JP7503547B2 - Selective deposition of metal silicides and selective removal of oxides - Google Patents

Description

Embodiments of the present disclosure generally relate to methods for depositing metal silicide and selectively etching native silicon oxide.

Precise placement of materials on nanoscale devices is critical for engineering atomic-scale properties for next-generation nanoelectronics. Semiconductor manufacturing utilizes detailed placement of materials with excellent conformality and stoichiometry to meet cost, yield, and throughput demands. As channel lengths in metal-oxide-semiconductor field-effect transistors (MOSFETs) continue to shrink, it is desirable to overcome limitations imposed by top-down processes, such as reactive ion etching damage and structural complexity in structure alignment.

Recently, as MOSFET devices are fabricated in 3D structures (FinFETs), there has been an increased interest in nanoscale area selective deposition while maintaining conformal film quality. One approach to area selective deposition is to utilize self-assembled monolayers (SAMs) as passivation layers in combination with atomic layer deposition (ALD) processes. The passivation layer blocks or eliminates surface functional groups that are reactive to the ALD precursors, so selectivity can be obtained. However, the SAM approach still utilizes selective deposition of a passivation layer. Furthermore, the passivation layer is selectively removed after selective deposition, which forces the process to become more complex and reduce throughput.

Furthermore, to enable highly selective area deposition, native oxide material must be removed to expose underlying materials for selective deposition thereon. However, at advanced nodes, removal of native oxide becomes increasingly complex and selective when other oxide materials are present on the substrate in addition to the native oxide material.

Thus, there is a need in the art for improved methods for selective deposition of materials and selective removal of oxides.

In one embodiment, a substrate processing method is provided, which includes heating a substrate having a silicon _- containing surface to a first temperature, exposing the substrate to a plasma containing hydrogen, exposing the substrate to a first supply of _MoF6 precursor, and exposing the substrate to a second supply of _Si2H6 precursor. The exposing of the substrate to the first supply and the exposing of the substrate to the second supply are cycled sequentially, and after each successive cycle, the substrate is exposed to a third supply of _{Si2H6 precursor} .

In another embodiment, a method for processing a substrate is provided. The method includes placing a substrate on a heater in a reaction chamber having a chamber wall, heating the substrate on the heater to a first temperature, maintaining the chamber wall at a second temperature less than the first temperature, and exposing a silicon-containing surface of the substrate to hydrogen. The substrate is exposed to a first supply of _MoF6 precursor, the substrate is exposed to a second supply of _Si2H6 precursor, and the substrate is exposed to the first supply and the substrate is exposed to the second supply sequentially _{cycled ,} and after the successive cycles, the substrate is exposed to a third supply of _Si2H6 precursor.

In yet another embodiment, a method for processing a substrate is provided, which includes heating a substrate to a first temperature, exposing a silicon-containing surface of the substrate to a hydrogen-containing plasma, exposing the substrate to a first supply of _MoF6 precursor, and exposing the substrate to a second supply _{of Si2H6} precursor. The exposing of the substrate to the first supply and the exposing of the substrate to the second supply are cycled sequentially, and after the successive cycles, the substrate is exposed to a third supply of _Si2H6 precursor, and after the exposing of the substrate to the third supply, the substrate is annealed at _a second temperature between about 500°C and about 550°C.

The patent or application file contains at least one drawing executed in color. Copies of this patent or patent application publication with color drawing(s) will be provided by the Office upon request and payment of the necessary fee.

So that the above-mentioned features of the present disclosure can be understood in detail, a more particular description of the present disclosure, briefly summarized above, will be obtained by reference to embodiments, some of which are illustrated in the accompanying drawings. It should be noted, however, that the accompanying drawings show only exemplary embodiments and therefore should not be considered as limiting the scope of the present disclosure, as other equally effective embodiments may be permitted.

1 shows X-ray photoelectron spectroscopy (XPS) data of MoSi _x film selectivity on silicon substrate according to embodiments described herein. 1 shows XPS data of MoSi _x film selectivity on silicon oxynitride substrate according to embodiments described herein. 1 shows XPS oxidation state data for Si and Mo on a silicon substrate according to embodiments described herein. 1 depicts XPS oxidation state data for Si and Mo on a silicon substrate according to embodiments described herein. 1 shows XPS chemical composition data of various elements present on different substrate types prior to ALD processing, according to embodiments described herein. 1 shows XPS chemical composition data of various elements present on different substrate types after five ALD cycles according to embodiments described herein. 13 shows XPS chemical composition data of various elements present on different substrate types after additional ALD cycles according to embodiments described herein. 1 shows XPS chemical composition data of various elements present on different substrate types prior to ALD processing, according to embodiments described herein. 1 shows XPS chemical composition data of various elements present on different substrate types after five ALD cycles according to embodiments described herein. 4C shows XPS chemical composition data of the substrate of FIG. 4B after an annealing treatment according to embodiments described herein. 13 shows XPS depth profiling data of MoSi _x films after Ar sputtering according to embodiments described herein. 1 shows XPS chemical composition data for MoSi _x films according to embodiments described herein. 1 shows data depicting chemical composition of MoSi _x films versus time according to embodiments described herein. 13 shows XPS depth profiling data of MoSi _x films after Ar sputtering according to embodiments described herein. 1 shows surface composition data of MoSi _x films according to embodiments described herein. 6C shows bulk composition data for the MoSi _x film of FIG. 6B according to embodiments described herein. 1 shows data depicting chemical composition of MoSi _x films versus time according to embodiments described herein. 1 is a cross-sectional tunneling electron micrograph (TEM) of a MoSi _x film selectively deposited on silicon relative to other materials present on the substrate according to embodiments described herein. 1 is a graph illustrating selective etching of native silicon oxide relative to bulk silicon oxide according to embodiments described herein. 2 is a cross-sectional schematic diagram of a portion of a contact structure according to an embodiment described herein.

