15

61

Research ArticleReceived: 16 April 2010 Accepted: 20 April 2010 Published online in Wiley Online Library: 22 June 2010

(wileyonlinelibrary.com) DOI 10.1002/sia.3606

Oxygen Influence on the Growth of Thin HfFilms on the 6H-SiC(0001) SurfacesPiotr Mazur and Leszek Markowski∗

Initial stages of Hf growth on the SiC(0001) surface and the influence of annealing on the evolution of microstructure of theformed Hf films were investigated by XPS and scanning tunneling microscopy (STM). The character of XPS spectra and the shiftsof the binding energies of Hf 4f, Si 2p, C 1s, O 1s levels, observed with increasing Hf depositions, indicate that even at roomtemperature a hafnium oxide in the form of HfO2 was created. The registered band bending created at the HfO2/SiC interfaceis about 0.4 eV. Hafnium tended to grow irregularly with a tendency to form nanocrystalline grains near terraces edges, andpreferably on the surface regions not covered by additional graphite carbon layers which together with oxygen were presenton the virgin SiC(0001) surface. The grain sizes were sensitive to annealing and efficiently increased during heating the sampleat 1100 ◦C. These surface morphology changes proceeded through merging smaller nanoparticels into clusters. For longer Hfdepositions, with average thickness over 10 monolayers, excess carbon segregated and migrated towards the surface, withsimultaneous formation of Hf–C and O-C-O bonds. Copyright c© 2010 John Wiley & Sons, Ltd.

Keywords: X-ray photoelectron spectra; high-k dielectrics; hafnium compounds; HfO2; scanning electron microscopy

Introduction

The current progress in semiconductors industry tends mainly infour parallel directions: miniaturization of circuit elements on an in-expressible scale; reducing its power consumption; increasing thespeed of operation; and the tolerance-temperature range. Owingto its excellent physical and electrical properties, silicon carbide(SiC) is well known as a wide band-gap semiconductor (3.0 eVfor the 6 H polytype) with high-power, high-temperature, high-voltage and high-frequency microelectronics applications.[1 – 2] Inparticular, its high thermal conductivity, increased tolerance toradiation damage and its wide band gap make this material veryuseful in all the fields, especially in high temperature regions(above 300 ◦C) as well as in nuclear industry and aerospace inves-tigations, where the conventional silicon-based electronic devicescannot work properly. It has also applications in optoelectronicsas well as in chemical and biological analysis.[3]

However, not only semiconductors with good properties butalso sufficiently good insulators are needed. Unfortunately, theinsulating layers based on SiO2 and used as field transistorgates have reached now the limit of the thickness for whichthe tunneling-current leakage has unacceptably high values.[4 – 5]

As a result, intensive efforts have started to search a materialhaving the dielectric constant higher than that of SiO2 for thegate oxide. Such a high-κ gate dielectric would allow for that, ata given capacitance, the gate oxide could be physically thicker,with a strongly reduced electron-tunneling effect. Among suchmaterials, oxides and silicates of Hf and Zr seem to be the mostpromising high-κ candidates of which, in particular, HfO2 hasdesirable attributes.[5] Its energy gap is equal to 5.26 eV[6] and thedielectric constants at 2 GHz is about 15.[7] Nevertheless, moreor less complex rare-earth and transition metal oxides attractconsiderable interest as well.[8]

In this work the growth of Hf (in actual fact HfOx) on the 6H-SiC(0001) surface, and its high temperature annealing behavior,

was investigated by X-ray photoelectron spectroscopy (XPS) andscanning tunneling microscopy (STM).

