Thin Solid Films 520 (2012) 4911–4915

Contents lists available at SciVerse ScienceDirect

Thin Solid Films

j ourna l homepage: www.e lsev ie r .com/ locate / ts f

Valence band offset at GaN/β-Si3N4 and β-Si3N4/Si(111) heterojunctions formed byplasma-assisted molecular beam epitaxy

Mahesh Kumar a,b, Basanta Roul a,b, Thirumaleshwara N. Bhat a, Mohana K. Rajpalke a,A.T. Kalghatgi b, S.B. Krupanidhi a,⁎a Materials Research Centre, Indian Institute of Science, Bangalore- 560012, Indiab Central Research Laboratory, Bharat Electronics, Bangalore-560013, India

⁎ Corresponding author.E-mail address: sbk@mrc.iisc.ernet.in (S.B. Krupanid

0040-6090/$ – see front matter © 2011 Elsevier B.V. Alldoi:10.1016/j.tsf.2011.10.051

a b s t r a c t

a r t i c l e i n f o

Article history:Received 12 March 2011Received in revised form 12 October 2011Accepted 21 October 2011Available online 28 October 2011

Keywords:NitridesMolecular beam epitaxyX-ray photo-emission spectroscopy

Ultra thin films of pure β-Si3N4 (0001) were grown on Si (111) surface by exposing the surface to radio- fre-quency nitrogen plasma with a high content of nitrogen atoms. Using β-Si3N4 layer as a buffer layer, GaN epi-layers were grown on Si (111) substrate by plasma-assisted molecular beam epitaxy. The valence band offset(VBO) of GaN/β-Si3N4/Si heterojunctions is determined by X-ray photoemission spectroscopy. The VBO at theβ-Si3N4 / Si interface was determined by valence-band photoelectron spectra to be 1.84 eV. The valence bandof GaN is found to be 0.41±0.05 eV below that of β-Si3N4 and a type-II heterojunction. The conduction bandoffset was deduced to be ~2.36 eV, and a change of the interface dipole of 1.29 eV was observed for GaN/β-Si3N4 interface formation.

© 2011 Elsevier B.V. All rights reserved.

1. Introduction

Gallium nitride (GaN) and its related materials have been widelystudied for their unique applications in optoelectronic and high tem-perature/ high power electronic devices with relatively low powerconsumption [1]. A number of studies are being performed to growhigh quality of GaN epilayers on different substrates such as Al2O3,SiC, GaAs and Si [2–3]. Silicon is considered as one of the most prom-ising substrates for the GaN epitaxy because of its many advantagessuch as high-quality, large size, low cost and a well known existingdevice technology [4]. However, there exist several hindranceswhich prevent formation of high quality layers. Much difference inthe lattice constant and the thermal-expansion coefficients of GaNand Si, the deformation fields and corresponding generation ofhigh-density dislocations at the interface of GaN layers grown on Si(111) is much higher than that of homoepitaxially grown films. Toovercome these difficulties, recently, it has also been reported thatthe use of an ultrathin β-Si3N4 interlayer is advantageous for growinghigh-quality GaN epitaxial films on Si (111) [5–9]. Silicon nitrides canalso be used as dielectric materials due to their unique physical andchemical properties, such as a large dielectric constant as well as ex-cellent thermal, chemical, and mechanical stabilities. For both dielec-tric and III-nitrides-on-Si heterojunction applications, deviceperformance critically depends on the interface properties, such as

hi).

rights reserved.

valence band offsets (VBOs) and conduction band offsets (CBOs)and interface quality in terms of roughness and absence of interfacedefects. The CBO and VBO at the interface of two semiconductors orsemiconductor and insulator are among the most important proper-ties of the heterojunction. The band offsets (BOs) have a large influ-ence on, among other things, the ease of carrier transport across theinterface and electron–hole pair recombination velocity upon irradia-tion with super band gap light. Accurate BO measurements are essen-tial to understand the relationship between interface physicalstructure, electronic structure, and carrier transport [10].

