Low-temperature magnetron sputter-deposition, hardness, and electrical
`resistivity of amorphous and crystalline alumina thin films
`Quan Li
`Department of Materials Science and Engineering, Northwestern University, Evanston, Illinois 60208
`Yuan-Hsin Yu
`Department of Materials Engineering, Tatung University, Taipei, Taiwan, Republic of China
`C. Singh Bhatia
`IBM Storage Systems Division, 5600 Cottle Road, San Jose, California 95193
`L. D. Marks
`Department of Materials Science and Engineering, Northwestern University, Evanston, Illinois 60208
`S. C. Lee
`Department of Materials Engineering, Tatung University, Taipei, Taiwan, Republic of China
`Y. W. Chunga)
`Department of Materials Science and Engineering, Northwestern University, Evanston, Illinois 60208
`共Received 7 October 1999; accepted 12 May 2000兲
`Aluminum oxide films were grown by reactive magnetron sputtering. In order to maintain a stable
`deposition process and high deposition rate, a pulsed direct current bias was applied to the
`aluminum target and the substrate. An external solenoid was used to form a magnetic trap between
`the target and the substrate. The influence of substrate temperature, substrate bias, and the magnetic
`trap on film growth and properties was studied by different surface and thin-film analysis techniques
`and electrical measurements. Normally, amorphous alumina films were produced. However, under
`optimum process conditions, crystalline alumina films can be obtained at temperatures as low as
`250 °C, with a hardness ⬃20 GPa and excellent electrical insulating properties. © 2000 American
`Vacuum Society. 关S0734-2101共00兲04605-4兴
`
`that can be coated with alumina. Therefore, low-temperature
`synthesis of crystalline alumina is of great technological in-
`terest. One method to reduce the crystalline growth tempera-
`ture is to enhance the mobility of surface species via low-
`energy ion bombardment. To explore this strategy, Schneider
`et al.8 incorporated a rf coil to enhance ionization in a stan-
`dard magnetron sputter-deposition system. Together with
`pulsed dc substrate bias to enhance ion bombardment of the
`growing film, they demonstrated the growth of crystalline ␬
`alumina at relatively low substrate temperatures 共400 °C兲.
`In this study, we report the synthesis and properties of
`aluminum oxide thin films grown by pulsed dc magnetron
`sputtering. Instead of using a rf coil to increase the ion con-
`centration in the plasma, we used a magnetron sputter source
`equipped with strong magnets instead. An additional mag-
`netic field with the correct polarity was applied near the sub-
`strate to form a magnetic trap. With proper substrate biasing,
`this results in markedly improved ion currents to the sub-
`strate during film growth, as demonstrated by Petrov et al.9
`and Engstrom et al.10
`
`I. INTRODUCTION
`Aluminum oxide thin films have many properties that are
`useful for applications in optical electronics, and cutting tool
`industries. Methods of synthesizing aluminum oxide thin
`films include chemical vapor deposition combined with ion-
`beam irradiation,1 molecular beam epitaxy using a solid alu-
`minum source and N2O,2 laser-induced deposition from con-
`densed layers of organoaluminum compounds and water,3
`and magnetron sputter deposition.4
`Traditional radio frequency 共rf兲 sputtering uses an alumi-
`num oxide target and generally results in low deposition
`rates. Direct current 共dc兲 reactive sputtering of aluminum in
`an oxygen/argon atmosphere can produce stoichiometric
`Al2O3 at high rates. The major requirement is careful control
`of the oxygen partial pressure5 so that the aluminum target
`will not be poisoned. In addition, pulsed dc magnetron sput-
`tering was used in recent years to minimize arcing during
`deposition of insulating materials such as alumina.6 In this
`case, the magnetron target voltage was bipolar at a certain
`frequency 共2–30 kHz typically兲. In the positive cycle, elec-
`trons travel towards the target and neutralize the positive
`charge accumulated on the target surface in the negative
`cycle. When this is done properly and at sufficiently high
`frequency, one obtains a stable, arc-free magnetron plasma.
`Generally, one needs to use substrate temperatures higher
`than 700 °C to obtain crystalline alumina.7 This high sub-
`strate temperature requirement limits the type of substrates
`
`a兲Electronic mail: ywchung@nwu.edu
`J. Vac. Sci. Technol. A 18„5…, SepÕOct 2000
`
`2333
`
`II. EXPERIMENTAL PROCEDURES
`Aluminum oxide films were grown using pulsed dc mag-
`netron sputtering in a single-cathode deposition chamber
`共Fig. 1兲. The base pressure of the chamber was below 1
`⫻10⫺7 Torr. We used a 2 in. aluminum target 共99.99% pu-
`rity兲 in an unbalanced magnetron. The target voltage was
`pulsed at 20 kHz 共50% duty cycle兲, with the positive voltage
`set at 10% of the negative voltage 共e.g., at a set target bias of
`0734-2101Õ2000Õ18„5…Õ2333Õ6Õ$17.00
`
`©2000 American Vacuum Society
`
`2333
`
`INTEL 1220
`
`