Experimental

Two kinds of 6H-SiC(0001) crystals were used as substrates in thisinvestigation. The substrates were N-doped 6H-SiC(0001) singlecrystals off oriented 3.5◦ from the basal plane with Si- or C-face and with epitaxial layer of SiC (P-doped, concentration 7.41015/ cm3 and of 0.035 �cm resistivity) of the thickness 10 µmdeposited on them (Cree Research Inc.). Typical size of sampleswas 3 × 7 mm2. All substrates, each time before inserting in thevacuum chamber, were cleaned by degreasing in acetone, boiled inpotassium dichromate, dipped in HF and finally washed in distilledwater. Such prepared substrates were placed in the ultrahighvacuum (UHV) apparatus (∼10−8 Pa), the one equipped with theVT STM/AFM system (Omicron) and used for surface morphologydetermination, and the other with the XPS analyzer Phoibos HSA-3500 equipped with monochromatic Mg Kα (1253.6 eV) and AlKα (1486.6 eV) X-ray sources and low-energy electron diffraction(LEED) spectrometer, for surface chemical and structure analysis.The X-ray impact angle was 60◦.

The XPS spectrometer was calibrated to yield the standardvalues of 87.63 eV for the Au 4f5/2 and 83.95 eV for the Au 4f7/2

lines. In addition to a survey spectrum, detailed scans (a passenergy of 10 eV) of the O 1s, C 1s, Si 2p, and Hf 4f regionswere acquired for each surface condition. All XPS scans weretaken at a photoelectron take-off angle of 90◦. Then, before

∗ Correspondence to: Leszek Markowski, Institute of Experimental Physics,Wroclaw University, pl. M. Borna 9, 50-205 Wroclaw, Poland.E-mail: lmark@ifd.uni.wroc.pl

Institute of Experimental Physics, Wroclaw University, pl. M. Borna 9, 50-205Wroclaw, Poland

Surf. Interface Anal. 2010, 42, 1561–1565 Copyright c© 2010 John Wiley & Sons, Ltd.

15

62

P. Mazur and L. Markowski

the Hf deposition procedure was started, each time the sampleswere annealed at 1000 ◦C by electron bombardment for severalminutes. Thin layers of Hf, up to about 10 monolayers (ML) thick,were formed in situ by evaporation from an Hf-wire spiral (99.9%purity, Good Fellow). Due to a different geometry in the STM andXPS chambers, the Hf deposition rates were also different. DuringHf depositions the pressure in the vacuum chambers dropped to2·10−7 Pa. All XPS and STM measurements were performed at roomtemperature. Additionally, some check measurements of particlesdesorbing from the sample surface, when XPS spectra were takenand annealing was performed, were made by quadrupole massspectrometer.

Results and Discussion

STM topographies of the 6H-SiC(0001) surfaces for different Hfdepositions times are shown in Fig. 1. As can be seen for a 15-minute of Hf deposition, which corresponds approximately to theaverage Hf thickness of 0.5 ML, the Hf covers the surface ratheruniformly and no island formation is observed. However, whenthe Hf evaporation is prolonged, small islands appear. They arealready seen for the Hf layer thickness as low as about 1 ML and itcan be easily noticed that they have tendency to be formed nearterrace edges (see Fig. 1b). Although the dimensions of terracesare much smaller for the samples with an SiC-epilayer formed onthe off-axis C-terminated surface, the island formation seems likethat of the previous case. If we look closer at the investigatedsurface by impoving the scanning resolution, we notice that theisland have a nanocrystalline structure indicating that during thegrowing process the small islands merge into clusters. The Hflayers are sensitive to the annealing procedure, which is clearupon a longer Hf deposition, for which HfOx, rather than Hf as willbe shown later, covers almost all the area of the sample surface.Figure 2 shows the changes in surface topography caused by aseries of 10 several-second electron beam flashes onto the sampleup to 1100 ◦C, for the 4 ML deposition of Hf on the samples withan epilayer grown on the C-terminated SiC. From these figures it isclearly seen that during heating the sample a mass redistributionoccurs, which decreases the surface roughness with simultaneousincrease of the HfOx island dimensions.

