Abstract
This invention provides a method and a system to deposit a thin layer of very reactive metals by plasma enhanced atomic layer deposition (PEALD). The very reactive metals, selected from the highly electropositive elements include alkaline earth metals, group III metals, and some transition and rare earth metals. The method is comprised of sequentially pulsing one of above mentioned metal containing organometallic precursors and a hydrogen plasma as a reducing agent into a high vacuum reaction chamber containing a substrate surface with pulsed or continuous flow of an inert purge gas between each pulsing step. The system comprising a very high efficiency H plasma source, the high vacuum reactor chamber, an anti-corrosion turbo pump and a high vacuum load lock is required for reducing contaminant gases such as O.sub.2, H.sub.2O, and CO.sub.2, and for increasing hydrogen plasma efficiency.
Claims
1. A method of forming a thin film of elemental metal magnesium (Mg) in a high vacuum plasma enhanced atomic layer deposition (PEALD) system, comprising: (i) sequentially pulsing vapor of a magnesium (Mg) containing organometallic precursor, and a hydrogen plasma as a reducing agent into a high vacuum reaction chamber containing a substrate surface; (ii) applying pulsed or continuous flow of an inert purge gas between each said pulsing step, thereby forming a single atomic layer of pure magnesium (Mg) metal; and (iii) repeating steps (i) and (ii) for a next atomic layer on top of the atomic layer in step (ii), thereby accumulating many atomic layers to form a thin film of pure magnesium (Mg); wherein said substrate surface is preheated to between 200-300° C.; and wherein a vacuum level of said high vacuum PEALD system is 4×10.sup.−7 Torr or lower.
2. The method according to claim 1, wherein said magnesium (Mg) containing organometallic precursor is bis(ethylcyclopentadienyl) magnesium.
Description
BRIEF DESCRIPTION OF THE DRAWINGS
(1) FIG. 1 shows a perspective right-side view of the plasma enhanced high vacuum ALD system for very reactive metal deposition.
(2) FIG. 2 is a process flow chart.
(3) FIG. 3 shows XPS result from an Al film deposited by high vacuum ALD.
(4) FIG. 4 is a graph showing growth rate per cycle (GRC) vs. Mg(CpEt).sub.2 exposure shows a clear saturation curve.
(5) FIG. 5 shows a trend curve of Mg GRC vs. the growth temperature.
(6) FIGS. 6A to 6C show XPS result from a relative thick Mg film ˜70 nm. FIG. 6A is a composition depth profiling by XPS. FIG. 6B is a general survey scan after the Mg film surface is thoroughly cleaned by sputtering. FIG. 6C shows fine scans of binding energy (Eb) around the Mg2p peak at different depth compared to the Mg film surface.
(7) FIGS. 7A to 7B
(8) FIG. 7A is a photo showing AFM surface morphology of a ˜70 nm thick Mg film with RMS roughness =13 nm; FIG. 7b shows a RHEED image from an as deposited Mg film surface indicating Mg film is mainly polycrystalline with some texture.
(9) FIG. 8 is a graph showing the measured Mg thickness vs. the cycle number at a deposition temperature of 200° C.
BRIEF DESCRIPTION OF THE SYMBOLS
(10) 110: Substrate transfer rod
(11) 120: High vacuum load lock
(12) 130: Matching box
(13) 140: Turbo pump
(14) 150: ALD reaction chamber
(15) 160: Substrate holder
(16) 170: Substrate
(17) 180: Remote plasma source
(18) 210: Precursor pulse
(19) 220: Inert gas purge
(20) 230: Hydrogen plasma pulse
(21) 240: The 2nd Inert gas purge
(22) 250: Repeat
DETAILED DESCRIPTION OF THE INVENTION
(23) FIG. 1 shows the main parts of a cross-flow high vacuum PEALD system that is used for very reactive metal depositions. The reactor (150) is evacuated by a turbo pump (140) with a backing rotary vane pump. The system is also equipped with a sample transfer rod (110), a vacuum load lock (120), and a sample holding/unloading mechanism to avoid frequent venting of the reactor and reduce the reactor exposure to ambient atmosphere. A remote plasma source (180) is located on top central part of the reactor. The base vacuum pressure of at least 4×10.sup.−7 Torr with O.sub.2<10.sup.−8 Torr, H.sub.2O and CO.sub.2<10.sup.−9 Torr as measured by a residual gas analyzer (RGA) can be achieved. Ultra-high purity Ar with O.sub.2 and H.sub.2O in 10 ppb levels (with a built in filter) is used as the carrier gas. High purity H.sub.2 (6N or higher) is used for generating hydrogen plasma. Bis(ethylcyclopentadienyl) magnesium (Mg (CpEt).sub.2) is used as the Mg precursor. The high vacuum ALD system has sufficiently high vacuum and low impurity gas levels, which is capable of preventing some very reactive metals from oxidation or carbonization. A con-flange sealed turbo pump is added thus the ALD is always in high vacuum conditions where impurity levels such as O.sub.2, H.sub.2O, N, C are significantly reduced; a load lock (LL) with auto or manual sample transfer mechanism is added, which avoids frequent exposure of the ALD reactor to ambient and thus further reduces introduction of impurity levels of above gases. This is crucial for depositions of the very reactive elemental metals.
