Ion and Photon Surface Interaction during Remote PlasmaALD of Metal Oxides

H. B. Profijt,* P. Kudlacek, M. C. M. van de Sanden, and W. M. M. Kessels*,z

Department of Applied Physics, Eindhoven University of Technology, 5600 MB Eindhoven, The Netherlands

The influence of ions and photons during remote plasma atomic layer deposition (ALD) of metal oxide thin films was investigatedfor different O2 gas pressures and plasma powers. The ions have kinetic energies of �35 eV and fluxes of�1012–1014 cm�2 s�1 to-ward the substrate surface: low enough to prevent substantial ion-induced film damage, but sufficiently large to potentially stimu-late the ALD surface reactions. It is further demonstrated that 9.5 eV vacuum ultraviolet photons, present in the plasma, candegrade the electrical performance of electronic structures with ALD synthesized metal oxide films.VC 2011 The Electrochemical Society. [DOI: 10.1149/1.3552663] All rights reserved.

Manuscript submitted October 13, 2010; revised manuscript received November 30, 2010. Published February 25, 2011. This wasPaper 1406 presented at the Las Vegas, Nevada Meeting of the Society, October 10–15, 2010.

The interest in plasma-assisted atomic layer deposition (ALD)has grown considerably over the last few years, as evidenced by theincreased number of plasma-assisted ALD reactors installed andprocesses developed. This is a result of several studies that demon-strated the merits offered by plasma-assisted ALD, such asimproved material properties, higher throughput, additional controlover material properties, more freedom in precursors and processes,and facilitated deposition at reduced substrate temperatures.1–3 Forthermal ALD of metal oxides typically H2O vapor, O2 gas, or O3 areemployed for the oxidation reactions, whereas during plasma-assisted ALD the oxidation reaction of the surface and ligands isdriven by reactive radical species present in the O2 plasma. How-ever, besides radicals, additional species such as electrons, ions, andphotons are present in the plasma. Their densities and energies varywith the reactor geometry, the plasma source type, the gas pressureand flow, and the plasma power.4 Plasma–surface interaction duringplasma-based thin film processing is a topic addressed extensivelyin the literature, but not for the specific case of plasma-assistedALD. Therefore, the presence of ions and photons during the plasmaexposure step of plasma-assisted ALD was studied to identify theinteractions of ions and photons with the surface and the potentialeffect on the ALD growth process. Experiments were carried outunder conditions typical for remote plasma ALD of metal oxides inthree inductively coupled plasma setups used for ALD: the OxfordInstruments FlexAL and OpAL reactors, and the home-built ALD-Ireactor. In this article it is demonstrated that, although not alwaystaken into account, the ions and photons present in plasmas duringplasma-assisted ALD can influence the deposition process and thematerial quality significantly.

Experimental

The kinetic energy Ei and flux Ci of ions accelerated to the sub-strate surface were studied using an Impedans Semion retardingfield energy analyzer (RFEA)5 and a tungsten planar current collect-ing probe, respectively. Using the RFEA the flux of ions passingthrough a system of biased grids was measured, from which the ionenergy distribution (IED) was determined. The planar probe with anarea of 1 cm2 was used to measure the total ion current to the sub-strate, from which the ion flux Ci was calculated. Additionally, dou-ble Langmuir probe measurements were conducted at 5 mm abovethe center of the substrate stage to nonintrusively measure the elec-tron temperature Te and the electron density ne in the plasma.

The visible emission in the 400–800 nm range was recorded withan OceanOptics USB2000þ spectrometer, whereas vacuum ultra-violet (VUV) and ultraviolet (UV) emission in the 100–400 nmrange was detected by a differentially pumped McPherson 234/302

monochromator. The relative emission intensities were recordedfrom the position directly above the substrate stage.

Most plasma measurements were carried out in the OxfordInstruments FlexAL and OpAL reactors, for O2 pressures of 3.8–187.5 mTorr and powers in the 100–500 W range. Only the emissionstudy was carried out in the home-built ALD-I reactor, at pressuresof 3.8–37.5 mTorr and powers of 50–300 W. An O2 gas pressure pof 7.5 mTorr and a plasma power P of 100 W were used as standardconditions.