To facilitate understanding, wherever possible, identical reference numbers have been used to designate identical elements common to the figures. It is contemplated that components and features of one embodiment may be beneficially incorporated in other embodiments without further description.

The embodiments described herein include methods that utilize the substrate-dependent reactivity of ALD precursors for area-selective deposition. More specifically, the embodiments of the present disclosure relate to selectively depositing MoSi _x on Si in preference to SiO ₂ , SiON, and SiN _x using the substrate selectivity of MoF ₆ and Si ₂ H _6. To achieve stoichiometric MoSi ₂ films, additional Si was introduced into the film after the ALD cycle of MoF ₆ and Si ₂ H ₆ by supplying Si ₂ H ₆ to the Mo-rich MoSi _x film. The methods described herein also provide selective removal of native oxide, allowing removal of native oxide material without etching the bulk oxide material.

Highly selective deposition of MoSi _x on Si in preference to SiO ₂ and SiN _x was achieved via atomic layer deposition (ALD) using MoF ₆ and Si ₂ H ₆ precursors. The selectivity of the deposition was made possible by the lack of chemical reactivity between the reactants (MoF ₆ and Si ₂ H ₆ ) and the SiO ₂ and SiN _x containing substrates. In contrast, MoF ₆ nucleated on H-terminated Si in a self-limiting manner, followed by exposure to Si ₂ H ₆ reduced MoF _x to Mo ⁰ , consistent with Mo-Si bond formation.

X-ray photoelectron spectroscopy (XPS) revealed that five ALD cycles of _MoF6 and _Si2H6 selectively deposited substoichiometric _MoSi2 films on Si substrates. In the ALD process, _MoF6 and _{Si2H6 precursors} were cycled sequentially, with repeated purging between each successive precursor exposure. By supplying additional _Si2H6 to the substoichiometric MoSi2 film, _more Si was introduced into the film without interfering with the deposition selectivity to _SiO2 and _SiNx . In one embodiment, the bulk of _{the MoSi x film} has a Si:Mo ratio between about _1.7 and about 1.9, and has less than about 10% F and O impurities. It is believed that the embodiments described herein are more advantageous for the formation of silicide materials, for example in the formation of source/drain contact structures, than conventional high pressure Si ALD cycles.

The deposition selectivity of MoSi _x was analyzed on patterned Si substrates containing three-dimensional (3D) nanoscale SiO ₂ and SiN _x features according to embodiments described herein. Cross-sectional transmission electron microscopy (TEM) showed that selective deposition of MoSi _x was achieved on nanoscale 3D structures. In one embodiment, there are less than about 10 nuclei per μm ² on SiO ₂ . Since SiO ₂ has approximately 10 ⁷ /μm ² of OH groups, this corresponds to a selectivity of about 10 ⁷ :1 between OH groups on SiO ₂ and Si-H groups on Si. Thus, it is believed that the substrate-dependent selectivity for silicide deposition allows for the elimination of the use of a passivant (i.e., a SAM).

Experimental Various substrate types were used for the MoSi _x silicide formation process described herein. Four types of substrates were used: P-type Si(100), thermally grown SiO ₂ /Si(100), SiON, and patterned substrates with S, SiO ₂ and SiN _x material surfaces on a single substrate. The SiON (silicon oxynitride) described herein is Si ₃ N ₄ unless otherwise stated, which was subjected to reactive ion etching and plasma ashing in oxygen during fabrication. Thus, the SiON substrate contains oxygen similar to the state of Si ₃ N ₄ after processing with integrated 3D nanoscale devices.

The substrates were diced into 12 mm x 3 mm pieces and degreased with acetone, methanol, and deionized (DI) _H2O . The degreased substrates were immersed in 0.5% HF(aq) solution for 30 seconds to remove the native oxide of Si. For consistency of the cleaning procedure, the _SiO2 , SiON, and patterned substrates were subjected to the same cleaning procedure. In one embodiment, the native oxide removal process is the SICONI® pre-clean process available from Applied Materials, Inc., Santa Clara, Calif., USA.

It is also contemplated that plasma-based native oxide removal processes may be used. For example, _NF3 / _H2 and/or NF3/NH3 plasma cleaning processes may be used to clean and hydrogen terminate the silicon-containing surface of the substrate. On SiON substrates, _NF3 plasma treatment is believed to prevent or substantially reduce the loss of deposition selectivity due to passivation of active hydroxyl nucleation sites.

8 is a graph 800 showing the rate of selective etching of native silicon oxide thickness and bulk silicon oxide thickness as a function of time during plasma processing. Data 802 represents the bulk silicon oxide thickness when exposed to a _NF3 / _NH3 plasma. Data 804 represents the native silicon oxide thickness when exposed to a _NF3 / _NH3 plasma. Time 806 represents when the _NF3 / _NH3 plasma is turned on and time 808 represents when the _NF3 / _NH3 plasma is turned off.

In one embodiment, the plasma for selectively etching native silicon oxide relative to bulk silicon oxide is formed in-situ in the process chamber. Alternatively, the plasma for selectively etching native silicon oxide relative to bulk silicon oxide is formed remotely, for example by a remote plasma source, prior to delivery to the process chamber. Precursors used to form the plasma include _NF3 and _NH3 . In one embodiment, an inert gas such as Ar is used to facilitate delivery of active species to the substrate for selective removal of native silicon oxide.

