5][6][7] Thereby, one key destination is the development of a single material liner for future copper interconnects.][10] Due to its high surface energy, ruthenium follows a 3D (Volmer-Weber) growth mechanism and hence leads to the formation of polycrystalline, columnar films. 11These polycrystalline structures allow copper diffusion at unacceptably low temperatures through the grain boundaries. 12,13or this reason the need for the development of nanocrystalline or amorphous ruthenium-based films arises, e.g.obtained by incorporation of phosphorus. 14,15Such layers have shown to provide better copper diffusion barrier properties than pure polycrystalline ruthenium coatings. 166][17][18] In this dual-source approach, the films contain 10-50 mol% C with decreasing C contents in the order PPh 3 4 PMe 3 4 PH 3 . 13,15,17,18Furthermore, precise control and reproducibility of the film stoichiometry are difficult to achieve, because of the different vapor pressures and reactivities of both components. 18Much better reproducibilities are obtained in the single-source approach that uses precursors containing both elements in the same molecule.Moreover, a better homogeneity is provided as the desired elements are premixed at the molecular level. 20The only single-source precursor that has been examined in the deposition of Ru(P) films so far is air-sensitive RuH 2 (PMe 3 ) 4 . 11,16Therein, the undesired incorporation of C still results in a significant increase of the film resistivity and reduces its applicability as a directly plateable diffusion barrier for Cu. 13 The quality of the films deposited from RuH 2 (PMe 3 ) 4 can be improved by the addition of H 2 as a reactive gas. 15e herein present the synthesis of ruthenium precursors of the general type Ru(CO) 2 (P(n-Bu) 3 ) 2 (O 2 CR) 2 (R = Me, Et, i-Pr, t-Bu, CH 2 OCH 3 , CF 3 , CF 2 CF 3 ) and their use as single-source CVD precursors for the preparation of thin and conformal phosphorus-doped ruthenium layers. 21The influence of the carboxylate ligands on the thermal behavior and vapor pressure is discussed.The layers obtained were characterized by SEM, EDX, AFM and XPS measurements.