The chemical composition and bonding environments of theHfOx island growing on the 6H-SiC(0001) surface were investigatedusing XPS. Figure 3 shows the Hf 4f, Si 2p, C 1s and O 1s XPS spectraobserved for virgin samples and for samples with increasing Hf

depositions. For virgin samples, in the C 1s peak four componentscan be distinguished. The two main components located at283.0 eV and 284.7 eV correspond to a C–Si bond (carbon in siliconcarbide) and a C–C bond (carbon in graphite), respectively (seeFig. 3).[9 – 11] These values correspond very well to those observedby other authors, e.g. by Sha et al.[12] who have measured thesetwo peaks at 283.5 and 284.9 eV, respectively. However, the widthof the peak corresponding to the C-C bonds seems to be toolarge to originate from an ordered graphite layer, therefore werefer to it as a graphite-like peak. It is well known that annealingan SiC sample at temperatures above 1150 ◦C under UHV causesnot only desorption of carbon oxides and hydrocarbons from thesurface (compare the O 1s and C 1s spectra for the as-loaded andannealed samples shown in Fig. 3) but also leads to the substratedecomposition and formation of the carbon nanomesh bufferlayer[13] or graphite on the SiC surface,[14 – 15] which upon furtherheating is subsequently transformed into carbon nanotubes[16]

or can result in graphene overlayer formation.[17 – 18] However,in our case the graphite-like carbon is already present on thesubstrate surface of virgin samples, without any additional heatingprocedure applied. Moreover, it should be noticed that after suchan annealing procedure oxygen is still present (see Fig. 3) onthe surface. To completely remove it and simultaneously avoidsilicon depletion, heating over temperatures of 1000 ◦C and undera low Si flux is necessary,[18 – 19] which in these experiments wasnot applied. This is one of reasons why for annealing we chosetemperature lower than 1150 ◦C. Nevertheless, after annealing aclear LEED pattern of 1x1 structure related to bulk SiC, below somedisordered oxide, was observed.

Beside two dominating C 1s components identified as originatedfrom SiC and the graphite-like carbon, for virgin samples there aretwo smallest ones located on the high energy side, with thebinding energy (BE) at 286.7 eV and 289.0 eV, which disappearafter using the first annealing procedure. Thus, they clearly arerelated to the native surface contamination such as hydrocarbons(like CH3, CH2, CH and carbon oxide CO, respectively).[10,15] Theyresult from the exposure of the sample to air during transferring itto the XPS apparatus after its cleaning. It is known that oxidationproducts on the SiC surface appear already at room temperatureand that an oxygen atom does not penetrate very much intosilicon carbide and its insertion takes place below the surfaceclose to the first carbon atomic plane, leaving the Si adatomsunaffected.[17] As it was observed by other authors, hydrocarbonsdesorb already after annealing at 650 ◦C.[10] To thermally desorbthe oxide from 6H–SiC substrates the temperatures higher than

Figure 1. STM topographies of the 6H-SiC(0001) surfaces with hafnium depositions. (a) 15-min deposition of Hf on an SiC epilayer grown on the off-axisand Si-terminated surface. (b) The same, but with additional 45′ deposition; the nanocrystalite grains are evidently grown near terrace edges. (c) Hafniumoxide particles merged into clusters and formed on an SiC epilayer on the offaxis C-terminated surface with the same deposition as for (b). (30 min of Hfdeposition corresponds approximately to the average Hf film thickness of 1 monolayer).

View this article online at wileyonlinelibray.com Copyright c© 2010 John Wiley & Sons, Ltd. Surf. Interface Anal. 2010, 42, 1561–1565

15

63

Oxygen influence on the growth of thin Hf films

Figure 2. Temperature-induced changes of HfO2 grains formed on 6H–SiC(0001) surfaces after 4 ml Hf deposition. (a) Before flashing. (b) After tenseveral-second long flashes up to 1100 ◦C. (c) After the next series of flashing. (d) Surface roughness analysis.

