Porous silicon: material properties, visible photo- and electroluminescence

,I

:

Applied Surface North-Holland

Science 65/66

(1993) 394-407

Porous silicon: material and electroluminescence G. Bomchil, G. Vincent

A. Halimaoui,

properties,

and

29 June

1992; accepted

.. ..

:.

,:

” ;.. ,.

,.

P.A.

visible photo-

Badoz,

I. Berbezier

‘, P. Perret,

B. Lambert

‘,

J.L. Regolini

Cmtre National d’Etudes des T&communications, Received

. : ..”

applied surface science

I. Sagnes,

s, L. Garchery

,.y

France T&kwn, BP 98, 38243 Meylan, France

for publication

2X September

1992

Following the recent discovery of visible photo and electroluminescence of high-porosity porous silicon layers this paper presents a review of the most relevant results and models proposed to explain the phenomena. Porous silicon fabrication techniques are presented including some recommendations to allow a meaningful comparison of results obtained hy different laboratories. Recent results of pore size. surface area measurements. crystallographic structure determination and microstructure observations are discussed. Detailed studies of optical absorption coefficients of porous layers of different porosities are pt-esentetf. A clear upshift toward the visible range is explained by a quantum confinement model. Intense visible photoluminesccnce ol porous silicon layers is discussed on the ground of both quantum confinement and surface-controlled phenomena. The caaential role played by surface passivation, for efficient luminescence. is analysed. Reported results of visible electroluminescence during anodic oxidation of porous silicon layers and visible light emission from solid-state porous silicon dcviccs are reviewed.

1. Introduction

encc dedicated to the subject of light emission from silicon provide the first collection of papers

Because of its relatively small (1 .l eV) and indirect band-gap, the band-to-band luminescence of bulk-crystalline Si occurs in the near-infrared. The recent discovery [l] that porous silicon (PS) layers emit visible light at room temperature has generated much interest due to its potential applications for fabrication of light emitting silicon devices integrated within the core of the silicon technology. An impressive number of groups at different laboratories all over the world started to work on the subject and more than 200 papers have been published within two years of the first report. Communications presented at a recent confer-

[21.

’ Present address: CRMCZ-CNRS. 901, 13288 Marseille, France. ’ CNET, France Telecom, Route nion, France. ’ Permanent affiliation: Universitt 38041 Grenoble, France. 0169-4332/93/$06.00

0 1993

Campus

Luminy,

de Tregastel, Joseph

Elsevier

22301

Fourier,

Science

Case Lan-

BP 53X.

Publishers

PS actually is not a new material. A few groups studied in the past several of its properties mainly for applications in silicon-on-insulator technologies and comprehensive reviews of this earlier work are available [3,4]. Since then, most of the work concentrated on high-porosity layers which display the luminescence properties. Howcvcr. under the generic name of PS there arc very different and complex materials whose propertics _ structural, optical, electrical, surface, and mcchanical - critically depend on many different parameters. Consequently, there are various different interpretations and an intense debate about the mechanisms responsible for the luminescence. It appears on the one hand, useful to propose some simple recommendations for matcrial fabrication, in order to allow a meaningful comparison of results from different laboratories. and on the other hand, it appears that there arc already enough reliable results to propose a criti-

B.V. All rights reserved

395

G. Bomchil et al. / Porous silicon: material properties Gsible photo- and electroluminescence

cal analysis, which is the objective of this paper, of the most relevant features about the material and its luminescence properties. (ii) 2. Fabrication

of PS layers (PSL)

PS is obtained by anodic dissolution of silicon in hydrofluoric (HF) acid solutions in an electrochemical set-up where the silicon substrate constitutes the anode of the cell. Anodic processes can be well controlled working either at constant current or potential. Galvanostatic conditions (constant current density) are generally preferred because they allow an easy control of the layer formation. It is also known that PSL can be obtained in solutions containing an oxidant agent like the well known NO,H/HF mixtures. However, in this case the process control is more difficult to achieve and the morphology of the layers is much different. During PS formation, there is anodic hydrogen evolution. Hydrogen bubbles must be readily eliminated from the surface, otherwise samples are not homogeneous either laterally or in depth. One of the most appropriate means to overcome this problem is to use HF(49% in H,O)-ethanol solutions where ethanol is believed to facilitate hydrogen release by increasing wetting of the surface. In order to fulfil this role ethanol concentrations should not be less than 15%. Mechanical means like ultrasonic stirring or forced solution circulation along the surface have also been used but results are not so good. PS can be formed on lightly (p-1 and highly (p’) doped p-type silicon and on highly (nf) doped n-type silicon. In the case of lightly (n-1 doped n-type silicon, illumination (e.g. visible light with a 500 W tungsten lamp is required). Layers formed on n- substrates in the dark are also porous, but of very different morphology. Although very simple in principle, there are some elementary precautions that must be taken when preparing PS samples, otherwise a discussion of the results could be meaningless: (i) Resistivity and type of the starting silicon substrate must be determined precisely as properties of porous samples prepared on

(iii)

(iv)

