Characterization of laser ablated silicon thin films

Characterization of laser ablated silicon thin ®lms

S. Vijayalakshmia, Z. Iqbalb, M.A. Georgec, J. Federicid, H. Grebela,*

aOptical Waveguide Laboratory, Electrical and Computer Engineering,New Jersey Institute of Technology, Newark, NJ 07102, USAbResearch and Technology, Allied Signal Inc., Morristown, NJ 07962, USA

cUniversity Of Alabama at Huntsville, Huntsville, AL 35899, USAdDepartment of Physics, New Jersey Institute Of Technology, Newark, NJ 07102, USA

Received 11 November 1997; accepted 24 July 1998

Abstract

Using laser ablation, we deposited silicon layers consisting of clusters and crystalline domains onto glass, quartz, aluminum, titanium,

copper, single-crystal silicon and single-crystal potassium bromide substrates. The microstructure and the morphology of the ®lms were

characterized by use of optical microscopy, laser scanning microscopy, atomic force microscopy, transmission electron microscopy, micro-

Raman spectroscopy, X-ray photoelectron spectroscopy and X-ray diffraction. The results indicated that the deposited material was

composed of microcrystalline droplets, typically 3.5 mm in diameter, separated by amorphous-like regions. The droplets were composed

of crystalline material at their centers and an outer halo of nanometer-size particles. q 1999 Elsevier Science S.A. All rights reserved.

Keywords: Nanostructures; Raman scattering; Transmission electron microscopy (TEM); X-ray diffraction

1. Introduction

The unique non-linear properties of laser ablated silicon

samples measured recently in our laboratory [1,2], led us to

characterize them extensively by use of various techniques:

optical microscopy (OM), laser scanning microscopy

(LSM), atomic force microscopy (AFM), transmission elec-

tron microscopy (TEM), micro-Raman spectroscopy, X-ray

photoelectron spectroscopy (XPS) and X-ray diffraction

(XRD). It is well known that nanometer size clusters show

an increase in the non-linear optical coef®cients compared

to bulk materials. This increase may be as large as two

orders of magnitude [3,4] and was largely attributed to

con®nement of electron and hole states within small dimen-

sions. As the clusters become of the order of 20 nm, exci-

tonic states possess energy values comparable to the inter-

band separation and, therefore, may take part in the non-

linear process. In contrast, we have measured an increase of

more than four orders of magnitude in the real part of the

non-linear susceptibility [1,2]. For example, the real part of

the non-linear susceptibility, Re{x (3)}, was found to be,

2.28 £ 1025 esu with non-linear lifetime, t , of 140 ns at

l � 355 nm. We measured, Re{x�3�} � 21:33 £ 1023 esu

and t � 5 ns at l � 532 nm. These values should be

compared with typical values of Re{x�3�} , 1028 esu for

nanoclusters prepared by other methods [3,4]. Thus, it is

important to know the morphology and size range of the

nanoclusters involved in the non-linear processes. In parti-

cular, one would like to ®nd out whether the large non-linear

values stem from the existence of closed-packed nanocrys-

tallites. If this is the case, useful non-linear, nano-optical

materials may be realized. Our ®lms were composed of

micrometer size droplets that were made of nanocrystalli-

ties. Non-linear optical as well as some of the characteriza-

tion techniques (namely, XRD and XPS) were averaged

over relatively large sample area. As a result, the droplets

may have contributed more to the signal in XRD and XPS.

Thus with some caution, we may conclude that the large

non-linearities observed are related to these crystallites. The

paper is organized as follows. An introduction is provided in

Section 1 and the growth of the samples is described in

Section 2. In Section 3 we describe our results from the

various techniques. These are discussed in Section 4 and

conclusions are provided in Section 5.

2. Experiment

The silicon ®lms were grown on various substrates by

laser ablation. A KrF excimer laser beam (l � 248 nm,

average power kIl � 3 W, pulse duration 8 ns, repetition

Thin Solid Films 339 (1999) 102±108

0040-6090/99/$ - see front matter q 1999 Elsevier Science S.A. All rights reserved.

PII: S0040-6090(98)01158-4

* Corresponding author.

rate 50 Hz, deposition time 10 min) was focused to a

100 mm spot-size on a Si wafer target (k100l, n-type, 1016

cm23). The Si target was cleaned by hydro¯uoric acid and

methanol while the substrates were cleaned by acetone and

methanol. The substrates were positioned 3 cm from the Si

target. The ambient pressure in the chamber was maintained

at 8 £ 1026 Torr. The substrates were maintained at room

temperature. The size of each substrate was about 1.5 cm £1.5 cm. The samples possessed concentric ring regions with

varying colors. Most of our ®lms could be classi®ed into

three major regions, I, II and III, with colors in transmission

that were faint yellow, yellow, and brown respectively.

