Dynamics and control of recombination process at semiconductor surfaces, interfaces and nano-structures

Solar Energy 80 (2006) 629–644 www.elsevier.com/locate/solener

Dynamics and control of recombination process at semiconductor surfaces, interfaces and nano-structures Hideki Hasegawa a,*, Taketomo Sato a, Seiya Kasai a, Boguslawa Adamowicz b, Tamotsu Hashizume a a

Research Center for Integrated Quantum Electronics (RCIQE), Graduate School of Information Science and Technology, Hokkaido University, N13, W8, Sapporo 060-8628, Japan b Department of Applied Physics, Silesian University of Technology, Krzywoustego 2, 44-100 Gliwice, Poland Received 16 June 2005; received in revised form 22 September 2005; accepted 18 October 2005 Available online 30 January 2006 Communicated by: Associate Editor Arturo Morales-Acevedo

Abstract Characterization methods and fundamental aspects of surface/interface states and recombination process in Si and III– V materials are reviewed. Various measurement considerations are pointed out for the conventional metal–insulator–semiconductor (MIS) capacitance–voltage (C–V) method, a contactless C–V method, and the microscopic scanning tunneling spectroscopy (STS) method, and general features of surface states are discussed. Surface states are shown to have U-shaped distributions of donor–acceptor continuum with a characteristic charge neutrality level, EHO. Rigorous simulation of dynamics of surface recombination process has shown that the effective surface recombination velocity, Seff, is not a constant of the surface, but its value changes by many orders of magnitude with the incident light intensity and the polarity and amount of fixed charge. From this, new methods of surface state characterization based on photoluminescence and cathodoluminescence are derived. Attempts to control surface states and Fermi level pinning at metal semiconductor interface and free surfaces of nano-structures are presented as efforts toward ‘‘nano-photovoltaics’’.  2006 Elsevier Ltd. All rights reserved. Keywords: Surface states; Surface recombination; Fermi level pinning; Capacitance voltage method; Tunneling spectroscopy; Nanostructures

1. Introduction Rapid progress continues in the research of solar cells for a viable, clean and renewable energy *

Corresponding author. Tel.: +81 11 706 7170; fax: +81 11 757 1165. E-mail address: [email protected] (H. Hasegawa).

source. Trends are toward use of new materials, toward use of thinner films and new structures, and even toward exploitation of nano-structures (Corkish et al., 2002). It is well known that surface recombination through surface/interface states is a major loss mechanism for photo-generated carriers. Obviously, its importance increases as the geometrical feature sizes of the solar cell structures are reduced.

0038-092X/$ - see front matter  2006 Elsevier Ltd. All rights reserved. doi:10.1016/j.solener.2005.10.014

630

H. Hasegawa et al. / Solar Energy 80 (2006) 629–644

Nomenclature Nss Eoj EHO Ec Ev Eg S Seff vth

surface state density (cm2 eV1) surface state distribution parameter (eV) sp3 hybrid orbital energy (eV) conduction band minimum (eV) valence and maximum (eV) energy band gap (eV) surface recombination velocity (cm s1) effective surface recombination velocity (cm s1) thermal velocity of carrier (cm s1)

The purpose of this paper is to review a series of work done by authors’ group on the fundamental aspects of surface/interface states and recombination process, including their characterization methods as well as the method to control the surface recombination process and Fermi level pinning. Although the data is mostly presented on Si and III–V materials, underlying physics, analysis methods and control technologies seem to be applicable to other new materials and new structures. We start from the electrical characterization methods of the properties of surface/interface states lying at the passivated semiconductor surfaces. In addition to the conventional metal–insulator–semiconductor (MIS) capacitance–voltage (C–V) method, a contactless C–V method (Takahashi et al., 1999), and the microscopic scanning tunneling spectroscopy (STS) method (Hasegawa et al., 2000) are discussed. Various measurement considerations are pointed out. Then, general features of energy distribution of surface state density (NSS) are discussed. It is shown that surface states have U-shaped distributions with a characteristic charge neutrality level EHO in accordance with the disorder induced gap state (DIGS) model for Fermi level pinning (Hasegawa and Ohno, 1986). Subsequently, dynamics of surface recombination process is discussed, using a rigorous computer simulation program (Adamowicz and Hasegawa, 1998). It is shown that the effective surface recombination velocity, Seff, is not a constant of the surface with a given NSS distribution as usually assumed. Its value changes by many orders of magnitude with the incident light intensity and the polarity and amount of fixed charge. This analysis has led to two novel contactless analysis methods for unknown NSS distributions, i.e., the photolumi-

Us r n p / a /B v

surface recombination rate (cm2 s1) capture cross-section of the surface state (cm2) density of electron (cm3) density of hole (cm3) photon flux intensity (cm2 s1) adsorption coefficient (cm1) Schottky barrier height (eV) electron affinity (eV)

nescence surface state spectroscopy (PLS3) method (Adamowicz et al., 2002) and the cathodoluminescence in-depth spectroscopy (CLIS) method (Ishikawa and Hasegawa, 2002a,b). Finally, efforts toward ‘‘nano-photovoltaics’’ at authors’ group are briefly presented. It is shown that Fermi level pinning at metal–semiconductor interfaces is greatly reduced at electrochemically prepared nanometer-sized contacts (Hasegawa and Sato, 2005), opening up hopes for forming electron-collecting and hole-collecting nano-contacts for semiconductor nano-structures without p–n junction. It is shown that the Si interface control layer (ICL)-based passivation is effective to arrays of MBE-grown quantum wires (QWRs) (Shiozaki et al., 2005). These efforts may lead toward the realization of new high efficiency solar cells based on QWR arrays to be utilized as power supply for intelligent quantum (IQ) chips (Hasegawa et al., 2004) for coming ubiquitous network society as well as in conventional photovoltaic applications. 2. Characterization methods of surface/interface states 2.1. Conventional capacitance–voltage (C–V) method for macroscopic characterization Surface passivation by a suitable dielectric film is an important step for fabrication of solar cells, since minimization of surface recombination is a critical issue for maximization of the conversion efficiency. The most frequently used method to evaluate energy distributions of surface states (interface states) lying at passivated semiconductor surfaces is to construct metal–insulator–semiconductor (MIS) capacitors and carry out either high frequency or quasi-static