Results and Discussion

Ions and their surface interaction.— A typical IED measured bythe RFEA is displayed in Fig. 1. The majority of the ions arrive atthe substrate surface with a kinetic energy Ei of 27.3 6 1.0 eV,which is the ion energy where the IED is at its maximum value.This kinetic energy obtained by the ions in the plasma sheath isdetermined by the difference between the plasma potential Vp andthe substrate potential Vs. The energy has a peak value equal to eVp,as the substrate stage is grounded, and it is distributed due to fluctu-ations in Vp and collisions between the ions and other plasmaspecies in the sheath. The corresponding average ion flux Ci is(5.0 6 0.3)� 1013 cm�2 s�1. The energy and flux of the ions can beattributed mainly to singly ionized O2 molecules, as reported byGudmundsson et al.6 The values of Te and ne are 3.4 6 1.0 eV and(3.0 6 0.6)� 108 cm�3, respectively. The electron density directlyabove the substrate stage is approximately two orders of magnitudelower compared to the density in the plasma bulk.6

The pressure p and power P dependence of Ei, Te, Ci and ne isillustrated in Fig. 2. For pressures �22.5 mTorr, measurements werecarried out in the FlexAL reactor, whereas for pressures � 85 mTorrdata was obtained from measurements in the OpAL reactor. Ei isgiven only for p � 22.5 mTorr, because at higher pressures the ioncurrent was too low to be measured by the RFEA. The change in thepeak ion energy and the electron temperature show a similar trendupon a change in p and P, because the plasma potential depends onthe electron temperature. Obviously, the electron density and the ionflux are also related. In electropositive processing plasmas, such asthe one used in this work, the electron density and ion density in thebulk plasma are equal and a change in the electron density directlyaffects the ion flux.6 Upon a pressure increase, both Ei and Te

decrease due to the fact that more electron–molecule collisions takeplace at higher pressures increasing the ionization rate. As a result,the plasma can be sustained at a lower electron temperature explain-ing its decrease with increasing pressure. The electron density andion flux also decrease at higher pressures for the position directlyabove the substrate stage, possibly due to a slightly reduced out-diffusion of the plasma species from the plasma source. At higherplasma powers, more power is coupled into the plasma, which alsoresults in a higher ionization rate and lower Te and Ei. Obviously,the ion flux and density increase with increasing power as well. The

* Electrochemical Society Active Member.z E-mail: w.m.m.kessels@tue.nl

Journal of The Electrochemical Society, 158 (4) G88-G91 (2011)0013-4651/2011/158(4)/G88/4/$28.00 VC The Electrochemical Society

G88

) unless CC License in place (see abstract).  ecsdl.org/site/terms_use address. Redistribution subject to ECS terms of use (see 128.123.35.41Downloaded on 2014-08-29 to IP

magnitude of Te indicates that, although the plasmas in the FlexALand OpAL are generated remotely, they are still ionizing above thesubstrate over the complete pressure and power range measured inthis study.

The amount of energy the ions provide to the substrate surfacecan be illustrated by comparing the growth-per-cycle, in terms ofatoms deposited per unit area, with the ion dose to the surface withinone cycle. Table I summarizes this data for four metal oxide systemsdeposited in the FlexAL reactor. These materials were depositedunder saturated and optimized plasma-assisted ALD conditions aspreviously reported by Potts et al.3 and Heil et al.4 The ion dose percycle and the average energy per atom deposited are provided for aplasma generated at 15 mTorr O2 gas pressure and 300 W plasmapower (Ei¼ 17.8 eV, Ci¼ 3.2� 1013 cm�2 s�1). Consequently, forevery few atoms deposited there is one ion leading to an average ionenergy dose of one or a few electron volts per atom. The dose ishighest for deposition processes such as TiO2 that provide only alow atomic growth-per-cycle and require a relatively long plasmadosing time. From the data and the fact that good material propertieswere reported by Potts et al. and Heil et al., it can also be concludedthat the ion doses and energies are not high enough to induce sub-stantial ion-induced damage. On the other hand, as reported byTakagi the ion energies and fluxes are within the range to potentiallystimulate the ALD surface reactions, e.g., through ligand desorptionand adatom migration.7 As illustrated in Fig. 2, the potential influ-ence of the ions can be limited mainly by increasing the O2 gaspressure.

Photons and their surface interaction.— Emission spectra in theVUV (100–200 nm), UV (200–400 nm), and visible (400–800 nm)range of the optical spectrum were recorded at standard conditions.Figures 3a and 3b show a number of emission peaks correspondingto photons emitted after decay of electronically excited atoms andions. In the visible range (Fig. 3b), the spectrum is similar to whatwas observed and described in more detail by Mackus et al.8 In theUV and VUV range (Fig. 3a), an intense peak was observed at130.5 6 0.1 nm, corresponding to 9.5 eV photons (transition: 3s 3S0

! 2p4 3P). Besides this emission peak, this spectral range does notreveal other intense emission lines (the peak at 261.0 6 0.1 nm is asecond order effect of the diffraction grating used in the monochro-mator). High energy VUV photons are likely to be re-absorbed bythe plasma,9 so it is expected that the photons emitted toward thesubstrate surface are produced directly above the substrate surface,e.g., by de-excitation of O atoms in the plasma or by ion–electron

recombination at or near the sample surface. The intensity of the130.5 nm peak, a qualitative measure for its photon flux, decreaseswhen going to higher pressures and increases at higher powers, asillustrated in Figs. 3c and 3d, respectively.