(v)

substrates of different resistivities might be very different. Careful measurements should be done when using highly doped samples in the range > 3 x 10” cm-3. Porosity (fraction that is void) of the prepared layers must always be measured. This can be easily done by a double weighing technique with the actual samples or a triple weighing in separate wafers prepared under the same conditions [.5]. For small porous layer thicknesses (< 10 Frn) relative errors might be important. Anodization times should be as short as possible, no more than a few minutes, and samples must be anodized in dark conditions to minimize silicon dissolution by chemical attack of the already formed porous layer. In thick samples this could provoke large indepth porosity gradients. These effects are much more important in layers formed from lightly doped substrates. As a rule of thumb: for a given resistivity the same porosity can be obtained with different values of the couple current density/HF concentration. When the current density (as calculated from the geometrical surface of the exposed silicon) increases, porosity increases. When the HF concentration decreases, porosity increases. Table 1 gives some examples to obtain layers of different porosities. Backside ohmic contact of lightly doped wafers requires previous metallization and/ or high-dose ion implantation. For large diameter wafers, double tank cells with back-

Table 1 Experimental values of different parameters of porous silicon layers formed on lightly doped pm type, 1 bI.crn resistivity silicon substrate (HF-ethanol solutions) HF concentration (vol%)

Current density

(mA/cm* )

15 20 25 25 35

10 20 25 20 20

Porosity (%o)

Coulombic charge (C/cm*)

Layer thickness

(km)

85 76 70 65 58

1.55 4.15 3.84 3.56 3.18

1 3 3 3 3

side electrolyte contact should be implemented. change when exposed to (vi> PS characteristics air, in particular under daylight illumination. Storage conditions of samples before characterization are therefore critical. An important study on this subject has been published [6]. The effect of room-temperature oxidation after several hours of ambient air exposure might alter the PS properties (electrical, optical,. . .I.

3. Formation

mechanisms

Models describing the formation mechanism and microstructure of PSL should account for the different morphologies of PSL formed on silicon substrates of different type and dopant concentration. PSL formed in p- silicon substrates consist of a network of very small pores and silicon crystallites with typical dimensions of the order of 2-4 nm. PS layers formed in p+ silicon substrates (10” cm-j) or n+ silicon substrates consists of a branched network of rather long pores and crystallites with an anisotropic orientation in the direction of the current flow and typical dimensions of the order of 10 nm. In the case of n substrates, PSL formed in dark conditions are completely different to the above cases. They consist of a completely anisotropic network of very long pores with few branches and thick silicon walls in between with typical dimensions of the order of 1 pm. If formed under illumination, the morphology of PS in no type substrates is similar to the p- case. At present there are no quantitative theoretical models to describe these different forms of PSL and the dependence of porosity and microstructure on the parameters of the electrochemical process: substrate type and resistivity, HF acid concentration and current density. There are however, several qualitative approaches to explain, at least partially, some of these properties. The basic formation mechanism, the localized formation of pores and the most relevant features

of the porous layer microstructure, can be qualitatively explained in terms of the exchanged charge at the semiconductor/electrolyte intcrface [7]. It has been well established that silicon oxidation and dissolution in HF acid solutions under anodic potentials requires either injection of holes from the bulk silicon valence band to the surface or injection of electrons from the surface into the bulk silicon conduction band. Interfacial impedance measurements demonstrate [8] that the surface of the clectrodc in contact with HF acid solution is depleted of majority carriers. In the case of p-type silicon over a large doping range (10” up to IO”’ cm ‘1 the anodic current responsible for silicon dissolution is determined by hole supply over the potential barrier following a Schottky diode barrier type dependence whcrc the electrolyte plays the role of the metal. The only difference is that in this case there exists at the electrolyte side 01 the interface (the so-called Helmholtz layer) a potcntial drop V,,. For low doping levels the potential drop and its variations, within a large range ot current densities, arc negligible compared to the semiconductor surface potential. For higher doping levels this is no longer valid and the applied potential is shared between the scmiconductot and the Helmholtz layer. A common feature to all models is that the initiation of the porous structure occurs at localized regions of the silicon surface where random inhomogeneities creates tiny concave depressions. The distribution of the potential diffcrencc bctwcen the semiconductor and the electrolyte and thus the height of the Schottky barrier dcpcnds on the geometry of the interface in the sense that band bending is smaller on the concave intcrf’acc Consequently, the current density is regions. higher in these concave regions, resulting in a further increase of the concavity. Concerning the mechanism of port propagation, it is often assumed in the case of pi suhstrates that due to charge depiction, once the thickness of the crystallites between pores cxcecds 2w, where NJ is the charge depiction width, the crystallite is electrically isolated and reaction can only occur at the pore tips. This assumption is not correct because the measured thickness ot

