Physica B 322 (2002) 146–153

Porous Si formation and study of its structural and vibrationalproperties

B.K. Patela,1, R. Mythilib, R. Vijayalaxmib, R.K. Sonic, S.N. Beheraa,*, S.N. Sahua

a Institute of Physics, Sachivalaya Marg, Bhubaneswar, 751 005, IndiabPhysical Metallurgy Section, Material Characterization group, Indira Gandhi Center of Atomic Research, Kalpakkam, 603102, India

cDepartment of Physics, Indian Institute of Technology, New Delhi, 110016, India

Received 18 April 2001; accepted 28 January 2002

Abstract

In situ current ðIÞ–ðV Þ voltage studies were carried out to get a better understanding of the mechanism of formation

of porous silicon (PS). It is observed that on decreasing the anodization current density below a critical value

(B75 mA cm�2) the size of the PS crystallites increases while for its values above 75 mA cm�2 electropolishing occurs.

Raman spectroscopic studies show that the sizes of the Si crystallites are small and change from 4:7 to 3:8 nm when the

current densities are increased from 20 to 50 mA cm�2: Transmission electron micrographs show preferential

propagation of pores whereas transmission electron diffraction (TED) patterns show typical crystalline Si with the

cubic structure. r 2002 Elsevier Science B.V. All rights reserved.

PACS: 81.20.�n; 78.30.�j; 81.60.Cp; 78.50.Ge

Keywords: Porous Si; Etching; Raman scattering

1. Introduction

Bulk crystalline Si (c-Si) is extremely inefficientin emitting radiation under either optical orelectrical excitations. The minimum of the con-duction band lies at the X ðð2p=aÞð1; 0; 0ÞÞ pointalong the Dð1; 0; 0Þ direction of the Brillouin zoneresulting in an indirect band gap, Eg ¼ EðX1Þ �EðG250 Þ ¼ 1:17 eV; which is responsible for non-

radiative recombination of charge carriers. Theproperties of bulk-Si get modified when Si isprepared such that it acquires low-dimensionalstructures. Porous silicon(PS) [1–25] is one suchlow-dimensional structure consisting of a networkof randomly spaced pores [6,13,16] in bulk siliconformed by electrochemical etching of a crystallineSi wafer in concentrated hydrofluoric acid (HF).Such porous structures have attracted consider-able attention in recent years due to its possibleapplication in optoelectronics [7], sensors [7,8] andsolar cells [8]. Although PS is one of the popularmaterials in Nanotechnology, the origin of theluminescence in it is still not understood. Differentmodels related to quantum confinement (QC) [2],

*Corresponding author. Tel.: +91-674-301172; fax: +91-

674-300142.

E-mail addresses: snb@iopb.res.in (S.N. Behera),

sahu@iopb.res.in (S.N. Sahu).1Now deceased.

0921-4526/02/$ - see front matter r 2002 Elsevier Science B.V. All rights reserved.

PII: S 0 9 2 1 - 4 5 2 6 ( 0 2 ) 0 1 1 7 5 - 4

and the presence of foreign impurities like siloxene[17], and chemisorbed molecules, e.g. SiO2; SiH [9]etc. have been invoked to explain the origin of theinteresting and efficient visible luminescence dis-covered by Canham in 1991 [1]. The presence offoreign impurities has been attributed to arisefrom the solution during the PS formation. Thesynthesis of porous Si under anodic bias results inboth Si dissolution and the formation of SiH andSiO2 at the interface of PS/HF-electrolyte [10]which in turn is expected to generate surfacetraps [25–27] of different types within the wideband gap of porous Si. The present workattempts to understand the PS surface morphol-ogy, its structure, and the mechanism of the PSformation, through the studies of in situ currentðIÞ–ðV Þ voltage characteristics, Rutherford backscattering (RBS) analysis which relates to theestimation of the thickness of the PS layer and thestudy of vibrational properties by laser Ramanscattering.

