Photoactivated Ru chemical vapor deposition using (η³-allyl)Ru(CO)₃X (X = Cl, Br, I): From molecular adsorption to Ru thin film deposition

Ruthenium has many applications from nanoelectronics to catalysis.^1–3 Recently, there has been increasing interest in using Ru as a sub-10 nm interconnect for nanoelectronic devices because Ru has better reliability and resistance than current Cu interconnects.³ Since nanopatterning at these length scales is very challenging, area selective deposition (ASD) techniques are being developed as deposition methods.^1,2 ASD is a method by which a desired material, such as Ru, is deposited only on target areas, the “growth surface,” while deposition does not occur on nontarget or “nongrowth” areas.⁴ To meet continued performance and scaling requirements of electronic devices and to accommodate new materials, there is a need to develop low temperature (<400 °C) ASD techniques for organic, dielectric, and metallic films. The deposition techniques used for ASD are typically chemical vapor deposition (CVD) or atomic layer deposition, and the substrate surfaces are modified through adsorption of layers, such as alkanethiolate self-assembled monolayers (SAMs).^4,5 CVD is an attractive technique to deposit a wide variety of materials because the deposition thickness can easily be controlled and it can be chemically selective.^6–8 However, traditional CVD processes utilize high temperatures (>200 °C) because reactive gaseous/and or surface species are needed for the deposition to proceed.^6–9 These temperatures are incompatible with most organic thin films, including SAMs. For example, alkanethiolate SAMs can “melt” at temperatures of >100 °C,^10,11 which can lead to reduced deposition selectivity.⁵ 

In this article, we employ photoassisted chemical vapor deposition (PACVD) or photochemical CVD to deposit Ru on organic thin films. In PACVD, photoinduced reactions dissociate the ligands from the precursor leading to the formation of reactive intermediates, e.g., coordinatively unsaturated metal complexes or free radicals, near room temperature.¹² Our previous studies investigated Ru PACVD on functionalized SAMs using (η³-allyl)Ru(CO)₃Br, CpRu(CO)₂Me, and (COT)Ru(CO)₃ as precursors.¹³ From those three compounds, three conclusions about the design of candidate precursors for Ru PACVD were drawn: (1) moderate quantum yields for carbonyl ligand loss (ϕ ≥ 0.4) are required for ruthenium deposition; (2) anionic polyhapto ligands such as cyclopentadienyl and allyl are less appropriate ligands since they were more difficult to remove than carbonyls, halides, and alkyls; and (3) acid-base reactions between the precursor and the substrate promoted deposition more effectively than nucleophilic reactions.

We now report the effect of the halide ligand on Ru PACVD on organic substrates using (η³-allyl)Ru(CO)₃X (X = Cl, Br, I). To assess the effect of the terminal functional group, three functionalized SAM substrates with —COOH, —OH, and —CH₃ terminal groups were employed. Among the three SAMs, hydroxyl- and carboxylic acid-terminated SAMs are generally reactive, but their intrinsic reactivity differs. The hydroxyl group is a better nucleophile while the carboxylic acid group is a stronger acid. In contrast, methyl-terminated SAMs have terminal groups that are generally unreactive.^10,11 The photochemistry of the precursor complexes (η³-allyl)Ru(CO)₃X (X = Cl, Br, I) has been previously reported.¹⁴ For all three halides, the primary photoprocess is the loss of a carbonyl group with ϕ ≥ 0.6, but the values are wavelength and halide ligand dependent. Additional differences between the compounds are their Ru—CO bond strengths. In the present study, these photochemical results¹⁴ enable the investigation of the effect of the quantum yield for carbonyl loss and relative Ru—CO bond strengths on Ru deposition. Our data show that Ru deposition is not strongly dependent on the quantum yield for carbonyl loss but is strongly dependent on secondary photoprocesses and the reaction of the resulting coordinatively unsaturated complexes with the SAM terminal groups.