These colors probably originated from differences in

absorption for the various regions. Region III extended

between 0±3 mm with respect to the center of the ring.

The radii of rings of regions II and I were 3±4.5 mm and

4.5±6 mm respectively.

3. Results

3.1. Optical microscopy

The ®lms were ®rst inspected by use of an optical micro-

scope and were found to be composed of micrometer-size

droplets as shown in Fig. 1. Smooth regions consisting of

nanometer-size clusters were scattered in between the

droplets as indicated by atomic force microscopy (AFM)

and micro-Raman spectroscopy. The average size of a

droplet was surprisingly similar in all three regions. The

average droplet diameter was 3:63 ^ 2:17 mm,

3:38 ^ 1:48 mm and 3:34 ^ 1:51 mm in regions I, II and

III respectively. The distributions of the droplets were 22,

54, and 76 droplets within a 50 £ 50 mm square in regions I,

II and III respectively. Similar droplet formation was

observed on glass, quartz, aluminum-coated glass, titanium,

and single-crystal KBr and silicon, when the ®lms were

deposited under the same conditions. In addition, most

droplets were circular in shape.

3.2. Scanning laser microscopy

The thickness in each region is not a well de®ned feature

owing to the discontinuous nature of the samples. Therefore,

we prefer to characterize each region by its equivalent thick-

ness. This was achieved with the help of a laser scanning

microscope (LSM, Carl Zeiss, axioplan microscope). We

performed a two-dimensional scan, as shown in Fig. 2, of

the sample which was positioned at various heights with

respect to a reference point (topographical sectioning). A

histogram of the sample height was generated and an esti-

mate of the average height of each droplet was made. The

reference point in Fig. 2 was taken as the level of a crater

S. Vijayalakshmi et al. / Thin Solid Films 339 (1999) 102±108 103

Fig. 1. Optical microscope pictures of the Si sample on a glass substrate in

(a) region I, (b) region II and (c) region III.

Fig. 2. Topographic lines around a clean crater as measured by use of LSM.

Also shown is a histogram of the height occurrence in the sample.

formed by laser ablation of the sample which was silicon

clusters on a glass substrate. The glass substrate was trans-

parent to the laser beam and was visually inspected to make

sure that most of the silicon material was removed and the

glass remained unharmed. As can be seen from Fig. 2, the

average height of a droplet in region II is about 0.8 mm. The

average height of the droplets in regions I, II, and III was

measured to be, 0:6 ^ 0:2 mm, 0:8 ^ 0:2 mm and

2 ^ 0:5 mm, respectively. These droplets were located on

an amorphous-like background ®lm. The average thick-

nesses of the material in regions I, II and III were estimated

from these measurements as 100 nm, 200 nm and 400 nm

respectively. The sample thicknesses were also measured by

use of a pro®lometer. Like the LSM, the pro®lometer probe

averages the features in the height of the layer. Both tech-

niques yielded similar results for the background as well as

for the clusters. The scan with the LSM was repeated with a

lower resolution. The spot-size of the laser was thus

increased, resulting in a larger height averaging. This

number was used when assessing the optical loss (see

below).

3.3. Optical absorption

Fig. 3 shows the optical absorption from a laser ablated

silicon ®lm on glass. The optical absorption spectrum was

obtained by measuring the transmission and re¯ection for

each region assuming an averaged dielectric slab with a

given effective thickness as well as multiple re¯ections

within the slab. The absorption shows features associated

with an effective band gap, particularly at wavelengths in

the blue spectral range. In assessing the absorption for each

region one needs to identify a characteristic ®lm thickness.

The large droplets contributed more to the ®lm optical loss

since absorption depends exponentially on the ®lm thick-

ness. On the other hand, the number of droplets was small

and most of the transmission occurred through the back-

ground layer. In addition, surface roughness effectively

increased absorption values owing to multiple re¯ections

between and within droplets [5]. A practical solution to

the above was to de®ne an effective thickness of the back-

ground layer obtained by a low resolution LSM scan. These

values were 100 nm, 200 nm and 400 nm for regions I, II

and III respectively. This procedure results in an effective

value of the loss, which includes multiple re¯ections.