Instruments and materials
All synthesis procedures were performed under an atmosphere of argon with the solvents degassed prior to use.All reagents were obtained from commercial suppliers and used without further purification.For column chromatography, silica with a particle size of 40-60 mm (230-400 mesh (ASTM), Fa.Macherey-Nagel) was used.
NMR spectra were recorded using a Bruker Avance III 500 spectrometer operating at 500.3 MHz for 1 H, 125.7 MHz for 13 C{ 1 H} and 202.5 MHz for 31 P{ 1 H} in the Fourier transform mode at 298 K.Chemical shifts are reported in d (ppm) downfield from tetramethylsilane with the solvent as a reference signal ( 1 H NMR, CHCl 3 d 7.26; 13 C{ 1 H} NMR, CDCl 3 d 77.16; 31 P{ 1 H} NMR, standard external relative to 85% H 3 PO 4 d 0.0).FT-IR spectra were recorded using a Thermo Nicolet IR 200 instrument.The melting points were determined using a Gallenkamp MFB 595 010 M melting point apparatus.Elemental analyses were performed using a Thermo FlashAE 1112 instrument.Highresolution mass spectra were recorded using a Bruker Daltonite micrOTOF-QII spectrometer using electro-spray ionization (ESI †).
TG experiments were performed using a Mettler Toledo TGA/DSC1 1100 system with a UMX1 balance.CVD experiments were carried out using a home-built vertical cold-wall CVD reactor with a heater dimension of 20 Â 20 mm (BACH Resistor Ceramics GmbH).Heating could be adjusted up to 773 K and was controlled by a Gefran 600 module connected with a Pt100 thermosensor.The carrier gas (N 2 ) was controlled by MKS type 247 mass flow controllers connected to the reactor by heated copper lines.The CVD system was attached to a rotary vane pump RZ 6 (Vacuubrand).The pressure of the reactor system was controlled by a Vacuubrand vacuum controller (CVC 3000) in combination with an external Pirani vacuum sensor (VSP 3000).
The surface morphology was investigated by field-emission scanning electron microscopy using a ZEISS Supra60 SEM.Cross-sectional SEM investigations were carried out to determine the film thickness.All AFM measurements were accomplished using NanoWizard I and II devices (JPK Instruments AG, Berlin, Germany) under ambient conditions using JPK SPM Control Software provided by the manufacturer.RMS values were determined using JPK Image Processing.The measurements were carried out in tapping mode using silicon AFM tips with a cantilever length of 125 mm and a spring constant of approximately 42 N m À1 (Pointprobe s NCH, NanoWorld AG, Neucha ˆtel, Switzerland).Energy-dispersive X-ray analysis using a Bruker Quantax 400 system attached to a SEM was applied to determine the chemical composition of the films.The composition of the Ru samples was investigated using a PREVAC XPS system.Monochromatic aluminum Ka radiation (1486.6 eV) was provided by a VG Scienta MX 650 X-ray source and a monochromator.The energy distribution of the photoelectrons was measured using a VG Scienta EW3000 XPS/UPS/ARPES analyzer.This analyzer was operated at 200 eV pass energy with a step size of 100 meV and a measurement time of 3.0 s for each data point.Casa XPS 2.3.16Pre-rel 1.4 software was used for the deconvolution of the XPS peaks.For the calculation of the atomic concentration, Scofield relative sensitivity factors (RSFs) were used.These RSFs were corrected for a monochromator-analyzer angle of 52.551.For the escape depth correction in Casa XPS, a value of À0.7 was applied.XRPD patterns were collected on a STOE-STADI P diffractometer using Cu K a (1.5405 Å) radiation and a Ge(111) monochromator.
Diffraction data were collected with an Oxford Gemini S diffractometer using graphite-monochromated Mo K a radiation (4b,c,f,g) (l = 0.71073 Å) or Cu K a radiation (4e) (l = 1.54184Å) at 110 K using oil-coated shock-cooled crystals.The structures were solved by direct methods and refined by full-matrix leastsquares procedures on F 2 . 22,23All non-hydrogen atoms were refined anisotropically, and a riding model was employed in the refinement of the hydrogen atom positions.Graphics of the molecular structures were created by using SHELXTL 23 and ORTEP. 24ecursor synthesis and characterization General synthesis procedure for ruthenium complexes 4a-g.Synthesis of Ru(CO) 2 (P(n-Bu) 3 ) 2 (O 2 CCH(CH 3 ) 2 ) 2 (4c).The title compound was synthesized according to the general procedure by using isobutyric acid (220 mg, 2.50 mmol).The raw product was subjected to silica-gel column chromatography using a mixture of n-hexane/diethyl ether (ratio 1 : 1, v : v) as an eluent to give the respective product as a colorless, crystalline solid.Yield: 626 mg (851 mmol, 91%  Synthesis of Ru(CO) 2 (P(n-Bu) 3 ) 2 (O 2 CC(CH 3 ) 3 ) 2 (4d).The title compound was synthesized according to the general procedure by using pivalic acid (256 mg, 2.50 mmol).The raw product was subjected to silica-gel column chromatography using a mixture of n-hexane/diethyl ether (ratio 2 : 1, v : v) as an eluent to give the respective product as a colorless, crystalline solid.Yield: 493 mg (645 mmol, 69%).
The identities of all compounds were confirmed by elemental analysis, IR and NMR ( 1 H, 13 C{ 1 H}, 31 P{ 1 H}) spectroscopy and high resolution ESI mass spectrometry (Experimental section).The molecular structures of 4b, 4c and 4e-g in the solid state were determined by single crystal X-ray structure analysis.In addition, the thermal behavior was studied by vapor pressure and thermogravimetric measurements.
The 31 P{ 1 H} NMR spectra of ruthenium complexes 4a-g exhibit one characteristic singlet for the P(n-Bu) 3 groups ranging from 14.7 to 16.7 ppm (Experimental section).As the respective resonance signals are shifted to higher fields as compared to 2 (À30.9 ppm), 31 P{ 1 H} NMR spectroscopy is suitable to monitor the progress of the reactions.The 1 H and 13 C{ 1 H} NMR spectra of 4a-g are in accordance with the proposed structures.Unique to the 13 C{ 1 H} NMR spectra is the splitting of the aand b-CH 2 groups of the P(n-Bu) 3 ligand into triplets (Experimental section), which is a common phenomenon for complexes containing trans-phosphine ligands. 26,27This finding is confirmed by the calculations of Metzinger 28 and Harris. 29n the IR spectra of ruthenium complexes 4a-g, two strong stretching vibrations for the terminal carbonyl groups are observed between 1969 and 2057 cm À1 (Experimental section).The number of CO vibrations reveals that the carbonyl groups have to adopt a cis-arrangement in the octahedral coordination sphere of the Ru(II) ion, as for a trans isomer just one strong carbonyl stretching is expected. 30From the difference of the characteristic asymmetric (ũ asym ) and symmetric (ũ sym ) carboxylate stretching vibrations (Dũ, Dũ = ũasym À ũasym ) one can estimate the structural bonding motif of the carboxylate ligands. 31For ruthenium complexes 4a-g large Dũ values of over 250 cm À1 indicate a monodentate coordination of the carboxylate groups to the Ru(II) ion, which was confirmed by single X-ray structure determination (see below).
The structures of 4b,c and 4e-g in the solid state were determined by single crystal X-ray diffraction analysis.Suitable crystals were obtained from a concentrated n-hexane solution at 5 1C.The ORTEP diagrams of complexes 4b,c,e are presented in Fig. 1.The molecular structures of complexes 4f,g along with key structural data can be found in the ESI.† The crystal and structure refinement data are presented in the Experimental section.
All compounds crystallize in monoclinic space groups (4b: P2/c; 4c,h: I2/a; 4e: C2/c; 4f: P2 1 /c) with one molecule in the asymmetric unit, except for 4b and 4e with one half of the compound and a C 2 -symmetry axis through the Ru atom.The complexes consist of a slightly distorted octahedral coordinated ruthenium atom with the two trans-positioned tri-n-butylphosphines (P1 and P2) in the apical positions, two cis-oriented carbonyls and two cis-monodentate O-bonded carboxylates in the equatorial plane (Fig. 1).