(24) In addition to high or ultrahigh vacuum requirement, the system has been proved capable of handling corrosive organometallic precursors, and working simultaneously in both molecular flow (low pressure) and viscous flow (high pressure) ranges. The system is also robust when subject to pumping and mechanical stress induced fatigue during numerous cycles. All vacuum gauges, valves and in situ monitoring tools are expected anti-corrosive too. All these requirements put a big challenge to the turbo/ backing pump system and the exhaust.
(25) An ALD process cycle as shown in FIG. 2 comprises pulsing Mg precursor vapor such as Mg (CpEt).sub.2 into the reactor (210), purging the reactor with Ar carrier gas after the Mg precursor (220), pulsing a mixture of hydrogen and Ar into the reactor followed by switching on RF plasma RF power fix a desired time (230), and purging the reactor by Ar carrier gas after the hydrogen and Ar pulsing and plasma off (240). The ALD process cycle can be repeated (250) until the metal layer reaches a desired thickness. We prove that deposition of very reactive metals requires extremely low H.sub.2O/O.sub.2/CO.sub.2 background pressure in ALD (10.sup.−8 Torr or lower), nothing or oxides will be deposited if vacuum or background impurities are higher. The suggested ALD surface chemistry is:
Mg(CpEt).sub.2+H.sub.2*.fwdarw.Mg+2HCpEt
EXAMPLE 1
(26) First of all, we reconfirm Al metal deposition process which has been reported over ten year ago but has not been reproduced by other groups until recently. See, e.g., Y. J. Lee, et al. Electrochemical and Solid-State Letters, 5˜10, C91-C93 (2002); Y. J. Lee et al., J. Vac. Sci. Technol. A 20, 6, 1983 (2002). By using this process as a test bed we are able to evaluate vacuum quality of the high vacuum PEALD system and H plasma efficiency of the remote plasma source. Since Al is very reactive metal similar to Mg, both metals are easily oxidized.
(27) Trimethyaluminium (TMAl)-Al(CH3).sub.3 and mixed H/Ar plasma are used for this deposition. The films deposited on neatly insulating Si water with native oxide become more mirror/metallic-like and very conductive. X-ray photoelectron spectrometry (XPS) analysis indicate pure Al metal is deposited as shown in FIG. 3 for distinguishable Al2p metal peak and oxide peak in their binding energy measurement. The result surely proves that pure Al metal is deposited. On the contrary previous effort to deposit Al by PEALD never succeeded and some AlOx films were always obtained. The main difficulty lies in easy oxidation of Al metal on film surface and low H plasma efficiency in a conventional ALD reactor. The surface chemistry of Al metal deposition has been suggested that if the plasma can effectively break up metal-carbon bonds and reduce TMAl to Al.
EXAMPLE 2
(28) We propose a process and surface chemistry that Mg(CpEt).sub.2 also has metal-carbon bonds and Mg sits only next to Al in the periodical table, it is thus possible that H plasma can reduce the Mg precursor to Mg metal as well similar to Al. To test this idea, Mg(CpEt).sub.2 and H plasma were used in the deposition. Films deposited on nearly insulating Si wafer with native oxide show mirror/metallic-like color and are very conductive.