The (η³-allyl)Ru(CO)₃X (X = Cl, Br, I) complexes were synthesized at the University of Florida as previously reported.¹⁴ Syntheses were conducted in an inert atmosphere (N₂) using standard Schlenk techniques. Allyl iodide and Ru₃(CO)₁₂ were purchased from Oakwood Chemical (Estill, SC) and used as received. Allyl chloride and allyl bromide were acquired from Acros Organics (Thermo Fisher Scientific, NJ) and used as received. Isooctane was purchased from Fisher Chemical and used as received. The (η³-allyl)Ru(CO)₃X complexes were purified by sublimation at 100–200 mTorr onto a 4 °C cold finger. Samples were packed in vials under an atmosphere of CO and shipped overnight on reusable cold packs to the University of Texas at Dallas. Samples were stored in a freezer at −20 °C until use. The samples remained unchanged as determined by IR and ¹H NMR spectroscopies for periods of greater than 1 year under these storage conditions.

1-Hexadecanethiol (99%; HDT) and 16-mercaptohexadecanoic acid (90%; MHA) were purchased from Sigma Aldrich (St. Louis, MO). 16-Hydroxy-1-hexadecanethiol (99%; MHL) was acquired from Frontier Scientific Inc. (Logan, UT). Ethanol (200 proof, undenatured) was obtained from Acros Organics (Fair Lawn, NJ). All were used without further purification.

The preparation of the functionalized alkanethiolate SAMs has been discussed in detail previously.^15–17 Briefly, ∼200 Å Cr followed by ∼1000 Å Au were thermally deposited onto silicon wafers (⁠$⟨111⟩$⁠; Addison Engineering Inc., San Jose, CA) using an e-beam evaporator (CHA Industries, Freemont CA). Well-ordered SAMs were formed by immersing the Au substrate into a 1 mM ethanolic solution of the desired functionalized alkanethiol (with —CH₃, —OH, or —COOH terminal groups) for 24 h at room temperature (22 ± 1 °C). Upon removal from the ethanolic solution, the samples were rinsed thoroughly with ethanol and dried under nitrogen gas. To ensure that the SAMs were free from chemical contamination, samples from each batch were characterized using single wavelength ellipsometry (Gaertner Scientific Corp., Skokie, IL), x-ray photoelectron spectroscopy (XPS), and time-of-flight secondary ion mass spectrometry (TOF SIMS).

The functionalized SAM samples were exposed to (η³-allyl)Ru(CO)₃X (X = Cl, Br, I) precursors in a vhv chamber with a base pressure of ≤4 × 10⁻⁸ Torr. The separate precursor doser had a base pressure of ≤8.9 × 10⁻⁶ Torr. During dosing, the (η³-allyl)Ru(CO)₃X precursors were heated to 65 (±3) °C and exposed to functionalized SAMs that were kept at room temperature. To prevent condensation of the (η³-allyl)Ru(CO)₃X precursors, the dosing line was also heated to 65 (±3) °C. The precursors were exposed to UV light from a 500 W Hg arc lamp (Newport Corporation, Stratford, CT) with and without a dichroic mirror (λ: 280–400 nm) in the gas phase above the sample surface. To minimize photooxidation of the functionalized SAMs, the light is introduced into the chamber parallel to the sample surface. To investigate the initial reaction of the (η³-allyl)Ru(CO)₃X (X = Cl, Br, I) with functionalized SAMs, a precursor dose of ∼10 000 L and a deposition time of 10 min were employed. To examine whether the PACVD process resulted in a metallic Ru film, two higher precursor doses were used ∼295 000 and ∼435 000 L (20 and 30 min exposure at higher pressures, respectively).

After deposition, all samples were analyzed using ex situ XPS and TOF SIMS. Each reaction condition was repeated at least three times and the results reported are representative of these experiments.

Ex situ x-ray photoelectron spectra were measured using a PHI VersaProbe II Scanning XPS Microprobe (Physical Electronics Inc., Chanhassen, MN) equipped with a monochromatic Al Kα source (E_b = 1486.6 eV). During analysis, the chamber pressure was maintained at ≤6 × 10⁻⁸ Pa (4.5 × 10⁻¹⁰ Torr). High-resolution photoelectron spectra were collected with an analysis angle of 45°, a pass energy of 23.5 eV, and a step energy of 0.1 eV. All spectra were collected using a charge compensation system equipped with an electron beam and an ion beam. The data were analyzed using CasaXPS 2.3.19 (RBD Instruments Inc., Bend, OR). The spectra shown are representative of the data obtained and were calibrated to the Au 4f_7/2 binding energy at 84.0 eV.

