Massive point defect redistribution near semiconductor surfaces and interfaces and its impact on Schottky barrier formation

ARTICLE IN PRESS Physica B 404 (2009) 4768–4773

Contents lists available at ScienceDirect

Physica B journal homepage: www.elsevier.com/locate/physb

Massive point defect redistribution near semiconductor surfaces and interfaces and its impact on Schottky barrier formation L.J. Brillson a,b,c,, Y. Dong a, D. Doutt b, D.C. Look d,e, Z.-Q. Fang d a

Department of Electrical and Computer Engineering, The Ohio State University, Columbus, OH 43210, USA Department of Physics, The Ohio State University, Columbus, OH 43210, USA c Center for Materials Research, The Ohio State University, Columbus, OH 43210, USA d Semiconductor Research Center, Wright State University, Dayton, OH 45435 USA e Air Force Research Laboratory, WPAFB, Dayton, OH 45435, USA b

a r t i c l e in f o

PACS: 71.55.Gs 78.60.Hk 72.40.+w 78.68.+m Keywords: Defects Cathodoluminescence spectroscopy Interfaces Segregation Schottky barriers

a b s t r a c t Nanoscale depth-resolved cathodoluminescence spectroscopy calibrated with deep level transient spectroscopy of native point defects and capacitance–voltage measurements of free carrier densities, all at the same metal–semiconductor interface, demonstrate that native point defects can (i) increase by order-of-magnitude in densities with tens of nanometers of the semiconductor surface, (ii) alter free carrier concentrations and band profiles with the surface space charge regions, and (iii) dominate the Schottky barrier formation for metal contacts to ZnO and many other single crystal compound semiconductors. The spatial redistribution of electrically active defects within the surface space charge can be understood in terms of temperature-dependent atomic diffusion enabled by low formation energies and driven by strain and electric fields as well as metal-specific chemical reactions near room temperature, consistent with first-principles calculations of interfacial segregation and migration barriers. These results underscore the importance of native point defects in charge transport and barrier formation at semiconductor interfaces. & 2009 Elsevier B.V. All rights reserved.

1. Introduction Until now, native point defects in semiconductors were not considered a significant factor in Schottky barrier formation due to their relatively low bulk densities. Instead, many theories have postulated localized states, including adsorption-induced defects, at the semiconductor surface with high enough densities to ‘‘pin’’ the Fermi level. It is now possible to measure the energies and densities of point defects below the semiconductor free surface and its metal interface with nanoscale precision. Using depthresolved cathodoluminescence spectroscopy (DRCLS) of deep level emissions calibrated with deep level transient spectroscopy (DLTS) as well as capacitance–voltage (C–V) measurements of free carrier densities—all at the same interface, we can now demonstrate that native point defects can increase by order-ofmagnitude in densities within tens of nanometers of the semiconductor surface, introducing localized charge sites at densities that can significantly alter free carrier concentrations and band profiles within the surface space charge region. In turn

 Corresponding author at: Department of Electrical and Computer Engineering, The Ohio State University, Columbus, OH 43210, USA. Tel./fax: +1 614 292 8015. E-mail address: [email protected] (L.J. Brillson).

0921-4526/$ - see front matter & 2009 Elsevier B.V. All rights reserved. doi:10.1016/j.physb.2009.08.151

these can dominate Schottky barrier formation at metal–semiconductor interfaces. Here we present results for metal contacts to ZnO and show how the native defect features characteristic of this interface can extend to many other single crystal compound semiconductors. Previously we used this combination of techniques to show that surface adsorbates, hydrogen donors, and sub-surface native point defects each contribute independently to interface charge transport and Schottky barrier formation [1]. Furthermore, metal deposition [2] and subsequent annealing [3] induce additional native point defects extending tens of nanometers or more below the free ZnO surface that increase tunneling, recombination, and hopping transport. These effects are orientation-dependent, with significantly higher defect and free carrier densities below the free ð0 0 0 1Þ O face versus (0 0 0 1) Zn face and their interfaces with many different metals [4]. Unlike many semiconductors, metal contacts to ZnO single crystal surfaces yield a wide range of Schottky barrier heights [5]. While this large variation has been attributed to ZnO’s relatively large ionicity [6]; numerous studies determine large variations in Schottky barrier height for the same metal on ZnO crystals prepared under different conditions [1,7,8]. Very recently, nanoscale DRCLS studies showed that native point defects near the metal–ZnO interface [1,9] could alter Schottky barriers from Ohmic to Schottky and vice-versa. Earlier DRCLS studies showed