Semiconductor or insulator interfaces can be organized into threetypes of heterojunctions: straddling gap (type I), staggered gap (typeII) and broken gap (type III). There are three relevant material prop-erties, band gap, electron affinity and work function which decidethe type of heterojunction. A approach to describing a heterostruc-ture interface is to apply the electron affinity model (EAM). Thismodel holds in an ideal case, where there is no potential created asthe heterostructure is formed. Alternatively, a deviation from theEAM can be represented as a change in the interface dipole. Theband alignment between two semiconductors is controlled by thecharge transfer across the interface and the resulting interface dipolein a fashion similar to Schottky barrier models. Recently, many re-searchers has been employed charge neutrality levels (CNL) to relatethe relative contribution of the electron affinity model and the inter-face dipole in determining the band offsets [11–12]. X-ray photoelec-tron spectroscopy (XPS) has been demonstrated to be a direct andpowerful tool for measuring the valence band (VB) discontinuitiesof heterojunctions [13–15]. In this article, we determined the VBO

9

4912 M. Kumar et al. / Thin Solid Films 520 (2012) 4911–4915

and CBO values of the GaN/β-Si3N4 /Si heterojunctions characterizedin [9] by XPS.

2. Experiment details

Four sampleswere used in this XPS experiment, namely, (a) Si (111)substrate, (b) ~2 nmof β-Si3N4 on Si (111), (c) 245-nm-thick GaN layergrown on β-Si3N4 /Si (111) and (d) 5 nm GaN on β-Si3N4/Si (111).Thegrowth system used in this study was a plasma-assisted molecularbeam epitaxy system equipped with a radio frequency plasma source.The base pressure in the system was below 1×10-8 Pa. The undopedSi (111) substrates were chemically cleaned followed by dipping in 5%HF to remove the surface native oxide. The substrates were thermallycleaned at 900 °C for 1 h in ultra-high vacuum. Schenk et al.[16]reported that the Si(111)-7×7 surface reconstruction which becomesvisible at 780 °C changes into an unreconstructed 1×1-surface withstrongKikuchi lines above 850 °C, but is recovered to the crystalline sur-face on cool down prior to the growth. Hence, the substrates have beencooled down at 530 °C before growth. For sample (b), the β-Si3N4 layerwas grown by nitridation–annealing–nitridation process, in which firstthe nitridation of the substrate was carried out at 530 °C for 30 min fol-lowed by annealing at 900 °C for 30 min and again nitridation at 700 °Cfor 30 min, the details of growth conditions can be found elsewhere [9].For sample (c), first the β-Si3N4 layer were grown then a low tempera-ture GaN buffer layer of 20 nmwere grown at 500 °C, where the Ga ef-fusion cell temperature was kept at 950 °C and corresponding beamequivalent pressure was maintained at 5.6×10-5 Pa. The GaN growthrate is 80 nm per hour. Thereafter, a GaN epilayer of thickness 225 nmwere grown on the buffer layer at 700 °C. For sample (d), a 2 nm bufferlayer at 500 °C followed by 3 nm GaN layer at 700 °C was grown on β-Si3N4 layer. Nitrogen flow rate and plasma power were kept to 0.5sccmand 350W, respectively for the nitridation, buffer layer and subse-quent GaN growth. The crystal structures were characterized using aBruker-D8 DISCOVER high-resolution X-ray diffraction (HRXRD) appa-ratus with CuKα1 source. The XPS measurements were performedusing a Thermo Scientific Multilab-2000 instrument with Al Kα

(energy of 1486.6 eV) as the X-ray radiation source, which had beencalibrated utilizingwork function and Fermi energy level (EF). The pres-sure in the XPS system was ~2x10-8 Pa during the measurements. Theenergy resolution of analyser was around 10meVwithmonochromaticelectron source. Core-level spectrawere obtained using a pass energy of30 eV and 0.01 eV steps. These conditions result in a spectra resolutionof ~500 meV. XPS peak-fit analysis was performed using XPSpeak4.1software application.