Time-of-flight secondary ion mass spectra were acquired using an ION TOF IV spectrometer (ION TOF Inc., Chestnut Hill, NY) equipped with a Bi liquid metal ion gun. The instrument contains three chambers: a load lock, a preparation chamber, and an analysis chamber. The preparation and analysis chambers were maintained at <4 × 10⁻¹⁰ mbar. The primary Bi⁺ ions had a kinetic energy of 25 keV and were contained in an ∼100 nm diameter probe beam. All spectra were acquired using an analysis area of (500 × 500) μm² and an ion dose of <<10¹⁰ ions cm⁻². The secondary ions were extracted using a potential of 2000 V and reaccelerated to 10 keV before reaching the detector. Each sample was measured in three different areas in positive and negative ion modes. The spectra shown are representative of the data obtained.

Calculations were carried out using Gaussian09.¹⁸ The exchange-correlation potential used was B3LYP, which is a hybrid potential that combines a portion of the exact exchange term calculated from the Hartree–Fock theory and the correlation term from other sources.¹⁹ For C, O, H, and Cl, the basis set employed was 6-311 + G(2d,p), which incorporates diffuse functions.^20–23 For Ru, the LANL2DZ basis set, which includes an effective core potential, was employed in the calculations.^24–26 For Br and I, the LANL2DZ basis set was also used but with additional diffuse and polarization functions.²⁷ All structures reported in this article are minima; frequency calculations have been made to confirm this in every case. Additionally, UV/Vis spectra were obtained using time-dependent DFT (TD-DFT)²⁸ and are in good agreement with experimental data.¹⁴ Furthermore, Gibbs free energies of reactions calculated using the LANL2DZ basis set for Br are within ±2 kJ mol⁻¹ of the calculated energies using the 6-311 + G(2d,p) basis set for Br.

The initial reaction of (η³-allyl)Ru(CO)₃Br with —COOH, —OH, and —CH₃-terminated SAMs was discussed in detail in Ref. 13 and is summarized here so that halide effects on PACVD using (η³-allyl)Ru(CO)₃X (X = Cl, Br, I) can be compared. We also note that the TOF SIMS spectra of —COOH-, —OH-, and —CH₃-terminated SAMs have been described in detail previously,^15–17 and here, we highlight the differences in the spectra observed upon (η³-allyl)Ru(CO)₃X (X = Cl, Br, I) exposure.

In agreement with previous studies,¹³ the XPS and SIMS data show that UV light is required for efficient deposition and that the deposition is strongly wavelength dependent. An example of the data is given in Fig. 1. After exposure of ∼10 000 L of (η³-allyl)Ru(CO)₃X (X = Cl, Br, I) on —COOH-terminated SAMs in the dark, we observe little or no Ru⁺ ion intensity [Figs. 1(a), 1(c), and 1(e)]. Using a 500 W Hg arc lamp that is fit with a dichroic mirror (λ: 280–400 nm), there is little Ru present on the surface; in the SIMS spectra, there is a small Ru⁺ ion peak. In contrast, after deposition using the Hg arc lamp with no dichroic mirror, Ru⁺ ions are clearly observed [Figs. 1(a), 1(c), and 1(e)]. Furthermore, the intensities of the Ru⁺ ions are largest using (η³-allyl)Ru(CO)₃Br, suggesting that more Ru-containing species are deposited from the bromide complex [Fig. 1(c)]. Consistent with Ru⁺ ion intensities, we also observe halide ions, Cl⁻, Br⁻, or I⁻, indicating that the precursor has either reacted or molecularly adsorbed on the surface [Figs. 1(b), 1(d), and 1(f)]. The data also show that the deposition is strongly dependent on the terminal group of the SAM. For all three precursors, Ru-containing species are deposited on —COOH- and —OH-terminated SAMs using UV light. Furthermore, the intensities of the Ru⁺ ions are higher after deposition on —COOH-terminated SAMs than on —OH-terminated SAMs, suggesting that more Ru-containing species are deposited on —COOH-terminated SAMs [Fig. 2(a)]. There is little or no Ru deposition on —CH₃-terminated SAMs; no Ru⁺ ions are observed in the mass spectra [Fig. 2(b)].