ARTICLE IN PRESS L.J. Brillson et al. / Physica B 404 (2009) 4768–4773

that hydrogen below the surface acts as an n-type dopant [10]. Furthermore, surface adsorbates could induce surface accumulation and high conductivity. These studies also found that exposure to a remote oxygen plasma (ROP) could reduce such sub-surface hydrogen [10], sub-surface defects [1], and surface adsorbates [1]. Thus the combination of DRCLS and ROP treatment can be used to separate the effects of hydrogen doping and defects in the ZnO depletion region near the metal–semiconductor interface. Recent DRCLS studies of clean, ordered metal–ZnOð0 0 0 1Þ surfaces show that Schottky barriers are reduced and ideality factors increased by high near-surface densities of native point defects [9]. Significantly, capacitance–voltage (C–V) measurements reveal a monotonic increase in carrier concentration for hydrogen-depleted ZnOð0 0 0 1Þ interfaces that correlates with DRCLS-measured deep level defect densities. These results suggest that Schottky barrier decreases may be due to both hopping transport through the depletion region of the barrier as well as tunneling through barriers that are substantially narrowed by high near-surface dopant densities. The spatial redistribution of electrically active defects within the surface space charge can be understood in terms of temperature-dependent atomic diffusion enabled by low formation energies and driven by strain and electric fields as well as metal-specific chemical reactions near room temperature. Selfconsistent electrostatic calculations based on sub-surface trap distributions, energies and carrier densities yield a wide range of effective Schottky barriers, in agreement with measured values for different metals on a wide array of ultrahigh vacuum clean ZnO crystals grown by various methods. These results are not unique to ZnO. Indeed, there is now considerable evidence to show that these sub-surface and near-interface native point defects dominate the Fermi level movement across the band gaps of most compound semiconductors. Here we discuss this dramatic native defect redistribution and its electronic effects using first-principles calculations of interfacial segregation [11] and migration barriers [12]. The massive accumulation and creation of native point defects at these interfaces underscore their importance in charge transport and Schottky barrier formation.

2. Experiment The DRCLS technique provides a tool to measure the emissions from defects as a function of depth on a nanometer scale. For a description of this technique, see for example, Ref. [13]. Basically an incident electron beam of energy EB excites secondary electrons that decay by X-ray emission, plasmon generation, and ultimately electron–hole pair creation. The free electron–hole pairs can recombine radiatively by photon or nonradiatively by phonon emission. For EB in the range of a few keV or less, the electron cascade extends into a semiconductor by a Bohr–Bethe range RB that is on the order of tens to hundreds of nanometers. The depth U0 of maximum rate of energy loss due to electron–hole pair creation is a fraction of this depth such that electron–hole creation and recombination across the semiconductor band gap or between band edges and gap states occur at correspondingly shallower depths. This capability enables measurements of defect and impurity emissions as a function of depth from the bulk to within a few nanometers of the surface. DRCLS has successfully probed localized states at buried quantum well interfaces [14] and gate oxide heterostructures [15], only a few nanometers thick, and even the variations in electronic states at different semiconductor surface reconstructions [16]. In general, we prepare the ZnO crystals by a combination of chemical cleaning and ROP plasma treatment to produce atomically clean surfaces in ultrahigh vacuum (UHV) [17]. DRCL spectra