G. Bomchil et al. / Porous silicon: material properties 1Cible photo- and electroluminescence

crystallites and the pore diameters are much smaller than the depletion width. There are however three models that address this issue correctly. The first one assumes [7l that the current flow is determined primarily by the height of the barrier which is reduced at the pore tips. Small tip radii of the order of 2 nm can produce substantial field intensification leading to barrier lowering and hence increase in current density at that point. As the pore grows and its radius is increased, the growth rate is reduced and enhanced dissolution occurs at locations where inhomogeneities will create new pore nucleation sites. The overall result of this mechanism is the formation of a random network of pores of very small dimensions. The second one [93 is based on the well known model of Witten and Sanders where formation of PS is supposed to be controlled by the diffusionlimited reactant (holes in this case> supply from the bulk to the surface. As a particle randomly walks towards the growing pores it is more likely to find a reaction site at the tip of the pores. Computer simulations of this process results in porous structure similar to what is observed in TEM cross section microphotographs. The third one [lo] proposes that hole depletion results from quantum confinement in the very small silicon crystallites. If the crystallite dimensions are of the order of l-3 nm there is an increase in the band-gap energy. If a hole from bulk silicon approaches the substrate/PS interface it will need an additional energy (due to confinement) to penetrate into a wall between two pores, whereas no additional energy would be required to reach a pore tip at the pore/bulk silicon interface. This is a self-adjusting process. Thinning of thick crystallites between pores continues until the energy is such that holes have a higher probability to move to the pore tip than to penetrate the wall. In the case of heavily doped silicon, as explained above, the potential distribution at the silicon/electrolyte interface is such that the potential drop on the electrolyte side can no longer be neglected. A substantial part of the potential drop occurs on the electrolyte side of the interface. In this case current flow is dominated by the

391

width of the depletion layer (and not by its height). It is suggested that hole or electron exchange, occurs by tunneling through the barrier 171. The tunneling current increases as the depletion layer width is reduced and therefore is concentrated at the pore tips where the depletion width is minimum. Typical dimensions of the order of 8 nm are predicted. The columnar anisotropic structure observed is also explained from geometrical effects, because in regions between pores due to the overlap of depletion layers the current density is greatly reduced.

4. Surface area and pore-size

distribution

Surface areas of PS layers were determined by the gas adsorption BET technique [5]. For the p+ samples surface areas are of the order of 200 m2/cm” for porosities up to 70%, while in the surface areas are of the case of p- substrates order of 650 m2/cm”. The surface area of the pf samples remains nearly the same for porosity values increasing from 36% to 70%, while the mean pore radii increase from 3.1 to 6.9 nm. This is valid for different couples HF concentration/ current density giving the same porosity. In the case of p - substrates the mean pore radius is 2.6 nm for samples of 60% porosity. In a recent study [ll] by X-ray small-angle scattering of PS samples with different porosities it was found that the surface area of p- samples decreases with increasing porosity from 50% to 85%. Similar results were reported by BET measurements [12]. It is often assumed, incorrectly, that increasing porosities are associated with increasing surface areas. If greater porosities occur by thinning of the silicon particles, thus by enlarging the pore size, the surface area should decrease. In the case of p- samples the number of pores is greater and the dimensions smaller. Increasing porosities up to 85% would lead to complete dissolution and disappearance of the smallest particles, large pore sizes by pore coalescence, and an important measurable decrease in surface area. A simple model can be used to calculate the mean value of the remaining silicon crystallites. If

398

G. Bomchil

el al. / Porous silicon:

material

a cylindrical geometry for voids is assumed (however, quantitative values of surface/ volume ratio can be model-dependent) the porosity is equal to rrd2/4a2 where d is the pore diameter and a the distance between pores. The particle size between pores is equal to L = 1.41a - d. Using 2.6 nm as the mean pore radius for a p- sample of 60% porosity, a becomes equal to 5.9 nm and L equal to 3.2 nm. When the porosity is increased to 85% there is pore coalescence (critical porosity for coalescence is 78.5%) and particle size can be obtained from the simple expression L = a(1 I’)‘/*. If the distance between pores is assumed nearly constant the mean particle size L is reduced to 2.3 nm. Assuming a distribution of silicon crystallite sizes around this value, the smallest crystallites will completely dissolve provoking pore coalescence and reducing in consequence the surface area.

5. Structural

properties

PS layers with porosities as high as 72% for p+ samples and 56% for psamples show in

Fig. I. High-resolution

transmission

electron

properties

~Chle photo-

and ele~trolumines~erl~~

double-crystal X-ray diffraction experiments a narrow Bragg peak well separated from the silicon substrate peak. Measured variations of the lattice parameter relative to the silicon substrate increase with porosity resulting in maximum values of Au/a = 10m3 and 4 X 10-j for the p+ and p- samples, respectively, and the above-indicated porosities [ 13,141. Thus, even for these porosities, PS keeps the monocrystalline nature of the silicon substrate and a quite similar lattice paramcter. There have not yet been reported values of Au/a using high-resolution techniques for samples of higher porosities (e.g. 85%). The origin of lattice expansion is attributed to the silicon-hydrogen bonds at the porous surface [ 151. Morphology and crystalline structure of highporosity (85%) p+ and p- samples have been studied using transmission electron microscopy (TEM) associated with electron diffraction [16181. Both plan-view and cross sections of the porous layers have been analysed. It appears that p’ samples display in cross section an anisotropic structure of a longitudinal-branched network of pores/ silicon crystallites. The high-resolution

microscopy cross section of a 85% porosity doped p+ silicon substrate.

