Transport mechanism of microcrystalline silicon thin films

F. Liu a,*, M. Zhu a, Y. Feng a, Y. Han a, J. Liu a, S. Kasouit b,R. Vanderhaghen b

a Department of Physics, Graduate School, Chinese Academy of Sciences, University of Science and Technology of China,

P.O. Box 3908, Beijing 100039, People’s Republic of Chinab Laboratoire de Physique des Interfaces et des Couches Minces, Ecole Polytechnique, 91128 Palaiseau cedex, France

Abstract

The dark conductivity, rdðT Þ, and the time-resolved microwave conductivity (TRMC) of hydrogenated micro-

crystalline silicon (lc-Si:H) films prepared by hot-wire chemical vapor deposition (HWCVD) and very-high-frequency

plasma-enhanced CVD have been investigated. The rdðT Þ simulations were carried out based on a simplified energy

band model. Fitting the rdðT Þ data indicates that the electrical transport mechanism in lc-Si:H film depends on the

nanocrystallite volume fraction (Xc). Around room temperature, thermionic emission dominates for samples with low

Xc, however, for high Xc, tunneling is the major conduction mechanism. In the low temperature range, the behavior of

films with low-Xc is fairly explained by variable range hopping conduction. An activation energy of 0.15 eV related to

oxygen was found in HWCVD films. TRMC measurements show that the mobility increases with Xc. � 2002 Elsevier

Science B.V. All rights reserved.

PACS: 72.80.Cw; 73.63.Bd; 73.90.þf

1. Introduction

Hydrogenated microcrystalline silicon films,combining the superior electronic properties ofcrystalline silicon with advantages of amorphoussilicon (a-Si:H), have attracted considerable at-tention for large-area electronic applications. Op-timized, undoped hydrogenated microcrystallinesilicon (lc-Si:H) films show good stability againstlight-induced degradation and improve the spec-tral response of solar cells in red and infrared light.

However, lc-Si:H films exhibit a mesoscopic and/or microscopic inhomogenity with a mixture ofgrains, grain boundaries, voids and an amorphousphase. Hence the properties of electronic transportare complicated, especially in undoped lc-Si:H,and remain poorly understood. Efforts to under-stand the essential transport process in lc-Si:Hhave been carried out [1–5], with the transportmodel being based on the structural features.Overhof and Otte [1] and Shimakawa [2] describelc-Si:H as a continuous mixture of crystallineand amorphous silicon. Computer simulationsby Overhof et al. reported a critical percolationthreshold for Xc � 0:32. Shimakawa found Hallmobility, lH, is sublinearly proportional to thesize of crystallites when Xc is kept the same and

Journal of Non-Crystalline Solids 299–302 (2002) 385–389

www.elsevier.com/locate/jnoncrysol

* Corresponding author. Tel.: +86-10 6822 6506; fax: +86-10

6821 0501.

E-mail address: liu_fz@hotmail.com (F. Liu).

0022-3093/02/$ - see front matter � 2002 Elsevier Science B.V. All rights reserved.

PII: S0022-3093 (02 )00915-8

discussed the Xc dependence of rd and lH in termsof an effectivemedium approximation. A sequentialtunneling model [3] and hopping in donor states [4]were proposed to explain the conductivity temper-ature dependence and the low-temperature con-duction in heavily doped lc-Si:H. Despite theincreasing number of investigations, both micro-structure and conduction mechanism of lc-Si:Hare far from being well-established.

In this paper, a simplified energy band model oflc-Si:H is proposed. Considering different trans-port mechanisms, rdðT Þ in lc-Si:H with differentXc is well fitted. The effect of post-oxidation on thetransport of lc-Si:H is discussed.

2. Experimental

Microcrystalline Si films were prepared byhot-wire chemical vapor deposition (HWCVD)and very-high-frequency plasma enhanced CVD(VHF-PECVD) with a gas mixture of SiH4 andH2. For the HWCVD set of samples, tungsten wireof 0.35 mm diameter was used as the filament andthe filament temperature, Tf , was fixed at 2000 �C.The flow ratio of the gases, RH ¼ H2=ðSiH4 þH2Þwas set at 0.83 and the deposition pressure, p, wasvaried. The VHF-PECVD set of samples wasprepared using a 95 MHz plasma at a depositionpressure of 40 Pa with varying RH and plasmapower, Pw. Sample PH29 was deposited by hotwire assisted PECVD (HW-PECVD) techniquewith plasma excited frequency of 13.56 MHz andTf ¼ 1800 �C. The films were deposited on c-Siand Corning 7059 glass at substrate temperaturesof 200 �C for VHF-PECVD samples and 300 �C

for HWCVD and HW-PECVD samples. The de-position parameters are summarized in Table 1.