4769

are obtained both in bare areas between diodes and by excitation and detection through the diodes. DRCLS has shown previously that bulk defect densities in commercially grown ZnO crystals can vary by many orders of magnitude between different growth methods or even between crystals from the same vendor and growth method [18]. In turn, crystals with high defect densities exhibit large changes in defect redistribution and creation with metal contacts [19]. For the studies reported here, we used ZnO with some of the lowest native point defect densities available. These crystals exhibited defect emission intensities many orders of magnitude below those of the near-band-edge (NBE) emission as verified by DRCLS. Metallization in the form of 400 mm, 25– 30 nm thick diode arrays immediately followed ROP cleaning in UHV to avoid any impurity contamination of the metal–ZnO interfaces. We calibrated the defect intensities measured by DRCLS with gap densities measured by DLTS of the same diodes. DLTS provides trap densities and cross sections at depths corresponding to the edge of the semiconductor surface space charge region. Similarly, C–V measurements provide carrier concentrations and their variation with depth. Current leakage in forward voltage limits the measurement of trap and free carrier densities very near the semiconductor interface. However, DLTS can calibrate DRCLS emission intensities at depths over which the two techniques overlap. Assuming a linear dependence of DRCLS with trap density, DRCLS can then provide densities at shallower depths up to a few nanometers of the surface. Similarly, DRCLS can provide densities for contacts with relatively low or ohmic contacts, for which DLTS cannot provide information.

3. Results DRCLS measurements of ZnO initially showed the one of the most common deep level defects, the ‘‘green’’ defect with emission centered at 2.5 eV, increased toward the surface from depths of 30–50 nm and that ROP annealing could reduce their density significantly [1]. As defect density decreased in stages by an order of magnitude, their corresponding Schottky barrier heights with the same metal contact increased from ohmic to Schottky-like and their ideality factors decreased. Similarly, ZnO with orders of magnitude decrease in bulk native point defect density produced order-of-magnitude decrease in reverse current–voltage (I–V) saturation current [19]. These results demonstrated that native point defects within the semiconductor bulk could have major effects on Schottky barrier formation. Combined 1/C2V, DLTS, and DRCLS measurements of a high barrier contact provide a correlation between carrier density, trap density, and deep level emission over the same range of depths. Fig. 1(a) illustrates DRCLS spectra obtained with three EB corresponding to depths from the bulk to 50 nm for a ROPcleaned Ir2ZnOð0 0 0 1Þ diode. These spectra show the NBE peak at 3.35 eV and defects with emission peaks at 2.1, 2.5, and 3 eV that vary strongly with depth. In particular, the 2.5 eV peak increases three-fold between 10 and 5 keV, corresponding to depths of 200 and 50 nm, respectively. For the same diode, Fig. 1(b) displays carrier concentration versus depth. Correspondingly, carrier densities increase by approximately a factor of 3 from depths of 200 to 90 nm. Absolute carrier densities reach 1017 cm3 at 90 nm, indicating a corresponding number of electrically active defects at this depth. Furthermore, the sharp rise in carrier density at shallow depths indicates that these densities may increase by at least another order of magnitude at even shallower depths. Deep level optical spectroscopy (DLOS) of this diode shows the presence of multiple deep levels with energies extending across

ARTICLE IN PRESS 4770

L.J. Brillson et al. / Physica B 404 (2009) 4768–4773

Ir Depth Dependence (Microscope CL) No Anneal O2 Plasma

11 Carrier Density (x1016 cm-3)

10000

CL Intensity (a.u)

2.5 eV

1000 2.10 eV 100 5 kV 10 kV 20 kV

3.0 eV

Ir-ZnO(000-1) Diode

10 Carrier Density 9 (Capacitance)

8

I-V ΦB = 0.69 eV

7 6 5 4 3

10 1.5

2.0

2.5 3.0 Energy (eV)

3.5

4.0

80

100

120

140 160 Depth (nm)

180

200

Intensity (arb. units)

Fig. 1. (Color online) Ir2ZnOð0 0 0 1Þ (a) DRCLS at depths corresponding to 50 (5 keV), 200(10 keV), and bulk (20 keV) depths and (b) carrier concentration versus depth. Both increase three-fold between 200 and 90 nm.

