Low-temperature polysilicon deposition by
`ionized magnetron sputtering
`
`Cite as: Journal of Vacuum Science & Technology A 18, 2006 (2000); https://doi.org/10.1116/1.582463
`Submitted: 28 September 1999 . Accepted: 27 March 2000 . Published Online: 10 July 2000
`
`Junghoon Joo
`
`ARTICLES YOU MAY BE INTERESTED IN
`
` Tutorial: Reactive high power impulse magnetron sputtering (R-HiPIMS)
`Journal of Applied Physics 121, 171101 (2017); https://doi.org/10.1063/1.4978350
`
`23% efficient p-type crystalline silicon solar cells with hole-selective passivating contacts
`based on physical vapor deposition of doped silicon films
`Applied Physics Letters 113, 061603 (2018); https://doi.org/10.1063/1.5037610
`
`High power impulse magnetron sputtering discharge
`Journal of Vacuum Science & Technology A 30, 030801 (2012); https://
`doi.org/10.1116/1.3691832
`
`Journal of Vacuum Science & Technology A 18, 2006 (2000); https://doi.org/10.1116/1.582463
`
`18, 2006
`
`© 2000 American Vacuum Society.
`
`Page 1 of 7
`
`APPLIED MATERIALS EXHIBIT 1014
`
`

`

`Low-temperature polysilicon deposition by ionized magnetron sputtering
`Junghoon Jooa)
`Department of Materials Science and Engineering, Kunsan National University, Kunsan 573-701 Korea
`~Received 28 September 1999; accepted 27 March 2000!
`
`Ionized magnetron sputtering was successfully applied to polycrystalline silicon thin-film deposition
`on glass substrate at temperatures lower than 250 °C maintaining a deposition rate of about 133
`Å/min. Hydrogen mixing was effective up to Ar:H2510:6 by mass flow rate. Prior to deposition,
`H2 inductively coupled plasma was used for precleaning the substrate with 240 V bias. During Si
`deposition, the substrate biasing scheme was in two steps; 120 V for an initial stage and 120 to
`240 V bipolar pulse bias for the rest of the deposition time. The crystallinity was evaluated by both
`x-ray diffraction analysis and Raman spectroscopy; the average crystalline fraction was calculated
`as 70%. Grain size was measured in plan-view scanning-electron micrographs after selective etching
`of the amorphous phase by chemical solution. In 800-nm-thick samples, grains are 500–700 Å in
`diameter. Optical emission spectroscopy was used as real-time diagnostics, and ionization of
`sputtered silicon atoms distinctly increased as the hydrogen partial pressure increased. The
`successful deposition of polycrystalline silicon was explained as being due to enhanced ionization
`of sputtered and reflected neutrals and resultant energy control by bipolar substrate bias. © 2000
`American Vacuum Society. @S0734-2101~00!13104-5#
`
`I. INTRODUCTION
`
`The thin-film transistor/liquid-crystal display ~TFT/LCD!
`industry has been experiencing a need for finer display capa-
`bility and higher charge-carrier mobility necessary in large-
`area projection devices. Currently, laser-annealed polycrys-
`talline silicon is used, but postannealing of plasma-enhanced
`chemical vapor deposited ~PECVD! amorphous silicon ~a-
`Si! requires high-power excimer laser recrystallization and
`has inherently low throughput. The cost is still high for mass
`production, and adsorbed gases during CVD will explode
`during high-power laser scaning to make blisters in addition
`to a rough surface. A new deposition technology for direct
`polycrystalline silicon should have manufacturability and
`high crystallinity. In CVD, mixed gases of SiH4 and H2 are
`used in PECVD for depositing poly- or microcrystalline Si