940 ◦C are necessary.[14,20] In our case the first surface cleaningby annealing the samples was performed at 1000 ◦C, which wasa sufficiently high temperature to remove both these unwantedchemical components from the surface. The progress of surfacecleaning reveals by changes in shape, intensity and position of theO 1s peak. Initially, for as-loaded sample, the O 1s peak is locatedat a BE of 532.7 eV and is relatively broad. This peak position isrelated to the adsorbed OH groups and CO groups.[21] Then, afterannealing it is still broad, but shifts to 531.8 eV with its intensitytwice decreased indicating that now O atoms bond mainly withatoms of a lower electro-negativity or that now O occurs in suchcomponents in which the charge transfer to O is larger thanpreviously. Similar shifts of the O 1s peak (to almost the same BEof 531.7 eV) after flashing were observed also by other authors.[11]

Note that the observed increase in intensity (by a factor of 1.5) ofthe C 1s peak corresponding to the SiC bond indicates removalof the attenuation by the oxide layer. Simultaneous decreaseof the C-C component indicates that during sample annealingoxygen diffuses into the surface, interacts with the carbon of thegraphite-like form and then desorbs as CO or CO2. This statementis confirmed by the results of quadrupole mass spectroscopyperformed during annealing – the main desorbed component isCO. There is one more argument. Before annealing the estimatedconcentration of the three considered elements in the surfacelayer are as follows. O 17%, C 54% and Si 29%, and after annealingthese values change to: O 7%, C 51% and Si 42%. The Si 2pspectrum shows a single peak located at 100. 6 eV and this peakis attributed to bulk SiC.[20] However, it can be noticed that theshape and position of the Si 2p peak (at 100.7 eV) is almostunaffected by the annealing procedure performed at 1000 ◦C (see

Fig. 3) and only the Si 2p emission intensity is increased (by afactor of 1.3). At this temperature no Si 2p peak shift to a lowerbinding energy is observed, which is consistent with the fact thatSi atoms are being released when SiC decomposes to the carbonbuffer layer[13] above 1150 ◦C. Therefore, we can deduce that, afterannealing, about 10% of the surface area is still covered by oxygenand carbon in the form of Si–O–C and Si–O–Si bridges and C-Cgraphite-like carbon, but upon removing the oxide layer there isno change observed in the relative Si and C concentration in theSiC substrate.

For submonolayer Hf coverage (i.e. for Hf deposition timesshorter than 120 minutes, which was estimated to correspond tothe average Hf film thickness of 1 ML), small increases in BE ofabout 0.4 eV for both substrate peaks and decreases in BE of about0.3 eV for O and Hf have been observed. These BE changes arerelated to creation of Hf-O-SiC bonds during the submonolayerstages of Hf deposition and accompanied by a small negativecharge transfer from the substrate to the HfO2 created on it.[22]

It is evident that both low BE energy peaks, i.e the Hf 4f7/2 peakobserved at 18.0 eV and separated by 1.7 eV from the Hf 4f5/2 peakat 19.7 eV originate from Hf in the 4+ state bonded with oxygenatoms.[23 – 24] However, these two peaks look like rather broad,which indicates formation of an amorphous film with irregularHfO2 grains instead of a uniform layer, which is indeed confirmedby our STM investigations.

Upon further Hf deposition, as HfO2 has covered the larger areaon the substrate and HfO2 get thicker the O 1s, C 1s and Hf 4f peaksstay steady. On the other hand, the BE of Si 2p decreases slightly byabout 0.1 eV. This can be explained by the fact that, for a longer Hfdeposition, the top layer becomes oxygen depleted and thus the

Surf. Interface Anal. 2010, 42, 1561–1565 Copyright c© 2010 John Wiley & Sons, Ltd. View this article online at wileyonlinelibray.com

15

64

P. Mazur and L. Markowski

Figure 3. Normalized comparison of the Hf 4f, Si 2p, C 1s and O 1s XPS spectra observed for virgin samples and for samples with increasing Hf depositionsamount. Peaks for the graphite-like carbon and for the carbon bound with Si are easily distinguished. For all depositions, hafnium oxide in the form ofHfO2 is created. For longer Hf depositions, carbon segregates and forms HfC (this can be easily noticed in the C 1s spectra) and CO bonds. For comparisonHf 4f peaks for metallic hafnium are also shown. (120 min of Hf deposition corresponds approximately to the average Hf film thickness of 1 monolayer).

diffusion process of the small amount of residual oxygen atomsstill remaining at the HfO2/SiC interface and bounded originallyto Si atoms goes through the HfO2 film to the surface. Such asituation can be clearly seen from the Hf 4f spectrum for 1200minutes of Hf deposition - two additional broad peaks located atthe BE 15.0 eV and 16.6 eV appear. These two peaks cannot beattributed to metallic Hf, because in this form the maxima of Hf 4fdoublet are located at a much lower energy.[24] We have measuredalso the XPS spectrum for metallic Hf to have comparison with thatof Hf 4f in question. As it is shown in Fig. 3, the Hf 4f7/2 peak andthe Hf 4f5/2 peak are observed at 14.1 eV and 15.7 eV, respectively.