In one embodiment, the ratio of _NF3 : _NH3 is between about 1:5 and about 1:20, e.g., about 1:10. In an embodiment using an Ar carrier gas, Ar is provided in an amount greater than _NF3 but less than _NH3 . For example, the ratio of _NF3 : _NH3 :Ar is 1:10:1.5. The pressure of the processing chamber environment in which the selective native oxide removal process is performed is between about 10 mTorr and about 1,000 mTorr, e.g., between about 100 mTorr and about 500 mTorr, e.g., about 200 mTorr. In one embodiment, the pressure is about 190 mTorr. The power used to generate the plasma is between about 10 W and about 500 W, e.g., between about 50 W and about 250 W, e.g., about 100 W. The temperature of the environment in which the native oxide removal process is carried out is between about 30°C and about 70°C, such as between about 40°C and about 50°C, for example about 45°C.

At time 806, the plasma is initiated and the native silicon oxide 804 experiences a reduction in thickness as indicated by a decrease in the thickness of the native silicon oxide material. In one embodiment, the plasma treatment is performed for a time period of less than 1 minute, for example, less than 40 seconds, such as between about 15 seconds and about 30 seconds. Within the first minute of plasma exposure, the native silicon oxide 804 is etched and the bulk silicon oxide experiences substantially no reduction in thickness. This indicates a high degree of selectivity for removing the native silicon oxide preferentially over the bulk silicon oxide. It is also believed that the native oxide removal process is selective to silicon nitride material as well, such that the native silicon oxide is removed preferentially over the silicon nitride.

Atomic force microscopy analysis of the substrate after selective removal of the native silicon oxide revealed that the exposed silicon surface (where the native silicon oxide was removed) exhibited sub-Angstrom surface roughness. Since etching of silicon material is expected to roughen the surface, such roughness is consistent with the underlying silicon material not being etched or not substantially etched after removal of the native oxide.

In some embodiments, residual materials such as ( _NH4 ) ₂ ) _SiF6 salts may remain on the substrate after performing the native oxide selective removal process. An optional annealing process is performed to remove the salts. In one embodiment, the annealing process is between about 80°C and about 160°C, e.g., about 100°C and about 140°C, e.g., about 120°C. The annealing is believed to remove the salts from the surface of the substrate, e.g., the silicon surface of the substrate, e.g., by volatilizing the salts.

9 is a cross-sectional schematic diagram of a substrate 900 having a contact structure 910 formed thereon, according to embodiments described herein. The substrate 900 includes a silicon material film 902 and a bulk silicon oxide 904 formed on the silicon material film 902. The contact structure 910 is formed on a surface 906 of the silicon material film 902. Prior to selective removal of the native oxide, the surface 906 has a thin layer of native oxide formed thereon. Using the above embodiments, the native oxide is removed from the surface 906 without substantially altering or removing the bulk silicon oxide 904 or the underlying silicon film material 902.

A contact structure 910 formed on the surface 906 includes a gate 916 surrounded by a gate oxide 914, a spacer 918, and a cap 920. In one embodiment, the gate 916 is a metal-containing material. The spacer 918 and the cap 920 include a nitride-containing material, such as a silicon nitride material. The native oxide selective removal process described herein can be used before or after the formation of the contact structure 910 to prepare the surface 906 for subsequent metal deposition. The metal deposition in the channel 912 formed between adjacent contact structures 910 extends from the surface 906 toward the cap 920. Selective removal of the native oxide from the surface 906 improves metal adhesion to the underlying silicon material film 902.

After removal of the native oxide, the substrates were blown dry using high purity _N2 gas. The Si, _SiO2 , SiON, and patterned substrates were loaded together on a single substrate holder to expose the substrates to the same ALD conditions. The substrates were loaded into a load lock chamber pumped by a turbomolecular pump and backed by a mechanical pump. The base pressure of the load lock was about 2.0× ^10-7 Torr. The substrates were then transferred in-situ to an ultra-high vacuum chamber with a base pressure of about 3.0× ^10-10 Torr. The chamber was pumped by an ion pump and a titanium sublimation pump. The ultra-high vacuum chamber was equipped with a monochromatic XPS instrument, a scanning tunneling microscope (STM), and an annealing system using a pyrolytic boron nitride (PBN) heater.

The substrates were first annealed at 120° C. in an ultra-high vacuum chamber and XPS was used to determine the chemical composition of the substrates. The substrates were transferred in-situ into a reaction chamber with a base pressure of about 5.0×10 ⁻⁷ Torrn. MoF ₆ (99% purity) and Si ₂ H ₆ (99.99% purity) precursors were used for the deposition of MoSi _x .

A constant purge of _N2 (80 mTorr) was used during _the ALD cycle, and the pressure of this purge was controlled using a leak valve. The supply of _MoF6 and _Si2H6 was controlled using a pneumatic valve. An expansion space was used to supply _MoF6 and _Si2H6 . The use of the expansion space _includes filling the secondary space with _MoF6 or _Si2H6 and supplying the precursor from the respective secondary space. The filling time of _MoF6 is between about 10 ms and about 10 ms, e.g. _, about 40 ms. The supply time of MoF6 is between about 10 ms and about 100 ms, e.g., about 50 ms. The filling time of _Si2H6 is between about 1 ms and about 50 ms, e.g., about 18 ms. The supply time of _Si2H6 is between about 1 ms and about 50 ms, e.g. _{, about} 18 ms _.

The exposure of _MoF6 and _Si2H6 was calculated in Langmuir (L), where 1L= ^1x10-6 Torrx1sec. Pressure spikes during exposure were monitored using a convectron gauge in the reaction chamber. The feed rate was about 1.8 MegaL for _MoF6 and about 4.2 MegaL for _Si2H6 . There was a 2 minute wait between feeds. The substrate was heated using a PBN heater and the temperature was maintained at a temperature between about 100°C and about 150 _{° C} , for example about 120°C. The chamber walls were maintained at a temperature between about 65°C and about 85°C. In one embodiment, the feed rate of _MoF6 was between about 1.0 MegaL and about 10 MegaL. In another embodiment, the supply _{of Si2H6} was between about 1.0 MegaL and about 10 MegaL.