H. Hasegawa et al. / Solar Energy 80 (2006) 629–644

capacitance voltage (C–V) measurements whose principles are described in the standard textbooks (for example, Sze, 1981b). The methods are straightforward and well applicable to high quality Si–SiO2 systems. However, difficulties can arise in other MIS systems which exhibit unexpected C–V behavior. An example is given in Fig. 1(a) where anomalous frequency dispersion behavior of C–V curves generally seen in n-type GaAs MIS capacitors is shown for a SiN passivated GaAs capacitor. Depending on how to interpret this dispersion, i.e., whether this is due to dispersion of permittivity of dielectric or due to semiconductor surface states, the result of MIS C–V analysis becomes very different as shown in Fig. 1(b) which is taken from an old paper of our group (Hasegawa and Sawada, 1983). This clearly illustrates that very careful interpretation and analysis are necessary, when C–V methods are applied to combinations of new semiconductors and new dielectrics, and some anomalous C–V behavior not described in the textbooks is observed. Another point is the measurement frequency. In the high frequency method, 1 MHz is usually high enough for Si MOS capacitors to obtain their high frequency limits. However, this may not be the case in other MIS systems. For example, some III-V MIS capacitors reached their high frequency limit at GHz frequencies (Iizuka et al., 1997) due to very short time constants of some parts of surface state continuum. For the quasi static method involving integration of charging currents, utmost care has to be made for the leakage current through the dielectric film.

631

By a detailed careful C–V study on various GaAs MIS systems, we came to the conclusion that the anomalous frequency dispersion shown in Fig. 1(a) is due to high density surface states that have a specific U-shaped continuous density distribution. Namely, the state density has a minimum at about 0.9 eV below the conduction band edge and sharply increases from near midgap towards the conduction band edge. 2.2. Contactless capacitance–voltage (C–V) method For characterization of processing steps, contactless and non-destructive characterization of semiconductor free surface without deposition of field plates is obviously desirable. For this purpose, we have developed an ultrahigh vacuum (UHV) contactless C–V measurement system (Takahashi et al., 1999) in order to measure NSS distributions on clean and processed free surfaces in the UHV environment as well as those in air. Such a technique is highly useful for optimization of each device processing step, since conventional MIS C– V method requires growth or deposition of insulator, and thereby surface properties are changed. The basic principle of the UHV contactless C–V measurement system developed by our group is shown in Fig. 2(a)–(c). Instead of depositing a gate electrode directly on the sample surface, C–V measurements are performed from the field electrode that is placed above the sample, being separated GaAs MIS capacitors 10

GaAs SiNx MIS capacitor

14

1kHz

10

13

10kHz

100kHz

10

10

12

11

1MHz 0

10

Gate bias (V) (a)

20

Ev 0.4

0.8 (b)

1.2

Ec

Fig. 1. (a) Capacitance–voltage anomaly observed in GaAs MIS capacitors and (b) surface state density distribution reported on GaAs MIS capacitors by various workers.

632

H. Hasegawa et al. / Solar Energy 80 (2006) 629–644

connection to other UHV chambers

piezo-unit sample sensor head

to vacuum pumps anti-vibration mechanism

UHV chamber

(a) bottom view of sensor head

parallelism electrodes

side view of sensor head photo diode

laser diode

UHV gap (Lg) (100-300nm)

field electrode

sample electrodes

(b)

evanescent wave (c)

Fig. 2. Ultrahigh-vacuum contactless C–V measurement system. (a) Schematic diagram of the system, (b) bottom view of the sensor head and (c) side view of the sensor head.

GaAs, having As-stabilized (2 · 4), As-stabilized c(4 · 4) and Ga-stabilized (4 · 6) reconstructions, are shown in Fig. 3 by solid curves. The distributions shown by the dashed curves are those obtained after MBE growth of 1 monolayer (ML) of Si which will be explained later. It is seen that surface states 1015

(2x4) & c(4x4) (2x4)+1ML Si

1014 Nss (cm-2eV-1)

from the sample surface by a thin ‘‘UHV-gap’’ (300– 400 nm). Thus, MIS assessments of MBE-grown clean free surfaces, ultra-thin-oxidized surfaces as well as free surfaces of devices processed in the air become possible by use of a ‘‘UHV-gap insulator’’. In this system, the ‘‘UHV-gap’’ is maintained to be constant by a piezo-mechanism with capacitance feedback from the three surrounding parallelism electrodes shown in Fig. 2(b). The UHV-gap distance was optically determined through reflectance change due to reflection of the evanescent wave caused by the presence of the semiconductor surface (Takahashi et al., 1999), using the set-up shown in Fig. 2(c). C–V measurements are performed between the center filed plate in Fig. 2(b) and the semiconductor surface. Cr was used as the field plate metal and the area of the electrode is 7.5 · 103 cm2. The frequency of the applied ac signal is 500 kHz. This technique has been applied to thin oxide covered and hydrogen passivated Si surfaces (Yoshida et al., 1997; Yoshida and Hasegawa, 2000), MBE grown and UHV processed GaAs (Negoro et al., 2003) and InP (Takahashi et al., 1999) surfaces. As an example, measured surface state distributions on MBE grown clean (0 0 1) surfaces of

1013 c(4x4) +1ML Si

1012 1011 (4x6) 1010 -1.2 Ev

(4x6)+1ML Si -0.8

Energy (eV)

-0.4

0 Ec

Fig. 3. Distribution of surface state densities on various GaAs surfaces measured by UHV contactless C–V measurement system. Dashed curves indicate the data on samples after 1 ML Si growth.