porous

silicon sample

formed

on a highly

399

G. Bomchil et al. / Porous silicon: material properties visible photo- and electroluminescence

TEM (HRTEM) photograph of fig. 1 shows a rather broad distribution of crystallites sizes (3-20 nm). Even for these high porosities the monocrystalline properties are not lost. The case of highporosity p- samples is rather different. Careful mechanical cleavage of porous layers, without ion miling and immediately after formation, allows one to get specimens displaying the structure of the as-formed layers. Cross-section HRTEM photographs shown in fig. 2 indicate that p- samples consists of an array of very small interconnected pores. Fig. 3 shows a plan-view HRTEM. It is observed that the silicon thickness between pores is in the range of 2-4 nm. Electron diffraction reveals, as shown in fig. 4, elongated spots indicating degradation of the crystalline structure. Amorphous material can also be found, in others regions of the specimen, and coexists with crystalline clusters. Only amorphous PSL have been found in other studies [19,20]. It has been proposed that in the limit of 2-3 nm the crystal lattice expansion of very small crystallites leads to an instability of the diamond structure and a

Fig. 2. High-resolution

transmission

electron

spontaneous phase [21].

transformation

to

an

amorphous

6. Optical properties Transmission optical spectra and the associated absorption spectra of PS constitute one of the most powerful tools to study the fundamental optical properties of PS structures. First experimental measurements of optical transmission [lo] on a PS sample of 60% porosity prepared on psubstrate indicate that there is an upshift in the knee of the transmission curve. In order to compare values of the optical transmission of PS samples of different porosities, the relevant parameter is the absorption coefficient which takes into account the total quantity of matter and is deduced from the transmission experiments if sample thickness and porosity are well determined. Measurements have been performed first using only one sample [221 and recently [23] using homogeneous free standing PS films of various,

microscopy cross section of a 85% porosity doped p- silicon substrate.

porous

silicon sample

formed

on a lightly

-_

3.8 nm

;: Fig.

3.

High-resolution

transmission

plan-view of a XSr/; porosity II lightly

doped p

clearly xen.

silicon

electron

microscopy

porous silicon sample formed on substrate.

Pore

coalescence i\

Silicon thickness between pore, 7-4 nm.

uniplc\. .1‘11t.IitJif-ix? I,.~i~~i-ii,ti, buih \tIic~c,tl ot>taitlcd tx c\tr;tpol,ill~rv lo trll~ =- 0 is I.1 cV. f:or p I’S ~mplc\ ilic~~ I\ a shift which incrcascs with porosity (arcrunciIO(I I’\ mcV for 78.5”5# samplc4). In the c’;l’rc t )i I’ samples the plot is not linear and ttlerc~torc ;III energy gap cannot be determined tln~liiit,iglic)li\l~. Nevertheless, thcrc is a clear and important \hilr (between X)0 and 500 meV depending OII po1-0\ity) that has to be explained. The most likely explanation for the incr-c;tsc III the band-gap energy is the existcncc ot ;t quantum confinement effect in the small silicon crystallites that constitute the silicon skeleton. (‘alculated L values using the well known. simple cxpression AE, = h’/4m*L2 where Ir is f’lanck’s constant, M* the effective electron or hole mass. and L the crystallite dimension, are in qualitative agreement with the experimental ones. 1. IS in the range 2-4 nm for the lightly doped samplch and 7-9 nm for the highly doped samples. Mot-c elaborate calculations have been rcportcd [24,X]. However, exact calculations using the effectivemass approximation to obtain the confinement energy arc model-dependent. toI the

1-1

CIICI.~,’

t’cjr

porous

40 I

7. Photoluminescence A strong red-like photoluminescence easily visible to the naked eye even at room temperature was first observed [I] from p PS samples of about 8OS; porosity obtained by electrochemical formation of a rclativc-low-porosity sample followed by chemical dissolution in HF during several hours. Since then different techniques were used to reproduce the effect [2h]. A first approach consists in the USC:of electrolyte conccntrations and current densities leading directly to

(a)

’

Wavelength 1

2

(pm) 0.5

0.7

1

B

x

‘;j .Y E 2 E b

0.8

Fig. 0. Square energy versus

root of the absorption coefficient times photon and p * porous photon energy for silicon, p layers (ref. [23]).

the photoluminescent layers (see fig. 7) that can be further thinned by chemical dissolution, leading to porous layers displaying photoluminescence in the green region of the visible spectra. In the second approach, the photoluminescent layers were obtained by thinning down the silicon crystallites of low-porosity layers through forma-

0.6 0.4 0.2

0 ^

0.5

1.5

1

2

2.5

Energy (eV)

04

Wavelength 1

(Km) 0.7

1 E-1 2 ‘8 .z E % E E-

0.8 0.6 0.4 0.2

Energy (eV) Fig. 5. Transmission coefficient versus photon energy for 40 pm thick free standing porous silicon films of different porosities and substrate doping: (a) p- samples, (b) pt samples. Measured transmission of free standing silicon films of ‘the same thickness is shown for comparison (ref. [23]).

Fig. 7. Photoluminescence of 85% porosity porous silicon layers under ultraviolet light illumination. Porous layer thickness 2 pm.

G. Bomchil et al. / Porous silicon: material properties risible photo- and electroluminescence

402

tion of a silicon dioxide layer on the crystallite walls by anodic oxidation at room temperature. This is a very convenient technique to control with good precision (by the coulombic charge) the thinning process and the thickness of the oxide layer; in addition, it improves the ageing stability of the porous layers. Thermal oxidation was also used 127,281 to achieve equivalent results. However, during thermal treatments even for temperatures as low as 350°C [29] the PS structure starts to suffer important restructuring. Finally, photoluminescent layers were also obtained by formation of PS films using silicon chemical stain etches

1301.