`

`2334
`
`Li etal.: Low-temperature magnetron sputter-deposition
`
`2334
`
`FIG. 1. Schematic diagram of the deposition chamber.
`
`⫺400 V, the actual target voltage is a rectangular wave cy-
`cling between ⫺400 and ⫹40 V兲. The sputtering gas was an
`argon-8% oxygen mixture at a total pressure of 7 mTorr.
`Si共100兲 wafers were used as substrates. They were first
`cleaned by acetone and 2-propanol, then transferred into the
`chamber and reverse-sputter etched in an argon plasma at 60
`mTorr. In selected runs, glass substrates were used to con-
`firm visually the stoichiometry of alumina films 共stoichio-
`metric alumina films should be transparent兲. A 2 kHz pulsed
`dc bias 共similar pulse shape as that applied to the target兲 was
`applied to the substrate during deposition for low-energy ion
`bombardment and eliminating charging. This bias was varied
`from ⫺200 to ⫺400 V. Substrate temperatures were main-
`tained at or below 300 °C, as measured by a thermocouple
`attached to the substrate holder. A magnetic field was ap-
`plied 共via an external solenoid兲 during deposition to further
`increase the substrate ion-current density. The field at the
`substrate surface was 42 G. Unless otherwise stated, the
`thickness of all films was 300 nm.
`Auger electron spectroscopy was used to examine the
`chemical composition of the films. Film structure was char-
`acterized by standard x-ray diffraction using the Cu K␣line
`共0.154 nm兲 and transmission electron microscopy 共TEM兲.
`The surface topography was obtained by atomic force mi-
`croscopy 共AFM兲. The film hardness was measured using a
`Hysitron nanoindentor,
`analyzed by the Pharr–Oliver
`method.11 In the latter case, the penetration was kept at less
`than 15% of the film thickness to minimize substrate effects.
`Film stress was measured using the wafer curvature method.
`Through-thickness dc electrical resistivity was also deter-
`mined.
`III. RESULTS AND DISCUSSION
`A. Appearance
`At a nominal target power of 100 W, deposition under
`conditions described in the previous section resulted in clear
`
`J. Vac. Sci. Technol. A, Vol. 18, No. 5, SepÕOct 2000
`
`FIG. 2. Auger electron spectrum of a typical alumina film deposited on Si.
`The Al Auger peak is 51 eV, characteristic of oxidized Al.
`
`alumina films at 5 nm/min on substrates 13 cm from the
`target. Auger electron spectroscopy showed that the alumi-
`num in the films gives an Auger peak at 51 eV, confirming
`that it is in an oxidized state 共Fig. 2兲.
`
`B. Substrate temperature effect
`To examine the effect of substrate temperature on the
`growth of alumina, we deposited three films at the same
`nominal power 共100 W兲, pulsed substrate bias 共⫺300 V兲 and
`with the external solenoid on, but with different substrate
`temperatures, viz., 200, 250, and 300 °C. X-ray diffraction
`共Fig. 3兲 showed that crystalline alumina films can be grown
`at 250 and 300 °C and appear to consist of mixed phases. For
`example, the peak at ⬃25° suggests the presence of ␣alu-
`mina, while the peak at ⬃30° can be due to either ␪or ␬
`phase. Because of stress-induced shifting of diffraction peaks
`and nonrandom texture, we are unable to make specific
`phase identification and composition analysis at this time.
`The reference powder patterns for four typical alumina
`phases are included in this and other x-ray diffraction
`patterns.12 The alumina film grown at 200 °C is amorphous
`关the Si共200兲 forbidden reflection appears because of growth-
`induced stress兴.
`For the film grown at 300 °C, high-resolution TEM im-
`ages were taken at the Si substrate/film interface 共Fig. 4兲, as
`well as regions away from the interface 共Fig. 5兲 with the
`electron beam along the substrate 关110兴 direction. Fringes in
`the aluminum oxide region are indicative of its crystalline
`nature. This is also confirmed by the ring diffraction pattern
`taken from the same region of the film. Figure 6 shows that
`some amorphous regions still exist.
`Crystalline films grown at ⭓250 °C have nanoindentation
`hardness 18–21 GPa, in the same range as crystalline ␣alu-
`
`