Thus this new doublet has to be attributed to the Hf-O bond inwhich Hf is in the 2+ state.

There is no indication of that low Hf depositions lead to theformation of hafnium carbide or hafnium silicide (Hf has a lowerelectronegativity than Si and thus pulls a weaker electron densityaway from C than in the case of SiC; therefore additional peakshould appear for HfC at a lower BE than that of the C-Si peak).Moreover, we have to notice that the C 1s peak position related tothe graphite-like carbon remains practically unchanged for all Hfdepositions made, although it maximum decreases with increasingamount of deposition. This means that, for small depositions, Hf

View this article online at wileyonlinelibray.com Copyright c© 2010 John Wiley & Sons, Ltd. Surf. Interface Anal. 2010, 42, 1561–1565

15

65

Oxygen influence on the growth of thin Hf films

does not adsorb on the SiC area covered by the graphite-likecarbon and that the HfO2 is initially formed at the places in whichoxygen is bonded with the surface Si atoms. This statement isconfirmed by the fact that, for HfO2 grown on epitaxial grapheneon the 4H-SiC substrates, a shift of core-level spectra from agraphene layer was observed, which evidently implies that chargetransfer takes place at the HfO/graphene interface.[18]

For longer Hf depositions (corresponding to the average Hfthickness of 10 ML) annealing the sample at 1100 ◦C led to theagglomeration of HfOx domains revealing bigger islands (STMevident) and Hf started to react with the rest of carbon located onthe surface to form hafnium carbide (see in Fig. 3 the new peakin the Hf 4f spectrum located at 282.2 eV).[25] For such samplesthe concentration of O, C and Hf elements in the surface layeris as follows: O 56%, C 15% and Hf 29%, and after annealing at1100 ◦C these values change to: O 58%, C 10% and Hf 32%. Thisindicates that some part of carbon reacts with oxygen to form CO2,which reduces the carbon content in the film, to finally evaporatefrom the sample. However, as can be seen this reaction is in directcompetition with the reaction between the Hf and C that forms theHfC grains. Indeed, similar observations have been already notedin the literature.[26]

Conclusion

It is shown that during Hf deposition under UHV conditions evenat room temperature an ultra-thin HfO2 layer is formed on the 6H-SiC(0001) surface. The results of the STM measurements indicatethat HfO2 is characterized by the island-like growth. Flashing thesample up to 1100 ◦C causes that some HfO2 clusters begin tocoalesce to form a bigger island, simultaneously smoothing thesurface. The flashing procedure used for a longer Hf depositionmakes the top surface layer be oxygen-depleted and beside HfO2

the HfO component is formed. Moreover, for such HfOx layers,the reaction between the Hf and C occurs, leading to formationof HfC grains. This reaction is in direct competition with theoxygen-carbon interactions, which creates the evaporating CO2,and finally reduces the carbon content in the film. The presenceof graphite-like carbon on the 6H-SiC(0001) surface does not needany additional heating procedure and it is already registered forthe virgin sample. The XPS spectra of the elements involved alsoexhibit a noticeable shift in the binding energy and are mainly

originated from the band bending effect. The observed bandbending at the HfO2/SiC interface amounts to 0.4 eV.