H. Hasegawa et al. / Solar Energy 80 (2006) 629–644

have roughly U-shaped continuous distributions in both cases rather than showing well defined discrete peaks. Particularly, surface states on the As-stabilized GaAs (2 · 4) surface has a narrow U-shaped distribution with high densities which will strongly pin the Fermi level at around 0.9 eV below the conduction band edge. Separate band bending measurements by X-ray photoelectron spectroscopy (XPS) technique clearly indicated a very strong Fermi level pinning. The Ga-stabilized (4 · 6) surface shows a wider U-shape, but the distribution sharply rises up at around 0.5 eV below the conduction band edge. It is interesting to note that the technologically most important As-stabilized GaAs (2 · 4) surface, which is usually used as the initial surface for GaAs device processing, has the largest surface state density, although the surface shows a well defined and regular (2 · 4) missing dimmer structure under the UHV in situ scanning tunneling microscope. Presence of large band bending on this surface was also detected by X-ray photoelectron spectroscopy (XPS) measurements (Ishikawa et al., 1998). 2.3. Scanning tunneling spectroscopy for microscopic characterization Recent progress of scanning tunneling microscopy (STM) has provided very powerful means of dI dV

measuring the surface state properties locally in nanometer scale areas by using the scanning tunneling spectroscopy (STS) mode. Such a technique seems to be useful for basic study as well as process optimization related to surface passivation of solar cells. In the STS technique (Martensson and Feenstra, 1989), the tunneling conductance through the vacuum gap is assumed to be proportional to density of states (DOS) within the energy gap as schematically shown in Fig. 4(a). Such an idea is now generally expected in the research community, and this technique is widely used to find out local DOS spectra of surface states on surfaces of various materials, including not only inorganic semiconductors, but also organic semiconductors. However, we have seen anomalous STS spectra schematically shown in Fig. 4(b) on clean GaAs (0 0 1) surfaces freshly grown by MBE that possess very well defined surface reconstruction patterns such as (2 · 4), c(4 · 4) and (4 · 6) patterns under RHEED observation and exhibit regular and periodic atomic images under UHV STM observation. Actual measured examples are shown in Fig. 5 (Negoro et al., 2003) for various surface reconstructions. Features of the observed spectra are a conductance gap defined by E0v  E0c which is much larger than the GaAs energy gap, and that there is no apparent DOS of surface states within the

dI dV no apparent gap states

gap state DOS EV

(a)

633

EC

V

EV' EV

EC E C '

V

(b)

charged surface states

surface state

qV

EF(tip)

EC EF charge neutrality point EHO (c)

EV

Fig. 4. (a) Usually expected STS spectrum, (b) actually observed STS spectrum on GaAs and (c) explanation for anomalous conductance gap.

634

H. Hasegawa et al. / Solar Energy 80 (2006) 629–644

2.0 (dI.dV)/(I/V)

very rare 1.5 1.0 0.5

Ev

Ec

0 -1.0 0 1.0 2.0 mostly Sample bias (V) (a)

(dI.dV)/(I/V)

2.0 very rare 1.5 1.0 0.5

Ev

Ec

0 -1.0 0 1.0 2.0 Sample bias (V) (b)

mostly

(dI.dV)/(I/V)

2.0 1.5

surface state DOS. However, in the case of abovementioned GaAs surfaces, surface states are deep in the gap, and current transport is limited, not by tunneling itself, but by the slow thermal emission process of electrons from surface states to the conduction band under positive sample bias. Thus, surface states remain partially filled, and thus locally charged. This leads to tip-induced local band bending, and increases the apparent conductance gap. The tunneling current becomes small in spite of a large surface state DOS due to partial filling of states. Obviously, this effect also influences the topological STM images. By a detailed investigation of STS spectra and bias-dependent STM images, we came to the conclusion that local DOS of the GaAs surfaces has a U-shaped continuous distribution rather than showing discrete peaks (Hasegawa et al., 2000; Negoro et al., 2003). The above example clearly indicates the importance of through understanding of current transport with a possible charging effect, whenever a local surface state DOS of a new material is investigated by the STS technique. 3. Distribution and recombination dynamics of surface states

1.0 0.5

3.1. Energy distributions and origin of surface/ interface states

0 -1.0 0 1.0 2.0 Sample bias (V) (c) Fig. 5. Anomalous STS spectra observed on clean GaAs surfaces with (a) (2 · 4), (b) c(4 · 4) and (c) (4 · 6) surface reconstructions.

observed conductance gap in spite of existence of high density surface states detected by UHV contactless C–V measurements. This unexpected result is explained in terms of the probe-induced local charging of surface states (Hasegawa et al., 2000), particularly that of high density acceptor-type surface states lying in the upper half of the band gap, by the STM tip, as schematically shown in Fig. 4(c). Namely, usual assumption of the tunneling conductance being proportional to gap state DOS is justified under the following two conditions. One is that the rate limiting process for the current transport is the tunneling process itself, and the other is that surface states into which electrons tunnel are empty so that the tunneling probability becomes proportional to

After extensive study on energy distributions of surface/interface states on Si and various III–V surfaces, using the characterization methods discussed in the previous section, we have come to the conclusion that the energy distribution of surface state density (NSS) at the insulator–semiconductor interface and on free surfaces under fresh MBE conditions or air-exposed condition constitutes a U-shaped donor/acceptor continuum with a characteristic charge neutrality level EHO as shown in Fig. 6(a). Here, EHO is the mean sp3 hybrid orbital energy (Hasegawa and Ohno, 1986), given by the following equation:   es þ 3ep EHO ¼  ð1Þ 4 av where es and ep are atomic terms energies of s and p electrons, respectively, and ‘‘av’’ denotes the average which has to be taken over anions and cations, when the semiconductor material under consideration is a compound. When the Fermi level coincides with EHO, charge of the surface state con-

H. Hasegawa et al. / Solar Energy 80 (2006) 629–644

acceptor -like

donor-like

EHO

Ev

635

Ec

(a)

EC EV

EV

EC EV

GaAs

Si

InP

EHO

EHO 0 0.32 1.12 [eV]

0

E C EV

GaN

EHO

0.44 1.42 [eV]

0 (b)

0.95 1.35 [eV]

EC

EHO 0

2.37 3.39 [eV]

Fig. 6. (a) Schematic illustration of surface state continuum for the disorder-induced gap state (DIGS) model, and (b) position of EHO for various semiconductor materials.