3.4. X-ray diffraction

X-ray diffraction was used to determine the average size

of the microcrystals and nanocrystals in the sample. A

Rigaku D/MAX-B system with a Cu Ka source and a thin

®lm attachment was used to obtain the diffraction data

displayed in Fig. 4. The data were taken with the X-ray

beam incident at a grazing angle of 28 to enhance the signal

from the thin ®lms, whereas the data for the reference wafer

was taken at 908 to the wafer surface. The k111l, k220l,k311l and k400l crystallographic planes of cubic Si corre-

sponding to 2u values of 28.458, 47.358, 56.28, and 69.38respectively, were identi®ed in the diffraction patterns. The

k400l re¯ection from the k100l oriented Si target was

observed at 69.38. The diffraction data from the laser ablated

samples showed the crystallites were predominantly

oriented along k111l and k220l planes. When different

regions of the ®lms were studied, the X-ray re¯ections

appeared at the same 2u positions, but peak intensities for

region III were observed to be larger owing to the larger

amount of material in this region. We have grown samples

on k100l double-polished Si substrates under similar condi-

tions. Here too, the microcrystallites in the sample were

predominantly oriented on k111l Si planes in contrast to

the k100l orientation of target or substrate crystals.

An attempt was made in each region to calculate the

crystallite size from the half-widths of the X-ray diffraction

lines. High resolution diffraction patterns were taken around

2u � 288 and the data obtained were ®tted to a Gaussian line

shape superimposed on a linear background. The width of

the diffraction line is proportional to the crystallite size. The

latter can be estimated from high resolution X-ray diffrac-

S. Vijayalakshmi et al. / Thin Solid Films 339 (1999) 102±108104

Fig. 3. (a) Transmission and re¯ection from region II. (b) Optical assess-

ment based on the equivalent ®lm thickness of 0.2 mm.

tion data using the Scherrer relation [6]:

B � 0:9l

tcosu�1�

where B is the linewidth at half maximum, l is the X-ray

wavelength, and t is the diameter of the crystal. The data are

displayed in Fig. 5, and as can be seen, the lines observed

from regions II and III have identical full widths at half-

maximum (FWHM) of 0.1678, corresponding to an average

crystallite size of 43 nm. Region I did not show signi®cant

X-ray diffraction line intensities. This estimate represents

the average size of the crystallites within the droplets in

each region as will be discussed below.

3.5. X-ray photoelectron spectroscopy

The average oxygen levels on the surface of the ®lms

were determined with X-ray photoelectron spectroscopy

(XPS) using a Kratos 8000 XPS system. The measured Si

core level data are displayed in Fig. 6. The peak at a binding

energy of 101 eV is attributed to Si 2p core electrons while

the peak at 106 eV is assigned to 2p core electrons from

SiO2. The oxide thickness on the ®lm surface was estimated

by angle resolved XPS data shown in Fig. 6. In this method,

the photoelectrons were detected at various take-off angles.

The escape depth of photoelectrons is known to be 1±2 nm

from the sample surface. Thus, at large take-off angles, the

data include signals from the surface and from a 2 nm thick

layer just below it. As the take-off angles became smaller,

the contribution of the surface photoelectrons remained the

same, but there was a substantial decrease in the contribu-

tion from the layer underneath. The reduction of the 2p Si

peak and the essentially constant oxide peak indicate that

only the surface layer (which is less than 2 nm and also less

than the typical size of the crystallites and clusters) is

oxidized, and that the bulk of the ®lm is mostly pure Si.

3.6. Atomic force microscopy measurements

Atomic force microscopy (AFM) was used to study the

size distribution of the clusters inbetween the droplets in

each region. The measurements were performed in contact

or tapping mode conditions using a Digital Instruments

Nanoscope III system with a piezoelectric tube scanner of

effective scan range 0.5±15 mm. Fig. 7 shows the surface

morphology of the samples in all the three regions. As is

evident from Fig. 7a, region I shows smaller and more

uniform clusters than the other two regions. The statistics

on each region was achieved by collecting the data of actual

heights in sample cross-sections rather than relying on

(often misleading) color chart. The statistics on all three

S. Vijayalakshmi et al. / Thin Solid Films 339 (1999) 102±108 105

Fig. 4. X-ray diffraction spectra of a ®lm containing nanoclusters on a

glass substrate

Fig. 5. High resolution XRD spectra from region II (solid circles) and

region III (open circles) respectively

regions were as follows: region I had a cluster size distribu-

tion from 3±43 nm with the mean at 7 nm and a standard

deviation of 3 nm; region II ranged between 5 and 42 nm

with the mean at 11 nm and a standard deviation of 8 nm;