After the deposition cycle, the substrates were transferred in-situ to an ultra-high vacuum chamber for XPS and STM analysis. For XPS measurements, X-rays were generated with an Al Kα anode (1486.7 eV). XPS data were obtained using constant analyzer energy (CAE) with a step width of 0.1 eV and a pass energy of 50 eV. The XPS detector was positioned at 60° to the substrate normal (take-off angle 30° from the substrate surface) and the detector acceptance angle was 7°. XPS spectra were analyzed using the Casa XPS v. 2.3 program after correcting each peak area with its respective relative sensitivity factor. All of the chemical components in this study were normalized to the sum of all components. Scanning tunneling microscopy was performed at a substrate bias of -1.8 V and a constant current of 200 pA.

Ar ⁺ sputtering was performed in conjunction with XPS to investigate the bulk elemental composition of the films. A lens voltage of 5 kV and a beam current of 1.2 μA at 6.0 × ^10-7 Torr of Ar were used. A raster was used to cover the entire substrate area, so the current density was approximately 1.2 uA/ ⁵⁰ mm2. The MoSi _x substrates were kept at 25 °C during sputtering to minimize thermal desorption.

Results FIG. 1A shows XPS chemical composition data of _a HF-cleaned Si surface before and after sequential delivery of _MoF6 and _Si2H6 at 120° C. Two sets of 5.4 MegaL of _MoF6 were delivered at 120° C. onto a HF-cleaned Si substrate. XPS showed saturation of MO at 16%. Then, 4.2 _{MegaL of Si2H6 and} an additional 42 _{MegaL of Si2H6 were} delivered onto the _MoF6- saturated Si surface at 120° C., resulting in 59% saturation of Si. In one embodiment, _MoF6 was delivered between about 1 MegaL and about 10 MegaL. In another embodiment, _SI2H6 was delivered between about 1 MegaL and about ₁₀ MegaL. In another embodiment, the amount _of additional _Si2H6 provided was between about 20 MegaL and about 50 MegaL.

After cleaning with HF, all the Si was in an oxidation state of 0 with 9% O and 12% C contaminants. The contaminants are believed to be caused by adventitious hydrocarbon adsorption during transfer of the substrate to vacuum. HF(aq) was used to remove the native oxide on the Si, leaving the Si surface H-terminated. Note that the Si 2p data in Figure 1 indicates the total amount of Si, whereas the Si(0) data indicates the amount of Si in an oxidation state of 0.

After 5.4 MegaL of _MoF6 at 120°C, 14% Mo and 38% F were deposited on the HF-cleaned Si surface. After supplying an additional 5.4 MegaL of _MoF6 at 120°C, the Mo concentration increased from 14% to 16%, and the F concentration increased from 38% to 42%. This small increase in Mo and F content after an additional 5.4 MegaL of _MoF6 indicates that the reaction of _MoF6 on HF-cleaned Si is self-limiting. After the Si surface was saturated with _MoFx , the ratio of F/Mo was 2.6, and all of the Si was in the 0 oxidation state. The successive supply of _4.2 MegaL of _Si2H6 and 42 _MegaL of _Si2H6 indicates that _{the Si2H6} reaction also saturates on the _MoFx -covered Si surface. It is believed that with thicker substoichiometric _MoSi2 films, more Si could be introduced onto the surface, however _{, Si2H6} reacts self-limitingly with thin films (monolayers) of Mo.

After saturation _{of Si2H6 ,} the Si content was 59% and F was reduced to 10%. Since the substrate is Si, this increase in Si content after _Si2H6 supply can be partially attributed to the substrate as F desorption occurred. However, a decay of Mo was observed after _Si2H6 supply, which is consistent with Si deposition. The reaction of _{MoF6 and Si2H6} on _{H -} terminated Si illustrates the feasibility of MoSi _x ALD on Si-H-terminated Si.

FIG. 1B shows XPS chemical composition data for the same series of _MoF6 and _{Si2H6 saturation} doses described above with respect to FIG. 1, but on a SiON substrate. As explained, no reaction was observed. Note that although the SiON substrate was nominally SiON, the XPS showed only small amounts of N on the surface, so this substrate is mostly ion-damaged _SiOx . After the first three pulses of _MoF6 , 8% F and very small amounts of Mo (<1%) were observed. At the remaining saturation dose, the SiON surface remained unreactive to both _MoF6 and _Si2H6 . Although the SiON used in this study is ion-damaged, the Si is in the +3 and ₊₄ oxidation states, and the data is consistent with strong Si-O, Si-N, and SiO-H bonds, thus substantially preventing Si from bonding with Mo.

Figures 2A and 2B show the Si 2p and Mo 3d XPS spectra of HF-cleaned Si substrates, shown to compare the oxidation states for each experimental run. Figure 2A shows the Si 2p peaks after sequential feeding of _{MoF6 and S2H6} , indicating that after 10.8 MegaL of _MoF6 at 120°C (blue line), Si remained at an oxidation state of 0, consistent with Mo-Si bond formation and no etching of Si by F. After _4.2 MegaL of _Si2H6 at 120°C (red line), the bulk of the Si remained at an oxidation state of 0, consistent with the formation of a monolayer of _MoSi2 . A slightly oxidized Si peak appeared at the surface at high bond energy, which could be _{SiHxF4 -x} (x=2 or 3) or _SiOx . FIG. 2B shows the Mo 3d peaks after sequential feeding of _MoF6 and _S2H6 , showing that the Mo 3d peaks were present in multiple oxidation states after saturation feeding of _MoF6 (black and blue lines). After feeding _{of Si2H6} (red line), all of the Mo was reduced and the peak was centered at 227.4 _eV , consistent with the formation of _MoSi2 .