tinuum becomes neutral. The position of EHO is determined by the band structure of the host semiconductor crystal (Hasegawa and Ohno, 1986). General NSS distribution shapes and EHO positions are shown in Fig. 6(b) for Si, GaAs, InP (Hasegawa and Ohno, 1986) and GaN (Kampen and Mo¨nch, 1997; Hasegawa et al., 2003) surfaces. For most semiconductors, EHO lies within the energy gap. Interestingly, however, EHO lies within the conduction band for InAs (Hasegawa and Ohno, 1986) and InN (Veal et al., 2004), leading to formation of electron layers near the surface even without doping due to Fermi level pinning. A recent theoretical calculation also confirms this for InN (Van de Walle and Neugebauer, 2003). As for the shape of the distribution, it has also been found that the distribution can be fitted by the following formula for each of the donor-type and acceptor type states: N SS ðEÞ ¼ N SS0 expðjE  EHO j=Eoj Þ

nj

ð2Þ

where j = d (donor) or a (acceptor) with a common NSS0. It is evident that the Fermi level is firmly pinned at a position close to EHO, when the minimum state density, NSS0, is high. This explains, for example, the well known tendency of large band bending near midgap seen on various surfaces of n-type GaAs. Although appearance of U-shaped continuous distributions is most generally observed in our experiences, there are cases in which a discrete peak appears on the U-shaped background distribution. Such cases include hydrogen passivated Si (0 0 1) and (1 1 1) surfaces (Yoshida and Hasegawa, 2000), annealed free surfaces of GaAs and InP (0 0 1) surfaces (Sawada et al., 1993) and free surfaces of GaN and AlGaN (Hasegawa et al., 2003). These peaks correspond to missing hydrogen atoms from the hydrogen passivated Si surface, As vacancies on the GaAs surface, P vacancies on the InP surface and N vacancies on GaN and AlGaN surfaces, respectively. Thus, when there are particular

636

H. Hasegawa et al. / Solar Energy 80 (2006) 629–644

types of dominant defects that are produced or remaining on the surface after insulator formation, processing or annealing, they appear as discrete peaks on the U-shaped background distribution. Appearance of discrete peaks were also reported on ultrathin oxide covered Si surfaces by other workers (Kobayashi et al., 1996; Yamashita et al., 1996), using an XPS-based new technique. Obviously, understanding the origin of surface states and related Fermi level pinning is an important issue for the progress of the surface passivation and other surface/interface related science and technology. One remarkable feature related to this issue is that there is a strong correlation between the energy position of Fermi level pinning on various free surfaces and insulator–semiconductor (I–S) interfaces and those at the metal–semiconductor (M–S) surfaces (Hasegawa and Ohno, 1986; Adamowicz et al., 2002), both being very close to the position of EHO. In order to explain this correlation, the unified defect model (UDM) (Spicer et al., 1980) and the disorder induced gap state (DIGS) model were proposed, as summarized in Table 1. The former ascribes the origin of the surface states to stoichiometry-related defects at the surface where the

Fermi level is pinned near the peak of the discrete DOS. On the other hand, the latter explains the origin in terms of loss of two-dimensional periodicity of the host crystal at the surface caused by bond disorder, producing a U-shaped donor/acceptor continuum where the Fermi level is pinned near the charge neutrality level, EHO. There is also a model called metal-induced gap state (MIGS) model (Heine, 1965; Tersoff, 1984) which is solely applicable to the metal–semiconductor interface. According to this model, evanescent tails of metal wavefunctions produce a donor/accepter gap state continuum with a charge neutrality level EHO which pins the Fermi level at EHO. This model cannot explain the above-mentioned correlation of the pinning position between I–S and M–S interfaces. The above-mentioned appearance of peaks on the U-shaped background in some surfaces as mentioned above, indicates that both situations assumed in the UDM and DIGS models can take place. However our wide range of experiments have indicated that the pinning position is near to EHO even when defect DOS peaks are observable. Our interpretation is that there is a strong tendency to avoid generation of energetically unfavorable isolated

Table 1 Major models for Fermi level pinning at semiconductor surfaces and interfaces Model

Origin of pinning

Unified defect model

DIGS model

Nss distribution and pinning position

Applicable interfaces

Nature of pinning

References

Deep level related to stoichiometry, especially, AsGa

V–S, S–S, I–S, M–S

Extrinsic

Spicer et al. (1980)

Loss of 2D periodicity due to disorder of bonds at interface

V–S, S–S, I–S, M–S

Extrinsic

Hasegawa and Ohno (1986)

M–S

Intrinsic

(Heine, 1965; Tersoff, 1984)

mean hybrid orbital energy

MIGS model

Penetration of metal wave function into semiconductor

midgap energy V–S: Vacuum–semiconductor interface; S–S: semiconductor–semiconductor interface; I–S: insulator–semiconductor interface; M–S: metal–semiconductor interface.

H. Hasegawa et al. / Solar Energy 80 (2006) 629–644

637

defects on the surface by screening these defects by small amount of bond disorder of host atoms. Such a tendency was clearly observed by detailed study of GaAs clean surface using STM (Negoro et al., 2003; Hasegawa et al., 2000).

S is usually assumed to be a characteristic constant of the surface, and is often related to the surface state density, NSS (states/cm2) by the following simple equation (Sze, 1981a):

3.2. Dynamics of surface recombination process

where rp is the hole capture cross-section of the surface state and vth is the thermal velocity of holes. However, the actual physical situation is very complicated, as schematically illustrated in Fig. 7(b). Generally, there exists a certain band bending reflecting Fermi level pinning producing a depletion layer which is often referred to as the dead layer by photonics researchers because it does not produce light. Under light illumination, quasi Fermi levels split, and these levels together with band banding determine the carrier densities, ns and ps, at the surface. These carriers participate in the non-radiative surface recombination via the surface state continuum through the following equation for surface recombination rate, Us, based on the SRH statistics:

ð4Þ

S ¼ rp vth N SS

As mentioned in Section 1, surface recombination at passivated semiconductor surfaces is one of the loss mechanisms of photo-generated carriers. Thus, accurate quantitative characterization of the surface recombination process is very important. For this, the surface recombination velocity, S, is used with reference to Fig. 7(a) where S satisfies the following boundary condition for the standard semiconductor drift-diffusion equations:  opðxÞ Sðpð0Þ  p0 Þ ¼ Dp ð3Þ ox  x¼0

Here, the standard notation is used and an n-type semiconductor is assumed.

surface states

S x

0

generation

(a)

SRH recombination

Auger recombination

Ec EFn

nonradiative surface recombination

EFp Ev (b)

surface states

Seff depletion layer (dead layer)

0

x

xw

(c)

Fig. 7. (a) Definition of surface recombination velocity, (b) actual physical situation and (c) definition of effective surface recombination velocity.