References

[1] E. Janzen, O. Kordina, Mater Sci. Eng. 1997, B 46, 203.[2] H. Matsunami, Microelectron. Eng. 2006, 82, 2.[3] S. Bet, N. Quick, A. Kar, Phys. Stat. Sol. (a) 2007, 204, 1147.[4] H. Wong, H. Iwai, Microelectron. Eng. 2006, 83, 1867.[5] G. D. Wilk, R. M. Wallace, J. M. Anthony, J. Appl. Phys. 2001, 89, 5243.[6] H. Y. Yu, M. F. Li, B. J. Cho, C. C. Yeo, M. S. Joo, D.-L. Kwong, J. S. Pan,

C. H. Ang, J. Z. Zheng, S. Ramanathan, Appl. Phys. Lett. 1997, 81, 376.[7] B. Lee, T. Moon, T.-G. Kim, Appl. Phys. Lett. 2005, 87, 12901.[8] T. Busani, R. A. B. Devine, J. Appl. Phys. 2005, 98, 44102.[9] I. Dontas, S. Ladas, S. Kennou, Diamond Relat. Mater. 2003, 12, 1209.

[10] C. M. Hollering, F. Maier, N. Sieber, M. Stammler, J. Ristein, L. Ley,A. P. J. Stampfl, J. D. Riley, R. C. G. Leckey, F. P. Leisenberger, F. P.Netzer, Surf. Sci. 1999, 442, 531.

[11] D. Schmeißer, D. R. Batchelor, R. P. Mikalo, P. Hoffmann, A. Lloyd-Spetz, Appl. Surf. Sci. 2001, 184, 340.

[12] Z. D. Sha, X. M. Wu, L. J. Zhuge, Y. D. Meng, Physica E 2006, 35, 38.[13] W. Chen, H. Xu, K. Ping Loh, A. T. S. Wee, Surf. Sci. 2005, 595, 107.[14] T. Maruyama, H. Bang, N. Fujita, Y. Kawamura, S. Naritsuka, M.

Kusunoki, Diamond Relat. Mater. 2007, 16, 1078.[15] Z. P. Guan, A. L. Cai, H. Porter, J. Cabalu, S. Huang, R. E. Giedd, Appl.

Surf. Sci. 2000, 165, 203.[16] T. Maruyama, H. Bang, Y. Kawamura, N. Fujita, K. Tanioku, T. Shiraiwa,

Y. Hozumi, S. Naritsuka, M. Kusunoki, Chem. Phys. Lett. 2006, 423,317.

[17] S. W. Poon, W. Chen, E. S. Tok, A. T. S. Wee, Appl. Phys. Lett. 2008, 92,104102.

[18] Q. Chen, H. Huang, W. Chen, A. T. S. Wee, Y. P. Feng, J. W. Chai,Z. Zhang, J. S. Pan, S. J. Wang, Appl. Phys. Lett. 2010, 96, 072111.

[19] K. Heinz, U. Starke, J. Bernhardt, J. Schardt, Appl. Surf. Sci. 2000,162–163, 9.

[20] M. Eremtchenko, J. Uhlig, A. Neumann, R. Ottking, S. I.-U. Ahmed,J. A. Schaefer, Surf. Sci. 2008, 602, 584.

[21] M. M. Beerbom, Z. Bednarova, R. Gargagliano, Y. N. Emirov, R. Schlaf,Appl. Surf. Sci. 2004, 236, 208.

[22] L. Muehlhoff, W. J. Choyke, M. J. Bozack, J. T. Yates Jr., J. Appl. Phys.1986, 60, 2842.

[23] G. He, M. Liu, L. Q. Zhu, M. Chang, Q. Fang, L. D. Zhang, Surf. Sci.2005, 576, 67.

[24] S. Lee, W.-G. Kim, S.-W. Rhee, K. Yong, J. Electrochem. Soc. 2008, 155,H92.

[25] J. H. Jang, T. J. Park, J. H. Kim, K. D. Na, W. Y. Park, M. Kim, Ch. S.Hwang, J. Electrochem. Soc. 2009, 156, H76.

[26] J. Zhang, Ch. Yang, Y. Wang, T. Feng, W. Yu, J. Jiang, X. Wang, X Liu,Nanotechnology 2006, 17, 257.

Surf. Interface Anal. 2010, 42, 1561–1565 Copyright c© 2010 John Wiley & Sons, Ltd. View this article online at wileyonlinelibray.com