After the first 5.4 MegaL of MoF ₆ , the Si 2p peak remained at 0 oxidation state, which is consistent with Si-Mo bond formation. The Mo 3d peak appeared in multiple oxidation states, indicating that the surface species was MoF _x with x = 4, 5, and 6 (black line). An additional 5.4 MegaL of MoF ₆ did not change the oxidation state of the Si 2p or Mo 3d peaks (blue line). The data suggest Si-Mo-F _x formation at the surface. Note that after the MoF ₆ saturation dosage, the F/Mo ratio was 2.6 (Figure 1A XPS data), but Mo was in oxidation states 4-6, so some Mo-O bonds are likely formed. After 4.2 MegaL of Si ₂ H ₆ dosage (red line), a small shoulder peak appeared at higher binding energy (103 eV) on the Si 2p XPS peak. This is consistent with Si-F or Si-O formation. The Mo 3d spectrum shows that after a single Si ₂ H ₆ supply, all of the Mo is reduced to Mo ⁰ with a bond energy of 227.4 eV. This is consistent with the formation of a single layer of MoSi _x and the transfer of residual oxygen or fluorine from Mo to S in the form of Si-O and Si-F bonds. The simplified reaction of MoF ₆ and Si ₂ H ₆ is:
It can be written as follows.

The ALD characteristics of MoSi _x on Si substrates and the selectivity to SiO ₂ and SiN _x substrates were verified via XPS of MoSi _x deposition on patterned substrates. FIG. 3A shows the chemical composition of a set of three substrates: HF-cleaned Si, HF-cleaned SiO ₂ , and HF-cleaned patterned substrate. FIG. 3B shows the chemical composition of each of the substrates in FIG. 3A after five ALD cycles of MoF ₆ and Si ₂ H ₆ at 120° C. The data showed that Si-deficient MoSi _x was selectively deposited on Si but not on SiO ₂ . The Si ⁰ component of the patterned sample was also selectively attenuated by MoSi _x deposition. FIG. 3C shows the chemical composition of each of the substrates in FIG. 3B after an additional 25.2 MegaL (between 3 and 10 pulses) of Si ₂ H ₆ . The _{additional Si2H6} introduced Si onto the MoSi _x surface. The selectivity to _SiO2 was maintained during the additional _{Si2H6 pulses} ( _SiO2 had 0% Mo and 0% ^Si0 throughout the ALD process).

The three substrates were loaded together on a single substrate holder to ensure that they were exposed to the same deposition conditions. The Si and _SiO2 substrates allowed for the verification of selectivity during deposition onto the patterned substrate. The patterned substrate has a _SiO2 layer sandwiched between _SiNx on top of the Si substrate. Note that the _SiNx on the patterned substrate was actually SiON since it was ion damaged and ashed to _O2 during fabrication. As shown in FIG. 3A, a 30s HF clean removed the native oxide on the Si. The thermally grown _SiO2 was 300 nm thick, and the 30s HF clean did not change the elemental composition or oxidation state of the _SiO2 . The HF cleaned patterned substrate consisted of a mixture of _SiNx , _SiOx , and ^Si0 .

As shown in FIG _. 3B, XPS was performed after five ALD cycles of _MoF6 and _Si2H6 at 120° C. XPS showed a surface composition of 32% Mo and 10% Si on the Si substrate, which corresponds to MoSi _x that is highly depleted in Si. No MoSi _x was deposited on the _SiO2 substrate, which was consistent with highly selective ALD. On the patterned substrate, XPS showed that 5% Mo was deposited and Si ⁰ decayed to 1%. The fraction of N and O on the surface does not change significantly during ALD on the patterned substrate. This data is consistent with Si-deficient MoSi _x being selectively deposited on 6% Si ⁰ on the patterned substrate.

The deposition selectivity on the patterned substrate is consistent with three aspects of the embodiment described herein: (1) MoSi _x deposition occurred on the Si substrate but not on the SiO ₂ substrate; (2) after MoSi _x deposition, the Si ⁰ (but not the higher oxidation state Si peaks from Si-N and Si-O) attenuated on the patterned substrate; and (3) numerically, the approximately 4% Mo deposition on the patterned substrate with 6% Si ⁰ is proportional to 32% Mo on the Si substrate and 54% Si ⁰ on the HF cleaned surface.

Although monolayers of _MoSi2 could be deposited on Si in the ALD saturation experiments described in Figures 1 and 2, successive ALD cycles did not produce stoichiometric MoSi2. The formation of Si-deficient _{MoSi x is} believed to be due to surface Si-H species that desorb during the _fluorosilane removal process and residual Mo-F bonds that are not easily removed with standard _Si2H6 supply. In the first 1-3 monolayers, there is excess Si from the substrate present to aid in fluorine desorption, but in thicker films, the only _Si available is gaseous _Si2H6 , so Mo-F surface bonds may persist. The overall fluorosilane removal chemistry using _{MoF6 and Si2H6} is consistent with one of two chemical reactions.
1:
2:

To form _MoSi2 , the three substrates were exposed to an additional 25.2 MegaL (between 3 and 10 pulses, e.g., 6 pulses) of _Si2H6 at 120°C (see Figure 3C _{). After the additional Si2H6 exposure, Si increased by 20% on the Si substrate, consistent with Si being incorporated into the film or on the surface of the substrate. The additional Si2H6 supply did not reduce} the deposition selectivity onto Si relative to _SiO2 .

4A-4C show XPS chemical composition data for selective deposition of MoSi _x on HF cleaned Si, SiO ₂ and SiON that were post-deposition annealed. FIG. 4A shows the XPS chemical composition of the Si, SiO ₂ and SiON substrates after HF cleaning. FIG. 4B shows XPS chemical composition data indicating that MoSi _x is selectively deposited only on Si after five ALD cycles of MoSi _x followed by an additional six pulses (25.2 MegaL) of Si ₂ H ₆ at 120° C. FIG. 4C shows XPS chemical composition data for a substrate that was post-deposition annealed (PDA) at 520° C. for 3 minutes. As shown, the PDA removed F from the MoSi _x film and reduced Mo to Mo ⁰ .