Z

Ec

Ev

rn rp vthn vthp ðps ns  n2i ÞN SS ðEÞ dE rn vthn ðns þ n1 Þ þ rp vthp ðps þ p1 Þ

ð5Þ

US pð0Þ

-5x1010 Nss(E)

(cm-2eV-1)

where standard notations are used. Although the above integral is calculated over the entire gap, the states actually participating in recombination are those bounded by the two quasi Fermi levels. Thus, change in light intensity change the number of states participating in surface recombination. It also changes the surface carrier concentration and flow of carriers for recombination. Thus, the charge balance is changed and hence the band bending changes. Thus, the whole problem requires to solve the drift diffusion equation and Poisson’s equation in a self-consistent way, using the surface recombination as the boundary condition. Obviously, other recombination processes such as bulk SRH recombination, Auger recombination and radiative recombination processes affect the situation, and therefore should be considered. We have developed a computer program based on a one-dimensional Scharfetter–Gummel-type vector–matrix algorithm to solve this problem, and the result was applied to analysis of the surface recombination processes taking place on passivated Si and GaAs surfaces (Adamowicz and Hasegawa, 1998), and InP surfaces (Miczek et al., 2001). In spite of the complicated situation, one can still define the effective surface recombination velocity, Seff, through the following equation, referring to Fig. 7(c). S eff ¼

depletion

-8x1010

106

ð6Þ

Here, an n-type semiconductor is assumed. It should be noted that Seff is defined, not at the surface, but at the edge of the depletion layer whose position is dependent on the light intensity. Calculated examples of Seff are shown in Fig. 8(a) and (b) for Si and GaAs (Adamowicz and Hasegawa, 1998), respectively. Typical NSS distributions shown in the inset of each figure were used. It is shown that the effective surface recombination velocity, Seff, is not at all a constant of the surface with a given NSS distribution, as usually assumed in the solar cell simulation. Its value changes by many orders of magnitude with the incident light intensity and the polarity and amount of fixed charge. There are many methods of measuring the surface recombination rate, and these are, rigorously speaking, measurements of the effective surface recombination velocity. Therefore, if the conditions measuring its value are different from

-2x1010

105

Seff (cm/s)

US ¼

H. Hasegawa et al. / Solar Energy 80 (2006) 629–644

12

10 -1x1011 0

104

1011

2x1010

103

EV EHO

-2x1011 inversion

102

EC

QFC=1x1011cm-2

accumulation

101 8 10

10

12

n-Si 16

20

10

10

24

10

-2 -1

Photon flux density (cm s ) (a)

Seff (cm/s)

638

10

7

10

6

10

5

QFC=0 cm-2

1x1013

depletion 5x1013 1.5x1014

-5x1011

-7x1011

2x1014

accumulation

inversion

10

4

10

3

10

2

n-GaAs

10

9

(cm-2eV-1) 14 10

Nss(E)

13

10

12

12

10

10

EV EHO

15

10

10

18

EC

21

10

10

24

-2 -1

Photon flux density (cm s ) (b) Fig. 8. Plots of calculated effective surface recombination velocity vs. photon flux density for (a) n-Si and (b) n-GaAs.

those for solar cell operation, measured value may not be applicable to solar cells. The calculation also shows that the value of Seff can be controlled by intentionally introducing fixed charge to the surface. 3.3. PLS3 and CLIS methods for surface state characterization The above computer analysis has then led to two novel contactless analysis methods for unknown surface state density (NSS) distributions, i.e., the photoluminescence surface state spectroscopy (PLS3) method (Adamowicz et al., 2002) and the cathodoluminescence in-depth spectroscopy (CLIS) method (Ishikawa and Hasegawa, 2002a,b). The principle of the PLS3 method is shown in Fig. 9(a). The intensity of the band edge photoluminescence under the presence of non-radiative surface recombination through a surface state continuum can be calculated from the following formula, using the above-mentioned computer program:

H. Hasegawa et al. / Solar Energy 80 (2006) 629–644

intrinsic

Nss

slo

pe

un

ity

N ss

EV EFP EFn EC pinned

slope < 1

EV EFP EFn EC

PL quantum efficiency YPL (arb.units)

Photon flux density (a)

10-1

NSS0 = 1010 eV-1cm-2

10-2

1011 1012

10-3

1013

1020

1021

1022

1023

(photon cm-2s-1)

Excitation light intensity, (b)

NSS (E) (eV-1cm-2)

10

10

10

1024

13

12

acceptorlike states donor-like states

11

EHO -1.2 EV

-0.8

-0.4

E – EC (eV) (c)

0 EC

Fig. 9. (a) Behavior of PL intensity as a function of photon flux density. (b) Calculated and experimental PL quantum efficiency vs. excitation intensity. Black dots indicate experimental data taken on a chemically polished InP (1 0 0) surface. (c) NSS(E) distribution which gave the best fit in (b).