In the XPS spectra after deposition on —COOH-terminated SAMs, we observe a small broad Ru 3d_5/2 peak (shoulder) at ∼282 eV, which we assign to the adsorbed molecular Ru(II) precursor, (η³-allyl)Ru(CO)₃X (X = Cl, Br, I) [Fig. 3(b)].^29,30 The data also show that the Ru 3d_5/2 peak is largest using (η³-allyl)Ru(CO)₃Br showing that more Ru-containing species are deposited in agreement with the SIMS data (Fig. 1). We also observe that the —CH₂ peak (285.0 eV) has broadened on the high energy side indicating that the precursor has reacted with the —COOH-terminated SAM [Figs. 3(a) and 3(b)].¹⁶ After deposition on —OH-terminated SAMs, the XPS data also show that there is a Ru(II) species present indicative of molecular precursor adsorption; a small broad Ru 3d_5/2 peak (shoulder) at ∼282 eV is observed [Fig. 3(d)]. After exposure to (η³-allyl)Ru(CO)₃I, we note that the C 1 s/Ru 3d photoelectron peak is best fit to a single peak with a binding energy of 285.2 eV [Fig. 3(d)].The peak area of the Ru 3d_5/2 is also smaller than on the —COOH-terminated SAM, indicating that less Ru species are deposited. Additionally, the —CH₂₋ peak (285.2 eV)¹⁷ has broadened on the high binding energy side, indicating that the —OH terminal groups and the precursor have also reacted [Fig. 3(d)].¹⁷ In contrast for —CH₃-terminated SAMs, no Ru 3d peaks are observed showing that there is little or no deposition of Ru-containing species [Figs. 3(e) and 3(f)]. There is a small broad Ru 3d_5/2 peak at ∼282 eV upon exposure to (η³-allyl)Ru(CO)₃I [Fig. 3(f)]. However, the binding energy of the —CH₂₋ peak (285.0 eV)¹⁵ slightly changes, suggesting that there is a charge transfer between the functionalized SAM surface and the polar (η³-allyl)Ru(CO)₃X (X = Cl, Br, I) [Fig. 3(e)]. In the SIMS spectra, we observe ions indicative of the Ru metal center interacting with the hydroxyl and carboxylic acid terminal groups, such as RuOC(CH₂)_x^± and RuO(CH)_x(CH₂)_y⁺ (Fig. 4). For —COOH-terminated SAMs, ions of the form RuO₂C(CH₂)^± and RuO₂C(CO)CH₂⁻ are also present. Additionally, ions of the form Ru(CH)_x⁺, e.g., RuC₃H₃⁺, are detected showing that η³-allyl ligands are present on the surface (Fig. 4). Since there are also halide ions present on the surface (Fig. 1), this suggests that either the precursor has adsorbed after loss of carbonyl, which is the first step in the photodecomposition of (η³-allyl)Ru(CO)₃X (X = Cl, Br, I),¹⁴ or there is an unreacted precursor present on the surface. The intensities of these ions are dependent on both the SAM terminal group and the chemical identity of the precursor. In general, the intensities of these ions are weaker on —OH-terminated SAMs than on —COOH-terminated SAMs. These ion intensities are also much smaller after exposure to (η³-allyl)Ru(CO)₃X (X = Cl, I) than after deposition using (η³-allyl)Ru(CO)₃Br (data not shown).

As deposition continues, the XPS data show that Ru PACVD is strongly dependent on the chemical identity of the precursor and the terminal functional groups of the SAM.

For —CH₃-terminated SAMs, after 30 min deposition (∼435 000 L) in UV light, there is little or no Ru species deposited using (η³-allyl)Ru(CO)₃X (X = Cl, I) [Figs. 5(a), 5(e), and 6]. In the XPS spectra, we observe a very low intensity, broad Ru 3d_5/2 photoelectron peak at ∼281.3 eV, which is characteristic of a Ru(II) species^29,30 (Fig. 6) and is consistent with the adsorption of the precursor after loss of the carbonyl ligand and/or adsorption of the intact molecular precursor. Additionally, after (η³-allyl)Ru(CO)₃ I exposure in the O 1s spectra, there is a very small peak observed at ∼532.5 eV, which we assign to the carbonyl groups present in the precursor [Fig. 5(f)].³⁰ No Cl 2p and weak I 3d peaks (BE ∼620 eV) are observed. We note that light halogen (F and Cl) containing polymers are known to degrade with extended exposure to x rays,³¹ so it is unlikely that Cl 2p peaks will be observed in the photoelectron spectra. In contrast, using (η³-allyl)Ru(CO)₃Br, the data indicate that initially there is little Ru present on the surface [Fig. 5(c)]. However, after 20 min deposition, we observe Ru 3d_5/2 and Ru 3p_3/2 photoelectron peaks at 280.7 eV (Fig. 5) and 461.9 eV (not shown), indicating that Ru is present on the surface as Ru(II) species.^29,30 In the O 1s spectrum, there is a photoelectron peak at 532.3 eV, which is consistent with RuO_x.^29,30 After 30 min deposition, two Ru 3d_5/2 peaks are observed, with binding energies of 281.5 and 280.0 eV, indicating that both RuO_x and Ru(0) have deposited. Furthermore, the Ru 3p_3/2 peak has a binding energy of 461.8 eV, which is consistent with RuO_x and Ru(0) present on the surface.^29,30 We also note that the C 1s —CH₂₋ photoelectron peak (285.0 eV) broadens on the high binding energy side, which is assigned to the presence of ruthenium oxides and ruthenium on the surface.^29,30 In the O 1s spectrum, we observe that the binding energy decreases to 531.6 eV [Fig. 5(d)] and can be fit to three peaks at 530.4, 531.9, and 533.5 eV. The peak at 533.5 eV is assigned to C—O from the precursor, while the peaks at 530.4 and 531.9 eV are assigned to RuO_x.²⁹ 