`

`2335
`
`Li etal.: Low-temperature magnetron sputter-deposition
`
`2335
`
`FIG. 3. High-angle x-ray diffraction patterns of aluminum oxide films de-
`posited at 200, 250, and 300 °C 共nominal target power⫽100 W, pulsed
`substrate bias⫽⫺300 V, magnetic field on兲. The Si 共200兲 forbidden reflec-
`tion at 2␪⫽33° appears because of stress.
`
`mina 共sapphire兲. For amorphous films grown at 200 °C, the
`hardness is 12 GPa or less.
`
`C. Substrate bias effect
`In this series of experiments, we kept the nominal target
`power and substrate temperature constant
`共100 W and
`
`FIG. 5. High-resolution TEM image taken at the film region away from the
`film/substrate interface for the film shown in Fig. 3 共substrate temperature
`⫽300 °C兲. Randomly oriented aluminum oxide grains can be seen. The ring
`diffraction pattern in the inset shows a presence of polycrystalline aluminum
`oxide.
`
`300 °C, respectively兲, with the external solenoid on. The sub-
`strate bias was set at ⫺200, ⫺300, and ⫺400 V. Crystalline
`alumina films were obtained at ⫺300 and ⫺400 V substrate
`bias, while films grown at ⫺200 V were amorphous 共Fig. 7兲.
`Alumina films grown at 300 V substrate bias appear to have
`better quality than those at ⫺400 V. The diffraction peaks
`are stronger, and the films are harder 共20.7 vs 17.6 GPa兲.
`AFM indicates 共Fig. 8兲 that films grown at ⫺300 V are
`smoother 关root-mean-square 共rms兲 surface roughness⫽0.36
`nm兴 than those grown at ⫺400 V 共rms surface roughness
`⫽1.78 nm兲.
`Substrate bias has two opposing effects on crystal growth.
`On the one hand, substrate bias increases ion bombardment
`of the growing film, enhancing mobility of surface species.
`This makes it possible to grow crystalline alumina films at
`reduced temperatures. On the other hand, excessive ion bom-
`bardments can create defects at higher rates than can be re-
`paired by enhanced mobility of surface species. Further, ex-
`cessive compressive stress can result. Therefore, some
`optimum bias must be found.
`
`FIG. 4. High-resolution TEM image taken at the interface of Si and alumi-
`num oxide with the electron beam along the Si 关110兴 direction for the film
`shown in Fig. 3 共substrate temperature⫽300 °C兲. Fringes in the aluminum
`oxide region are evident, indicative of crystalline growth.
`
`D. Magnetic field effect and amorphization
`The next series of films were grown at a nominal target
`power of 100 W, substrate bias of ⫺300 V, and substrate
`
`JVST A - Vacuum, Surfaces, and Films
`
`