region III had a range of 15±85 nm with the mean at 50 nm

and a standard deviation of 22 nm. We should emphasize

that the cluster width which may be inferred from Fig. 8 is a

result of the convolution between the entire tip pro®le and

the cluster and, therefore, is not indicative of the actual

width of a cluster [7]. The cluster size was determined by

measuring the height of the clusters, assuming spherical

shaped clusters, and was calibrated against standards of

gold colloids.

3.7. Micro-Raman spectroscopy

Micro-Raman spectroscopy was used to determine the

size of the crystalline domains in each portion of the ®lm.

An Instruments SA micro-Raman system with a cooled

charged coupled device (CCD) array detector, was used to

collect the Raman data. Backscattered Raman spectra were

typically excited with 10 mW of l � 514:5 nm radiation

from an argon ion laser source with a spot size of 1±

2 mm. A SPEX U-1000 spectrometer with a 30±50 mm

spot size was used to study the average spectra from larger

areas. Raman spectra from a droplet, the region inbetween

the droplets and a k100l oriented silicon target are shown in

Fig. 8. Similar spectra were obtained for the different

regions in the samples. The ®rst-order Si phonon line

from the droplets was slightly asymmetric and relatively

sharp. The linewidth at half height was 5.6 cm21 relative

to 4 cm21 for the k100l oriented Si crystal. At 30±50 mm

spatial resolution, the Si phonon line, which is at 520 cm21

under ambient conditions, was down-shifted approximately

by 4 cm21. The crystalline-size estimated from the

frequency shift is 5.8 nm, but the size obtained from the

linewidth is greater than 15 nm [8,9]. By focusing down

to 1 mm we were able to scan regions within each droplet

showing crystallite sizes below and above the average value

of about 5.8 nm based on the observed shifts. A broad peak

around 475 cm21 was observed for the region between the

droplets, which can be assigned to the one phonon density-

of-states of amorphous silicon, consistent with the TEM

data discussed below. In some spots, crystalline peaks corre-

sponding to ordered domains were also superimposed on the

broad peak due to amorphous silicon. No amorphous-like

scattering was obtained from the particles within the

droplets. From the Raman results we conclude that the

droplets are composed of nanometer size crystalline

domains, while the region between the droplets essentially

shows no long-range order except for some spots of short- to

medium-range order that were found to be below 3 nm in

size. No indication of a photoluminescence background was

observed in the Raman spectra.

3.8. Transmission electron microscopy

In Fig. 9 we show an image taken with a transmission

electron microscope. The sample was grown on KBr,

checked with micro-Raman spectroscopy and thinned by

ion milling. It is evident from Fig. 9 that a droplet consists

of large Si crystallites at its center and smaller nanocrystals

in its peripheral regions. Selected area electron diffraction

from the droplet is consistent with that of cubic silicon. The

region between the droplets is amorphous.

4. Discussion

We have studied the structure and morphology of laser

ablated Si ®lms on various substrates. The ®lms were

composed of micrometer-size droplets and nanometer size

clusters scattered between the droplets. From the XPS and

Raman data we conclude that the bulk of the ®lm is Si and

the surface is composed of SiO2. Electron diffraction and

Raman data revealed that the micrometer size droplets

consist of crystallites. The region between the droplets

was mainly amorphous. In addition, the X-ray diffraction

data showed that the crystallites possessed an orientation

different from that of the target. The ®lms were predomi-

nantly k111l oriented while the target was oriented in the

k100l plane. This result was independent of the substrate; Si

®lms deposited onto glass, polycrystalline quartz, and k100l

S. Vijayalakshmi et al. / Thin Solid Films 339 (1999) 102±108106

Fig. 6. Angle resolved XPS spectra shown at take-off angles of (a) 908, (b)

458 and (c) 158.

oriented Si substrates all exhibit predominantly k111l orien-

tation. By itself, this difference of crystallographic orienta-

S. Vijayalakshmi et al. / Thin Solid Films 339 (1999) 102±108 107

Fig. 7. AFM intensity maps: (a) region I (brightness range 0±50 nm); (b)

region II (brightness range 0±200 nm); (c) region III (brightness range 0±

200 nm). Scan was made within a 1.5 mm £ 1.5 mm area between the

droplets. In the paper, size information was obtained from cross-sectional

height histograms rather than from these intensity maps.