4

Pi

60

65

70

Fig. 9. Wavelength

at maximum

peak versus porosity for p

The most striking feature of the luminescence phenomena is the blue shift towards increasing energies at maximum intensity, when the porosity is increased. Photoluminescent porous layers are also obtained from n- (when PS is formed under illumination), n+ and p+ silicon substrates. Results of a detailed comparative study between photoluminescence of p- and p+ porous silicon are shown in fig. 8. It appears that although the structure is very different, PL spectra of both samples present a maximum at about the same wavelength. Photoluminescence of p+ samples is however much less intense. When the porosity of the layers is increased, there is a blue shift of the PL peak for both p- and pf samples as shown in fig. 9.

1

r

0 12

16

14

I+lerg> Fig. 8. Photoluminescence and p’

18 le\‘l

2

intensity versus wavelength

porous silicon samples.

: for p

75

80

I'ww~il\

C 'i

85

90

95

1

of the photoluminescence

and p+ porous silicon samples.

7.1. Quantum confinement The luminescence has been interpreted as resulting from quantum confinement in small-size crystallites. Very small crystallites are indeed present in PS layers and the variations of crystallite size with porosity are consistent with the observed energy blue shift of the emission but do not prove the existence of quantum confinement Optical absorption properties presented previously in this paper on the contrary support the hypothesis that a fundamental change in the energy gap of bulk silicon occurs. Theoretical calculations of the carrier confincment energy (which is the sum of the electron and hole confinement energy) and the optical gap predicts variations as a function of the crystal size in good agreement with the experimental results [24,25,31]. A recent detailed study [32] compares experimental data of photoluminescence from p layers of various well know porosities from 60% to 85% with a theoretical model which assumes that the observed luminescence is due to a quantum confinement effect. Fig. 10 shows a good agreement between the calculated confinement energy and the wavelength at maximum intensity of the emitted light. Support for a quantum confinement model is provided by visible photoluminescence obtained from: silicon nanoparticles prepared by techniques of gas-phase synthesis [33,341 and germanium nanostructures fabricated by plasma-as-

G. Bomchil et al. / Porous silicon: material properties uisible photo- and electroluminescence

G

O0 .6

0.7 I’orosit).

0.X

0.9

p (%)

Fig. 10. Energy of confinement versus porosity. Experimental confinement energies obtained at 2 K (squares). Experimental confinement energies at 300 K (circles). Full line: theoretical confinement energy (ref. [32]).

sisted etching [35]. A recent paper [36] demonstrates that cubic silicon clusters consisting of eight silicon atoms each terminated by a ter-butyl substituent has an absorption edge 0.8 eV larger than the bulk silicon, and show visible photoluminescence. Also recent studies [37,38] show that PSL prepared by anodization of boron-doped amorphous hydrogenated silicon films and also silicon-based sub-oxides yield visible luminescence indicating that light emission is not necessarily associated with a crystalline substrate as base for PS formation. Also, although luminescent PS structures formed from single-crystal silicon substrates do contain crystalline material, it has not been proved that amorphous material also present (in particular in p-porous layers) does not contribute to the luminescence process. The difference in PL intensity of p- and pf porous layers can be explained on the basis of differences in the density of silicon crystallites of the appropriate size. In fact, p+ layers display at least two different distributions of crystallite sizes. The one of the order of 3 nm could be responsible for the observed PL. In the case of p- samples the crystallite sizes are mainly in the 3 nm range and thus the density of “light emitting” crystallites is much larger. 7.2. Surface properties Many optical properties of luminescent PSL cannot be easily explained on the basis of quan-

403

turn confinement: (i) Excitation spectra show that the excitation edge is emission-wavelength-dependent revealing an inhomogeneous nature of the process and suggesting that the absorptive states are not the states through which the electron-hole recombination occurs [39,40]. (ii) Temperature dependence of the luminescence peak has a very complex behaviour [41-431. (iii) Photoluminescence decay times are non-exponential and decrease with emission energy. Typical radiative times are very long, of the order of 1 ms, far from those expected for a direct transition in the range of nanoseconds to microseconds indicating that the assumption “k” translational invariance breakdown is not correct [39,42]. However, the dependence of the non-radiative decay rates on confinement energy has been recently explained by a model of escape of carriers from the confinement zone. This model accounts well for experimental results obtained with anodically oxidized porous layers [46]. Some of the above complex properties have already been observed in a well established literature on the luminescence properties from microcrystallites, polysilane, hydrogenated amorphous silicon and siloxene. In this last case, similarities with PS luminescence are striking [43-4.51. At present, it is difficult to identify the formation mechanism of such compounds in the HF solutions. The state of PS surface plays an important role in the efficiency of the photoluminescence. Quenching of the photoluminescence for annealing temperatures below 400°C has been attributed by some authors to the desorption of H, from SiH, species [47-491. Others authors [50] found that quenching of the PL significantly precedes hydrogen desorption from the SiH, surface species, and propose that surface traps provide a non-radiative path recombination. Studies of PL degradation 151,521 show that degradation is caused primarily by oxidation of the material (strongly enhanced under illumination). All these experiments show that hydrogen at

the

surfacc

cence directly

but

is do

csxntial not

involved

to

prove

obtain

that

in the intrinsic

the

Si-H

lumincs-

bonds

optical

arc

prowa\;

its role

could be more related to surface passi\k tion and decrease of non-radiative rccombinatic~ri paths.