Fig. 1. (Color online) X-ray 2θ-ω diffraction pattern of the thick GaN films on Si (111)substrate.

3. Results and discussion

Fig. 1 shows the HRXRD 2θ-ω scans of thick GaN films (sample(c)). Beside the Si (111) and (222) diffracted peaks at 2θ=28.45°and 2θ=58.87°, respectively, only a strong GaN (0002) peak at2θ=34.59° and a weak GaN (0004) peak at 2θ=73° can be identi-fied, indicating the epitaxial GaN film to be highly oriented alongthe [0001] direction of the wurtzite GaN. The intensity of Si (222) re-flection is observed less than the Si (111) and this can be explained byB. E. Warren et al. [17] report that states, the intensity of the

Fig. 2. (Color online) Si 2p core-level spectra taken for (i) Si (111) substrate and(ii) β-Si3N4/ Si (111). (iii) Valence-band spectrum of β-Si3N4 / Si (111). Two lead-ing edges of valence bands from β-Si3N4 and Si can be observed in this spectrum.

Fig. 3. (Color online) XP spectra of (i) Ga2p3/2 and (ii) valance band spectra for thick GaN. Experimental data points are fitted by Voigt (mixed Lorentzian–Gaussian) line shapesafter the application of a Shirley background (BG).

4913M. Kumar et al. / Thin Solid Films 520 (2012) 4911–4915

forbidden Si (222) reflection should be very less than the signals ofthe allowed (111) diffraction peak.

In Fig. 2(i)–(ii) Si 2p core-level spectra (CLS) were recorded for Si(111) and crystalline β-Si3N4. For Si with β-Si3N4 overlayer, the thick-ness of β-Si3N4 layer is about ~2 nm as estimated by XPS depth pro-file. The Si 2p signal of silicon nitride has a higher binding energyrelative to that of bulk silicon because of a different bonding environ-ment. In these measured spectra, there is no 2p signal resulting fromsilicon oxide [18]. The CLS of β-Si3N4/Si(111) have been numericallyfitted using Shirley backgrounds (BG) and Voigt (mixed Lorentzian–Gaussian) functions. In stoichiometric silicon nitride a Si atom isbonded to 4N atoms (Si4+) and an N atom is bonded to 3 Si atoms[19], thus only the component with Si4+ is expected. At the interfacetwo more coordinations are required (Si3+ and Si1+) for an idealmatch between the silicon nitride and Si (111) lattices [20].Fig. 2(ii) also shows the compositions: one bulk and three other com-ponents assigned to silicon in Si1+, Si3+ and Si4+ coordination areneeded. The corresponding binding energies are at +0.66, +2.20,+3.05 eV respectively (with respect to the bulk position). Fig. 2(iii)shows the valence-band (VB) spectrum from an ultrathin β-Si3N4

layer and this VB spectrum contains both the photoelectron spectros-copy signals from the β-Si3N4 film and from the Si (111) substrate. Asa result, there are two slopes of the valence band edges, as reported inthe literature for the Si3N4 / Si films [21–22]. The VBO value can be de-termined to be 1.84±0.05 eV from the binding energy differencefrom two valence-band leading edges. Taking the bandgap values ofSi and β-Si3N4 as 1.1 eV and 5.3 eV, respectively, the band alignment

Table 1XPS core level spectra fitting results and VBM positions obtained by linear extrapola-tion of the leading edge to the extended base line of the VB spectra.

Samples States Binding energy(eV)

Bonding

(a) Si Si 2p 99.1 Si–Si(b) β-Si3N4/Si Si 2p (Si+4)

Si 2pVBM (β-Si3N4)

102.199.12.6

Si–N (Si+4)Si–Si

(c) GaN(245 nm)/β-Si3N4/Si Ga2p3/2

VBM1117.61.9

Ga–N

(d) GaN(5 nm)/β-Si3N4/Si Ga2p3/2

Si 2p (Si+4)Si 2p

1118.7102.099.1

Ga–NSi–N (Si+4)Si–Si

between β-Si3N4 and Si is shown in Fig. 5 and the corresponding CBOis 2.36±0.05 eV.