1.8x105

Energy (keV) 3 5 8

3.35 eV (NBE) 30 nm Pd-ZnO(0001) x 1000 2.45 eV <20 nm

9.0x10

4

20-70 nm >100 nm

0.0 2.0

2.5

3.0

3.5

Photon Energy (eV) Fig. 2. (Color online) DRCL spectra of 30 nm Pd on ZnO(0 0 0 1) obtained with 3,5, and 8 keV corresponding to the U0 values shown. Defect emission intensity at 2.45 eV increases by 42x in the o20–100 nm depth range. After Fang et al. [38].

the band gap. In particular, a trap level is evident at 2.7 eV above the valence band EV, i.e., 0.6 eV below the conduction band EC. DLTS activation energy for this diode shows a defect located at EC—0.53 eV, in good agreement with the DLOS position and an optical transition from a state 2.5 eV above EV. These results show that there are high near-surface defect and carrier densities that correlate semi-quantitatively between optical and electronic techniques. A second example of this correlation appears in Fig. 2. Here, 30 nm thick diodes of Pd on ZnO(0 0 0 1) exhibit DRCLS spectra taken at 3, 5, and 8 keV, corresponding to U0 ¼ o20, 20–70, and 4100 nm, respectively. For the same NBE emission intensity, deep level emission at 2.5 eV grows by more than a factor of 2 in this o20–100 nm depth range, even greater if the spectra are recalibrated to subtract lower energy contributions to the higher energy spectra. Fig. 3 shows the corresponding C–V and DLTS measurements as a function of depth of an array of these Pd diodes on the same ZnO(0 0 0 1) surface. Fig. 3(a) shows that carrier concentration remains constant at 7–8  1016 cm3 in the bulk, increasing by 2–5 times in the 70–80 nm depth range. Even higher densities at shallower depths cannot be confirmed due to leakage effects.

Fig. 3(b) illustrates DLTS spectra taken from these diodes as a function of bias voltage and temperature to gauge trap densities and energies at different depths. As with Ir2ZnOð0 0 0 1Þ diodes, DLTS exhibits a defect with energy ES ¼ EC0.5 eV that grows by 42x in the 60–90 nm depth range and by over an order of magnitude relative to the well-known E3 defect. Figs. 2 and 3 demonstrate that near-surface defects with dramatically higher densities than the bulk introduce new donors within 100 nm of the Pd–ZnO(0 0 0 1) junction. In general, nanoscale depth-resolved probes reveal sub-surface native point defects that vary by orders of magnitude as measured by deep level optical emissions, carrier densities, and trap densities. Figs. 2 and 3 show that, at 80–90 nm distances from the interface, carrier concentrations can be well above 1017 cm3. These carrier concentrations and defect densities can extrapolate to values at least an order of magnitude higher at shallower depths, as the growth in defect emission between 4100 nm and o20 nm suggests. For trap densities in the 1018 cm3 range within 10 nm of the surface, equivalent surface trap densities are 41012 cm2, sufficient to introduce dipole voltages of 0.1 eV or more. Thus sub-surface and interface defect densities are large enough to impact Schottky barriers. The growth of defects near metal–semiconductor interfaces displays both a strong chemical dependence and a dependence on ZnO crystal polarity. With EB adjusted to probe similar ZnO depths versus next to the metal diodes, Fig. 4 illustrates the difference in defect formation at interfaces where metals react with the oxygen versus the zinc in ZnO. Fig. 4(a) shows the growth of the 2.5 eV ‘‘green’’ defect at the Al–ZnO interface. There is a strong thermodynamic driving force given by the heat of reaction DHR as [20]

DHR ¼ ð1=xÞDHðMx AÞ  DHðCAÞ

ð1Þ

(neglecting free energy differences) for metal M reacting with semiconductor cation–anion CA M þ ð1=xÞCA-ð1=xÞ½Mx A þ C

ð2Þ

In this case, for Al to react with oxygen in the ZnO lattice resulting in an enthalpy gain of 3.23 eV/molecule. As a result of oxygen diffusing out of the lattice, oxygen vacancies VO and/or their complexes form within a few nanometers of the metal interface. Thus the 2.5 eV CL emission is associated with oxygen vacancies. For the case of a Au2ZnOð0 0 0 1Þ diode in Fig. 4(b), relatively few changes are evident for the CL spectra normalized to the NBE intensity maximum for the interface without annealing and with

ARTICLE IN PRESS L.J. Brillson et al. / Physica B 404 (2009) 4768–4773

4771

OSU-ZnO ZNT-2 Pd-SBDs 30/08-06/09/07 ΔC (b1)/C@300K (10-3)

Carrier concentration (cm-3)