The structure of the lc-Si:H films was charac-terized by X-ray diffraction (XRD). Fourier trans-form infrared spectroscope Nicolet 560 (FTIR).The crystallinity of lc-Si:H films was determinedusing a Raman scattering spectroscope JYU1000.The void volume fraction, XV, of HWCVD sampleswas determined by spectro-ellipsometry. The rdðT Þwas measured in a coplanar configuration in anAPCI DE202 cryogenic refrigeration system overthe temperature range of 100–300 K.

The carrier mobility was measured by usingtime-resolved microwave conductivity (TRMC)set-up: K band microwaves (100 dBm, 30 GHz)are incident and reflected on a lc-Si:H samplewhich is exposed to a 4 ns pulsed, 532 nm YAGlaser. The photo-generated carriers induce a tran-sient variation in the microwave reflectivity, andhence reflectivity power detected by a diode. De-tailed descriptions of the TRMC measurement andcarrier mobility analysis can be found in Ref. [6].

3. Results

The average lc-Si:H grain size, L, was deter-mined from XRD. The Xc was evaluated by a ratioof an integrated Raman intensity of the crystallinephase to a total integrated intensity. In the calcu-lation, the experimentally determined value of 0.88[7] was used for the ratio of integrated Ramancross sections of c-Si to a-Si. The film structuralparameters are listed in Table 1.

The rdðT Þ for HWCVD and VHF-PECVDsamples is plotted in Fig. 1(a) and (b), respectively.

Table 1

Deposition parameters and film structure features

Name p (Pa) RHð%Þ � 0:5 PW (W) LðnmÞ � 0:5 Xcð%Þ � 0:5 Xvð%Þ � 0:5

H55 1 83 25.6 57 1

H53 3 83 26.8 77 6

H56 10 83 29.2 79 13

PH29 100 98 40 46.0 93

A98 40 98 5 88

A95 40 95 5 35.2 75

A94 40 94 5 20.0 68

B30 40 95 30 63

B40 40 95 40 69

386 F. Liu et al. / Journal of Non-Crystalline Solids 299–302 (2002) 385–389

It is clear that the conductivity increases withXc. The conductivity values of HWCVD films areabout two orders magnitude larger than the onesfrom VHF-PECVD films. The slope of the rdðT Þcurves for HWCVD films varies continuously. TherdðT Þ curves of VHF-PECVD films show a kinkaround 220 K. Leaving out the detailed analysis ofthe TRMC signals, the effective mobility of lc-Si:Has a function of Xc are plotted in Fig. 2. It can beseen that the TRMCmobility is proportional to Xc.

4. Discussion

To understand the temperature and Xc depen-dence of the conductivity, a model has to be es-tablished. Microcrystalline silicon film shows verycomplex structure. However, considering the col-umnar crystal growth [8] and the coplanar mea-surement, we can regard, as a first approximation,lc-Si:H as an array of crystalline and amorphous

silicon regions with interfaces between them. Thegrain size can be taken as the average grain size L.The amorphous barrier width S was set as a sim-ulation parameter.

The band offset of an a-Si:H/c-Si heterojunctionis a critical parameter for establishing an energyband model. According to Sebastiani et al. [9], thevalence band offset of the a-Si:H/c-Si junction,DEV, is 0:44� 0:02 eV. The energy gaps for c-Siand a-Si:H are 1.1 and 1.7 eV, respectively, there-fore, the conduction band offset, DEC, equals0:16� 0:02 eV. Dangling bonds may be created atthe interfaces. It is assumed that the danglingbonds can be neglected because of the passivationby hydrogen as suggested by Overhof and Otte [1].A simplified energy band structure of lc-Si:H issketched in Fig. 3. EC and EV represent edges ofthe conduction and valence bands, respectively. EF

is the Fermi level. E1 and E2 represent the activa-tion energy, Ea, of c-Si and a-Si:H, respectively,and E2 ¼ E1 þ DEC.