It has been reported that high energy VUV photons have the abil-ity to affect the performance of electronic devices with metal oxidethin films negatively during plasma processing.10–15 To investigatethe influence of VUV photons on the material properties of filmsprepared by ALD, a number of plasma exposure experiments wereconducted in the FlexAL reactor. During the experiments the minor-ity charge carrier lifetime of c-Si wafers passivated by Al2O3 wasmonitored using a Sinton WCT-100 tool.16 This parameter is verysensitive to electrical defects near the c-Si/Al2O3 interface,11–13

whereas Al2O3 is a commonly applied dielectric prepared by ALD.For the experiment, double side polished float-zone<100> c-Siwafers (3.5 X cm, n-type) were used, double side deposited with a30 nm Al2O3 passivation layer and annealed for 10 min at 400�C.These Al2O3 layers were deposited by thermal ALD to avoid theinfluence of plasma photons at the initial stage of the experiment.The resulting effective lifetimes seff, i.e., the minority charge carrierlifetimes at a charge injection level of 1015 cm�3, in the Si sub-strates were measured to be �6 ms. The high lifetime is a result of

Figure 1. (Color online) Ion energy distribution (IED) as measured for anO2 plasma at standard conditions in the FlexAL reactor, with the RFEAplaced at the substrate stage. The dashed line indicates the peak ion energyEi of 27.3 6 1.0 eV. The inset shows the data from the original I–V measure-ment from which the IED was derived.

Figure 2. (Color online) Pressure and power dependence of (a) the peak ionenergy Ei, (b) the electron temperature Te, (c) the ion flux Ci, and (d) theelectron density ne in an O2 plasma used for plasma-assisted ALD. Forp� 22.5 mTorr, measurements were performed in the FlexAL reactor,whereas for p� 85 mTorr, measurements were carried out in the OpAL.Lines serve as a guide to the eye.

Journal of The Electrochemical Society, 158 (4) G88-G91 (2011) G89

) unless CC License in place (see abstract).  ecsdl.org/site/terms_use address. Redistribution subject to ECS terms of use (see 128.123.35.41Downloaded on 2014-08-29 to IP

the excellent passivation of silicon by the ALD-prepared Al2O3

films.17 The effect of the high energy VUV radiation was demon-strated by exposing the samples to an O2 plasma for different timeintervals. The resulting effective lifetime versus the plasma expo-sure time is illustrated in Fig. 4, for three samples exposed at differ-ent plasma conditions. The results clearly show a decrease in theeffective lifetimes for samples under plasma exposure. The decreasein lifetime depends on the intensity of the VUV radiation (see Figs.3c and 3d) as it is evident that the lifetime decreases faster for expo-sures at higher plasma powers, but slower when higher gas pressuresare employed. This automatically also implies that the effect ofplasma radiation on the lifetime can be controlled by choosing theright plasma parameters.

The experiments were repeated under standard conditions for thesituation in which samples were exposed to the radiation through5.0 mm thick MgF2 and quartz windows (blocking plasma radiation<110 nm and <140 nm, respectively). In this way the potentialinfluence of ion bombardment on the lifetime degradation can beexcluded and the role of the 9.5 eV photons can be identified. Thetransparency of the MgF2 window for 130.5 nm radiation is approxi-mately 60%, and after correcting for the resulting lower photon fluxthrough this window the results were included in Fig. 4. For sampleexposure through the MgF2 window, the lifetime as a function ofthe plasma exposure time is comparable to the situation in which nowindow was used at all. For the case with the quartz window virtu-ally no lifetime degradation was observed. Therefore, it is obviousthat the VUV photons are responsible for the plasma exposuredamage.