5x1017 2x1017 1017 5x1016

Pd-SBD-5 (06/09) Pd-SBD-8 (30/08) Pd-SBD-7 (31/08) Pd-SBD-6 (30/08) Pd-SBD-2 (31/08)

SBD area: 0.125mm2 T = 300K

2x1016 1016 0.07

E3

0.8

-0.25 V -0.5 V -1.0 V -1.5 V -2.0 V

UR

0.6

Es

0.4

110 nm

0.2

130 nm 150 nm

0 0.08

0.09

0.10

0.11

0.12

60 nm 90 nm

E4

0.13

100

150

200

250

300

350

T (K)

Depth (μm)

Fig. 3. (Color online) (a) 1/C2V plots of carrier concentration versus depth for an array of Pd–ZnO(0 0 0 1) diodes. (b) DLTS spectra of traps E3 and ES at various diode bias voltages and corresponding probe depths. After Fang et al. [38].

ZnT O2 Plasma No Anneal 100000 2kV Bare 5kV Through Al Luminescence Intensity

100000

10000

1000

100

10

ZnO Band Gap 3.36 eV

650°C 550°C 450°C 350°C

10000 VZn-related

VO-related

1000 100 10 No anneal

1

1 1.5

2.0

2.5

3.0

3.5

1.5

Energy (eV)

2.0

2.5

3.0

3.5

Energy (eV)

Fig. 4. (Color online) DRCLS through 30 nm metal diodes on ZnOð0 0 0 1Þ versus of the bare surface next to the diode at EB ¼ 2 keV for (a) Al without any annealing and (b) Au as a function of annealing temperature. Arrows show metal-induced features. After Brillson et al. [9].

annealing up to 550 1C. At 650 1C, however, strong emission appears at 2.0 eV, consistent with the Au–Zn eutectic that forms at 642 1C. The effect of such eutectic formation is to extract Zn from ZnO within a few nanometers of the metal–ZnO interface. Hence the 2.0 eV defect emissions are associated with Zn vacancies VZn and/or their complexes. The 2.5 and 2.0 eV assignments to defects associated with VO and VZn, respectively, are consistent with other techniques. Thus electron paramagnetic resonance, luminescence, and absorption support the 2.5 eV ‘‘green’’ defect emission with VO [21], while the correlation of DRCLS and positron annihilation spectroscopy (PAS) depth profiles on a micron scale provides strong evidence for associating the 2.0 eV ‘‘red’’ defect with VZn and VZn clusters [22]. The polarity of the ZnO surface has a strong effect on the distribution of defects and the resultant changes in interface electronic behavior. DRCLS measurements of the Zn versus O faces of the same ZnO crystals show that the predominant deep level defect emissions at 2.5 eV are consistently higher within the outer 100 nm of the O versus the Zn polar surface [4]. This difference in defect concentration manifests itself in major differences in free carrier concentration. Fig. 5 shows the C2–V measurements of barrier height and doping density as a function of depth for Au and Pd diodes on both Zn-polar and O-polar surfaces. CVmeasured Schottky barrier heights FCV SB are calculated from

FCV SB ¼ Vi þ V0 þ kT=q, with Vi the intercept, V0 ¼ (kT/q)ln(NC/ND), and NC/ND the ratio of conduction band to donor concentrations. Au and Pd produce relatively high barrier heights that enable capacitance measurements without excessive current leakage. As shown by the intersection of the intersection of the extrapolated curves with the (A/C)2 baseline, the barrier heights for the Znpolar diodes are 50 and 130 meV higher than for the O-polar diodes for Au and Pd, respectively. Hence this polar asymmetry is evident regardless of the metal and the different absolute barrier heights. The higher Zn-polar diodes are consistent with the lower free carrier concentrations shown in Fig. 5 inset. The Zn-polar ZnO diodes exhibit free carrier concentrations that gradually decrease from bulk values to densities 30% lower at depths of 75–80 nm. In contrast, O-polar ZnO diodes display carrier concentrations that are relatively constant over the same depth range. The effective donor concentration Ndeff is described in terms of donor and acceptor densities exponentially decaying away from the surface as Ndeff ¼ Ndbulk þ Ndsurf expðz=d1 Þ  Nasurf expðz=d2 Þ Ndbulk