Fig. 8. Raman spectrum from Si droplet on glass substrate (a). The Raman

signature of the amorphous-like region in between the droplets (b). The Si

peak from a k100l wafer is also shown for comparison (c). The arbitrary unit

scale is different for every curve.

Fig. 9. TEM image of a droplet. Clearly seen are the nanocrystallites in the

droplet.

tion for the deposited layers as compared to that of the target

may not be of signi®cance. However, we have also found

that the droplets of Si were similar in size across the entire

®lm independent of the substrate material. This is quite

surprising in view of the directional nature of the laser-

induced plasma plume, suggesting that droplet formation

occurred in the gas phase during deposition.

The fact that the X-ray diffraction lines from both regions

II and III have identical linewidths suggests that they origi-

nated from regions of similar average crystallite sizes. We

estimated this size to be 43 nm. This value is within the

range of crystallite sizes observed in the TEM images of

the droplets.

The existence of crystalline domains within the droplets

was indicated by the micro-Raman and TEM experiments.

Based on the frequency shift we estimate the size of the

nanocrystals to be 5.8 nm, while a crystallite size of

15 nm was obtained from the Raman linewidth. Similar

results are found for all three regions as well as for ®lms

deposited onto different substrates. One may argue that

these shifts may be explained by residual strain on the crys-

tallites developed during droplet formation. If this is the

case, the down-shifted frequency values would vary as a

function of substrate material because of the different

wetting properties of each surface. Surprisingly, the

Raman lines from the crystallites were remarkably narrow.

This is not fully understood at this point, but may result from

the compact arrangement of small and relatively large crys-

tallites within the droplets. Our measurements of isolated

clusters formed by ion-implantation featured a much

broader Raman line and a larger degree of asymmetry [10].

No photoluminescence (PL) background was observed in

the Raman data from the ®lms and no PL spectra were

detected by exciting the sample with the l � 514:5 nm

argon ion line. We believe that this is due to a lack of

substantial oxygen/silicon interface or oxygen-de®cient

sites in our system. Non-linear absorptions possess lifetimes

of the order of 5 ns at l � 532 nm in these ®lms. This leads

us to believe that PL, if present, may be very short lived and

thus could not be detected. Annealing of the samples

appears to exhibit a small increase in the PL intensity; this

effect is still under investigation.

5. Conclusions

Thin ®lms containing Si clusters and nanocrystalline

regions were grown on various substrates by laser ablation

from a k100l Si target. The crystalline regions in the ®lms

were predominantly oriented in the k111l plane. We

observed that the ®lms were composed of micrometer-size

droplets with a nanocrystalline to microcrystalline structure.

The droplets consisted of nanocrystalline domains and were

not embedded in an amorphous matrix material. The AFM

measurements indicated that regions between the droplets

consisted of nanometer size clusters. Raman spectroscopy

and TEM showed that the regions between the droplets are

amorphous. We believe that the very large non-linear opti-

cal coef®cients measured for these samples arose from the

closed-packed nanocrystallites within the micrometer-sized

droplets.

References

[1] S. Vijayalakshmi, M. George, H. Grebel, Appl. Phys. Lett. 70 (1997)

708.

[2] S. Vijayalakshmi, F. Shen, H. Grebel, Appl. Phys. Lett. 71 (1997)

3332.

[3] R. Jain, R. Lind, J. Opt. Soc. Am. 73 (1983) 647.

[4] L. Banyai, S.Koch, Semiconductor Quantum Dots, World Scienti®c,

Singapore, 1993.

[5] H. Grebel, K. Fang, J. Appl. Phys. 77 (1995) 367.

[6] Z. Iqbal, S. Veprek, J. Phys. C 15 (1982) 377.

[7] M.A. George, et. al, J. Appl. Phys. 76 (1994) 4099.

[8] O. Cullity, Elements of X-ray Diffraction, 2nd edn., Addison-Wesley,

New York, 1978, Chapter 3.

[9] Z. Iqbal, S. Veprek, A.P. Webb, P. Capezzuto, Solid State Commun.

37 (1981) 993.

[10] S. Vijayalakshmi, H. Grebel, Z. Iqbal, C. White, J. Appl. Phys. 84

(1998) 6502.

S. Vijayalakshmi et al. / Thin Solid Films 339 (1999) 102±108108