2. Experimental details

The in situ I–V characteristics measurementswere carried out at 3001 K using a three electrodesingle compartment electrochemical cell compris-ing of (i) an aluminium back coated p-Si ð1 0 0Þsample, having resistivity of around 0:01 O cm asan anode having a surface area B1 cm2; (ii) asaturated calomel electrode (SCE) as referenceelectrode and (iii) a Pt cathode of area 2 cm2: Theelectrolyte was a mixture of HF acid(49%)þethanol in equal proportion and thevoltage was applied with a Tacusol potentiostat.Porous Si samples were prepared with fourdifferent current densities of 20, 30, 40 and50 mA cm�2; respectively, maintained for a con-stant duration of 30 min each. Transmissionelectron microscopy (TEM) was performed onthe samples using the Philips CM200 AnalyticalTEM operating at 200 kV and fitted with anEDAX analyser with a super ultra thin windowsystem. For TEM studies, the porous Si layer waspealed off from the sample with a sharp blade andwas collected on a carbon-coated Cu grid. Beforelifting the grid, the layers were given a dilute HF

ðHF : H2O ¼ 1 : 3Þ acid treatment to removepossible SiOx traces. Raman spectra of the PSsamples were recorded at room temperature inthe backscattering configuration, using the488 nm line of the Ar-ion laser and a doublemonochromator with the standard detectionsystem. Because of the low thermal conductivityof porous Si and the expected temperaturedependence of the Raman shifts and line widths,care was taken to avoid laser heating by keepingthe laser power low ðE30 mWÞ with beam size ofabout C10 mm:

3. Results and discussion

The results of various measurements performedon p-Si samples are presented and discussedbelow. The measurements carried out were thestructure and morphology studies using the TEM,the current–voltage (I–V ) characteristics, theRutherford back scattering (RBS) studies todetermine the size of the silicon nanostructure,and the Raman scattering studies which againprovides information on the nanostructure ofporous silicon. The results of these measurementswere analysed in order to gain a better under-standing of the mechanism of pore formation insilicon.

3.1. TEM study

In order to study the surface pore morphologyof porous Si under high-current etching condi-tions, the sample was synthesized at 50 mA cm�2:The porous silicon sample thus prepared wasimaged using the transmission electron microscope(TEM). The TEM micrograph which is the brightfield (BF) image is shown in Fig. 1(a). The surfaceshows anistropy without branching and is rela-tively smooth and porous. Note that the directionof the current flow during the preparation ofsample was from the bottom to the top. We arenot able to measure the individual pore diametersdue to the overlapping phenomena which iscommon in less porous material (porosityo70%). However, voids separated by rods run-ning perpendicular to the surface can be seen

B.K. Patel et al. / Physica B 322 (2002) 146–153 147

clearly. Due to large current density at the poretips, the voids propagate perpendicular to thesurface as seen in Fig. 1(a). The crystallite sizescould not be estimated accurately due again tooverlap of adjacent pores normally associated withlow porosity samples [24]. The nano-porous Sisample with the above morphology, when excitedwith the 369 nm wavelength radiation from theHg–Xe lamp, emits red light in the 600–640 nmrange (details will be published elsewhere [28])which can be seen with the naked eye, whichsuggests the porosity to be less than 70%. In orderto identify the structure and phase of the crystal-lites, a selected area diffraction (SAD) pattern wastaken (Fig. 1(b)) using TEM for the same region ofthe sample for which the BF image is shown inFig. 1(a). A single-crystalline pattern typical of thediamond structure is observed. However, streakingof the Bragg spots is also clearly visible, suggestingthe possibility of a small amount of disorientationof the crystal planes. The diffraction patternfurther suggests that the porous Si is in the cubicphase. It can be seen from Fig. 1(b) that no broadand diffuse rings are present in the diffractionpattern indicating the absence of an amorphousphase. This has been further confirmed by Ramanstudies to be discussed in a later sub-section.

3.2. Analysis of current ðIÞ–ðV Þ voltage

characteristics

The I–V characteristic of the p-Si/electrolyteinterface under bias is shown in Fig. 2. Differentregions in the figure are labelled as R1, R2, R3 andR4. R1 denotes the reverse bias regime where verylow current flows through the samples as shownand no silicon dissolution occurs in this region. Incontrast, R2, R3, R4 are the forward bias regimes

Fig. 1. TEM micrograph (Bright Field (BF) image) of PS prepared at 50 mA cm�2 and (b) Selected Area Diffraction (SAD) pattern

of PS.