FIG. 4A shows that after HF cleaning, the SiON surface is composed primarily of SiN _x . After five ALDs of _{MoSi x} followed by an additional 25.2 MegaL of Si ₂ H ₆ , less than 1% Mo was detected on the SiO _x and SiN _x surfaces, as shown in FIG. 4B, compared to 24% Mo and 18% Si present on the HF-cleaned Si. The three substrates were then annealed at 520° C. for 3 minutes, which reduced F from 25% to 3% on the Si substrate. PDA at 520° C. also reduced Mo to Mo ⁰ on the Si substrate, lowering the Si:Mo ratio at the surface from about 0.75 to about 0.5. This is consistent with the desorption of surface F in the form of SiHF ₃ or SiF _4. XPS analysis of the PDA indicates that PDA removes F from the film, thereby reducing the possibility of F diffusion into adjacent MOSFET device structures.

In-situ STM and ex-situ atomic force microscopy (AFM) were used to investigate the surface topography after deposition and PDA on Si and _SiO2 substrates. Separate substrates of HF-cleaned Si after 20 cycles _{of MoF6} and _Si2H6 were prepared for in-situ STM. STM data showed that the MoSi _x films were atomically flat and conformal with an RMS roughness of about 2.8 Å. The above substrates were annealed in-situ at 500°C for 3 minutes in an ultra-high vacuum chamber at a pressure of about 5.0× ^10-10 Torr. After annealing at 500°C, the films were much flatter with an RMS thickness of about 1.7 Å.

Another substrate of MoSi _x /HF cleaned Si after five ALD cycles at 120 °C followed by in-situ anneal at 550 °C was placed in an ex-situ furnace for spike annealing at 900 °C in N balanced 5% H _{2 .} After the 900 °C spike anneal, the surface morphology was obtained using AFM. The film retained a sub-nanoscale RMS roughness of 4.75 Å. This demonstrated that the MoSi _x film has high thermal stability up to about 900 °C.

Ex-situ AFM imaging data of the _SiO2 substrate surface after five ALD cycles at 120°C followed by in-situ annealing at 550°C for 3 minutes was performed by counting the number of nuclei on the substrate surface to confirm the selectivity. The density of nuclei was about 9 nuclei/ ^μm2 , confirming the preferential deposition of Si over _SiO2 . It is believed that the high deposition selectivity of the embodiments described herein can be _further improved by controlling the temperature of the reaction chamber walls and by using short high pressure _Si2H6 pulses and long purge cycles to facilitate ALD and avoid CVD deposition regimes.

Depth profile studies were also performed to determine the internal composition of the MoSi _x films. Figure 5A shows the XPS chemical composition data after Ar ⁺ sputtering on HF-cleaned Si after five cycles of MoF ₆ and Si ₂ H ₆ at 120 °C. Figure 5B shows the Si 2p peaks after sequential Ar ⁺ sputtering, indicating that the bulk of the MoSi _x film consists of mostly Si ^0. Figure 5C shows the chemical composition data of the deposited film plotted against Ar ⁺ sputtering time on Si after five cycles of MoF ₆ and Si ₂ H ₆ at 120 °C.

The XPS data shown in FIG. 5A is from a MoSi x film deposited on a HF-cleaned Si substrate at 120° C. using five ALD cycles of MoF ₆ and Si ₂ H ₆ without additional Si ₂ H ₆ introduction. As the sputtering time was increased, the MoSi _x film thinned until the underlying Si substrate was exposed. In the first 10 minutes of sputtering, F decreased from 35% to 8%, while Mo changed from a mixture of Mo oxide and Mo ⁰ to pure Mo ^0. This data is consistent with the surface _F being primarily bound to Mo.

After successive sputtering cycles, the amount of Si increased and the amount of Mo decreased. Furthermore, the amount of ^Si0 increased with the total amount of Si, reaching a maximum value of 43% after a total sputtering time of 100 min. The ratio of ^Si0 to ^Mo0 was used to distinguish the pure MoSi _x phase, because in the pure MoSi _x phase, both Mo and Si are bonded to each other and have an oxidation state of 0. After removing the silicon oxide and MoF _x species at the substrate surface, the percentage of ^Si0 exceeded the percentage of ^Mo0 . The ^Si0 : ^Mo0 ratio in the bulk of the MoSi _x film was 1.41, which corresponds to a Si-deficient MoSi _x film. It is noted that in the center of the film, the Si:Mo ratio was 1.77, and therefore the ^Si0 : ^Mo0 ratio could be close to 2 in the absence of background _O2 / _H2O .

Figure 5B shows the raw XPS spectrum of Si 2p corresponding to each XPS measurement in Figure 5A. After the fourth sputtering cycle, the Si peak at 99.2 eV increased and extended to higher binding energies. In contrast, after each sputtering cycle, the energy of the Mo peak corresponded to Mo ^0. Thus, it is believed that the bulk MoSi _x film is primarily Si ⁰ and Mo ⁰ in the form of MoSi _x , while the top and bottom interfaces are rich in SiO _x . The top SiO _x is consistent with contaminants from the chamber environment, while the oxide at the bottom interface is consistent with an incomplete ex-situ HF clean.

The substoichiometric oxide at the bottom interface did not affect the deposition and film quality. This indicates that the selectivity of MoSi _x ALD is sensitive to the quality of SiO ₂ . Figure 5C shows the percentages of the chemical constituents obtained from the XPS measurements in Figure 5A. After the second sputtering cycle (40 min of total sputtering time), F decreased to less than 3% and eventually reached 0%. O in the bulk of the film was <10% but gradually increased to 15% at the MoSi _x -Si interface, consistent with the presence of an interface oxide layer.