The

origin

of the efficient

taincd by anodic oxidation degradation

produced

pasivation

[46] (compared

by O? oxidation)

oh-

tc) the of the

porous layers remains unclear.

Fig. 11. In situ electroluminescence

of porous silicon layers during anodic oxidation. layers, 2 grn thick.

Starting material

XO’%~porosity pc,rou\ silicon

l-Ic~troluniinc~c~Iicc

of

I’S

was

first

ohwrvcd

during the anodic oxidation 01‘ lightlq doped p I’S I;I~cI-s 1531. Once the PS is formed, the HF acid ~lectrolytc is immcdiatcly replaced in the same clcctrochcmical cell by an electrolyte without HI; (HU or NO,K aqueous solutions have been used). Anodic oxidation is pcrformcd thereafter, gencrally at constant current density (I- 10 mA/cm’). During anodic oxidation an intense red-orange light emission is visible over the entire surface of the silicon wafer (see fig. 11). During this process t hc clcctrode potential remains nearly constant before an abrupt increase. Emission starts shortly after the onset of the anodic process and lasts until the abrupt increase in the anodic potential. The electroluminescence appears at t = 10 for high-porosity samples - 85%’ - that are already photoluminescent under illumination, but is dclayed by several hundreds of seconds for samples of lower porosities. For these samples the appearance of electroluminescence is nearly coincident with the appearance of photoluminescence. It has been proposed that the appearance of the luminescence requires crystallite sizes within the quantum size range. They are already present in the 85% porosity sample and result from thinning of the silicon walls during the anodic oxidation for the low-porosity samples. In situ spectral distribution of the emitted light during the anodic process shows a net blue shift of the wavelength at maximum intensity from 830 down to 700 nm. This is in agreement with the assumption that silicon crystallites are thinned down. It has been proposed that holes involved in the radiative recombination are supplied by the p--

silicon s~lh~tratc. Oxidation of the adsorbed hydrc)gcn i. the most probable source of electrons. Solid :;tatc Lisible elcctroluminesccllce was earlier obscrvcd t’rcm SiOl layers containing small \i[icon prcc.ipitalc\. and ;I quantum confinement n~oclcl U’..IS prop044 for the first time to explain the w;l\,~~lcngth shift 15.41. I’lcctroluminescencc in the \.isiblc range from PS layers has also hccn rcportcd f-irk1 results [.i5] were obtain4 from 75 Toni thicl, photc)luminescent porous layers prepared on n suhstrates. Iliodcs uc’rc obtained by thin-film gold deposits. Visible light emission was ohen cd for both IX fi>rward and rcvcrsc pola-it!, ;ind high voltages. Elcctroliiminesccncc w;is not stable and only lasted t’or about one hour. Electroluminescence using p type PS samples .T-7 pm thick were also reported [%)I. After formation of a porous silicon layer. samples were chcmicatly etched in HF under illumination or by immersion in KOH solution for a few seconds followed by deposition of thin gold films or IT0 (indium tin oxide). A stable and uniform visible (orange) light emission was observed through the transparent electrode only under forward polarization. The electroluminescence intensity increases nearly in proportion to the diode current density. Integrated quantum efficiency is tow. estimated to be of the order of lo-‘. Etectrotuminescence was also observed [S7] under forward bias from devices where PS was formed laterally in the p type buried layer of a structure At/n+Si/PS/p+ substrate. Typical turn-on volt-

I

I

200

400

600

WA\‘EI,EN(;TH

x00

IOOO

(nmr

Fig. 12. Spectral distribution from an electroluminescent porous silicon solid-state device taken at 30 V forward voltage. Porous silicon layer 85% porosity, 0.2 wrn thick.

406

G. Bomchil et al. / Porous silicon: material properties cisible photo- and electroluminescence

age of the diodes is 0.7 V. Electroluminescence was measured at 12.5 A/cm2. Recently [58] electroluminescent devices were fabricated using a thin p- 85% porosity porous layer and a rather thick (0.1 pm> metal contact. Stable and reproducible visible electroluminescence (red) was observed with the naked eye in the dark under forward voltage. The spectrum taken at 30 V is shown in fig. 12. The maximum intensity corresponds roughly with the photoluminescent spectra of the same sample. However, the integrated quantum efficiency although slightly larger than previously reported, is still low, of the order of 10d5 to 10e6. Because of their formation mechanisms, PSL are depleted of carriers and behave like an insulator. The very high resistivity of the porous layers confirms this hypothesis. Carriers can be injected in such kind of layers (i) when a bias voltage is applied between the metal contact and the silicon substrate, the potential drop occurring within the porous layer. For high enough forward voltages (substrate positive, metal negative) band bending will allow injection of electrons from the metal and holes from the substrate, (ii) by breakdown which could explain the first reported results of electroluminescence at high reverse voltage in thick layers. In all cases, the recombination of electron-hole pairs (holes supplied by the substrate electrons by the metal counter electrode) in the small crystallites that compose the PS skeleton would produce visible light emission in the same manner as for the photoluminescent phenomena. All these results indicate that solid state PS light emitting devices are feasible.