I PL / g

Z

1

ðpðxÞnðxÞ  p0 n0 Þ  expðaxÞ dx

ð7Þ

0

where a is the absorption coefficient. It can be shown that the external quantum efficiency of photoluminescence, YPL = IPL//, where / is the incident photon flux intensity, is not a constant, but

639

behaves as shown in Fig. 9(a) as a function of /. The reason why the efficiency increases with / is basically due to saturation of the capability of the surface state as the recombination center under a large supply of carriers. The slope in this rise-up transition region should be unity for a discrete surface state peak as shown in Fig. 9(a), because the number of surface states participating surface recombination becomes constant after the peak becomes enclosed in a region bounded by the two photo-split quasi Fermi levels. For continuous distribution, the slope becomes smaller than unity due to increase of participation state with increase of light intensity. By calculating the dependencies of the PL efficiency, YPL, on the excitation intensity / for various possible surface state distributions, and comparing the result with the measured YPL  / curve, one can determine the most likely surface state distributions. For this fitting procedure, we have a fully computerized multi-parameter fitting procedure, which is based on Genetic Algorithm concept (Adamowicz et al., 2002). The best fit is determined by minimizing the so-called fitting error function (FEF). An example of application of such an analysis to a published result on InP surface is shown in Fig. 9(b) and (c) where an excellent fitting has been obtained. Similarly, one can use cathodoluminescence (CL) as the probe to detect surface states instead of photoluminescence. The penetration depth of electrons into semiconductor is strongly dependent on the acceleration voltage and this allows depth-resolved measurements. Paying attention to this, we have recently proposed (Ishikawa and Hasegawa, 2002a) the cathodoluminescence in-depth spectroscopy (CLIS) as a non-destructive characterization method of surfaces and buried heterointerfaces. Here, a plot where the CL intensity for a particular radiative transition is plotted vs. the acceleration voltage, is defined as a CLIS spectrum. If the CLIS spectra for an ideal structure without any defects are known by a detailed computer simulation, then any deviations in measured spectra from the ideal CLIS spectra such as change in spectrum shapes and intensities, and/or appearance of new CLIS spectra from unexpected CL peaks, indicate presence of unwanted anomalies. Thus, one should be able to identify the causes of the anomalies, if a suitable computer simulation program is available. For the calculation of the CLIS spectra, we have modified our program for PLS3, and the CLIS method has

640

H. Hasegawa et al. / Solar Energy 80 (2006) 629–644

been successfully applied to various heterostructures. It was found that it is vitally necessary to include surface recombination in the calculation in order to reproduce the experimental spectra. For this, we included SRH recombination through a U-shaped surface state continuum similarly to the case of the PLS3 method (Ishikawa and Hasegawa, 2002b). Thus, the method can be used for the surface state analysis. 4. Controls of surface states in nano-structures toward nano-photovoltaics 4.1. IQ-chip and nano-photovoltaics Recent rapid progress of nanotechnology has created interest in the photovoltaic area, as already mentioned previously. According to literature (Corkish et al., 2002), advantages expected in solar cells utilizing nano-structures is reduction of amount of materials, increased light absorption, higher efficiency, improvement of spatial uniformity, reduced contact area, tunability of effective band gap due to quantum confinement etc. These features may be combined to produce high efficiency cells with lower cost as the next generation, clean and renewable energy sources. Another possibility we are currently pursuing is concerned with a new type of application of nanostructure-based solar cells to the ubiquitous network era. Namely, recent revolutionary progress of the internet and wireless technologies has created a concept of the ‘‘ubiquitous network society’’ for the new century where new information technology (IT) is being developed not only for further advance of internet, but also formation of new wireless networks such as those including smart dusts, RFID (radio frequency identification) chips and sensor networks. With this background, we are currently developing a new smart chip called an intelligent quantum (IQ) chip (Hasegawa et al., 2004) schematically shown in Fig. 10(a). An IQ chip is a III-V semiconductor chip with sizes of millimeter square or below where nanometer scale quantum processors and memories are integrated on chip with capabilities of wireless communication, wireless power supply and various sensing functions. It is an attempt to endow ‘‘more intelligence’’ than simple identification (ID) like in RFID chips to semiconductor chips so that they can be utilized as versatile tiny ‘‘knowledge vehicles’’ to be embedded anywhere in the society. Here, we plan

to use nano-structure-based solar cells as the wireless power source for the chip. We use ultra-low power III–V quantum LSIs based on nano-structures, because the power density of the current and future Si CMOS technology is too high. Its value is several 10–100 W/cm2, and is several orders of magnitude larger than the solar energy of 100 mW/cm2. The actual structure of a nano-photovoltaic cell we are trying to integrate on-chip is schematically shown in Fig. 10(b). Here, photo-generated electrons and holes in a quantum wire (QWR) array are collected, respectively, by low work function and high work function metal nano-contacts to achieve a high conversion efficiency without utilizing p–n junction. The actual structure of the QWR grown by our selective MBE growth method (Hasegawa, 2003) is shown at the bottom of Fig. 10(c) together with a plan view SEM image of its array shown on the top. 4.2. Unpinning of Fermi level at nanometer-scale metal contacts The operation principle of the solar cell based on metal contacts shown in Fig. 10(b) is the following. Operation of a solar cell includes two elementary processes. One is to generate electron-hole pairs by energy conversion from solar energy photons, and the other is to separate them before they recombine so that their energy difference is consumed in the outside circuit. Usually, the work function difference of the p–n junction is utilized to achieve the second step. However, this should be possible by work function difference of metal contacts, provided that the Schottky barrier heights (SBHs) for electrons, /Bn and that for holes, /Bp, can be adjusted by the metal work function, /m, through the following equation in the so-called ideal Schottky limit: /Bn ¼ /m  vs ; /Bp ¼ Eg  ð/m  vs Þ

ð8Þ

where vs and Eg are the electron affinity and the energy gap of the semiconductor, respectively. However, it is well known that such a limit never takes place in the ordinary metal contact. Due to strong Fermi level pinning at the metal–semiconductor interface, SBHs are independent of, or only very weakly dependent on, the metal work function, being given approximately by the following equations in the so-called Bardeen limit: /Bn ¼ Ec  EHO ;

/bp ¼ EHO  Ev

ð9Þ

H. Hasegawa et al. / Solar Energy 80 (2006) 629–644

Ultra-low power III-V quantum LSIs (nanoprocessor, nanomemory)

Wireless communication circuit microwave millimeter wave THz light

sun light room light nanophotovoltaics

641

communication circuits

CPU, memory

rectenna, solar cell

sensors

a few 100-1,000 µm Ultra-high sensitive sensor utilizing nanostructures

Power supply (a)

GaAs QWR arrays on (001) low workfunction metal nano-contacts QWR array

plan view 3 m

GaAs QWR

<110>

AlGaAs

side QW high workfunction metal nano-contacts

vertical stacking of QWRs with different wire dimensions

(b)

GaAs buffer (001) GaAs sub. (c)

bottom QW

Fig. 10. (a) Concepts of the IQ chip, (b) nano-photovoltaic device based on a quantum wire (QWR) array and metal nano-contacts. (c) QWR structure and a plan view SEM image. QWR arrays are fabricated by the selective MBE growth method.