According to Kraut's method, the VBO value of GaN/β-Si3N4 can becalculated from the formula [23]

ΔEν ¼ ΔECL− EGaNGa2p3=2−EGaNVBM

� �þ Eβ−Si3N4

Si2p −Eβ−Si3N4VBM

� �ð1Þ

where ΔECL=(EGa2p3/2

GaN/β− Si3N4−ESi2pGaN/β− Si3N4) is the energy difference

between Ga 2p3/2 and Si 2p (β-Si3N4) core levels which are measuredin the heterojunction sample (d). (EGa2p3/2

GaN−EVBMGaN)and (ESi2pβ− Si3N4−

EVBMβ− Si3N4) are the GaN thick film and β-Si3N4 constants, and VBM de-

notes for valance band maximum. The Ga 2p3/2 and VBM spectra ofthe sample (c) are shown in Fig. 3. The Ga 2p3/2 fitted using Shirleybackgrounds (BG) and Voigt (mixed Lorentzian–Gaussian) functions,and the values of the parameters are shown in Table 1. Since consid-erable accordance of the fitted line to the original measured data hasbeen obtained, the uncertainty of the CL position should be lowerthan 0.03 eV, as evaluated by numerous fittings with different param-eters. The existence of impurities on the surface should prevent theprecise determination of the position of the valence band maximum(VBM). To reduce the contamination effect, all the samples were sub-jected to a surface clean procedure by Ar+bombardment in thevacuum chamber equipped in the XPS instrument and reduced athickness of ~1–2 nm calculated by the sputtering time. There is pos-sibility that surface composition may change due to Ar+ bombard-ment but we have minimized it by using low energy (0.5 KeV) ionbombardment for short time (~1–2 min) [24]. The valence bandmaximum (VBM) positions in the VB spectra were determined bylinear extrapolation of the leading edges of VB spectra to the baselines in order to account for the finite instrument resolution[25–26]. The scatter of the data with respect to the fit are estimatedas uncertainty in VBM positions of less than 0.05 eV, which are alsoestimated from fitting with different parameters. The Ga2p3/2 and Si2p XPS spectra of GaN/β-Si3N4 (sample (d)) are shown in Fig. 4.

The VBO value of GaN/β-Si3N4 heterojunction is calculated to be0.41±0.05 eV by substituting those values obtained in experi-ments into Eq. (1). The CBO (ΔEc) can be estimated by the formulaΔEC=(Eg

GaN–Egβ-Si3N4)–ΔEv. Taking the recently reported band gap

value (EgGaN=3.42 eV, Egβ−Si3N4=5.3 eV), the ΔEC is calculated tobe ~2.29 eV. The band alignment diagram of the β-Si3N4/Si(111)and GaN/ β-Si3N4 heterojunctions are shown in Fig. 5. Accordingto these measured results, a type-II band alignment is seen to

Fig. 4. (Color online) XP spectra of (i)) Ga2p3/2 and (ii) Si2p of β-Si3N4 for GaN/β-Si3N4/Si sample.

4914 M. Kumar et al. / Thin Solid Films 520 (2012) 4911–4915

have formed at the β-Si3N4/Si (111) heterojunction. These calcu-lated values of GaN/β-Si3N4/Si interferences are close to valuesobtained previously by Cook et al. [27] for both Si/SiN and SiN/GaN interfaces.