To understand the effect of additional _Si2H6 supply on the Si:Mo ratio of MoSi _x films, XPS depth profiling was performed on MoSi _x with additional Si introduced. Five _ALD passes of _MoF6 and _Si2H6 at 120 °C were followed by an additional six pulses (25.2 _{MegaL) of Si2H6 at the} end of a 3 min anneal at 530 °C on dry cleaned Si. The post-anneal dry clean process described herein utilizes a plasma of _NF3 and _NH3 with Ar as the carrier gas.

Figures 6A-6D show XPS profile data of MoSi _x films after exposure to additional Si ₂ H ₆ supply. Figure 6A shows XPS chemical composition data after Ar ⁺ sputter dry cleaned Si after 5 cycles of MoF ₆ and Si ₂ H ₆ followed by 6 additional pulses (25.2 MegaL) of Si ₂ H ₆ at 120°C. Figure 6B shows XPS surface composition data after 5 ALD cycles of MoF ₆ and Si ₂ H ₆ , but without additional Si ₂ H ₆ pulses. The Si:Mo ratio was 0.33 for 5 ALD and 0.89 for 5 ALD + 6xSi ₂ H ₆ , which is consistent with Si incorporation on the surface. Figure 6C shows the XPS bulk composition data of MoSi _x with and without a Si ₂ H ₆ pulse after Ar ⁺ sputtering was used to remove surface contaminants. The Si:Mo ratio was 1.77 for 5 ALD and 1.96 for 5 ALD + 6 × Si ₂ H _6. Figure 6D shows the XPS chemical composition data of MoSi _x films plotted against Ar ⁺ sputter time on Si after 5 cycles of MoF ₆ and Si ₂ H ₆ followed by an additional Si ₂ H ₆ pulse at 120 °C.

Figure 6A shows _a series of depth profile XPS after each run performed on a dry cleaned substrate. After _6xSi2H6 /5ALD cycles, 28% F, 20% Si, and 28% Mo were present on the substrate surface. The surface F was almost completely removed after annealing at 530°C, and all Mo was reduced to ^Mo0 . This was consistent with the desorption of F from the surface as shown in Figure 4C. In this run, the Si:Mo ratio was 0.89. In comparison, the Si:Mo ratio of the MoSi _x film without additional _{Si2H6 supply} was only 0.33, as shown in Figure 6B.

After removing the surface _oxide contaminants, the bulk ^Si0 : ^Mo0 was 1.32 (Si:Mo=1.96) for MoSi _x with the additional _Si2H6 pulse, which was comparable to the ^Si0 :Mo0=1.41 (Si:Mo= ^1.77 ) in the bulk of MoSi _x without additional _Si2H6 introduction shown in FIG. 6C. Thus, _the additional _Si2H6 pulse is believed to increase the Si content of the Si-deficient MoSi _x surface after the ALD cycle. In contrast, the Si:Mo ratio in the _bulk of the MoSi _x film was close to stoichiometric _MoSi2 . FIG. 6D shows the XPS percentage of each chemical component as a function of Ar ⁺ sputter time, which is consistent with the formation of MoSi _x in the bulk of the MoSi _x film.

In one embodiment, 4.2 MegaL of _Si2H6 was introduced _into the reaction chamber using a pneumatic valve for a duration of ₆ seconds. The _Si2H6 process profile uses approximately 3 times _{the Si2H6 exposure over} a delivery duration that is approximately 10 times shorter than conventional _Si2H6 delivery parameters. Thus, the embodiment described herein uses a 30 times higher partial pressure during the ALD delivery compared to conventional delivery schemes. The 30 times higher instantaneous pressure during delivery is believed to allow the precursor-mediated _{Si2H6 chemisorbed} layer to remain on the surface long enough to react with Mo and incorporate more Si into the MoSi _x film. The Si incorporation is also believed to be self-limiting, which allows the MoSi _x growth rate to be approximately 1.2 nm/cycle.

The resistance of the MoSi _x films was measured using four-point probe measurements. For electrical measurements, up-doped Si(001) with a resistivity of >10000 ohm-cm was used as the substrate. For electrical measurements, 10 cycles of MoSi _x ALD at 120° C. were deposited on HF-cleaned intrinsic (semi-insulating) Si substrates, followed by an in-situ anneal at 550° C. for 3 minutes and a spike anneal at 900° C. in 5% H ₂ balanced with N _2. A small amount of Ni was deposited as the probe contact. The resistance was 110 Ohm and the resistance was calculated using the infinite sheet approximation as follows:
where k is a constant, t is the thickness, and _Rmax is the maximum resistance measured.
was calculated as follows:

To confirm the selectivity of MoSi _x on the nanostructured patterns, cross-sectional TEM studies were performed on the patterned substrate. Figure 7 shows a cross-sectional TEM image of a MoSi _x /HF cleaned patterned substrate. The HF cleaned patterned substrate was subjected to 5 x 5 cycles of MoSi _x ALD followed by an additional 25.2 MegaL of Si ₂ H ₆ at 120°C. The elemental composition of this substrate at each deposition step is shown in Figures 3A-3C. The TEM images show the complete selectivity of the deposition of MoSi _x on Si, not on SiN _x or SiO _2. The thickness of the MoSi _x film deposited on Si after 5 ALD cycles followed by an additional 25.2 MegaL was about 6.3 nm. This achieved a growth rate of about 1.2 nm/cycle. Due to the growth rate per cycle of MoSi _x ALD, 5 ALD cycles are considered sufficient for contact materials and contact device structures.