From the viewpoint of the MIGS model in Table 1, the Fermi pinning is an intrinsic one caused by the metal wavefunction tail, and SBHs can not be changed by the metal work function. However, we believe from the viewpoint of the DIGS model that a situation near the ideal Schottky limit will be realized if a nano-meter scale contact is produced by a low energy metal deposition process (Hasegawa and Sato, 2005), since stress associated with metal deposition is much smaller in a small area contact. From such an idea, we have been characterizing various nano-contacts on III–V materials formed by our pulsed in situ electrodeposition process. As shown in Fig. 11(a) for a Pt/InP system, the electrodeposition process produces metal nano-particles on the surface. The measured work function dependence of SBH is shown in Fig. 11(b) for elec-

trodeposited nano-contacts (contact diameter 50 nm) on n-GaAs. A remarkably strong work function dependence is seen here, although it is well known that macroscopic metal contacts on GaAs show particularly strong Fermi level pinning in the Bardeen limit. This is a very promising result for further development of nano-photovoltaics. 4.3. Si interface control layer-based surface passivation of nanowires Another very critical issue in nano-photovoltaics is the surface passivation of nano-structures, since the surface-to-volume ratio becomes large in nanostructures. For our nano-structure solar cell, we have been trying our Si interface control layer (Si ICL)-based passivation scheme on the QWR arrays

642

H. Hasegawa et al. / Solar Energy 80 (2006) 629–644

shown in Fig. 10(c) (Shiozaki et al., 2005). The basic passivation structure is shown in Fig. 12(a) for a planar structure where Si ICL is grown be MBE. A UHV STS spectra obtained after MBE growth of monolayer level of Si on Ga-rich (4 · 6) is shown

in Fig. 12(b). It is seen that anomalous STS behavior seen in Fig. 5 disappeared completely, showing a normal conductance gap. The PL spectra from the QWR array obtained before and after Si-ICL passivation are in Fig. 12(c). Here, we see a large

1.0

Pt

(eV)

Bardeen limit 0.8

SBH, φ

Bn

S = 0.38 Ni

0.6 Sn

Schottky limit

0.4 4.0

(a)

4.5 5.0 5.5 Workfunction, φm (eV)

6.0

(b) Fig. 11. Nano Schottky contacts formed by the electrodeposition process. (a) Plan view of the Pt nano particles on InP and (b) metal workfunction dependent Schottky barrier heights of nano Schottky contacts formed on n-GaAs.

Si3N4 (1.5nm)

SiO2 (50nm) GaAs or AlxGa1-xAs epitaxial layer

Si ICL (0.5 nm)

GaAs buffer n+-GaAs substrate (a)

PL intensity (arb. units)

2.0 STS

(dI/dV)/(I/V)

1.5 1.0 Ev

0.5

Ec

(001) QWR 25K GaAs after Si-ICL passivation

d=10nm

QWR

before passivation

0 -2

-1 0 1 Sample bias (V) (b)

2

1.4

1.5 1.6 Photon energy (eV)

1.7

(c)

Fig. 12. Si-ICL based surface passivation technique. (a) Passivation structures, (b) STS spectra and (c) increase of PL intensity from GaAs QWR after Si-ICL passivation.

H. Hasegawa et al. / Solar Energy 80 (2006) 629–644

increase of PL intensity, indicating successful passivation. According to a theoretical analysis, surface recombination caused by carrier tunneling from QWR region is almost completely suppressed due to removal of surface states. 5. Summary In this paper, a series of our work on characterization methods and fundamental aspects of surface/interface states and recombination process in Si and III–V materials are reviewed. In addition to the conventional metal–insulator– semiconductor (MIS) capacitance–voltage (C–V) method, a contactless C–V method and the microscopic scanning tunneling spectroscopy (STS) method are discussed, and advantages and measurement pitfalls are pointed out. Then, general features of surface states are discussed. Surface states are shown to have U-shaped distributions of donor–acceptor continuum with a characteristic charge neutrality level EHO. Rigorous simulation of dynamics of surface recombination process has shown that the effective surface recombination velocity, Seff, is not a constant of the surface but its value changes by many orders of magnitude with the incident light intensity and the polarity and amount of fixed charge. This led to two related characterization techniques called the photoluminescence surface state spectroscopy (PLS3) and the cathodoluminescence in-depth spectroscopy (CLIS). Finally, attempts to control surface states and Fermi level at metal semiconductor interface and free surfaces of nano-structures were presented, giving promising results and hope for future ‘‘nanophotovoltaics’’. Acknowledgements This work is supported by a 21st Century COE (Center of Excellence) program at Hokkaido University on ‘‘Meme-media technology approach to the R&D of next generation ITs’’ from MEXT, Japan. References Adamowicz, B., Hasegawa, H., 1998. Computer analysis of surface recombination process at Si and compound semiconductor surfaces and behavior of surface recombination velocity. Jpn. J. Appl. Phys. 37, 1631–1637.