The electron affinities for GaN and β-Si3N4 are represented inFig. 5. The electron affinity model of heterojunction formation as-sumes that the vacuum levels would align at the interface. The resultshows a deviation from the electron affinity model (EAM) of ~1.29 eVand this deviation is determined directly from the difference in mea-sured ionization energy minus the measured band offset. The EAMmodel is based on the premise that the interface is formed withoutdisruption of the surface electronic states of either of the two mate-rials. The measured difference between the prediction of the EAMand the experimentally observed band offset represents a change inthe interface dipole. As an alternative to the electron affinity model,it may be considered that the heterojunction band offsets are deter-mined by alignment of the charge neutrality levels (CNLs) of thetwo materials. The charge neutrality levels represent the branchpoint of the surface (or interface) states since they are related to thevalence or conduction band. Charge can transfer between the inter-face states of the two materials, which will cause an interface dipole.The band offset will be determined by the relative position of the CNL

Fig. 5. (Color online) Schematic energy band diagram of GaN/β-Si3N4/Si interfaces. Thedashed line at the top corresponds to the electron affinity of the β-Si3N4 surface, whichis common in both interfaces. The deviation from the electron affinity model is shownby Δ, and the charge neutrality level is indicated by a dotted line within the band gap.The VBOs are determined from the measurements.

of the two materials if the density of states is high. The modelwas used to analyze silicon/oxide and Si3N4/GaN interfaces is givenby [27]

φCBO ¼ φCNL;a−φCNL;b

� �− Eg;a−Eg;b� �

þ S χa−χbð Þ þ Eg;a−Eg;b� �

− φCNL;a−φCNL;b

� �n oð2Þ

where φCBO is the CBO, χ and φCNL are the electron affinities andcharge neutrality levels of each semiconductor (a and b), respectively,and S is a pinning factor based on the dielectric properties of the ma-terials. A value of S=1 represents the EAM while a value of 0 repre-sents pinning at the CNL levels. Using the value of S=0.51 and φCNL,

Si=0.2 eV for the Si3N4–Si interface [28] and the measured CBO ofβ-Si3N4/Si heterojunction, we can deduce the CNL of β-Si3N4 to be1.56 eV above the valence band maximum. Monch et al. [29] hasreported the CNL of GaN to be 2.37 eV above the valence band maxi-mum and by using Eq. (2) the dielectric screening parameter can bededuced for the GaN/β-Si3N4 interface using the CNL of GaN andSi3N4. A value of S=0.24 is obtained for the GaN/β-Si3N4 interface.

4. Conclusion

An ultra thin β-Si3N4 layer was grown on Si (111) surface by ex-posing the surface to radio-frequency nitrogen plasma and using β-Si3N4 layer as a buffer layer, GaN epilayers were grown on Si (111)substrate. The VBO of the GaN/β-Si3N4/Si heterojunctions has beenmeasured by XPS. The valence band of β-Si3N4 is found to be 1.84±0.05 eV below the Si valance band and 0.41±0.05 eV above theGaN. The conduction band offset of ΔEc~2.36 eV and 2.29 eV areobtained for β-Si3N4/Si and GaN/β-Si3N4 heterojunctions, respective-ly. Knowledge of the band alignment parameters of a GaN/β-Si3N4/Siheterojunction is believed to facilitate the design of GaN and Si baseddevices. The pinning factor for the GaN/β-Si3N4 interface wasdeducted 0.24.

References

[1] S. Nakamura, Science 281 (1998) 956.[2] S. Tripathy, S.J. Chua, P. Chen, Z.L. Miao, J. Appl. Phys. 92 (2002) 3503.[3] W.Y. Uen, Z.Y. Li, S.M. Lan, S.M. Li, J. Cryst. Growth 280 (2005) 335.[4] A. Dadgar, F. Schulze, M. Wienecke, A. Gadanecz, J. Blasing, P. Veit, T. Hempel, A.

Diez, J. Christen, A. Krost, New J. Phys. 9 (2007) 389.[5] C.L. Wu, J.C. Wang, M.H. Chan, T.T. Chen, S. Gwo, Appl. Phys. Lett. 83 (2003) 4530.[6] S. Vezian, A. Le Louarn, J. Massies, J. Cryst. Growth 303 (2007) 419.[7] E. Arslan, M.K. Ozturk, S. Ozcelik, E. Ozbay, Curr. Appl. Phys. 9 (2009) 472.

4915M. Kumar et al. / Thin Solid Films 520 (2012) 4911–4915

[8] N. Yamabe, H. Shimomura, T. Shimamura, T. Ohachi, J. Cryst. Growth 311 (2009)3049.

