formation of uniform high density and small size ge si quantum dots by scanning pulsed laser annealing of pre deposited ge si film

Figure 1 shows the AFM images, for various positions a–e inside or outside the irradiated region as denoted in the upper-left inset, of an SPLA sample produced with a Ge layer thickness

pulsed laser annealing of pre-deposited Ge/Si film

Hamza Qayyum, Chieh-Hsun Lu, Ying-Hung Chuang, Jiunn-Yuan Lin, and Szu-yuan Chen,

Citation: AIP Advances 6, 055323 (2016); doi: 10.1063/1.4953057

View online: http://dx.doi.org/10.1063/1.4953057

View Table of Contents: http://aip.scitation.org/toc/adv/6/5

Published by the American Institute of Physics

Formation of uniform high-density and small-size Ge/Si quantum dots by scanning pulsed laser annealing

of pre-deposited Ge/Si film

Hamza Qayyum,1,2,3Chieh-Hsun Lu,1,2Ying-Hung Chuang,1,4

Jiunn-Yuan Lin,4and Szu-yuan Chen1,2,3, a

1Institute of Atomic and Molecular Sciences, Academia Sinica, Taipei 106, Taiwan

2Department of Physics, National Central University, Zhongli, Taoyuan 320, Taiwan

3Molecular Science and Technology Program, Taiwan International Graduate Program,

Academia Sinica, Taipei 115, Taiwan

4Department of Physics, National Chung Cheng University, Chiayi 621, Taiwan

(Received 11 March 2016; accepted 19 May 2016; published online 25 May 2016)

The capability to fabricate Ge/Si quantum dots with small dot size and high dot density uniformly over a large area is crucial for many applications In this work, we demonstrate that this can be achieved by scanning a pre-deposited Ge thin layer on Si substrate with a line-focused pulsed laser beam to induce formation of quantum dots With suitable setting, Ge/Si quantum dots with a mean height of 2.9 nm, a mean diam-eter of 25 nm, and a dot density of 6×1010cm−2could be formed over an area larger than 4 mm2 The average size of the laser-induced quantum dots is smaller while their density is higher than that of quantum dots grown by using Stranski-Krastanov growth mode Based on the dependence of the characteristics of quantum dots on the laser parameters, a model consisting of laser-induced strain, surface diffusion, and Ostwald ripening is proposed for the mechanism underlying the formation

of the Ge/Si quantum dots The technique demonstrated could be applicable to other materials besides Ge/Si C 2016 Author(s) All article content, except where otherwise noted, is licensed under a Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).[http://dx.doi.org/10.1063/1.4953057]

I INTRODUCTION

In recent years, the growth of quantum dots (QDs) embedded in a solid-state device has become

an active research field because of their novel electrical and optical properties Among various materials, Ge/Si quantum dots is of particular interest due to its small band gap, non-toxicity, and compatibility with the well-developed Si technology Many applications of Ge/Si QDs have been explored, such as optoelectronic devices,1 solar cells,2 and thermoelectricity.3 Convention-ally, Ge/Si QDs are grown by depositing Ge atoms on a Si substrate under the condition for Stranski-Krastanow (SK) growth mode The 4.2% lattice mismatch between the Si and Ge layers

is the driving force for the dot formation In the beginning of Ge deposition, strain builds up but still accommodated Once the Ge layer thickness exceeds the critical thickness, the strain relaxes by forming Ge dots.4The unassisted SK growth exhibits poor control over the evolution of quantum dots, since the dot nucleation is fluctuation-driven in nature This results in unsatisfactory and uncontrollable QD size and density, inadequate for fabricating optoelectronic devices.5Therefore, it

is critical to have a way for a high degree of control over formation of QDs

To date, the most promising approach for control of the size and density of QDs appears

to be assisted SK growth by templating the underlying Si layer using various methods such as nano-indentation and lithography with electron or ion beams.68In these methods, lithographically

or mechanically produced pits serve as preferred nucleation sites for deposited atoms However, the

a E-mail: sychen@ltl.iams.sinica.edu.tw

QD size and density are limited by the resolution of the lithographic instruments, and the process is time-consuming It has also been reported that QD size can be reduced and QD density increased

by surfactant-mediated growth of QDs using sub-monolayer Sb or C,9 11or ultra-thin SiO2layer.12

Nevertheless, such methods inevitably introduce interfacial species which may modify the elec-tronic structure of Ge/Si QDs On the side of top-down approaches, Dais et al has recently reported the growth of densely packed QD arrays with a density of 8.16 × 1010 cm−2 by using extreme ultraviolet interference lithography.13 However, this method relies on large synchrotron radiation facility and thus are not time- and cost-effective The close proximity between adjacent QDs also results in electronic coupling between QDs, rendering this method unsuitable for many applications Pulsed laser annealing is an emerging material fabrication method based on localized photo-chemical and photothermal reaction Unlike conventional thermal annealing, which heats up the whole device, the irradiation with a pulsed laser exerts thermal annealing only in the region requir-ing the treatment without causrequir-ing a detrimental effect in other regions of the device The selective heating of the surface by pulsed laser enables processing of the targeted layer/region on any kind of substrate Motivated by these advantages, pulsed laser annealing has been used for producing nan-odots or nanocones on a semiconductor surface.14 – 18However, in all of the previous works reporting

on pulsed laser annealing of semiconductors, the fabrication of nanostructures is limited to a small area Uniformity over a large area is hard to achieve in these methods because it is hard to get a large laser beam profile with the required smoothness Therefore, these methods lead to bad uniformity of the nanostructures on chip-size scale and thus are not ideal for fabricating optoelectronic devices