For —OH-terminated SAMs, the Ru PACVD process appears to be similar to that observed on —CH₃-terminated SAMs for (η³-allyl)Ru(CO)₃Br, but using (η³-allyl)Ru(CO)₃X (X = Cl, I), the deposition process is different (Fig. 7). After 20 min exposure to (η³-allyl)Ru(CO)₃X (X = Cl, I), there are Ru 3d_5/2 and Ru 3p_3/2 with binding energies of 281.5 and 461.8 eV, respectively, which indicate that Ru(II) species are present and are consistent with the molecular adsorption of the precursor and deposition of RuO_x^29,30 [Figs. 7(a) and 7(e)]. Upon further exposure to (η³-allyl)Ru(CO)₃X (X = Cl, I), both Ru peak intensities increase but do not change binding energy, indicating that these species continue to deposit. The O 1s spectra are altered upon exposure to (η³-allyl)Ru(CO)₃X (X = Cl, I) [Figs. 7(b) and 7(f)]. Initially, the —CH₂OH O 1s peak is observed at 533.3 eV,¹⁷ and after deposition, the O 1s peak broadens and decreases in binding energy to 533.1 eV, which attribute to C—O from the precursor.²⁹ No Cl 2p peaks are observed after deposition, but upon exposure to (η³-allyl)Ru(CO)₃I, I 3d peaks are present with binding energies of 619.8 and 631.2 eV and are consistent with molecular adsorption of the precursor.³⁰ In contrast, using (η³-allyl)Ru(CO)₃Br after 20 min, the Ru 3d_5/2 peak has two components at 281.5 and ∼280 eV, which we attribute to Ru/RuO_x on the surface.^29,30 Furthermore, we also observe a Ru 3p_3/2 peak with a binding energy of 461.8 eV (data not shown) and an increase in intensity and broadening of the O 1s peak, which are also indicative of a Ru/RuO_x film. After 30 min, the Ru 3d_5/2 peak at 280.0 eV greatly increases, indicating the Ru(0) is being deposited.^29,30 There is also a second photoelectron peak at 281.5 eV, indicating that RuO_x is present on the surface. The O 1s and Ru 3p_3/2 photoelectron spectra are also consistent with the Ru 3d_5/2 spectrum.

On —COOH-terminated SAMs for all precursors studied, there is little or no Ru(0) deposited at longer deposition times (Fig. 8). After exposure to (η³-allyl)Ru(CO)₃X (X = Cl, I), we observe that the Ru 3d_5/2 peak increases with deposition time, and the binding energy remains constant at 281.5 eV, which is consistent with Ru(II) of the adsorbed molecular precursor.^29,30 Similar to other functionalized SAMs, no Cl 2p peaks are observed, but I 3d peaks are present, confirming that (η³-allyl)Ru(CO)₃X (X = Cl, I) molecularly adsorbs to the surface.³⁰ In the O 1s spectra, initially, there is a peak at 532.0 eV, which is assigned to —COOH.¹⁶ We note that the best fit to the photoelectron spectrum is to two peaks at 532.0 and 533.4 eV, which we assign to O—C=O and O=C—O.³⁰ Upon exposure to the precursors, the peak intensity decreases and the peak broadens indicating the molecular adsorption of the precursor.³⁰ Upon exposure to (η³-allyl)Ru(CO)₃Br, the data indicate that there is a small increase in the Ru 3d_5/2 peak intensity. After 20 min, the Ru 3d_5/2 peak has a binding energy at 281.4 eV, which is consistent with Ru in the +2 oxidation state.^29,30 As the deposition continues, the Ru 3d_5/2 peak has two components with binding energies of 281.5 and 280.0 eV, which are assigned to Ru(II) and Ru(0) and are consistent with the deposition of Ru/RuO_x.^29,30 There are also two Ru 3p_3/2 binding energies at 462.0 and ∼465 eV, which are assigned to Ru and RuO_x, respectively. Finally, the O 1s peak (532 eV) broadens on the low binding energy side and increases in intensity, indicating that ruthenium oxides are present (Fig. 8).