`film on various substrates such as quartz, glass, and single-
`crystalline silicon. Recently, Murata et al. reported that a
`two-step process could increase the crystallinity of the Si
`film at the fairly low substrate temperature of 300 °C.1 They
`maintained a very slow deposition rate of 0.2 Å/s at the
`initial nucleation stage, equivalent to one-tenth of the total
`film thickness, and made a charge-free state by installing
`strong magnets in front of a substrate. In the following step,
`normal electron cyclotron resonance PECVD was used. The
`idea of applying a magnetic field in a large-area deposition
`machine will not be easily realized; and CVD uses toxic or
`explosive gases such as SiH4 that are not environmentally
`friendly. In the magnetron sputtering arena, Yang and Abel-
`son reported that polycrystalline silicon with a mean grain
`diameter of ;400 Å was successfully deposited on a 100-Å-
`thick microcrystalline hydrogenated silicon seed layer.2 They
`also performed a kinetic study of incoming fluxes toward a
`substrate in a reactive magnetron sputtering process; fast
`
`a!Electronic mail: jhjoo@ks.kunsan.ac.kr
`
`neutral hydrogen atoms both implant and recoil, releasing
`surface H.3 In a heavily ionized process, such as pulsed laser
`ablation, there was a report from Trusso and Vasi;4 in appro-
`priate hydrogen partial pressure, polycrystalline Si film was
`successfully deposited on a glass substrate, which means de-
`positing ions with kinetic energy of several eVs that should
`assist the adatoms to crystallize without creating severe dam-
`age. In reactive magnetron sputtering of damage-sensitive
`materials, where fast reflected neutrals ~;100 eV! cause
`damage, we can control their energy either by ionizing and
`retarding with positive substrate bias, or by collision slow-
`down in high background gas pressure. In this study, we
`applied an ICP-based ionized magnetron sputtering method
`to deposit polycrystalline Si on glass substrate, where rela-
`tively high gas pressure ~.4 Pa! is used for efficient ioniza-
`tion of sputtered neutrals by electron-impact collisions. Ion-
`ization of depositing atoms ~like Si! and proper acceleration
`would increase the adatom’s surface mobility so that crystal-
`lization at low temperature becomes easier. It is very inter-
`esting that in high-density plasmas, the angular distribution
`of incoming ions has some correlation with defect formation
`and enhanced crystallization. The balance between those two
`processes can be steered by energy control of the incoming
`particles. The benefit from the ion-assisted deposition pro-
`cess results from: ~1! the parallel component of the incoming
`particle’s momentum that is induced either from collisions in
`a substrate sheath, or ~2! succeeding elastic collisions with
`adatoms present on the substrate surface, where the trans-
`ferred energy should be lower than the minimum lattice dis-
`placement energy of the growing film surface. In the case of
`using an ion gun, the incident angle determines this effect; in
`a plasma-immersed process, this occurs because of the gas
`pressure. For an example, the above ~Trusso and Vasi! re-
`ports that laser ablation of silicon deposits poly-Si on a glass;
`most of the depositing particles are ionized by high-power
`
`2006
`
`J. Vac. Sci. Technol. A 18(cid:132)4(cid:133), Jul(cid:213)Aug 2000
`
`0734-2101(cid:213)2000(cid:213)18(cid:132)4(cid:133)(cid:213)2006(cid:213)6(cid:213)$17.00
`
`©2000 American Vacuum Society
`
`2006
`
`Page 2 of 7
`
`