(29) FIG. 4 shows the typical self limiting curve of Mg GRC vs. Mg(CpEt).sub.2 exposure with GRC of a 1.5-2.0 Å/cycle after saturation. FIG. 5 shows a trend curve of Mg GRC vs. the growth temperature. At ≦160° C., a much larger growth rate was obtained with a non-metallic and insulating film deposited indicating the deposition is most likely physi-sorption dominated thus poor quality Mg is obtained; at ≧200° C., Mg films show typical metallic color with GRC decreasing with increasing growth temperature. This may result from increased volatility and reduced sticking coefficient of Mg at elevated temperatures.
(30) Most of as deposited Mg films showed strong O signals. To identify where exactly the O contaminant conies from, we deposited a relative thick Mg film ˜70 nm, and a composition depth profiling by XPS was measured as shown in FIG. 6A. The O peak intensity keeps decreasing with increasing sputtered thickness indicating O is most likely from surface contaminant. C contamination in the film is negligible. The fact that Si content keeps increasing may result from Si substrate used and island formation of Mg. FIG. 6B shows a general survey scan after the Mg film surface is thoroughly cleaned by sputtering. Again the film is dominated by Mg peaks. O and other signals are significantly weaker. FIG. 6C shows fine scans of binding energy (Eb) around the Mg2p peak at different depth compared to the Mg film surface. The surface Mg2p Eb=50.2 eV is very close to that of an oxide Mg while at 10 to 25 nm depth, Mg2p Eb is very close to 49.5 eV from a pure Mg metal. See, e.g., J. S. Corneille, et al., Surface Science, 306 (1994) 269-278; S. Rajput, et al., Bull. Mater. Sci., Vol. 29, No. 3, June 2006, pp. 207-211. Indian Academy of Sciences; http:/srdata.nist.Gov/xps/. This confirms that pure Mg has been deposited.
(31) FIG. 7A shows AFM surface morphology of a ˜70 nm thick Mg film with RMS roughness =13 nm. Typical island morphology with very large surface roughness indicates a rough surface and severe discontinuity. This is confirmed by measured Mg thickness vs. the cycle number at a deposition temperature of 200° C. as shown in FIG. 8. The fact that the line does not intersect at zero suggests an incubation/nucleation period when Mg shows no physical thickness. This island morphology has been observed on silicon wafers, either with native SiO.sub.2 or HF etching cleaned surface, glass slides, and c-sapphire wafers. Island formation of metals deposited by ALD is very common such as Cu, Pt, etc. See, e.g., Z. W. Li, et al, J. Electrochemical Society, 153(11) C787-C794 (2006), The main reason is low surface energy of the metals and their low adhesion to many material surfaces. The adhesion can be improved by using a variety of metallic glue materials as proved by R. Gordon's group. See, e.g., B. Han, et al., Angew. Chem. Int. Ed., 2010, 49, 148-152. Mg happens to have very low surface energy. Nb however is listed as a good glue material.
(32) Measured resistivity of as deposited films by a standard 4 probe station turns out to be sensitive to substrate material used. We got 10.sup.−3 to 10.sup.−6 Ωcm on Si either with native SiO.sub.2 or HF etched surface. Si wafers used are high resistivity type in the range of 100 Ωcm. However as deposited films on c-sapphire wafers are always insulating. We believe this result is related to island formation of Mg, especially when films are thin, islands are mostly isolated, no conductivity path is formed and thus no percolation happens. This phenomenon has been reported for Cu by ALD. No conductivity is measured even when the film is relatively thicker due to island formation.
(33) FIG. 7B shows a reflection high energy electron diffraction (RHEED) image from an as deposited Mg film surface indicating Mg film is mainly polycrystalline with some texture. However, since Mg is very reactive to air, it is also possible that the RHEED pattern comes from some Mg oxides.
(34) We demonstrated pure Mg can be deposited by H plasma ALD in a self limiting behavior. This result has been confirmed by XPS and atomic force microscopy (AFM) measurements as described below. This is the first time pure Mg metal is deposited by ALD. It further proves that high vacuum ALD and H reduction method is certain a way to deposit very reactive metals.
(35) This invention provides a method of forming a thin film of elemental metal magnesium Mg in a high vacuum plasma enhanced atomic layer deposition (PEALD) system configured for realization of this method.