Selective atomic layer deposition of substoichiometric _MoSi2 was achieved by selective processing on hydrogen-terminated Si relative to thermally grown _SiO2 , ion-damaged _SiON , and _SiNx . The selectivity was based on the preferred reactivity of _MoF6 and _Si2H6 on H-Si, but not _SiO2 or _SiNx , because the Si-O, Si-N, and SiO-H bonds are strong enough that they cannot be broken by either precursor at 120°C. Both _MoF6 and _Si2H6 exhibited self-limiting behavior. This allowed the deposition of highly conformal, smooth films with a root-mean-square (RMS) roughness of 2.8 _Å . Three minutes of PDA in ultra-high vacuum at temperatures of about 500°C to 550°C further reduced the RMS roughness to 1.7 Å. The quality of the MoSi _x films is preserved even after a spike anneal at 900° C. in a H ₂ /N ₂ environment, which is consistent with high thermal stability.

Depth profiling XPS studies reveal that the bulk of the MoSi _x is close to stoichiometric MoSi ₂ (Si:Mo=1.7-1.9) with <10% oxygen and fluorine. The surface of the MoSi _x film after five ALD cycles shows a highly Si-depleted MoSi _x surface with a Si:Mo ratio of 0.33, which is improved to 0.89 by pulsing additional Si ₂ H _6. Cross-sectional TEM images show that selectivity is preserved at the nanoscale, allowing selective deposition of MoSi _x on Si without consuming the substrate.

The growth rate of the MoSi _x film of about 1.2 nm/cycle makes less than 10 ALD cycles, e.g., 5 ALD cycles, sufficient to use the MoSi _x film as a contact material. Thus, compared to conventional ALD processes, process throughput is improved using the embodiments described herein. The selective deposition of MoSi _x is believed to eliminate or substantially reduce the dependency on lithographic processing for complex 3D MOSFET structures (e.g., FinFETs). The selectivity of Si-H bonds over SiO-H bonds is greater than 10 ^6. Thus, high nanoscale selectivity is possible, even without the use of additional passivation layers. The embodiments described herein also show that ALD of silicide over metal can be easily switched while retaining selectivity by changing the partial pressure during the ALD pulse of the reducing agent.

While the above description is directed to embodiments of the present disclosure, other and further embodiments of the present disclosure may be devised without departing from the basic scope of the present disclosure, the scope of which is determined by the following claims.

Claims (13)

A method for processing a substrate, comprising:
The method includes exposing a silicon-containing substrate, the silicon-containing substrate including bulk silicon oxide and native silicon oxide, to a plasma formed from a _NF3 precursor and a _NH3 precursor and Ar to selectively and preferentially remove the native silicon oxide from the bulk silicon oxide and native silicon oxide contained in the substrate, the exposing comprising:
heating the substrate to a temperature between about 40° C. and about 50° C.; and exposing the substrate to the plasma and Ar for a period of less than about 40 seconds.
Including,
The method wherein the ratio of the NF3 _precursor , the NH3 _precursor , and Ar (NF3 _: NH3 _: Ar) is 1:10:1.5 . A method for processing a substrate, comprising:
1. Exposing a silicon-containing substrate, comprising bulk silicon oxide and native silicon oxide, to a plasma formed from a _NF3 precursor and a _NH3 precursor to selectively and preferentially remove the native silicon oxide from among the bulk silicon oxide and native silicon oxide contained in the substrate,
heating the substrate to a temperature between about 40° C. and about 50° C.; and exposing the substrate to the plasma for a period of less than about 40 seconds.
selectively and preferentially removing the native silicon oxide, comprising:
heating the substrate to a first temperature;
exposing the substrate to a plasma comprising hydrogen;
exposing the substrate to a first supply of _MoF6 precursor;
exposing the substrate to a second supply of _{Si2H6 precursor} ;
exposing the substrate to the first supply and exposing the substrate to the second supply in successive cycles;
exposing the substrate to a third supply of _{Si2H6 precursor} after the successive cycles;
A method comprising: The method of claim 2 , further comprising annealing the substrate after exposing the substrate to the third dose at a second temperature between about 500° C. and about 550° C. The method of claim 2 , wherein the first temperature is between about 100° C. and about 150° C. 3. The method of claim 2 , wherein the successive cycles are performed less than 10 times. The method of claim 2 , wherein the plasma containing hydrogen is formed from a precursor selected from the group consisting of NF ₃ , NH ₃ , and H. 3. The method of claim 2 , wherein a nitrogen purge process using _N2 is performed during said successive cycles. The method of claim 2 , wherein the first delivery is performed for a duration between about 10 ms and about 100 ms. 9. The method of claim 8 , wherein the first supply comprises a _MoF6 flow rate between about 1 MegaL and about 10 MegaL. 9. The method of claim 8 , wherein the second delivery is performed for a duration between about 1 ms and about 50 ms. The method of claim 10 , wherein the second supply rate comprises a Si ₂ H ₆ flow rate between about 1 MegaL and about 10 MegaL. The method of claim 11 , wherein the third supply rate comprises a Si ₂ H ₆ flow rate between about 20 MegaL and about 50 MegaL. A method for processing a substrate, comprising:
placing a silicon-containing substrate comprising bulk silicon oxide and native silicon oxide on a heater in a reaction chamber having chamber walls;
exposing the substrate to a plasma formed from a _NF3 precursor and a _NH3 precursor to selectively and preferentially remove the native silicon oxide from among bulk silicon oxide and native silicon oxide contained in the substrate;
heating the substrate to a temperature between about 40° C. and about 50° C.; and exposing the substrate to the plasma for a period of less than about 40 seconds.
selectively and preferentially removing the native silicon oxide, comprising:
heating the substrate on the heater to a first temperature;
maintaining the chamber walls at a second temperature lower than the first temperature;
exposing the silicon-containing surface of the substrate to hydrogen;
exposing the substrate to a first supply of _MoF6 precursor;
exposing the substrate to a second supply of _{Si2H6 precursor} ;
exposing the substrate to the first supply and exposing the substrate to the second supply in successive cycles;
exposing the substrate to a third supply of _{Si2H6 precursor} after the successive cycles;
A method comprising:

Images