Deposition experiments performed in the dark also support the interpretation that most, if not all, of the (η³-allyl)Ru(CO)₃X (X = Cl, I) adsorbs molecularly on the functionalized SAMs. After 20 min deposition (∼295 000 l) of (η³-allyl)Ru(CO)₃I in the dark, we observe Ru 3d_5/2 (∼281.3 eV), I 3d (619.8 and 631.2 eV), and O1s (532.5 eV) photoelectron peaks on all SAMs, which are indicative of the molecular adsorption of the precursor (Fig. 9). In the I 3d photoelectron spectra, there is also a broad peak at ∼645 eV [Fig. 9(c)], which is shake-up/shake-off satellites.³² 

The XPS and SIMS data clearly show that Ru PACVD is dependent on both the chemical identities of the precursor and the SAM terminal group. Any reaction pathway must account for the following observations:

1. (η³-allyl)Ru(CO)₃X (X = Cl, I) adsorbs molecularly on all functionalized SAMs;

2. On —CH₃-terminated SAMs, there is very little deposition of (η³-allyl)Ru(CO)₃Cl;

3. Initially, more Ru species are adsorbed on —COOH-terminated SAMs, but at later deposition times, little Ru-containing adsorbates are observed;

4. On —CH₃-terminated SAMs, no interaction is observed between the precursor and the SAM terminal group during initial stages of deposition; and

5. On —CH₃- and —OH-terminated SAMs, Ru(0) deposition is observed using PACVD with (η³-allyl)Ru(CO)₃Br.

An effective PACVD process is dependent on both the formation of coordinatively unsaturated intermediates generated during the gas phase process and the facile reaction of these intermediates with the SAMs. In the deposition chamber, the exposure time to the precursor is relatively short, so high quantum yields are required for efficient deposition. Brewer et al.¹⁴ measured the quantum yields (ϕ) for the primary photoprocess of (η³-allyl)Ru(CO)₃X (X = Cl, Br, I). The results showed that ϕ_Cl > ϕ_Br > ϕ_I for loss of a single carbonyl ligand and can be related to the strength of the Ru—CO bond as evaluated by trends in the carbonyl stretching frequencies for the series. In the IR spectra, the ν_CO frequency decreased from (η³-allyl)Ru(CO)₃Cl to (η³-allyl)Ru(CO)₃I due to more π backbonding to the CO, which increases the metal-carbon bond strength in the complexes with lower period halogens.¹⁴ This suggests that the (η³-allyl)Ru(CO)₃Cl precursor will be the most reactive toward ligand loss and the (η³-allyl)Ru(CO)₃I precursor the least reactive. In agreement with this, DFT calculations show that the Ru—CO bond energy is the smallest for (η³-allyl)Ru(CO)₃Cl and highest for (η³-allyl)Ru(CO)₃I (Fig. 10). Thus, PACVD is very inefficient using (η³-allyl)Ru(CO)₃I both in the dark and under UV light because the Ru—CO bonds do not readily dissociate at room temperature.