Fig. 2. I–V characteristic of p-Si/HF acid electrolyte junction.

Porous silicon is prepared only in the shaded transition region.

B.K. Patel et al. / Physica B 322 (2002) 146–153148

where silicon dissolution takes place. Note the twomaxima in the current designated as J1 and J2 inthe forward bias I–V characteristic which areclearly seen. The exponentially increasing currentdensity 0oJoJ1 in regime R2, starting from0:0 V (SCE) corresponds to the pore formation[15,16] region. The second exponential increase ofcurrent density between J1oJoJ2 corresponds tothe transition regime R3, where pore/oxide for-mation and electropolishing takes place. Theresulting structure in this later region is generallyporous in nature and pore diameters are compara-tively larger as shown in the idealized model(Fig. 3b), than the pore diameter in the Regime R2(Fig. 3a). On further increasing the forwardpotential, the porous structure vanishes with theSi dissolution from all over the surface. This is theelectropolishing regime R4, for which one would

observe a planar Si surface as shown in Fig. 3c.From the above arguments, it is clear that in orderto achieve large pore diameter vis-"a-vis small Sicrystallite one should make a judicious choice ofthe applied potential to control the current flow.On synthesizing PS at low current/potential(Region R3), one would expect the pore diametersto be large and the Si crystallites to be small.Besides, even at lower current density voids aregenerated in the PS which would penetrate deepinto bulk Si. Hence, large diameter pores and smallSi crystallites can be achieved with the choice ofthe anodic etching current density in the rangebetween 20 and 50 mA cm�2 as shown in Fig. 2.Considering the pore tip to be hemispherical witha radius of curvature r; the electric field experi-enced at the interface (PS/electrolyte) is givenby [9]

E ¼fs

r; ð1Þ

where fs is the barrier height of the PS/HF-electrolyte junction. Thus, we can see that there isa considerable field enhancement at regions ofhigh curvature i.e, r: From Eq. (1) it is evident thatthe electric field at the pore tips can be sufficientlyhigh, to result in a high current flow which wouldlead to elongation as well as widening of the poresas seen in Fig. 3(b). As a consequence of such anexcess current flow there may be a large chargetransfer at the interface which is facilitated by thelocalized trap states present within the PS bandgap. The other more likely possibility may be thetunneling of carrier across the small depletionwidth in the heavily doped ð1019 cm�3Þ p-typesilicon and the HF-electrolyte system. The sche-matic pictures of pore propagation for differentcurrent densities are shown in Figs. 3(a)–(c). Thesepictures in Figs. 3(a)–(c) are the depictions ofidealized models whereas in actual practice it isexpected that there can be pore branching andvariation in pore diameter as well as fractal type ofpore growth.

3.3. RBS analysis

The small thickness of the PS layer on the Sisubstrate may interfere with the bulk optical

Fig. 3. Idealized cross-sectional morphology induced by differ-

ent anodization current densities: (a) low anodization current

density; (b) relatively higher anodization current density; and

(c) very high anodization current density. The arrow mark

shown at the pore tip denotes the local electric field. The width

of the arrow denotes the strength of the electric field.

B.K. Patel et al. / Physica B 322 (2002) 146–153 149

properties of the PS sample. It is thereforeessential to study the thickness of PS layer whichis determined using the RBS technique. These RBSmeasurements were carried out using a Heþ ionbeam of Energy 3 MeV from the pelletron accel-erator at the Institute of Physics in Bhubaneswar.The channel width (DE) of the RBS spectrum isdirectly related to the thickness of the porous Silayer (Dt) and the energy loss factor [S] by

Dt ¼DE

½S�and ½S� ¼ K

dE

dtþ

1

cos ydE

dt; ð2Þ

where K is the kinematic factor:

K ¼ðm2

2 � m21 sin

2 yÞ12 þ m1 cos y

m1 þ m2

24

35

2

; ð3Þ

m1 and m2 are atomic masses of the projectile (He)and target (Si), respectively, and y is the scatteringangle. Relations (2) and (3) are taken into accountfor the thickness measurements within the simula-tion program itself. Note that in Fig. 4 the RBSyield of PS is much lower than that of bulk Si. Thelowering of RBS yield in PS implies Si depletionfrom bulk Si under anodic etching. Further, theexperimental RBS spectrum of both PS and bulkSi agree very well with the simulated spectra withthe Si edge remaining the same for both. From thesimulated RBS spectrum of PS, the thickness ofthe PS layer was estimated to be B3:2 mm; whichis fairly large. For such a thick PS layer, the opticalproperties are not expected to interfere with bulkSi as will be discussed in the next sub-section.Fig. 4 also suggest the absence of any other

impurity excepting O2 in PS. Unfortunately, RBSwill not be able to detect low Z (atomic number)elements such as Hydrogen which is supposed tobe present in PS.

3.4. Raman scattering analysis

The TEM study of the microstructure of porousSi reveals a wire-or dot-like morphology with sizesof these wires (dots) being in the range of a fewnanometers and isolated from each other. Becauseof their low-dimensional character, the motion ofthe carriers is confined in either one dimension(1D) or zero dimension (0D). Thus, the transla-tional degree of freedom of the quasi-particles andcollective excitations of the system is partially lostdue to localization. Hence, the q ¼ 0 selection rulefor the phonon-Raman scattering correspondingto the phonons in the center of the Brillouin zonefor the bulk Si is relaxed, leading to a downwardfrequency shift, broadening and asymmetry in thefirst-order Raman peak. It is expected that all thephonons with wave vectors in the range 0pqo_=l

will participate in Raman scattering of PS where l

is the average diameter of spherical Si crystallitesin 0D or the diameter of the cylindrical wire in 1Dcase. A quantitative model [18,19] describing thefirst-order Raman intensity to estimate the averagesize (or correlation length) l of the spherical orcylindrical Si nanocrystals is given by

IcðoÞ ¼ cons:

Z q

0

jcð0; qÞj2 d3q

½o� oðqÞ�2 þ ðGcÞ2; ð4Þ

where oðqÞ is the LO phonon dispersion curve forbulk silicon. The size-dependent parameter cð0; qÞbeing

jcð0; qÞj2 ¼ exp �q2l2

4a2

� �; ð5Þ

where a is the lattice spacing for Si. From Eq. (5)one can see that for qa0; as l-N; cð0; qÞ-0;indicating that for bulk samples only the q ¼ 0 LOphonon will contribute to the Raman intensity. Inthe case of a nanomaterial, the contribution of theqa0 phonons to the Raman intensity decreasesexponentially with increasing size.

The Raman data of the p-Si samples wereanalysed by fitting the intensity of the peak using

Fig. 4. Rutherford back scattering spectroscopy study of c-Si

and Porous Si.

B.K. Patel et al. / Physica B 322 (2002) 146–153150

the spatial correlation function, of Eq. (4). Thefrequency-dependent background is fitted to apolynomial and subtracted from the raw data. Atypical example of the background subtractionalong with Raman peak for the Si sampleanodically etched at 20 mA cm�2 current flowingfor 30 min duration is shown in the inset ofFig. 5(c). In order to numerically simulate the first-order Raman intensity of differently sized nano-crystallites of Si, using the model given by Eqs. (4)and (5), the phonon dispersion around q ¼ 0 isassumed to be oðqÞ ¼ A � Bq2 where A is the near-zone-center LO-mode frequency ðA ¼ 520:5 cm�1Þand the value of the fitting parameter B ¼ 120 cmfor Si. This approximation holds near the zone-centre so long as lba: For the zone-centre ðq ¼ 0ÞLO phonon of bulk Si we choose o0 and Gc to be521.9 and 4:7 cm�1; respectively. The wave vector

q is expressed in units of 2p=a where a ð5:4 (AÞ isthe Si-lattice spacing. The observed Raman shiftfor c-Si is 520 cm�1 and it has a symmetric lineshape with a width of 7–9 cm�1 (FWHM) [21].