Based on Fig. 3, the rdðT Þ of lc-Si:H was an-alyzed by various transport mechanisms includingthermionic emission of carriers over the potentialbarriers, quantum tunneling through the barrier,conduction through the band-tail states and vari-able range hopping (VRH) between localizedstates at low temperatures.

It is known that the thermionic emission currentdensity Je and tunneling current density Jt have thesimilar expression as

Fig. 1. Temperature dependence of dark conductivity for

samples H56, H53, H55 (a) and A94, A95 (b). The symbols are

experimental data. The dashed lines are the fitting curves at

different temperature ranges. The solid lines are the whole fit-

ting result.

Fig. 2. Xc dependence of the TRMC mobility for lc-Si:H films

prepared by different techniques. The uncertainties of the data

are �0.5 cm2 V�1 s�1.

F. Liu et al. / Journal of Non-Crystalline Solids 299–302 (2002) 385–389 387

JeðJtÞ ¼ J0eðJ0tÞ exp�� E1

kT

�: ð1Þ

The prefactors J0e and J0t are given by

J0e ¼ A�T 2fse exp�DEc

kT

� �; ð2Þ

J0t ¼ A�T 2t expV2kT

� ��� exp

�V2kT

� ��: ð3Þ

Here, k is the Boltzmann constant, A� is theRichardson–Fermi constant to be 260 Acm�2K�2

for electrons in silicon. fse is a factor related tothe reduction in current density due to electron–phonon scattering and quantum reflection at theinterface. V is the bias voltage and t represents thetransmission coefficient of electron tunneling. AsDEc kT , it can be expressed as [10]

t ¼ 16kTDEc

exp

�� 2S

�h2m�DEc½ �1=2

�: ð4Þ

Here, m�, the electron effective mass, is 0.2m0. Eq.(2) indicates that J0e increases monotonically withT but is independent of S as shown in Fig. 4(a), inwhich the factor fse was set as 0.1. Fig. 4(b) showsthat J0t is almost independent of T for a fixed Sbut decreases rapidly with increasing S. At roomtemperature and in comparison with J0e, J0t can beneglected when S > 5 nm. Based on the aboveanalysis, the simulations of rdðT Þ around roomtemperature were carried out; they fitted the ex-perimental data well, as show in Fig. 1. The fittingparameters are listed in Table 2. It is noted in

Table 2 that E1 and E2 are smaller than the Ea ofintrinsic c-Si and a-Si:H. This is probably dueto the oxygen contamination. We will discuss itlater. The parameter fse ¼ 0:07–0:15 is in agree-ment with the report of Lombardo et al. [11].

In the low-temperature range, the ln rd �ðT0=T Þ1=4 dependence VRH conduction was usedto fit the rdðT Þ in which T0 ¼ 24a3=ðpkgðEÞÞ. Thefitting curves are plotted in Fig. 1. The value of 1/a, the extended length of the localized states, is1:1–1:3 nm which are consistent with the resultsfrom Muller et al. [12]. The density of localizedstates, gðEÞ, for HWCVD samples is higher thanthat of the VHF-PECVD samples. The results showthat, based on the transport mechanisms men-

Fig. 3. Sketched energy band structure of lc-Si:H thin film.

Fig. 4. (a) The temperature dependence of J0e with S ¼ 4 nm

and barrier width dependence of J0e calculated at room tem-

perature. (b) The temperature dependence of J0t with S ¼ 4 nm

and barrier width dependence of J0t at different temperatures.