Our experimental results indicate that Ru PACVD using (η³-allyl)Ru(CO)₃Cl is far less efficient than the quantum yield measurements and calculated Ru—CO bond energies would imply. Instead, there is little molecular precursor adsorption on the functionalized SAMs. DFT calculations suggest that a small proportion of (η³-allyl)Ru(CO)₃Cl can decompose thermally to form (η³-allyl)Ru(CO)₂Cl and CO, but the back reaction is strongly thermodynamically favored (Fig. 9). This is consistent with the observation that (η³-allyl)Ru(CO)₃Cl underwent a slow thermal reaction in the dark with trimethylphosphite P(OCH₃)₃.¹⁴ Upon UV exposure, the (η³-allyl)Ru(CO)₃Cl initially decomposes to form a coordinatively unsaturated reactive intermediate, (η³-allyl)Ru(CO)₂Cl, which can then undergo a secondary photoprocess to form (η³-allyl)Ru(CO)Cl or react in the gas phase with a second molecule of (η³-allyl)Ru(CO)₃Cl, leading to the formation of an unreactive Cl-bridged dimer, [(η³-allyl)Ru(CO)₂Cl]₂. Our DFT calculations indicate that the formation energies for [(η³-allyl)Ru(CO)₂Cl]₂ and (η³-allyl)Ru(CO)Cl are accessible under our experimental conditions (Fig. 10). However, neither of these species are likely to persist because the back reactions are strongly thermodynamically favored. In agreement with this hypothesis, experimental data indicate that dimer formation occurs in solution upon exposure to UV light at long times. However, the dimer is unstable and readily decomposes to form (η³-allyl)Ru(CO)₃Cl and unidentified Ru species. Taken together, the data suggest that under PACVD conditions, the predominant gaseous species is the unreacted precursor (η³-allyl)Ru(CO)₃Cl, which weakly adsorbs on the functionalized SAM surfaces but does not undergo conversion to a Ru(0) deposit (Figs. 5–8).

For PACVD using (η³-allyl)Ru(CO)₃Br, we observe that Ru/RuO_x is deposited after 30 min on —OH- and —CH₃-terminated SAMs in UV light, but little Ru species are deposited on —COOH-terminated SAMs (Figs. 5–7). Furthermore, little or no Ru deposition is observed after deposition in the dark using (η³-allyl)Ru(CO)₃Br (Figs. 1 and 2). The strong dependence of deposition on light exposure indicates that the Ru—CO bond strength in (η³-allyl)Ru(CO)₃Br is sufficiently high that it does not decompose in the dark as proposed for (η³-allyl)Ru(CO)₃Cl. However, upon exposure to UV light, (η³-allyl)Ru(CO)₃Br decomposes to the coordinatively unsaturated complex, (η³-allyl)Ru(CO)₂Br. TD-DFT calculations also indicate the (η³-allyl)Ru(CO)₂Br also absorbs light at similar wavelengths to (η³-allyl)Ru(CO)₃Br and so can further photolyze to (η³-allyl)Ru(CO)Br if there is a sufficient population of the dicarbonyl complex in the gas phase for absorption. Loss of a second carbonyl from (η³-allyl)Ru(CO)₃Br has a similar calculated reaction energy to the first loss of the CO ligand (Fig. 10). We note that both DFT calculations and experiments show that (η³-allyl)Ru(CO)₂Br can also react to form a dimeric species. However, the dimer is also unstable with respect to the formation of the precursor, (η³-allyl)Ru(CO)₃Br, and unidentified Ru species. Thus, the dominant deposition pathway is the loss of two carbonyl ligands to form the coordinatively unsaturated complex (η³-allyl)Ru(CO)Br followed by reaction with the functionalized SAM surfaces leading to the deposition of Ru(0).

The data also indicated that the doubly unsaturated intermediate (η³-allyl)Ru(CO)Br (or possibly the singly unsaturated (η³-allyl)Ru(CO)₂Br) reacts with the —OH and —COOH SAM terminal groups (Fig. 4). In the SIMS spectra, we observe ions of the form, RuOC(CH₂)_x^± and RuO(CH)_x(CH₂)_y⁺, indicating that the metal center of the (η³-allyl)Ru(CO)Br interacts with one terminal group. However, the data also show that there is little Ru deposition on —COOH-terminated SAMs (Fig. 8). For —COOH-terminated SAMs, in the SIMS spectra, we also observe ions of the form RuO₂C(CH₂)^± and RuO₂C(CO)CH₂⁻, indicating that the Ru metal center of the gaseous (η³-allyl)Ru(CO)_xBr (x = 1, 2) forms a carboxylate complex with some terminal groups leading to an unreactive species on the surface. Hence, as deposition progresses, on —COOH-terminated SAMs, the CVD reaction cannot proceed on the unreactive carboxylate complexes and no Ru(0) is deposited. On —OH-terminated SAMs, the reaction proceeds via the reaction of the gaseous (η³-allyl)Ru(CO)_xBr (x = 1–3) with the Ru-surface complexes. It is possible that the difference in behavior between the —COOH- and —OH-terminated SAMs is the ability of the carboxylate functional group to chelate as a bidentate ligand, occupying two coordination sites and shutting down reactivity of the Ru center. The alcohol can only be a monodentate ligand and reaction of the resulting Ru species eventually yields Ru(0)/RuO_x. For —CH₃-terminated SAMs, the deposition proceeds via the adventitious adsorption of the reactive (η³-allyl)Ru(CO)Br complexes on the sample surface followed by subsequent reaction with further (η³-allyl)Ru(CO)_xBr (x = 1–3) impinging on the substrate.