The fitted Raman spectra of PS samplessynthesized by impressing anodic currents of 20,30 and 40 mA cm�2 are shown in Fig. 5(a)–(c).Using Eqs. (4) and (5) and the spherical phonondispersion curve, the crystallite sizes of thedifferent PS samples have been estimated andfound to be 4.7, 4.1 and 3:9 nm; corresponding tothe current densities 20, 30 and 40 mA cm�2;respectively. Along with the observed red shift ofthe Raman frequencies with increasing current/decreasing crystallite size, the phonon line widthchanges from 7.0 (bulk) to 14, 18.6 and 20 cm�1

for samples prepared with current densities of 20,30 and 40 mA cm�2; respectively. Note that a cleardemonstration of Raman shift towards lowfrequency has been consistently observed forincreasing anodic etching current densities. Thisalso suggests that the Raman contribution isexclusively from the thick PS layer and not frombulk Si. A shift of 514 cm�1 compared to the bulk-phonon frequency of 520 cm�1 has been observedin case of the 40 mA cm�2 anodically etched PSsample whereas the observed shift is 516 cm�1 for20 mA cm�2 anodically etched sample. Thisamounts to a maximum change in frequency of6 cm�1 from the bulk-phonon mode observed inthe present experiment. One can raise the questionwhether the phonon mode observed at 514 cm�1 isnot a consequence of the other contaminants in PSsuch as the presence of siloxene. If it were so, onewould expect the same 514 cm�1 mode to show upfor all the anodically etched PS samples. On thecontrary, the observation of a consistent red shiftof the phonon mode in samples prepared withincreasing anodic current suggests that the514 cm�1 peak is not due to the presence ofsiloxene or other contaminants. Furthermore, itshould be noted that a broad and weak phononmode at 480 cm�1 supposed to arise due to thepresence of an amorphous component in thestructure is not observed in our Raman spectraimplying thereby that the PS still retains itscrystallinity as confirmed from the earlier TEMresults [2,24].

Fig. 5. Raman spectra of PS prepared with different anodic

etching current ðJÞ: The peak at 519 cm�1 is the plasma line

from the Arþ ion laser. Note that the Raman spectra overlap

with the plasma line. The dot-dashed lines are for the fitted

plasma line whereas the dotted curves are fitted Raman lines.

The solid line is the total fit to the experimental Raman

spectrum. Inset: Representative case of background substrac-

tion of first-order Raman spectrum of PS prepared with anodic

current of 20 mA cm�2 flowing for 30 min duration.

B.K. Patel et al. / Physica B 322 (2002) 146–153 151

The PS samples exhibit a structure whichpredominantly consists of Si wires as depicted bythe TEM micrograph. The dimension of thesewires decreases with increasing anodic current,leading to a decrease in crystallite size. If the size iscomparable to Bohr’s exciton radius, aB

(aB ¼ 5 nm for bulk Si [1]) one would expect thequantum confinement effect (QCE) which wouldlead to band-gap widening. In the present work,the size of the crystallites in the PS samples linearlydecreases with increasing current density as shownin Fig. 6(a) and falls in the range of 4.7–3:8 nm:Hence, it is expected that these samples shouldshow the QCE. Along with the QCE withdecreasing size there is an increase in the asym-metry in the low-frequency part of the Ramanspectra. As can be further noted from Fig. 5 the

Raman intensity increases with decreasing crystal-lite size and the asymmetry also increases.Furthermore, no splitting in the Raman peakcorresponding to the longitudinal optic (LO) andtransverse optic (TO) modes has been observed inthe present study in contrast to the report by otherworkers [23]. Fig. 6(b) shows the size dependenceof the Raman half-width ðGÞ and shift ðDoÞ withrespect to that of bulk Si. The solid lines are resultsof the calculation of the quantities for sphericalcrystallites using different weight factors such as (i)exponential, (ii) sine and (iii) Gaussian functions.The data points marked s1, s2, s3, s4 and s5correspond to the spherical Si crystallites of sizesof 3.8, 4.1, 4.4, 4.7 and 4:8 nm; respectively. Theexperimental curves for the Raman widths andpeak shifts agree well with the theoretical curveobtained with the Gaussian weight function.

4. Conclusion

The in situ I–V studies identify the pore/oxideformation and electropolishing regimes for the p-Si/HF acid-electrolyte system. TEM studies in-dicate that the porous Si retains its crystallinityand the pores propagate perpendicular to the Sisurface. Raman-scattering studies show a red shiftof the Raman frequencies with decreasing crystal-lite size.