[9] M. Kumar, B. Roul, T.N. Bhat, M.K. Rajpalke, P. Misra, L.M. Kukreja, N. Sinha, A.T.Kalghatgi, S.B. Krupanidhi, Mater. Res. Bull. 45 (2010) 1581.

[10] A. Franciosi, C.G. Van de Walle, Surf. Sci. Rep. 25 (1996) 1.[11] J. Robertson, J. Vac. Sci. Technol. B 18 (2000) 1785.[12] T.E. Cook Jr., C.C. Fulton, W.J. Mecouch, R.F. Davis, G. Lucovsky, R.J. Nemanich, J.

Appl. Phys. 94 (2003) 7155.[13] A.L. Yang, H.P. Song, X.L. Liu, H.Y. Wei, Y. Guo, G.L. Zheng, C.M. Jiao, S.Y. Yang, Q.S.

Zhu, Z.G. Wang, Appl. Phys. Lett. 94 (2009) 052101.[14] H.P. Song, A.L. Yang, H.Y. Wei, Y. Guo, B. Zhang, G.L. Zheng, S.Y. Yang, X.L. Liu, Q.S.

Zhu, Z.G. Wang, T.Y. Yang, H.H. Wang, Appl. Phys. Lett. 94 (2009) 222114.[15] P.D.C. King, T.D. Veal, C.E. Kendrick, L.R. Bailey, S.M. Durbin, C.F. McConville, Phys.

Rev. B 78 (2008) 033308.[16] H.P.D. Schenk, G.D. Kipshidze, U. Kaiser, A. Fissel, J. Kräuβlich, J. Schulze, W. Rich-

ter, J. Crystal Growth 200 (1999) 45.[17] B.E. Warren, X-Ray Diffraction, Dover, New York, 1990.[18] J.H. Oh, H.W. Yeom, Y. Hagimoto, K. Ono, M. Oshima, N. Hirashita, M. Nywa, A.

Toriumi, A. Kakizaki, Phys. Rev. B 63 (2001) 205310.

[19] G.L. Zhao, M. Bachlechner, Phys. Rev. B 58 (1998) 1887.[20] J. Kim, H. Yeom, Phys. Rev. B 67 (2003) 035304.[21] M. Higuchi, S. Sugawa, E. Ikenaga, J. Ushino, H. Nohira, T. Maruizumi, A. Teramoto,

T. Ohmi, T. Hattori, Appl. Phys. Lett. 90 (2007) 123114.[22] H.M. Lee, C.T. Kuo, H.W. Shiu, C.H. Chen, S. Gwo, Appl. Phys. Lett. 95 (2009)

222104.[23] E.A. Kraut, R.W. Grant, J.R. Waldrop, S.P. Kowalczyk, Phys. Rev. Lett. 44 (1980)

1620.[24] M. Petravic, P.N.K. Deenapanray, V.A. Coleman, K.-J. Kim, B. Kim, G. Li, J. Appl.

Phys. 95 (2004) 5487.[25] P.F. Zhang, X.L. Liu, R.Q. Zhang, H.B. Fan, H.P. Song, H.Y. Wei, C.M. Jiao, S.Y. Yang,

Q.S. Zhu, Z.G. Wang, Appl. Phys. Lett. 92 (2008) 042906.[26] T.D. Veal, P.D.C. King, S.A. Hatfield, L.R. Bailey, C.F. McConville, B. Martel, J.C. Mor-

eno, E. Frayssinet, F. Semond, J.Z. Perez, Appl. Phys. Lett. 93 (2008) 202108.[27] T.E. Cook Jr., C.C. Fulton, W.J. Mecouch, R.F. Davis, G. Lucovsky, R.J. Nemanich, J.

Appl. Phys. 94 (2003) 3949.[28] J. Robertson, J. Vac. Sci. Technol. B 18 (2000) 1785.[29] W. Monch, J. Appl. Phys. 80 (1996) 5076.