`

`2007
`
`Junghoon Joo: Low-temperature polysilicon deposition
`
`2007
`
`FIG. 1. Experimental setup. ICP is powered by a 2 MHz, 2500 W generator,
`and sputtering is done by dc power supply. The substrate biasing voltage is
`generated at square pulse signal generator ~max. 10 Vpp! and fed into a
`bipolar op-amp, where it can be amplified in the range of 2100 to 1100 V
`and also has variable dc offset. During pretreatment, a continuous 220 V is
`supplied, and in an initial stage of deposition, 120 V, and during main
`deposition stage, 120 V ~25% duty! and 240 V ~75% duty! of bipolar
`pulsed voltage is used for biasing.
`
`laser absorption, and the background gas pressure is as high
`as 88.6 Pa. So the depositing particles should have highly
`randomized angular distribution, where the perpendicular
`momentum component effectively enhances the adatom’s
`mobility. In this study, we used two strategies: one is to use
`high gas pressure and the other is to use bipolar pulse bias of
`the substrate to cause the incoming ions to relax their direc-
`tionality by oscillating the sheath at several tens of kHz. We
`expect that fast reflected neutrals should be slowed down
`either by high gas pressure or ionization, followed by appro-
`priate substrate bias; depositing atoms of Si should be ion-
`ized and accelerated to optimal energy for low-temperature
`crystallization on the substrate surface.
`
`II. EXPERIMENTAL METHOD
`Figure 1 shows our ionized magnetron sputtering system:
`a 50-mm-diameter magnetron cathode, and a 250-mm-
`
`FIG. 2. X-ray diffraction data show microcrystalline Si is deposited with ICP
`sputtering in pure Ar environment ~sputtering power dc 400 W; ICP 2 MHz,
`500 W, 1.33 Pa, 250 C, 60 min!.
`
`JVST A - Vacuum, Surfaces, and Films
`
`FIG. 3. X-ray diffraction results show that addition of hydrogen into sput-
`tering gas environment enhanced crystalline-silicon formation during depo-
`sition ~250 C, 60 min, sputtering dc 400 W, ICP 2 MHz, 400 W, 120–40
`V bipolar pulsed substrate biasing!, ~a! Ar:H2510:2, ~b! Ar:H2510:4, ~c!
`Ar:H2510:6.
`
`diameter, four turn, inductively coupled plasma-generating
`coil are installed in a 380-mm-diameter stainless-steel cham-
`ber. Substrate biasing power is supplied by a pulse signal
`generator, a bipolar operational amplifier ~Kepco, BOP™-
`100-1M!, with a low-pass filter unit to cut off interference
`from a 2-MHz-ICP power source ~ENI, GMW-2500™,
`2500W!. The pulse width can be varied from a few ns to ms,
`and the repetition rate can be increased up to 18 kHz. A
`substrate made of conventional soda-lime glass was posi-
`tioned 120 mm above the Si target, and the temperature was
`controlled from room temperature to 400 °C by a resistive
`heater block. High-purity gases ~99.999%! of Ar and H2 are
`introduced through a gas dose ring with small holes ~0.8 mm
`diameter! on the top surface. The chamber was evacuated to
`
`Page 3 of 7
`
`

`

`2008
`
`Junghoon Joo: Low-temperature polysilicon deposition
`
`2008
`
`FIG. 4. X-ray diffraction analysis and SEM micrograph show that films have good structural property and grain size from 300 to 700 Å. ~a!, ~b!, and ~c! are
`deposited in high-Ar1 bombardment condition, and ~d!, ~e!, and ~f! are deposited in low Ar1 bombardment condition.
`
`about 2.6631024 Pa by a turbomolecular pump ~Alcatel,
`5400CP™, 400 L/s! and baked at 150 °C for about 1 h. Prior
`to main deposition, hydrogen discharge cleaning using ICP
`was done for 20 min with a negative bias of 240 V. In the
`deposition stages, 120 V bias was applied for 10 min and
`switched to bipolar pulse bias ~120 V to240 V! continu-
`ously. Deposited samples were investigated by x-ray diffrac-
`tion analysis ~Brucker analytical, D-5005RA™, 18 kW rotat-
`ing anode, Cu Ka!. To observe the grain structure by
`scanning electron microscopy, the amorphous phase was se-
`
`lectively etched away by a chemical solution. Grain size dis-
`tributions were measured by software developed by Leica
`~Q-win™!, and Raman spectroscopy was done in a micro-
`Raman system using a 632.8-nm-He–Ne laser ~Renishaw,
`System-2000™!. The factors that affect the crystallinity and
`structural properties were investigated: substrate tempera-
`ture, hydrogen mixing ratio, ratio of ICP to sputtering power,
`thickness, substrate bias, and total pressure. Crystalline frac-
`tion was determined from Raman spectroscopy data after de-
`coupling of the amorphous and crystalline peaks by Lorent-
`
`J. Vac. Sci. Technol. A, Vol. 18, No. 4, Jul(cid:213)Aug 2000
`
`Page 4 of 7
`
`