ð3Þ Ndsurf

Nasurf

where is the bulk donor concentration, while and are surface donor and acceptor densities decaying away from the surface with decay lengths d1 and d2, respectively. Hydrogen

ARTICLE IN PRESS 4772

L.J. Brillson et al. / Physica B 404 (2009) 4768–4773

Fig. 5. (a) C2V curves at 1 MHz of Au and Pd diodes on O- and Zn-polar faces of the same ZnO crystal. (b) Net carrier concentrations versus depth for the corresponding diodes [4].

incorporation and possible impurity segregation [23] may account for the rapid rise of Ndsurf term and hence Ndeff near the interfaces of the Pd–ZnO diodes. This higher carrier density decreases the barrier width for Pd diodes, enhancing tunneling and lowering the surf near-surface effective FCV SB relative to the Au–ZnO diodes. A Na acceptor term with d2100 nm accounts for the Ndeff decrease in both Zn-face Au and Pd diodes. Depletion widths are 100– 120 nm, depending on the metal, polarity, and carrier concentration, so that the decreased carrier concentration falls across the entire surface space charge region. Our DRCLS measurements of ZnO crystals grown by a wide range of techniques and sources show that such electrically active point defects are present in the bulk with densities that can vary by order-of-magnitude as well as with depth. Such variations depend on growth and subsequent thermal and surface mechanical treatments [18], indicating that these defects are mobile below growth temperatures. For example, Janotti and Van de Walle [24] used first principle methods based on density functional theory and pseudopotentials to show that the migration barrier for Zn interstitials Zni is only 0.55 eV, consistent with experiment [25]. In turn, Zni’s high mobility infers that these defects will rapidly diffuse out of the crystal at elevated temperatures. Oxygen interstitial diffusion is calculated to have a 1–1.3 eV migration barrier, depending on the charge state with comparable energy barriers for oxygen vacancies [26]. Kohan et al. showed that Zn and O vacancies are the most stable point defects under both O-rich and Zn-rich conditions [27]. Oba et al. [28] used first-principles calculations based on hybrid Hartree–Fock density functions to show that oxygen vacancies can form with high concentration in n-type ZnO. These results are consistent with the dominant DRCLS defect emissions described here, electron paramagnetic resonance [21], and positron annihilation spectroscopies [22,29]. Furthermore, the predominant 2.4–2.5 eV emission associated with V0 is consistent with the high formation enthalpy of VZn [30] under Zn-rich conditions for all but the highest Fermi level positions in the band gap. The surface segregation of both acceptors and donors may be due to a number of driving forces, including electromigration, thermodynamic enthalpy change, and strain. Both piezoelectric and surface space charge fields can induce electromigration of native point defects, depending on their charge state. Thus for metal–oxide interfaces, Duffy et al. [31] took into account both the short range interactions between the ions and the metal cores, the energy of imbedding the ions in jellium, and the image force

energies associated with the ionic charges and the metal to show significantly lower formation energies for anion and cation vacancies within a few atomic planes of the interface. Similarly, Cho et al. [32] showed using first principles methods that oxygen vacancies in oxides are strongly attracted to metal interfaces with binding energies of several eV. Impurity and defect diffusion induced by electric fields has been widely observed. See for example, Ref. [33] the high electric fields associated with the polar ZnO surfaces can thus be an important factor in forming VO versus VZn near the interface, given their opposite charge states. Likewise, strain at these piezoelectric semiconductor interfaces can induce large electric fields. Impurity segregation to the ZnO surface may also induce energetically favorable complex formation, for example, VO complexes with near-surface hydrogen to form shallow donors [28]. Furthermore, the different chemistry of the Zn versus O surface promotes preferential H incorporation near the O-polar surface [34] consistent with the higher VO densities below this surface described above. A preferential segregation of impurities to the O-polar surface may also account for the asymmetry of defects at the two polar surfaces [35]. Finally, the massive defect segregation to the metal–semiconductor interface is not unique to ZnO. Previous near-surface CLS measurements have shown that metals induce native point defects in many other semiconductors [36]. Indeed at high enough concentrations, these defects can pin the Fermi level, limiting barrier heights to values below those expected on the basis of work function differences. Thus Fig. 6 illustrates band diagrams for six semiconductors with the energy levels of their predominant native point defects. Also shown in each diagram are the ranges of Fermi level movement obtained from Schottky barrier measurements [37]. The correspondence between these energy levels and maximum range Schottky barrier heights indicates that native point defects can segregate in high enough densities near metal–semiconductor interfaces to form localized charge states and dipoles that limit further Fermi level movement.