Thermal behavior and vapor pressure measurements
To gain first information on the thermal behavior of the designed precursor complexes, compounds 4a-g were studied by thermogravimetric (TG) analysis.The physical properties of 4a-g are summarized in Table 1.The melting points of the complexes could be tuned by changing the carboxylate ligands.In this respect, the melting points decrease with the increasing chain length or branching of the carboxylate ligand from 108 1C in 4a to 48 1C in 4d.TG measurements were carried out in a nitrogen carrier gas flow of 60 mL min À1 .From the respective TG traces it can be seen that all compounds show a weight loss between 220 and 350 1C which results from an overlapped process of precursor evaporation and decomposition (Fig. 2).
Thereby, the appearance of the TG traces is strongly dependent on the applied heating rate (ESI †).A heating rate of 10 K min À1 was employed to ensure that the precursor does not evaporate completely and hence information on the decomposition temperature of the precursor can be received.A lower heating rate leads to the evaporation of the preferred compounds, which is beneficial for the vapor pressure measurements (see below). 32,33The TG traces also indicate that marginal changes in the periphery of the complex, for example, the modification of the carboxylate ligands, do not have a significant effect on the evaporation and decomposition process.This observation is also confirmed by the similarity of the respective onset temperatures ranging from 244 1C for 4b to 277 1C for 4f (Table 1).
Vapor pressure measurements of ruthenium complexes 4a-g were carried out over a temperature range from 100 to 240 1C under atmospheric pressure using a TGA system with a horizontal balance.The details of the applied method have been published previously. 2 All vapor pressure measurements were performed thrice to provide a statistic validation of the experimental data.From the respective vapor pressure traces it can be seen that the different carboxylate ligands influence the vapor pressure of the resulting precursor complex (Fig. 3).The highest volatility was observed for compound 4e, featuring a CH 2 OCH 3 substituent.As expected, the CF 3 groups in 4f also increase the volatility due Fig. 1 ORTEP diagrams (30% probability level) of the molecular structures of 4b (left), 4c (middle) and 4e (right, 50% probability level) with the atom numbering scheme.All hydrogen and disordered atoms have been omitted for clarity (symmetry code for 4b: A = Àx + 1, y, Àz + 3/2; for 4e: to the low polarizability of the C-F bonds, which leads to reduced intermolecular interactions between the complexes. 34owever, the prolonged CF 2 CF 3 substitution in 4g resulted in a decrease of the volatility.The vapor pressures of the remaining compounds 4a-d hardly differ from each other.In comparison to the previously reported vapor pressures of substituted ruthenocenes and half-open ruthenocenes, 2 the reported ruthenium complexes 4a-g possess quite similar volatilities.