`

`2009
`
`Junghoon Joo: Low-temperature polysilicon deposition
`
`2009
`
`zian fit.5 As a real-time plasma diagnostic, optical emission
`spectroscopy over the plasma volume was done through a
`quartz view port by an optical fiber and a CCD detector
`~Ocean optics, SQ-2000™!.
`
`III. RESULTS AND DISCUSSION
`
`A. Effects of hydrogen mixing
`
`Sputtered silicon in an Ar gas background only produced
`amorphous or nanocrystalline Si film at a moderate tempera-
`ture range ~less than 500 °C!. This was the same for ionized
`physical vapor deposition ~I-PVD!, as shown in Fig. 2. The
`ICP power and some other process variations were tried, but
`in most cases, they gave only poorly crystalline film. In this
`study, hydrogen mixing was tried by varying the mass flow
`ratio of Ar and H2 from 10:2 to 10:10. Total pressure
`changes from 4 to 4.4 Pa, respectively, and this change is not
`believed to change the slowdown distance of the fast neutrals
`significantly. In Fig. 3, the x-ray diffraction data show that
`the mixing ratio of Ar:H2510:6 is optimal in depositing
`polycrystalline silicon film. Mota et al. report that the origin
`of hydrogen mixing effects are that the dangling bonds be-
`tween Si and Si are reduced, based on interatomic potential
`calculation.6 The halo centered around 2u522° appears to
`be from the glass substrate. A broad band from the amor-
`phous Si phase is reported to be around 2u530°. In our
`experimental setup, the substrate and target surface is in-
`clined about 15°, and in each run, three slide glasses are
`positioned at
`the substrate holder, where three different
`samples ~left, center, and right position! were made at differ-
`ent distances from the target; so, they had different thick-
`nesses and bombarding ion flux compositions ~i.e., Ar1 and
`1!. In the L position, Ar1 is the main bombarding species;
`H2
`but in the R position, the portion of Ar1 should be decreased,
`with a lower deposition rate of silicon. As noted earlier, fast
`neutrals such as hydrogen atoms are believed to break Si–H
`bonds on the surface and make them desorb as a form of H2.
`As conventional magnetron sputtering is done in gas pres-
`sure lower than 0.2 Pa reflected neutrals fly a few centime-
`ters to a substrate, which is not enough distance to slowdown
`to room-temperature neutrals. As I-PVD is done in relatively
`high gas pressure of 5 Pa or higher, the fast neutrals must be
`slowed down first, and some of them are ionized by an elec-
`tron impact collision or a Penning process. Substrate biasing
`
`FIG. 5. Raman spectroscopy measurements show that optimal hydrogen
`mixing ratio is around Ar:H2510:6, and heating up to 400 C does not show
`much improvement compared to case of 250 C.
`
`can change the incoming particles energy distribution signifi-
`cantly. In the present study, a bipolar pulse bias scheme was
`used to solve the charging of the insulating substrate and to
`control the incoming ion’s energy. Energy-selective quadru-
`pole mass spectrometry is planned for future study to con-
`firm this hypothesis.
`
`B. Grain-structure observation
`
`To measure the crystalline fraction and the grain size, the
`amorphous phase in the deposited specimen was selectively
`etched by a chemical solution. In Fig. 4, the plan view of
`samples deposited in Ar:H2510:6 at different positions re-
`veals well-developed polycrystalline silicon surfaces. The
`sample in Fig. 4~a! has a grain-size distribution of about
`300–500 Å, and Fig. 4~b! has a grain-size distribution of
`about 300–700 Å. The fraction of bright crystalline area was
`measured by an image analysis program, and the average
`result was about 73%. From the cross-sectional views, nor-
`mal columnar structure appeared, and there was no abrupt
`change in microstructure between the the initial stage and the
`main deposition stage. The positive bias of 120 V in initial
`nucleation stage on crystalline formation is thought to be due
`to the effective kinetic energy control in high-density multi-
`component plasma, i.e., Ar, H, H2, Si, and their ions. As
`mass spectrometric study is scheduled, the plasma species in
`
`TABLE I. Crystalline fraction calculated from Raman data with y50.9 in I c /(I c1yI a) after deconvolution with
`Lorentzian fit.
`
`Deposition
`conditions
`
`Specimen i.d.
`
`Ar:H2
`
`T sub ~ C!
`
`990809a-R-2
`990809b-C
`990809b-L
`990810a-C
`990811a-C
`990821a-R
`
`10:4
`10:6
`10:6
`10:8
`10:6
`10:6
`
`250
`250
`250
`250
`150
`400
`
`a-Si peak
`area
`
`c-Si peak
`area
`
`Crystalline
`fraction ~x c)
`
`2.6904
`3.1983
`1.2396
`1.7088
`1.0645
`2.2426
`
`8.1074
`6.393
`2.8873
`2.1867
`2.3799
`6.4447
`
`0.77
`0.69
`0.72
`0.58
`0.71
`0.76
`
`c-Si
`position
`~cm21!
`
`513.6
`515.6
`517.8
`514.6
`508.3
`513.6
`
`JVST A - Vacuum, Surfaces, and Films
`
`Page 5 of 7
`
`