4. Conclusions The ability to probe native point defects, sub-surface traps, and carrier concentrations on a nanometer scale has enable us to identify their dramatic increase near surfaces and their impact on Schottky barrier formation. Native point defects have densities near interfaces that are much higher than previously believed. As

ARTICLE IN PRESS L.J. Brillson et al. / Physica B 404 (2009) 4768–4773

ΦBn

CBM 0.9 eV

VP ΦBn 0.75 eV 0.75 eV 1.14 eV Si+VP EEGG==1.34 1.34 eV eV 1.21 eV

ΦBn

VO Zni

VIn

EEGG==3.36 3.36 eV eV

0.92 eV

EVV+2.5 +2.5 eV eV

4773

VS E+1.72 EV eVeV V+1.72

InP CdS

VBM ZnO PGa ΦB

n

0.9-1.0 eV

ΦBn

= 2.27eV eV EEG G=2.27 EEVV +1.1 +1.1eV eV 1.41-1.68 eV P3+/P4+

GaP

EGG=1.52 = 1.52eV eV E +1.05 eV E V+1.05 eV V CdTe

Fig. 6. Energy band diagrams for ZnO, InP, CdS, GaP, and CdTe with the energy levels of their predominant native point defects corresponding to the range of Fermi level energies observed in Schottky barrier formation.

a result, bulk, surface, and metal-induced native point defects can dominate interface charge transport. Metal interactions, surface polarity, and surface topography all matter, and these in turn shape the defect thermodynamics and diffusion that alter the profile of electrically active sites below the semiconductor surface. Overall, these interface studies provide new reasons for understanding the nature and dynamics of native point defects.

Acknowledgments The authors gratefully acknowledge support from the National Science Foundation, Grant no. DMR-0513968 and DMR-0803276 (Verne Hess), Dr. Steven Ringel and Maria Gonzalez for DLTS measurements, and Dr. Wu Lu, and Michael Schuette for C–V measurements. References [1] H.L. Mosbacker, Y.M. Strzhemechny, B.D. White, P.E. Smith, D.C. Look, D.C. Reynolds, C.W. Litton, L.J. Brillson, Appl. Phys. Lett. 87 (2005) 012102. [2] L.J. Brillson, H.L. Mosbacker, M.J. Hetzer, Y. Strzhemechny, G.H. Jessen, D.C. Look, G. Cantwell, J. Zhang, J.J. Song, Appl. Phys. Lett. 90 (2007) 102116. [3] H.L. Mosbacker, et al., Appl. Phys. Lett. 91 (2007) 072102. [4] Y. Dong, Z.-Q. Fang, D.C. Look, G. Cantwell, J. Zhang, J.J. Song, L.J. Brillson, Appl. Phys. Lett. 93 (2008) 072111. [5] C.A. Mead, Solid State Electron. 9 (1966) 1023. [6] S. Kurtin, T.C. McGill, C.A. Mead, Phys. Rev. Lett. 22 (1969) 1433. [7] B.J. Coppa, R.F. Davis, R.J. Nemanich, Appl. Phys. Lett. 82 (2003) 400. [8] S.J. Pearton, D.P. Norton, K. Ip, Y.W. Heo, T. Steiner, Progr. Mater. Sci. 50 (2005) 312 and references therein. [9] Y. Dong, Z.-Q. Fang, D.C. Look, D.R. Doutt, M.J. Hetzer, L.J. Brillson, J. Vac. Sci. Technol. 27 (2009) 1710. [10] Y.M. Strzhemechny, H.L. Mosbacker, D.C. Look, D.C. Reynolds, C.W. Litton, N.Y. Garces, N.C. Giles, L.E. Halliburton, S. Niki, L.J. Brillson, Appl. Phys. Lett. 84 (2004) 2545. [11] P. Bogus"awski, N. Gonzalex Szwacki, J. Bernholc, Phys. Rev. Lett. 96 (2006) 185501.