9. Conclusion Intense visible light emission from PS is now obtained in many different laboratories. Strict experimental conditions for preparation and storage of samples are required in order to compare results from different sources. Luminescence well above the band-gap of silicon is attributed to electron-hole recombination in the very small particles of the silicon skeleton. Optical absorp-

tion coefficient measurements support the model of quantum confinement. Surface passivation is required for efficient luminescence. More studies are however necessary for a better understanding of the passivation and the radiative recombination process. Visible light emission from solidstate devices has also been demonstrated. At present, the efficiency is low.

References [l] L.T. Canham, Appl. Phys. Lett. 57 (lY90) 1046. [2] Light Emission from Silicon, Vol. 256 of the MRS Symp. Proc. Ser. (1991 MRS Fall Meeting, Boston. MA), Eds. S. Subramanian, T.J. Iyer, R.T. Collins and L.T. Canham (Materials Research Society, Pittsburgh, PA. 1992). [3] G. Bomchil, A. Halimaoui and R. Hkrino, Microelectron. Eng. 8 (1988) 293. [4] G. Bomchil, A. Halimaoui and R. HCrino, Appl. Surf. Sci. 41/42 (1989) 604. 151 R. Hirino, G. Bomchil, K. Barla, C. Bertrand and J.I Ginoux, J. Electrochem. Sot. 134 (1987) 1994. [h] L.T. Canham, M.R. Ho&on, W.Y. Leong, C. Pickering and J.M. Keen, J. Appl. Phys. 70 (1991) 422. [7] M.I.J. Beale, J.D. Benjamin, M.J. Uren, N.G. Chew and A.G. Cullis, J. Cryst. Growth 73 (1985) 622. [X] I. Ronga, A. Bsiesy. F. Gaspard. R. Herino, M. Ligeon. F. Muller and A. Halimaoui. J. Electrochem. Sot. 13X (19Yl) 1403. L91 R.L. Smith, S.F. Chuang and S.D. Collins. J. Electron. Mater. 17 (198X) 533. [I()1 V. Lehmann and U. Giisele, Appl. Phys. Lett. 5X (1YYl) X56. of porous silicon hy X-ray Ill1 V. Vezin, Characterization hmall angle scattering, Thesis University of Poitier\. October 1991: V. VeLin, P. Goudeau. A. Naudon. A. Halimaoui and G Bomchil, Appl. Phys. Lett. 60 (IYY?) 2625. [I. Sugiyama and 0. Nittono. In: Proc. Int. LIZI K. Takemoto. Conf. on Solid State Devices and Materials. Yokohama, IYYl, p. 252. G. Bomchil. J.C‘. Pfister and A. [I.?1 K. Barla. R. Htrino, Freund, J. Cryst. Growth 68 (19X4) 727. (141 D. Bellet, G. Dolino, M. Ligeon. P. Blanc and M. Krisch. J. Appl. Phys. 71 (1992) 145. H. Watahe and ,\. [I51 T. Ito, H. Kiyama. T. Yasumatsu, Hiraki, Physica B 170 (1991) 535. [lhl A.G. Cullis and L.T. Canham. Nature 353 (IYYI 1 I. B. Jobst. T. Muschik, A. Kux and V iI71 V. Lehmann, Petrova-Koch. Appl. Phys. Lett., suhmitted. and P. Perret, In: Proc. EU[IN 1. Berhezier. A. Halimaoui REM Y2, Granada, Spain: see also R. Tsu, H. Shen and M. Dutta, Appl. Phy\. Lett. 60 (1992) 112.

G. Bomchil et al. / Porous silicon: material properties visible photo- and electroluminescence (191 T. George, M.S. Anderson, W.T. Pike, T.L. Lin, R.W. Fathauer, K.L. Jung and D.L. Kwong, Appl. Phys. Lett. 60 t 1992) 2359. [20] N. Noguchi, I. Suemune, M. Yamanishi, G.C. Hua and N. Otsuka, Jpn. J. Appl. Phys. 31 (1992) 229. [21] S. Veprek, Z. Igbal and F.A. Samot, Phil. Mag. B 45 (19821 137. [22] H. Koyama, M. Araki, Y. Yamamoto and N. Koshida, Jpn. J. Appl. Phys. 30 (1991) 3606. [23] I. Sagne, A. Halimaoui, G. Vincent and P.A. Badoz, Appl. Phys. Lett., submitted. [24] G.D. Sanders and Y.C. Chang, Phys. Rev. B 45 (1921 9202. [25] J.P. Proot. C. Delerue and G. Allan, Appl. Phys. Lett. 61 (1992) 1948. [26] A. Bsiesy, J.C. Vial, F. Gaspard, R. Herino, M. Ligeon, F. Muller, R. Romestain, A. Wasiela, A. Halimaoui and G. Bomchil, Surf. Sci. 254 (1991) 195. [27] T. Ito, T. Ohta and A. Hiraki, Jpn. J. Appl. Phys. 31 (1992) Ll. [28] S. Shih, C. Tsai, K.H. Jung, J.C. Campbell and D.L. Kwong, Appl. Phys. Lett. 60 (1992) 633. [29] R. HCrino, A. Perio, K. Barla and G. Bomchil, Mater. Lett. 2 (1984) 519. [30] R.W. Fathauer, T. George, A. Ksendzov and R.P. Vasquez, Appl. Phys. Lett. 60 (1992) 995. [31] MS. Hybertsen, in: Light Emission from Silicon (see ref.