`

`2010
`
`Junghoon Joo: Low-temperature polysilicon deposition
`
`2010
`
`FIG. 6. Extended scan of Raman spectroscopy. Crystalline peak is shifted
`from 520 to 517.8 cm21 and also contains amorphous Si contribution around
`low-frequency shoulder.
`
`the ICP volume will be clarified, and we will confirm the
`results by OES. As for the ion energy distribution in I-PVD,
`Kusano et al. measured the ion kinetic energy in a metal
`sputtering system.7 In this study, Ar1 and Ti1 have peak
`energy higher than 150 eV with zero termination resistance
`of the ICP coil ~as in our study!, whereas Ti1 has a broader
`peak than Ar1. They used an unshielded metal antenna that
`draws a large electron current to the ground that induces
`large negative coil potential but in our study, the Cu tube
`antenna was shielded from plasma by an insulating material,
`so the plasma potential should not be as high as in their case.
`
`C. Raman spectroscopic measurements
`
`In Fig. 5, the frequency shift in Raman spectroscopy rep-
`resents crystallinity, and the crystalline fraction can be de-
`duced. A sample deposited at 150 °C showed a broad amor-
`phous band around 480 cm21, and the crystalline peak was
`also shifted toward 500 cm21. As a reference, the Raman
`shift of a silicon wafer was measured and gave a sharp peak
`at 520.4 cm21. Considering the peak between 500 and 520
`cm21 as a crystalline silicon peak, the crystalline fraction can
`
`FIG. 7. Optical emission spectra show sputtered Si atom is ionized and
`hydrogen atoms are present in the ICP volume.
`
`J. Vac. Sci. Technol. A, Vol. 18, No. 4, Jul(cid:213)Aug 2000
`
`FIG. 8. Optical emission intensity shows that Ar ionization is reduced as the
`amount of hydrogen addition in ~a!, but the ionization ratio of sputtered Si is
`steadily increasing as in ~b!. Hb/Hg remains constant while each intensity
`decreases in ~c!.
`
`be calculated as I c /(I c1yI a); y50.9. The resultant fractions
`are shown in Table I. Our samples have crystalline fractions
`between 0.58 and 0.77. Other deposition conditions were
`chosen from our previous experiences on this experimental
`system: plasma diagnostics,
`tuning network modification,
`ICP frequency effect, and sputtering power to ICP power
`ratio.8–10 The positions of the crystalline peaks range from
`508.3 to 517.8 cm21 whereas the Si wafer has a main peak at
`520.4 cm21 in this experiment. The offset from the bulk Si
`peak would indicate the quality of crystallization of the film
`and the residual stress that is common in plasma-deposited
`films. This can be confirmed from the shift of the x-ray dif-
`fraction peak. The lattice parameter of
`standard sili-
`
`Page 6 of 7
`
`