Acknowledgements

Thanks are due to Prof. V.S. Ramamurthy,DST, Delhi, for his constant encouragement. S.Tripathy of IIT Delhi is gratefully acknowledgedfor help in carrying out Raman measurements.The authors acknowledge S.N. Sarangi for his helpwith the preparation of the manuscript.

References

[1] L.T. Canham, Appl. Phys. Lett. 57 (1991) 1046.

[2] A.G. Cullis, L.T. Canham, Nature 353 (1991) 335.

[3] V. Lehman, U. G .osele, Appl. Phys. Lett. 58 (1991)

856.

Fig. 6. (a) Particle size as a function of current density. Sizes

are estimated from experimental Raman data. The solid line is a

polynomial fit with slope �0:04 and intercept 5:44 nm: (b) The

calculated relationship between the Raman width, (G) and peak

shift ðDoÞ with respect to bulk Si, s1 to s5 are the average sizes

of the PS crystallites in the range 3.8–4:8 nms:

B.K. Patel et al. / Physica B 322 (2002) 146–153152

[4] R.T. Collins, P.M. Fauchet, M.A. Tischler, Phys. Today

50 (1997) 31.

[5] K.D. Hirschman, L. Tsybeskov, S.P. Duttagupta, P.M.

Fauchet, Nature 384 (1996) 338.

[6] Y. Kanemitsu, Phys. Rep. 263 (1995) 1 and references

therein.

[7] Porous Si and its application in Optoelectronics and

Biosensing, http://www.spie.org/web/meetings/programs/

pw01/courses/SC346.html.

[8] Nano-porous Si for Sun Sensors and Solar cells, http://

esapub.esrin.esa.it/pff/pffv8n1/martv8n1.html.

[9] G.C. John, V.A. Singh, Phys. Rep. 263 (1995) 93 and

references therein.

[10] C. da Fonseca, F. Ozanam, J.-N. Chazalviel, Surf. Sci. 365

(1996) 1.

[11] A.I. Belogorokhov, L.I. Belogorokhova, A. Perez-Rodri-

guez, J.R. Morante, S. Gavrilov, Appl. Phys. Lett. 78

(1998) 2766.

[12] C. Levy-Clement, A. Lagoubi, R. Tenneand, M. Neu-

mann-Spallart, Electrochim. Acta 37 (1992) 877.

[13] W. Thib; Surf. Sci. Rep. 29 (1997) 91 and references

therein.

[14] N. Koyama, N. Koshida, J. Electrochem. Soc. 138 (1991)

254.

[15] H. F .oll, Appl. Phys. A 53 (1991) 8.

[16] R.L. Smith, S.D. Collins, Appl. Phys. Lett. 71 (1992)

R1.

[17] M.S. Brandt, H.D. Fuchs, M. Stutzmann, J. Weber, M.

Cardona, Solid State Commun. 81 (1992) 307.

[18] I.H. Campbell, P.M. Fauchet, Solid State Commun. 58

(1986) 739.

[19] H. Richter, Z.P. Wang, L. Ley, Solid State Commun. 39

(1981) 625.

[20] G. Bomchil, A. Halimaoui, R. Herino, Appl. Surf. Sci. 41/

42 (1989) 604.

[21] S. Guha, P. Steiner, W. Lang, J. Appl. Phys. 79 (1996)

8664.

[22] Z. Sui, P.P. Leong, I.P. Herman, G.S. Higashi, H. Temkin,

Appl. Phys. Lett. 60 (1992) 2086.

[23] R. Tsu, H. Shen, M. Dutta, Appl. Phys. Lett. 60 (1992)

112.

[24] I. Berbezier, Halimaoui, J. Appl. Phys. 74 (1993) 5421.

[25] B.K. Patel, S.N. Sahu, Appl. Phys. A 71 (2000) 1.

[26] B.K. Patel, S.N. Sahu, Mater. Sci. Eng. A 304–306 (2001)

914.

[27] B.K. Patel, S.N. Sahu, Appl. Phys. A 71 (2000) 695.

[28] B.K. Patel, S.N. Sahu, in preparation.

B.K. Patel et al. / Physica B 322 (2002) 146–153 153