m. [32] M. Voos, Ph. Uzan, C. Delalande, G. Bastard and A. Halimaoui, Appl. Phys. Lett. 61 (1992) 1213. [33] H. Takaji, H. Ogawa, Y. Yamazaki, A. Ishizaki and T. Nakagini, Appl. Phys. Lett. 56 (19901 2379. [34] D. Zhang, R.M. Kolbas, P. Mehta, A.K. Singh, D.J. Lichtenwalner, K.Y. Hsieh and A.I. Kingon, in: Light in Silicon (see ref. [2]1. [35] R. Venkatasubramanian, D.P. Malta, M.L. Timmons and J.A. Hutchby, Appl. Phys. Lett. 59 (1991) 1603. [36] K. Furukawa, M. Fujino and N. Matsumoto, Appl. Phys. Lett. 60 (1992) 2744. [37] E. Bustarret, M. Ligeon and L. Ortega, Solid State Commun. 83 (1992) 461. [38] E. Bustarret, M. Ligeon, J.C. Bruyire, F. Muller, R. Herino, F. Gaspard, L. Ortega and M. Stutzmann, Appl. Phys. Lett. 61 (1992) 1552. [39] Y.H. Xie, W.L. Wilson, F.M. Ross, J.A. Mucha, E.A. Fitzgerald, J.M. Macaulay and T.D. Harris, J. Appl. Phys. 71 (1992) 2403.

407

[40] T. Motohiro, T. Kachi, F. Miura, Y. Takeda, S. Hyodo and S. Noda, Jpn. J. Appl. Phys. 31 (19921 L207. [41] Z.Y. Xu, M. Gal and M. Gross, Appl. Phys. Lett. 60 (1992) 1375. [42] X.L. Zheng, W. Wang and H.C. Chen, Appl. Phys. Lett. 60 (1992) 986. [43] M. Stutzmann, J. Weber, M.S. Brandt, H.D. Fuchs, M. Rosenbauer, P. Deak, A. Hiipner and A. Breitschwerdt, Adv. Solid State Phys. 32 (19921. [44] M.S. Brandt, A. Breitschwerdt, H.D. Fuchs, A. Hiipner, M. Rosenbauer, M. Stutzmann and J. Weber, Appl. Phys. A, submitted. [45] M.S. Brandt, H.D. Fuchs, M. Stutzmann, J. Weber and M. Cardona, Solid State Commun. 81 (1992) 307. [46] J.C. Vial, A. Bsiesy, F. Gaspard, R. Herino, M. Ligeon, F. Muller, R. Romestain and R.M. Macfarlane, Phys. Rev. B 45 (1992) 14171. [47] C. Tsai, K.H. Li, D.S. Kinosky, R.Z. Qian, T.C. Hsu, J.T. Irby, S.K. Banerjee, A.F. Tasch and J.C. Campbell, Appl. Phys. Lett. 60 (1992) 1700. [48] C. Tsai, K.H. Li, J. Sarathy, S. Shih, J.C. Campbell, B.K. Hance and J.M. White, Appl. Phys. Lett. 59 (1991) 2814. [49] SM. Prokes, O.J. Glembocki, V.M. Bermudez, R. Kaplan, L.E. Friedersdorf and P.C. Searson, in: Light in Silicon (see ref. [2]). [50] M.B. Robinson, A.C. Dillon, D.R. Haynes and SM. George, Appl. Phys. Lett. 61 (1992) 1414. [51] M.A. Tischler, R.T. Collins, J.H. Stathis and J.C. Tsang, Appl. Phys. Lett. 60 (1992) 639. [52] H.C. Chen, W. Wang, K.N. Manjularani, L.C. Snyder and X.L. Zheng, in: Light in Silicon (see ref. [2]). [53] A. Halimaoui, C. Oules, G. Bomchil, A. Bsiesy, F. Gaspard, R. Hkrino, M. Ligeon and F. Muller, Appl. Phys. Lett. 59 (1991) 304. [54] D.J. Dimaria, J.R. Curtley, E.J. Pakulis, D.W. Dong, T.S. Kuan, F.L. Pasavento and J. Cutro, J. Appl. Phys. 56 (1984) 401. [55] A. Richter, P. Steiner, F. Kozlowski and W. Lang, IEEE Electron. Device Lett. 12 (1991) 691. [56] N. Koshida and H. Koyama, Appl. Phys. Lett. 60 (1992) 347. 1571 E. Bassous, M. Freeman, J.M. Halbout, S.S. Iyer, V.P. Kesan, P. Munguia, S.F. Pesearcik and B.L. Williams, IBM Res. Rep. RC 17516, July, 1992. [58] L. Garchery, Rapport de stage (ruin 19921, DEA Optoelectronique, Universite Joseph Fourier de Grenoble.