According to the literature 9.5 eV photons have enough energyto depassivate hydrogen-passivated Si dangling bonds at the

interface between the Si and the Al2O3 and/or to create charge trapsin dielectric materials whose bandgap is lower than the VUV photonenergy.10,11 For 9.5 eV photons, the latter holds for the Al2O3, butalso for the 1–2 nm thick interfacial SiOx layer that generally formsbetween the Si and the Al2O3. The decrease in the minority chargecarrier lifetime, observed during the experiments reported on in thiswork, can therefore be attributed to an increased density of defectstates at the interface, induced by VUV photons. These resultsexplain the observations by Dingemans et al. who compared the life-time of as-deposited c-Si substrates coated with Al2O3 films depos-ited by thermal ALD and by plasma-assisted ALD.12,13 In theirstudy, lifetimes in the microsecond range were found for substratescoated with Al2O3 deposited by plasma-assisted ALD, whereas life-times in the millisecond range were observed for Al2O3 depositedby thermal ALD. The difference in lifetimes can therefore be attrib-uted to the fact that the plasma-assisted ALD samples were exposedto VUV radiation. After a post-deposition anneal the lifetimes weresimilar for the plasma-assisted and thermal ALD deposited sub-strates,12 which implies that the photon-induced lifetime degrada-tion can be repaired by such a postdeposition anneal. It was verifiedthat this holds for the thermal ALD Al2O3 samples (shown in Fig. 4)degraded by the VUV radiation as well. This confirms that theresults in Fig. 4 reflect the influence of VUV photons duringplasma-assisted ALD, despite the fact that the experiments reportedin this work were carried out on substrates with thermally depositedAl2O3 films. More evidence for the detrimental influence of VUVphotons during plasma-assisted ALD of Al2O3 (as-deposited) onsilicon is also provided by the recent observation that the lifetimedegradation during plasma-assisted ALD is lower for shorter plasmaexposure steps.13

Table I. Comparison between the growth-per-cycle in terms of atoms deposited and the ion dose for four metal oxides deposited under opti-

mized ALD conditions in the FlexAL reactor. The precursor [Me is methyl, CH3; Et is ethyl, C2H5; and iPr is isopropyl, CH(CH3)2], the plasma

dosing time per cycle (tplasma), the growth-per-cycle, the ion-to-atom ratio, and the average energy provided to an atom deposited (Ei-per-atom)

are given for depositions performed at p 5 15 mTorr and P 5 300 W. References to publications in which the optimization of the ALD processes

and the resulting material properties are described in more detail are indicated. The data demonstrates that the energy provided by the ions can

be significant, depending on the growth-per-cycle obtained and/or plasma dosing time required for the process.

Material Precursortplasma

(s)Growth-per-cycle(cm�2 cycle�1)

Ion dose per cycle(cm�2 cycle�1)

Ion-to-atomratio (–)

Ei-per-atom(eV) Reference

Al2O3 Al(Me)3 2 1.1� 1015 6.4� 1013 0.06 1.1 3

HfO2 Hf(NMeEt)4 3 8.5� 1014 9.6� 1013 0.11 2.0 4

TiO2 Ti(O iPr)4 12 3.7� 1014 3.8� 1014 1.03 18.4 3

Ta2O5 Ta(NMe2)5 5 6.3� 1014 1.6� 1014 0.26 4.6 3

Figure 3. (Color online) Emission spectrumof an O2 plasma as measured in (a) the VUVand UV region (100–400 nm) and (b) thevisible region (400–800 nm). The wave-length k and the photon energy Ep are givenon the lower and upper horizontal axis,respectively. The second order peak iscaused by the diffraction grating in themonochromator and is associated with the130.5 6 0.1 nm emission line. Additionally,the emission intensity at k¼ 130.5 nm peakis given as a function of (c) the O2 gas pres-sure p and (d) the plasma power P.

Journal of The Electrochemical Society, 158 (4) G88-G91 (2011)G90

) unless CC License in place (see abstract).  ecsdl.org/site/terms_use address. Redistribution subject to ECS terms of use (see 128.123.35.41Downloaded on 2014-08-29 to IP

Conclusions

The potential influence of ions and photons during plasma-assisted ALD was investigated for different O2 gas pressures andplasma powers. The ions present in the remote O2 plasma have ki-netic energies of �35 eV and fluxes of �1012–1014 cm�2 s�1. Theenergy provided by the ions can vary from one up to a few electronvolts per atom deposited, depending on the growth-per-cycle of theprocess and the plasma exposure time required to reach saturation ofthe ALD cycle. The energy dose is low enough to prevent substan-tial ion-induced film damage, but sufficiently large to potentiallystimulate the ALD surface reactions. The electron temperature andelectron density reveal that for the remote plasma systemsemployed, the plasma is still ionizing at the substrate stage level for

the pressure (3.8–187.5 mTorr) and power (100–500 W) range stud-ied. The emission spectrum of the plasma revealed the presence ofphotons with energies of 9.5 eV in the plasma. These VUV photonsare energetic enough to create electrical defects during plasma-assisted ALD processes, which have unambiguously been demon-strated in this work. The plasma-induced damage can be reduced bytuning the gas pressure and plasma power.