`

`2336
`
`Li etal.: Low-temperature magnetron sputter-deposition
`
`2336
`
`FIG. 6. High-resolution TEM image taken at the film region away from the
`film/substrate interface for the film shown in Fig. 3 共substrate temperature
`⫽300 °C兲. Both crystalline and amorphous phases are observed.
`
`temperature 300 °C, one set with the solenoid off and one set
`with the solenoid on. Separate measurements under the same
`process conditions showed that the ion-to-neutral arrival ra-
`tio at the substrate is 2.7 with the solenoid off and 5.2 on.13
`
`FIG. 8. AFM images of aluminum films deposited at different substrate bias
`voltages 共nominal target power⫽100 W, substrate temperature⫽300 °C,
`magnetic field on兲. 共a兲 Substrate bias⫽⫺300 V, rms surface roughness
`⫽0.36 nm and 共b兲 Substrate bias⫽⫺400 V, rms surface roughness⫽1.78
`nm.
`
`This approximate doubling of the ion-to-neutral arrival ratio
`has a dramatic effect on film crystallinity. Figure 9 shows
`that without the magnetic field provided by the solenoid,
`there is no crystalline growth.
`Together with the substrate bias, formation of a magnetic
`trap above the substrate surface increases ion bombardment
`of the growing film. As discussed earlier, this can enhance
`crystalline growth under optimum bias conditions. At the
`same time, such ion bombardment results in significantly in-
`creased film stress. For example, films grown with the sole-
`noid on 共Fig. 9兲 have internal compressive stress ⬃5 GPa,
`whereas films grown with the solenoid off have internal
`stress ⬃1 GPa. While thin crystalline alumina films appear
`to be stable 共from electrical resistivity measurements兲, thick
`共⬎100 nm兲 crystalline films grown under these conditions
`experienced partial amorphization 5–7 days after film depo-
`sition 共as shown by the marked decrease of x-ray diffraction
`intensity兲. This occurs even for films stored in a dessicator,
`suggesting that moisture is not a significant factor here.
`
`FIG. 7. High-angle x-ray diffraction patterns of aluminum oxide films de-
`posited at ⫺200, ⫺300, and ⫺400 V substrate bias 共nominal target power
`⫽100 W, substrate temperature⫽300 °C, magnetic field on兲. Crystalline
`phases are obtained at ⫺300 and ⫺400 V.
`
`E. Resistivity
`Through-thickness electrical resistivity of amorphous alu-
`mina films was measured as a function of the applied electric
`
`J. Vac. Sci. Technol. A, Vol. 18, No. 5, SepÕOct 2000
`
`

`

`2337
`
`Li etal.: Low-temperature magnetron sputter-deposition
`
`2337
`
`FIG. 11. Variation of through-thickness electrical resistivity with applied
`electric field for amorphous and crystalline alumina films. The two crystal-
`line films were grown at 100 W nominal target power, ⫺300 V substrate
`bias, 300 °C substrate temperature and with the magnetic field on.
`
`talline兲. The crystalline films were grown at 100 W nominal
`target power, ⫺300 V substrate bias, 300 °C substrate tem-
`perature and with the solenoid on. Except at the lowest elec-
`tric fields, crystalline alumina films generally have higher
`and more stable resistivity than amorphous ones with in-
`creasing electric field. This may be the result of crystalline
`alumina films having lower defect concentrations. These re-
`sults indicate that ultrathin alumina films may be useful for
`low-voltage electrical isolation applications.
`
`IV. CONCLUSIONS
`We demonstrate that stoichiometric alumina thin films
`can be synthesized by reactive magnetron sputtering with
`pulsed dc bias applied to the target and the substrate. Nor-
`mally, amorphous films were produced with hardness ⬍12
`GPa and high electrical resistivity. Under optimum process
`conditions and the formation of a magnetic trap between the
`target and the substrate, crystalline alumina films can be
`grown at substrate temperatures as low as 250 °C. The hard-
`ness is comparable to crystalline ␣alumina. Excellent elec-
`trical properties are maintained down to a thickness of 3 nm.
`
`ACKNOWLEDGMENTS
`This work was supported by the MRSEC program of the
`National Science Foundation 共DMR-9632472兲 at the Materi-
`als Research Center of Northwestern University and the IBM
`Storage System Division.
`
`1H. Nakai, J. Shinohara, T. Sassa, and Y. Ikegami, Nucl. Instrum. Methods
`Phys. Res. B 121, 125 共1997兲.
`2H. Wado, T. Shimizu, and M. Ishida, Appl. Phys. Lett. 67, 2200 共1995兲.
`3J. F. Moore, D. R. Strongin, P. B. Comita, M. W. Ruckman, and Myron
`Strongin, Appl. Phys. Lett. 65, 368 共1994兲.
`4C. Deshpandey and L. Holland, Thin Solid Films 96, 265 共1982兲.
`5W. D. Sproul, M. E. Graham, M. S. Wong, S. Lopez, D. Li, and R. A.
`Scholl, J. Vac. Sci. Technol. A 13, 1188 共1995兲.
`6M. S. Wong, W. J. Chia, P. Yashar, J. M. Schneider, W. D. Sproul, and S.
`A. Barnett, Surf. Coat. Technol. 86–87, 381 共1996兲.
`7B. Hirschauer, S. Soderholm, G. Chiaia, and U. O. Karlsson, Thin Solid
`Films 305, 243 共1997兲.
`
`FIG. 9. High-angle x-ray diffraction patterns of aluminum oxide films grown
`with and without the extra magnetic field 共nominal target power⫽100 W,
`substrate bias⫽⫺300 V, substrate temperature⫽300 °C兲. Crystalline alu-
`mina appears only when the field is on.
`
`field, film thickness ranging from 10 to 50 nm 共100 W nomi-
`nal target power, ⫺250 V substrate bias, with the solenoid
`off兲. There was no deliberate substrate heating. The substrate
`temperature was in the range of 100 °C. Figure 10 shows two
`general trends. First, the resistivity decreases with increasing
`film thickness. This is most likely due to the classical size
`effect. Second, the resistivity decreases with increasing elec-
`trical field. We speculate that this may be due to tunneling
`between conducting 共defective兲 regions in the film. The de-
`tailed mechanism is beyond the scope of our investigation.
`Another way to view the data is that all alumina films exhibit
`no through-thickness breakdown and maintain resistivity
`⬎109 ⍀ cm upon application of 10 V.
`Figure 11 shows the resistivity data for three alumina
`films 共10 nm amorphous, 3 nm crystalline, and 10 nm crys-
`
`FIG. 10. Variation of through-thickness electrical resistivity with applied
`electric field for amorphous alumina films 共nominal target power⫽100 W,
`substrate bias⫽⫺250 V, substrate temperature⫽100 °C, magnetic field off兲.
`
`JVST A - Vacuum, Surfaces, and Films
`
`