`

`2011
`
`Junghoon Joo: Low-temperature polysilicon deposition
`
`2011
`
`con powder is 5.4309 Å,11 and calculated values of our cur-
`rent study range from 5.4668 to 5.4459 Å. Based on this
`lattice spacing, I-PVD silicon films of this study have a ten-
`sile strain of about 0.66%–0.28%. This can be a factor to
`shift the crystalline Raman peak as in Fig. 6.
`
`D. Plasma diagnostics: Optical emission
`spectroscopic measurements
`
`In I-PVD, gas pressure should be high enough to effec-
`tively thermalize and ionize fast-sputtered neutrals. Once
`ionized, the kinetic energy can be controlled by appropriate
`substrate bias. Optical emission spectra can be a good quali-
`tative measure in estimating the composition of volume
`plasma. In the wide-range spectrum of Fig. 7, Si0 ~576.29
`nm!, Si1 ~597.89 nm!, Ar0 ~675.28 nm!, Ar1 ~668.42 nm!,
`and Hg ~656.52nm! are detected. In other channels of an
`optical multichannel analyzer, Ha~434nm!, Hb~486nm!, and
`SiH* ~414 nm! lines are detected, as opposed to Bae et al.;
`they reported that in capacitively coupled rf plasma, Ha and
`Hb lines were not noticeable.12 The tendency of emission
`intensities was followed as hydrogen partial pressure in-
`the ratio of ionic silicon @5Si1/~Si11Si0)]
`in-
`creases;
`creases steadily whereas the argon ionization ratio remains
`constant in Figs. 8~a! and 8~b!. Among hydrogen derivatives,
`H was emitting the strongest line, and the ratio of Hb to Hg
`decreased slightly and does not change much beyond 2 sccm
`of hydrogen mixing in argon @Fig. 8~c!#. In I-PVD, the main
`advantage is thought to be the ionization of depositing sili-
`con flux and retarding the ionized fast-reflecting neutrals
`such as hydrogen atoms. This explains why the x-ray diffrac-
`tion results are insensitive to substrate temperatures between
`150 and 250 °C.
`
`E. Effects of magnetron discharge voltage
`
`In most of the cases in this study, a continuous dc power
`source was used to sputter silicon in Ar1H2. As a test, a
`bipolar pulse power supply ~ENI, RPG-10™, 10 kW, 100–
`250 kHz! was used to maintain an equivalent film deposition
`rate as the dc power source, but at most it gave smaller x-ray
`diffraction peaks. Pulsing in the several hundreds of kHz
`range tends to modulate the ion kinetic energy by sheath
`oscillation and might alter incoming ions’ energy toward
`
`higher values. The discharge voltage was found to be a sen-
`sitive parameter in depositing crystalline silicon in I-PVD.
`From a series of experiments, samples with magnetron dis-
`charge voltage higher than 600 V did not show well-
`developed polycrystalline peaks in x-ray diffraction data.
`With a 6-mm-thick target, the discharge voltage at 400 W
`was about 754 V at the beginning. Eroding the sputtering
`target ~so that the leakage magnetic flux over the racetrack in
`the target becomes strong! reduced the discharge voltage
`down to a final value of 378 V in the lifetime. This is closely
`related to the reflected fast neutrals ~mainly consisting of H!.
`In a future study, the effects of magnetic field configuration
`~field intensity, balanced, or unbalanced! should be ad-
`dressed in detail.
`
`IV. CONCLUSION
`Polycrystalline-silicon film was successfully deposited by
`ionized magnetron sputtering in a mixed gas environment of
`Ar and H2 at a substrate temperature as low as 250 °C, where
`the crystalline fraction based on the Raman data was around
`70% and the grain size was 500–700 Å.
`
`ACKNOWLEDGMENT
`This work was supported by the G7 Project for Flat-Panel
`Display of Korea and the ministry of education of Korea,
`1998.
`
`1K. Murata, M. Ito, M. Hori, and T. Goto, J. Vac. Sci. Technol. B 17, 1098
`~1999!.
`2Y. H. Yang and J. R. Abelson, Appl. Phys. Lett. 67, 11~1995!.
`3J. R. Abelson, L. Mandrell, and J. R. Doyle, J. Appl. Phys. 76, 1856
`~1994!.
`4S. Trusso and C. Vasi, J. Vac. Sci. Technol. A 17, 921 ~1999!.
`5A. T. Voutsas, M. K. Hatalis, J. Boyce, and A. Chiang, J. Appl. Phys. 78,
`6999 ~1995!.
`6F. Mota, J. F. Justo, and A. Fazzio, J. Appl. Phys. 86, 1843~1999!.
`7E. Kusano, T. Saitoh, T. Kobayashi, K. Fukushima, N. Kikuchi, H.
`Nanto, and A. Kinbara, J. Vac. Sci. Technol. A 17, 2360~1999!.
`8J. Joo, J. Vac. Sci. Technol. A 17, 2368 ~1999!.
`9J. Joo, J. Korean Vac. Soc. 7, 225 ~1998!.
`10A. A. Mayo, S. Hamaguchi, J. H. Joo, and S. M. Rossnagel, J. Vac. Sci.
`Technol. B 15, 1788 ~1997!.
`11B. D. Cullity, Elements of X-Ray Diffraction ~Addison-Wesley, Reading,
`MA, 1978!, p. 507.
`12S. H. Bae, Y. C. Lee, and R. T. McGrath, 46th American Vacuum Society
`International Symposium 1999, Seattle, WA.
`
`JVST A - Vacuum, Surfaces, and Films
`
`Page 7 of 7
`
`