Acknowledgments

G. Dingemans, C. van Helvoirt, M. J. F. van de Sande, J. J. L. M.Meulendijks, J. J. A. Zeebregts, and H. M. M. de Jong are acknowl-edged for assisting with the experiments. This work is carried outwithin the Thin Film Nanomanufacturing (TFN) Programme and issupported financially by the Dutch Technology Foundation STW.

Eindhoven University of Technology assisted in meeting the publicationcosts of this article.

References

1. S. B. S. Heil, E. Langereis, F. Roozeboom, M. C. M. van de Sanden, and W. M.M. Kessels, J. Electrochem. Soc., 153, G956 (2006).

2. W. M. M. Kessels, S. B. S. Heil, E. Langereis, J. L. van Hemmen, H. C. M.Knoops, and M. C. M. van de Sanden, ECS Trans., 3, 183 (2007).

3. S. E. Potts, W. Keuning, E. Langereis, G. Dingemans, M. C. M. van de Sanden,and W. M. M. Kessels, J. Electrochem. Soc., 157, P66 (2010).

4. S. B. S. Heil, J. L. van Hemmen, C. J. Hodson, N. Singh, J. H. Klootwijk, F. Roo-zeboom, M. C. M. van de Sanden, and W. M. M. Kessels, J. Vac. Sci. Technol. A,25, 1357 (2007).

5. D. Gahan, B. Dolinaj, and M. B. Hopkins, Rev. Sci. Instrum., 79, 033502 (2008).6. J. T. Gudmundsson, I. G. Kouznetsov, K. K. Patel, and M. A. Lieberman, J. Phys.

D: Appl. Phys., 34, 1100 (2001).7. T. Takagi, J. Vac. Sci. Technol. A, 2, 382 (1984).8. A. J. M. Mackus, S. B. S. Heil, E. Langereis, H. C. M. Knoops, M. C. M. van de

Sanden, and W. M. M. Kessels, J. Vac. Sci. Technol. A, 28, 77 (2010).9. A. C. Fozza, A. Kruse, A. Hollander, A. Ricard, and M. R. Wertheimer, J. Vac.

Sci. Technol. A, 16, 72 (1998).10. C. Cismaru, J. L. Shohet, J. L. Lauer, R. W. Hansen, and S. Ostapenko, Appl.

Phys. Lett., 77, 3914 (2000).11. A. Stesmans and V. V. Afanasev, Appl. Phys. Lett., 80, 1957 (2002).12. G. Dingemans, R. Seguin, P. Engelhart, M. C. M. van de Sanden, and W. M. M.

Kessels, Phys. Status Solidi (RRL), 4, 10 (2010).13. G. Dingemans, N. M. Terlinden, D. Pierreux, H. B. Profijt, M. C. M. van de

Sanden, and W. M. M. Kessels, Electrochem. Solid State Lett., 14, H1 (2011).14. A. W. Flounders, S. A. Bel, and D. W. Hess, J. Electrochem. Soc., 140, 1414

(1993).15. G. S. Upadhyaya and J. L. Shohet, Appl. Phys. Lett., 90, 072904 (2007).16. R. A. Sinton and A. Cuevas, Appl. Phys. Lett., 69, 2510 (1996).17. B. Hoex, S. B. S. Heil, E. Langereis, M. C. M. van de Sanden, and W. M. M.

Kessels, Appl. Phys. Lett., 89, 042112 (2006).

Figure 4. (Color online) Effective lifetime seff measured for a float-zone<100> c-Si wafer (3.5 X cm, n-type), polished and deposited with 30 nmAl2O3 at both sides, as a function of the O2 plasma exposure time tplasma. Thedecrease in lifetime is demonstrated for different pressure and power levels.Results are also given with the substrate covered by windows blocking(Quartz) and not blocking (MgF2, �60% transmission corrected) the VUVemission at 130.5 nm. The Al2O3 film was deposited by thermal ALD andannealed for 10 min at 400�C before plasma exposure. Lines serve as a guideto the eye.

Journal of The Electrochemical Society, 158 (4) G88-G91 (2011) G91

) unless CC License in place (see abstract).  ecsdl.org/site/terms_use address. Redistribution subject to ECS terms of use (see 128.123.35.41Downloaded on 2014-08-29 to IP