`

`2338
`
`Li etal.: Low-temperature magnetron sputter-deposition
`
`2338
`
`8J. M. Schneider, W. D. Sproul, A. A. Voevodin, and A. Matthews, J. Vac.
`Sci. Technol. A 15, 1 共1997兲.
`9I. Petrov, F. Adibi, J. E. Greene, W. D. Sproul, and W. D. Munz, J. Vac.
`Sci. Technol. A 10, 3283 共1992兲.
`10C. Engstrom, T. Berlind, J. Birch, L. Hultman, I. P. Ivanov, S. R. Kirk-
`patrick, and S. Rohde, Vacuum 56, 107 共2000兲.
`11G. M. Pharr and W. C. Oliver, Mater. Res. Bull. 17, 28 共1992兲.
`12Standard powder x-ray diffraction patterns in Figs. 3, 7, and 9 include
`peaks with intensity ⬎10% of the strongest peak for that phase. Refer-
`ences for ␣, ␪, and ␥alumina are 1997 JCPDS 46-1212, 1997 JCPDS
`
`35-0121, and 1998 JCPDS 47-1308. Reference for ␬alumina is M. Hal-
`varsson, V. Langer, and S. Vuorinen, Powder Diffr. 14, 61 共1999兲.
`13The ion-to-neutral arrival ratio was calculated as follows. The deposition
`rate is 300 nm/h, or 1000 monolayers/h 共assuming 0.3 nm⫽1 monolayer兲.
`Assuming a surface atom density of 1⫻1015 atoms/cm2, this is equivalent
`to an atom arrival rate of 2.78⫻1014 atoms/cm2 s. The substrate ion cur-
`rent density was measured to be 0.12 mA/cm2 with the solenoid off, and
`0.23 mA/cm2 with the solenoid on. Assuming singly charged ions, we
`calculated ion arrival rates of 7.5⫻1014 and 1.44⫻1015 ions/cm2 s respec-
`tively. The corresponding ion-to-neutral arrival rates are then 2.7 and 5.2.
`
`J. Vac. Sci. Technol. A, Vol. 18, No. 5, SepÕOct 2000
`
`