Chemical vapor deposition experiments
Ruthenium complexes 4a-f have been successfully applied in CVD experiments for the deposition of phosphorus-doped ruthenium layers.The experiments were carried out in a home-built vertical cold-wall reactor equipped with a continuous evaporation system.Deposition was performed using nitrogen as a carrier gas (30-50 mL min À1 ; 0.6-0.8mbar working pressure).The ruthenium films were deposited on Si wafers, which were covered by a continuous 100 nm thick thermal SiO 2 layer.Additionally, precursor 4e was deposited on a trench patterned SiO 2 wafer in order to study the step coverage properties.
For appropriate evaporation of the precursor complexes, temperatures of 120-135 1C in the vaporizer unit were needed.Furthermore, the glass lines were heated to approximately 100 1C.The substrate temperatures were selected according to the results of the TG analyses.Decomposition of the precursor complexes starts at around 270 1C, but a minimum substrate temperature of 350 1C is required in all cases for complete precursor deposition and formation of ruthenium layers.Therefore, deposition was carried out using substrate temperatures between 350 and 400 1C.The MOCVD deposition parameters of the obtained layers A-H are summarized in Table 2.The deposition was monitored visually and terminated after the formation of metallic layers.Depending on the experimental conditions, 25-65 nm thick uniform layers were obtained.The 40-55 nm thin films are reflective metallic layers, whereas thinner films possess a slightly bluish appearance and thicker films show a yellowish tinge.

Layer characterization
The scanning electron microscopy (SEM) images of the films deposited from 4a-f evidence the formation of dense and conformal layers for all precursors (Fig. 4).In general, the morphology of the produced layers A-F is very similar, so that no optical differences in surface roughness or homogeneity could be observed.The film thicknesses between 25 and 65 nm were determined by cross-sectional SEM images (ESI †).Thereby, the highest growth rates were obtained for 4e and 4f, possessing the highest volatilities according to the vapor pressure measurements (Fig. 3).The deposition rate can be enhanced by higher deposition temperatures and carrier gas flow rates (layers E and F, Table 2).The step coverage property of ruthenium precursor 4e was investigated by the deposition on a trench patterned Si/SiO 2 wafer with an aspect ratio of 2.5 (ESI †). 35The SEM images reveal that a complete and conformal coverage of the patterned substrate was achieved for the deposition of ruthenium complex 4e (layer H, Table 2).
In order to study the surface roughness, the deposited ruthenium films A-F were examined by atomic force microscopy (AFM).A representative AFM image of a ruthenium film obtained from complex 4e is depicted in Fig. 5.In comparison, the AFM images taken for all samples show very similar structures with only minor variations in height.In all cases, the resulting layer topography is very homogeneous and characterized by wellinterconnected globular grains.However, the consistent bead-like shapes indicate that the actual structures are too small to be accurately imaged using a standard AFM tip.][38][39][40] The film composition was analyzed by energy-dispersive X-ray spectroscopy (EDX) and X-ray photoelectron spectroscopy (XPS).In EDX spectroscopic measurements, the presence of the characteristic pattern of ruthenium was found in all samples (ESI †).In addition, the presence of phosphorus, silicon, oxygen and carbon was detected.Silicon and parts of oxygen originate from the applied Si/SiO 2 substrate.Hence, the intensity of these signals decreases by reducing the energy of the electron beam (penetration depth) and increasing the film thickness.
In order to determine the film composition without penetration of the Si/SiO 2 wafer surface sensitive XPS measurements were carried out.Thereby no silicon could be detected.Instead the presence of ruthenium, phosphorus, carbon and oxygen was confirmed.It is also necessary to note that layer F (Table 2) is fluorine-free as no fluorine could be detected by EDX or XPS analyses.In liner materials, the absence of fluorine is of particular importance as it may penetrate into the oxide layer and hence lead to device damaging. 41PS measurements were performed on the surface as well as in the layer after argon ion sputtering (4.0 keV; 330 s) with the intention to remove surface contaminations (Fig. 6).The elemental contributions are summarized in Table 4. Due to the spin-orbit Fig. 4 SEM images (magnification: 80 000Â) of ruthenium films A-F deposited on SiO 2 using the parameters given in Table 2. A: 4a, B: 4b, C: 4c, D: 4d, E: 4e, F: 4f.Fig. 5 AFM height of layer E (Table 2) deposited from 4e.The RMS roughness value is 0.9 nm.interaction two peaks for Ru 3d were found (Ru 3d 3/2 at 284.2 eV and Ru 3d 5/2 at 280.0 eV), which are in agreement with XPS binding energy databases and correspond to metallic ruthenium. 42hese ruthenium peaks overlap with the C 1s peak at 284.5 eV, 43 preventing the direct measurability of the carbon content. 44o determine the carbon content a peak deconvolution was carried out.The applied deconvolution parameters regarding the peak shape and peak position are summarized in the ESI.† For all peaks a maximum full width at half maximum (FWHM) of 2.0 eV was allowed.Furthermore, the area ratio of both metallic and non-metallic Ru 3d 3/2 and Ru 3d 5/2 peaks was fixed to 2:3 and the peak separation between 3d 3/2 and 3d 5/2 to 4.2 eV. 45or the identification of peak splitting with respect to P 2p and O 1s a Gaussian-Lorentzian peak shape was applied with a Lorentzian contribution of 25%.The comparison of the surface and film composition reveals that the ruthenium and phosphorus contents increase in all layers, whereas the carbon and oxygen contents decrease.In general, the layers themselves consist of approximately 48-58 mol% ruthenium, 11-16 mol% phosphorus, 7-31 mol% carbon and 4-29 mol% oxygen (Table 4).The nature of the ruthenium exhibits that beside metallic species ruthenium oxide compounds are also present which can be attributed to RuO 2 (RuO 2 3d 5/2 at 281.0 eV). 43The RuO 2 in the surface layer most probably arises mainly from oxygen adsorption and surface oxidation of ruthenium during the transport of the sample from the CVD reactor to the vacuum system for XPS analysis.This estimation is supported by the observation that the oxygen content clearly decreases during sputtering and the films B-F then only contain 4-7 mol% oxygen.Since the deposition was carried out in an inert gas atmosphere, the remaining oxygen most likely originates from the dissociation of the carboxylate or CO ligands. 46,47For the phosphorus content different species in the layers compared to the surface were found (Fig. 6).On the film surface signals, which can be assigned to P(III) of undecomposed phosphine ligands (P 2p at 130.2 eV) 42 and P(V) of phosphine oxide (P 2p at 133.3 eV) 48,49 were observed.However, within the layers only one signal was monitored, which refers to elemental phosphorus (P 2p at 129.9 eV). 42Due to the small energy difference of P(0) and P(III) species of 0.3 eV, it cannot be excluded that P(III) is present within the layers. 42The carbon impurities arise from the catalytic properties of the ruthenium surface, which lead to C-H and C-C activation of the adsorbed ligands resulting in the incorporation of elemental carbon. 50,51o evaluate the crystallinity of the deposited films X-ray powder diffraction (XRPD) measurements were performed (Fig. S6, ESI †).The XRPD pattern shows some diffraction arising from the Si/SiO 2 substrate.As no peaks originating from the Ru(P) film could be determined the layers are most probably amorphous.