388 F. Liu et al. / Journal of Non-Crystalline Solids 299–302 (2002) 385–389

tioned above, the rdðT Þ of VHF-PECVD samples ismuch better fitted than that of HWCVD samples.In the latter case, a 0.15 eV activation energy, whichcorresponds to oxygen as a donor in silicon, mustbe included in the middle temperature rangeshowing in Fig. 1(a). Oxygen does exist in HWCVDsamples from FTIR spectra [13]. For example, theoxygen concentration in sample H56, deduced fromthe FTIR spectrum, is 1:9� 1019 cm�3. This oxygencontamination was mainly caused by post oxida-tion due to the porosity of lc-Si:H film. The con-ductivity contributed by active oxygen ions at roomtemperature, rO(300 K), can be deduced by ex-tracting the middle dashed lines, Ea ¼ 0:15 eV inFig. 1(a), to 300 K. Combining the rO (300 K) withTRMC mobility data, Fermi levels of samples H55and H53 were calculated to be 0.33 and 0.49 eV,respectively, which is fairly consistent with experi-mental data E1 in Table 2. And the correspondingcarrier concentration nO from rO (300 K) is1� 1013, 1:7� 1011 and 9� 109 cm�3 for H56, H53andH55, respectively. Data XV, nO and E1 imply thehigher porosity, the higher oxygen content andlower Ea. Simulation suggests that HWCVD sam-ples are more porous than VHF-PECVD samplesand may be the reason for higher conductivity inHWCVD samples than in VHF-PECVD samples.

5. Conclusion

The conductivity temperature dependence oflc-Si:H has been well simulated by thermionicemission, tunneling and VRH conduction when Xc

is smaller than 80%. Around room temperature,electrical transport in lc-Si:H depends on the Xc.The thermionic emission and tunneling currentdominate the conduction process. However, forfilm with low Xc, larger barrier width, the tunnel-ing current can be neglected. In the low-tempera-

ture range, although tunneling current is muchlarger than thermionic emission current, the VRHbecomes the major conduction mechanism. It issuggested that oxygen contamination contributedto the activation energy of 0.15 eV and muchenhanced the conductivity of HWCVD samples.

Acknowledgements

The authors are grateful to Dr. Finger of In-stitute of photovoltaic for supply the VHF sam-ples. This work was supported by G2000028208,NSFC and PRA-98MX-01.

References

[1] H. Overhof, M. Otte, in: Proceeding of the 9th Interna-

tional School on Condensed Matter Physics, World Scien-

tific, Singapore, 1997, p. 23.

[2] K. Shimakawa, J. Non-Cryst. Solids 266–269 (2000) 223.

[3] A.D. Nocera, A. Mittiga, A. Rubini, J. Appl. Phys. 78

(1995) 3955.

[4] R. Carius, J. Muller, F. Finger, N. Harder, P. Hapke, in:

Thin Film Materials and Devices, World Scientific, Singa-

pore, 1999, p. 157.

[5] F. Liu, M. Zhu, Y. Feng, Y. Han, J. Liu, Thin Solid Films

395 (2001) 97.

[6] R. Brenot, R. Vanderhaghen, B. Drevillon, I. French,

P. Roca, I. Cabarrocas, Thin Solid Films 296 (1997) 94.

[7] R. Tsu, J. Gonzaicz-Hernandez, S.S. Chao, S.C. Lee,

K. Tanaka, Appl. Phys. Lett. 40 (1982) 534.

[8] M. Zhu, X. Guo, G. Chen, H. Han, M. He, K. Sun, Thin

Solid Films 360 (2000) 205.

[9] M. Sebastiani, L. Digaspare, C. Bittencourt, F. Evangelisti,

MRS Symp. Proc. 377 (1999) 209.

[10] C.R. Wronski, P.D. Persans, B. Abeles, Appl. Phys. Lett.

49 (1986) 569.

[11] S. Lombardo, S.U. Campisano, F. Baroetto, Phys. Rev. B

47 (1993) 13561.

[12] J. Muller, F. Finger, R. Carius, H. Wagner, MRS Symp.

Proc. 507 (1998) 751.

[13] C. Guo, MS thesis, Graduate School, Chinese Academy of

Sciences, 1995.

Table 2

Simulation parameters

Name E1ðeVÞ � 0:05 E2ðeVÞ � 0:02 fse � 0:05 SðnmÞ � 0:5 gðEÞ (cm�3 eV�1)

H55 0.45 0.61 0.07 7.1 1:9� 1018 � 0:5H53 0.44 0.60 0.13 2.8 1:2� 1019 � 0:5

H56 0.34 0.50 0.15 2.5 3:4� 1019 � 0:5

A95 0.49 0.65 0.14 3.0 4:3� 1018 � 0:5

A94 0.47 0.63 0.11 5.0 6:2� 1018 � 0:5

F. Liu et al. / Journal of Non-Crystalline Solids 299–302 (2002) 385–389 389