[12] A. Janotti, C.G. Van de Walle, J. Crystal Growth 287 (2006) 58. [13] L.J. Brillson, J. Vac. Sci. Tech. B 19 (2001) 1762. [14] L.J. Brillson, T.M. Levin, G.H. Jessen, F.A. Ponce, Appl. Phys. Lett. 75 (1999) 3835. [15] S. Walsh, L. Fang, J.K. Schaeffer, W. Weisbrod, L.J. Brillson, Appl. Phys. Lett. 90 (2007) 052901. [16] I.M. Vitomirov, A.D. Raisanen, R.E. Viturro, S. Chang, L.J. Brillson, P.D. Kirchner, G.D. Pettit, J.M. Woodall, J. Vac. Sci. Tech. A 10 (1992) 749. [17] Y.M. Strzhemechny, J. Nemergut, P.E. Smith, J. Bae, D.C. Look, L.J. Brillson, J. Appl. Phys. 94 (2003) 4256. [18] L.J. Brillson, H.L. Mosbacker, D. Doutt, M. Kramer, Z.L. Fang, D.C. Look, G. Cantwell, J. Zhang, J.J. Song, Superlattices Microstructures 45 (2009) 206. [19] H.L. Mosbacker, S. El Hage, M. Gonzalez, S.A. Ringel, M. Hetzer, D.C. Look, G. Cantwell, J. Zhang, J.J. Song, L.J. Brillson, J. Vac. Sci. Tech. B 24 (2007) 1405. [20] L.J. Brillson, Phys. Rev. Lett. 40 (1978) 260. [21] K. Vanheusden, C.H. Seager, W.L. Warren, D.R. Tallant, J.A. Voigt, Appl. Phys. Lett. 68 (1996) 403. [22] Y. Dong, F. Tuomisto, B.G. Svensson, A. Yu Kuznetsov, L.J. Brillson, unpublished. [23] D.C. Look, B. Claflin, H.E. Smith, Appl. Phys. Lett. 92 (2008) 122108. [24] A. Janotti, C.G. Van de Walle, J. Crystal Growth 287 (2006) 58. [25] D.G. Thomas, J. Phys. Chem. Solids 3 (1957) 222. [26] P. Erhart, K. Albe, Phys. Rev. B 73 (2006) 115207. [27] A.F. Kohan, G. Ceder, D. Morgan, C.G. Van de Walle, Phys. Rev. B 61 (2000) 15019. [28] F. Oba, A. Togo, I. Tanaka, J. Paier, G. Kresse, Phys. Rev. B 77 (2008) 245202. [29] F. Tuomisto, V. Ranki, K. Saarinen, D.C. Look, Phys. Rev. Lett. 91 (2003) 205502. [30] S.B. Zhang, S.-H. Wei, A. Zunger, Phys. Rev. B 63 (2001) 075205. [31] D.M. Duffy, J.H. Harding, A.M. Stoneham, J. Appl. Phys. 76 (1994) 2791. [32] E. Cho, B. Lee, C.-K. Lee, S. Han, S.H. Jeon, B.H. Park, Y.-S. Kim, Appl. Phys. Lett. 92 (2008) 233118. [33] R.D. Chang, P.S. Choi, D.L. Kwong, D. Wristers, P.K. Chu, Appl. Phys. Lett. 72 (1998) 1709. [34] M.G. Wardle, J.P. Goss, P.R. Briddon, PRL 96 (2006) 205504. [35] S. Lautenschlaeger, J. Sann, N. Volbers, B.K. Meyer, A. Hoffmann, U. Haboeck, M.R. Wagner, Phys. Rev. B 77 (2008) 144108. [36] R.E. Viturro, M.L. Slade, L.J. Brillson, Phys. Rev. Lett. 57 (1986) 487. [37] Y. Dong, D.C. Look, Z.-Q. Fang, L.J. Brillson, unpublished. [38] Z.-Q. Fang, B. Claflin, D.C. Look, Y.F. Dong, H.L. Mosbacker, L.J. Brillson, J. Appl. Phys. 104 (2008) 063707.