Conclusion
Ruthenium complexes Ru(CO) 2 (P(n-Bu) 3 ) 2 (O 2 CR) 2 (4a, R = Me; 4b, R = Et; 4c, R = i-Pr; 4d, R = t-Bu; 4e, R = CH 2 OCH 3 ; 4f, R = CF 3 ; 4g, R = CF 2 CF 3 ) were successfully applied as single-source precursors for the deposition of phosphorus-doped ruthenium layers by the CVD process.The ruthenium compounds were synthesized by a ''one-pot'' synthetic methodology reacting Ru 3 (CO) 12 with P(n-Bu) 3 and the respective carboxylic acid.All precursors are stable against air or moisture and allow a liquid precursor delivery in a continuous CVD process due to their low melting points.Additional advantages of 4a-g are that their thermal properties including melting points, vapor pressures and decomposition behaviors can easily be modified by the introduction of different carboxylate ligands.Deposition was carried out in a cold-wall CVD reactor at deposition temperatures between 350 and 400 1C in an inert gas atmosphere without the need of any additional phosphorus source or reactive gas.The highest growth rates were observed for 4e and 4f, which also exhibited the highest vapor pressures.All received films were dense and conformal as proven by SEM images and were also uniform in the deposition on patterned wafers with an aspect ratio of 2.5.Furthermore, they possess a very low surface roughness with an RMS value of approximately 1.0 nm as determined by AFM.The elemental composition of the layers was analyzed by EDX and XPS measurements.
In conclusion, ruthenium complexes 4a-f show promising properties for the fabrication of homogeneous and conformal  phosphorus-doped ruthenium layers, as all complexes are easy to synthesize, stable in air and do not need a reactive gas or an additional phosphorus source during deposition.The relatively high phosphorus contents of up to 16 mol% make the resulting layers attractive as single material liners for copper interconnects in the damascene process.In the future, the application of phosphine ligands with shorter alkyl chains will lead to significantly decreased carbon contents.Investigations concerning this matter are currently in progress.

Table 1
Physical properties of Ru precursors 4a-g

Table 2
Deposition parameters of Ru layers A-H deposited from 4a-f

Table 3
Film properties of Ru layers A-F from Table2 a Measured by AFM.

Table 4
Elemental contribution of layers A-F from Table2 a After argon ion sputtering (330 s, 4.0 keV).