643

Adamowicz, B., Miczek, M., Arabasz, S., Hasegawa, H., 2002. Rigorous analysis of photoluminescence efficiency for characterization of electronic properties of InP(0 0 1) surfaces. Vacuum 67, 3–10. Corkish, R., Kettemann, S., Nelson, J., (Guest Editors), 2002. see papers in special issue of Physica E 14, issue 1–2 on ‘‘Nanostructures on Photovoltaics’’. Hasegawa, H., 2003. Formation of III–V low dimensional structures and their applications to intelligent quantum chips. Microelectronics Journal 34, 341–345. Hasegawa, H., Sawada, T., 1983. On the electrical properties of compound semiconductor interfaces in metal/insulator/semiconductor structures and the possible origin of interface states. Thin Solid Films 103, 119–140. Hasegawa, H., Ohno, H., 1986. Unified disorder induced gap state model for insulator–semiconductor and metal–semiconductor interfaces. J. Vac. Sci. Technol. B 4, 1130–1138. Hasegawa, H., Negoro, N., Kasai, S., Ishikawa, Y., Fujikura, H., 2000. Effects of gap states on scanning tunneling spectra observed on (1 1 0)- and (0 0 1)-oriented clean surfaces and ultrathin Si layer covered surfaces of GaAs prepared by molecular beam epitaxy. J. Vac. Sci. Technol. B 18, 2100– 2108. Hasegawa, H., Inagaki, T., Ootomo, S., Hashizume, T., 2003. Mechanism of current collapse and gate leakage currents in AlGaN/GaN heterostructure field effect transistors. J. Vac. Sci. Technol. B 21, 1844–1855. Hasegawa, H., Kasai, S., Sato, T., 2004. Hexagonal binary decision diagram quantum circuit approach for ultra-low power III–V quantum LSIs. IEICE Trans. Electron. E87-C, 1757–1768. Hasegawa, H., Sato, T., 2005. Electrochemical processes for formation, processing and gate control of III–V semiconductor nanostructures. Electrochimica Acta 50, 3015–3027. Heine, V., 1965. Theory of surface states. Phys. Rev. 138, 1689– 1696. Iizuka, K., Hashizume, T., Hasegawa, H., 1997. Small-signal response of interface states at passive InGaAs surfaces from low frequencies up to microwave frequencies. Solid State Electron. 41, 1463–1468. Ishikawa, F., Hasegawa, H., 2002a. Depth-resolved cathodoluminescence characterization of buried InGaP/GaP heterointerfaces. Appl. Surf. Sci. 190, 508–512. Ishikawa, F., Hasegawa, H., 2002b. Self-consistent computer analysis of cathodoluminescence in-depth spectra for compound semiconductor heterostructures. Inst. Phys. Conf. Ser. 170, 461–466. Ishikawa, Y., Tsurumi, N., Fukui, T., Hasegawa, H., 1998. Scanning tunneling microscopy and X-ray photoelectron spectroscopy studies of atomic level structure and Fermi Level pinning on GaAs (1 1 0) surfaces grown by molecular beam epitaxy. J. Vac. Sci. Technol. B 16, 2387–2394. Kampen, T.U., Mo¨nch, W., 1997. Barrier heights of GaN Schottky contacts. Appl. Surf. Sci. 117/118, 388–393. Kobayashi, H., Yamashita, Y., Nakato, Y., Komeda, T., Nishioka, Y., 1996. Interface states at ultrathin oxide/ Si(1 1 1) interfaces obtained from X-ray photoelectron spectroscopy measurements under biases. Appl. Phys. Lett. 69, 2276–2278. Martensson, P., Feenstra, R.M., 1989. Geometric and electronic structure of antimony on the GaAs (1 1 0) surface studied by scanning tunneling microscopy. Phys. Rev. B 39, 7744–7753.

644

H. Hasegawa et al. / Solar Energy 80 (2006) 629–644

Miczek, M., Adamowicz, B., Szuber, J., Hasegawa, H., 2001. Computer analysis of photoluminescence efficiency at InP surface with U-shaped surface state continuum. Vacuum 63, 223–227. Negoro, N., Anantathanasarn, S., Hasegawa, H., 2003. Effects of Si deposition on the properties of Ga-rich (4 · 6) GaAs (0 0 1) surfaces. J. Vac. Sci. Technol. B 21, 1945–1952. Sawada, T., Numata, K., Tohdoh, S., Saitoh, T., Hasegawa, H., 1993. In-situ characterization of compound semiconductor surfaces by novel photoluminescence surface state spectroscopy. Jpn. J. Appl. Phys. 32, 511–517. Shiozaki, N., Sato, T., Hasegawa, H., 2005. Effects of surface states and Si-interlayer based surface passivation on GaAs quantum wires grown by selective molecular beam epitaxy. J. Vac. Sci. Technol. B 23, 1714–1721. Spicer, W.E., Lindau, I., Skeath, P.R., Su, C.Y., 1980. Unified defect model and beyond. J. Vac. Sci. Technol. 17, 1019–1027. Sze, S.M., 1981a. Physics of Semiconductor Devices, second ed. John Wiley & Sons, New York, pp. 7–60, Chapter 1. Sze, S.M., 1981b. Physics of Semiconductor Devices, second ed. John Wiley & Sons, New York, pp. 362–430, Chapter 7. Takahashi, H., Hashizume, T., Hasegawa, H., 1999. X-ray photoelectron spectroscopy and ultrahigh vacuum contactless

capacitance–voltage characterization of novel oxide-free InP passivation process using a silicon surface quantum well. Jpn. J. Appl. Phys. 38, 1128–1132. Tersoff, J., 1984. Schottky barrier heights and the continuum of Gap states. Phys. Rev. Lett. 52, 465–468. Van de Walle, C.G., Neugebauer, J., 2003. Universal alignment of hydrogen levels in semiconductors, insulators and solutions. Nature 423, 626–628. Veal, T.D., Mahboob, I., Piper, L.F.J., McConville, C.F., Lu, H., Schaff, W.J., 2004. Indium nitride: evidence of electron accumulation. J. Vac. Sci. Technol. B 22, 2175–2178. Yamashita, Y., Namba, K., Nakato, Y., Nishioka, Y., Kobayashi, H., 1996. Spectroscopic observation of interface states of ultrathin silicon oxide. J. Appl. Phys. 79, 7051–7057. Yoshida, T., Hashizume, T., Hasegawa, H., 1997. Characterization of interface electronic properties of low-temperature ultrathin oxides and oxynitrides formed on Si(1 1 1) surfaces by contactless capacitance–voltage and photoluminescence methods. Jpn. J. Appl. Phys. 36, 1453–1459. Yoshida, T., Hasegawa, H., 2000. Ultrahigh-vacuum contactless capacitance voltage characterization of hydrogen-terminatedfree silicon surfaces. Jpn. J. Appl. Phys. 39, 4504–4508.