These studies provide the proof-of-concept that photochemical CVD can be employed to deposit metals on thermally sensitive substrates, such as SAMs, and also give insights into its use for low-temperature ASD. Our data show that using (η³-allyl)Ru(CO)_xBr that Ru(0) (and RuO_x) can be deposited at room temperature on functionalized SAMs with no or very little damage to the SAM. We also demonstrate that the amount of Ru deposited is dependent on the SAM terminal group chemistry, suggesting that selective deposition can occur using photochemical CVD; more Ru is deposited on —OH- and —CH₃-terminated SAMs than on —COOH-terminated SAMs. However, other blocking layer chemistries may need to be developed because current ASD techniques often employ —CH₃-terminated SAMs as the nongrowth or blocking layer.^4,5 For —COOH-terminated SAMs, Ru deposition is significantly reduced because (η³-allyl)Ru(CO)_xBr forms a nonreactive carboxylate complex with some terminal groups and suggests that the use of complex formation may provide a new route for blocking deposition. Although preliminary studies are encouraging, further improvements in the selectivity of photochemical CVD will be required for ASD.

The efficacy of Ru PACVD using (η³-allyl)Ru(CO)₃X (X = Cl, Br, I) on functionalized alkanethiolate SAMs is not strongly dependent on the quantum yield of the precursor but rather is strongly dependent on secondary photoprocesses and the interaction of the resulting coordinatively unsaturated precursor complexes with the SAM terminal groups. For (η³-allyl)Ru(CO)₃Cl, the bond energy of Ru—CO is smallest, while for (η³-allyl)Ru(CO)₃I, the bond energy of Ru—CO is highest. Consequently, (η³-allyl)Ru(CO)₃I molecularly adsorbs on all SAM substrates investigated because in UV light, the complex does not readily decompose due to its high Ru—CO bond energy. In contrast, for (η³-allyl)Ru(CO)₃Cl, the precursor does readily decompose to (η³-allyl)Ru(CO)₂Cl, but the back reaction is strongly thermodynamically favored. Thus, (η³-allyl)Ru(CO)₃Cl also molecularly adsorbs and no Ru(0) is deposited.

For (η³-allyl)Ru(CO)₃Br at long deposition times under UV light, Ru(0) and RuO_x are deposited on —CH₃- and —OH-terminated SAMs. In contrast, for —COOH-terminated SAMs, little or no Ru is deposited. DFT calculations show that the secondary photoprocess to form (η³-allyl)Ru(CO)Br has a similar reaction energy to the loss of the initial CO ligand. The results are consistent with the reaction of the resulting (η³-allyl)Ru(CO)_xBr complex (x = 1, 2) with —OH and —COOH terminal groups. On —OH-terminated SAMs, the surface complex acts as a nucleation site for the deposition of Ru(0). In contrast, for —COOH-terminated SAMs, an unreactive coordinatively saturated carboxylate complex forms and so the deposition cannot proceed. For —CH₃-terminated SAMs, the deposition proceeds via the adventitious adsorption of (η³-allyl)Ru(CO)Br as the nucleation sites for Ru(0) deposition.

Together, these data suggest that PACVD can be employed for ASD by exploiting the interaction of the precursor with the substrate functional groups and the reactivity of the precursor, as evidenced by the quantum yield for ligand loss(es).

We gratefully acknowledge support from the National Science Foundation (NSF) [Grant Nos. DMR 1609081 (A.V.W.), DMR 1608873 (L.M.-W.), and CHE 1708259 (A.V.W.)].

The data that support the findings of this study are available from the corresponding authors upon reasonable request.