In this paper, we report a method termed scanning pulsed laser annealing (SPLA) for fabricat-ing Ge/Si QDs with small dot size and high dot density uniformly over a large area It is achieved

by scanning a pre-deposited Ge thin layer on Si substrate with a line-focused pulsed laser beam

to induce formation of Ge/Si QDs With suitable setting, Ge/Si quantum dots with a mean height

of 2.9 nm, a mean diameter of 25 nm, and a dot density of 6×1010cm−2could be formed over an area larger than 4 mm2 The average size of the laser-induced QDs is smaller while their density is higher than that of QDs grown by using Stranski-Krastanov growth mode Based on the dependence

of the characteristics of QDs on the laser parameters, a model consisting of laser-induced strain, surface diffusion, and Ostwald ripening is proposed for the mechanism underlying the formation of the Ge/Si QDs

II EXPERIMENTAL

The Ge/Si QD samples were made by first depositing thin Ge layers on Si substrates and then subjected the Ge/Si films to scanning pulsed laser annealing of various conditions to induce formation of Ge dots on the Si substrates The two processes were carried out in the same vacuum chamber, which was evacuated to < 2 × 10−4torr for both processes

Pulsed laser deposition (PLD) was used to deposit the Ge layers A Ge disk (purity >99.99%)

of 10-mm diameter and 5-mm thickness was used as the PLD target The Si (100) substrates were cleaned by the sequence of dipping in a 0.56 M HF solution for 5 min, dipping in a 0.28 M HF solution for 4 min, flushing with deionized water, rinsing with isopropyl alcohol, and drying with nitrogen gas, right before being loaded into the vacuum chamber The Ge target was mounted on a motorized holder, which was controlled by a computer program to automatically rotate or translate the target after each laser shot This provided a different location for the next ablation pulse in order to avoid formation of large craters and thus change of on-target laser alation parameters

A third-harmonic Q-switched Nd:YAG laser of 355-nm wavelength, 8-ns pulse width, and 10-Hz repetition rate was used as the ablation laser The p-polarized ablation beam was focused with a spherical lens onto the Ge target at 45◦incidence angle The on-target beam size was set to 300 µ m (vertical)×420 µ m (horizontal) in clear aperture, resulting in a peak laser fluence of 120 J/cm2 The 1-cm×1-cm substrates were mounted on a motorized carousel-type holder which contained six substrate mounts, allowing for switching of the substrates under vacuum A mask was installed in front of the substrate holder to shield the substrates except for the one being coated The PLD sub-strate temperature was set at 400◦C for all the samples used in this experiment The substrates were positioned in the normal direction of the target, and the target-to-substrate distance was set at 4 cm

The deposition rate was 0.03 nm/s, which was measured with a quartz microbalance (SQM-160, Sigma Instruments) calibrated by using a surface profiler for film thickness measurement

SPLA were carried out by using a second-harmonic Q-switched Nd:YAG laser of 532-nm wavelength, 10-ns pulse width, and 10-Hz repetition rate The p-polarized annealing beam was expanded with a concave cylindrical lens in the vertical direction and shrunk with a plano-convex cylindrical lens in the horizontal direction to produce a line-shaped beam of 20 mm (vertical)×0.5 mm (horizontal) in full width at half maximum (FWHM) on the surface of the PLD-produced Ge/Si films with an incidence angle of 45◦ The Ge/Si films were mounted on a motorized holder that could move in the horizontal direction to allow scanning of the Ge/Si films by the annealing laser beam at a variable scan speed The Ge/Si films were kept at room temperature during pulsed laser annealing Atomic force microscopy (AFM) (NanoWizard II, JPK Instruments) with a cantilever of <12 nm tip radius of curvature was used to measure the surface morphology of the Ge/Si films and Ge/Si QD samples Raman spectroscopy (NTEGRA Spectra Probe NanoLabo-ratory, NT-MDT) with an excitation wavelength of 473 nm was used to characterize the crystallinity and composition of the QDs

III RESULTS AND DISCUSSION

All the SPLA samples presented in this paper were scanned by the laser beam for only a dis-tance of 2 mm in the center of the whole substrate, as illustrated in the upper-left inset of Figure1, since it satisfied the purpose of demonstrating the capability of SPLA while leaving a non-irradiated

FIG 1 AFM images, for various positions denoted in the upper-left inset, of an SPLA sample produced with a Ge layer thickness of 0.9 nm, a peak SPLA laser fluence of 115 mJ/cm 2 , and a scan speed of 1.0 µm/s The scale for all images is

1 µm×1 µm.

region for comparison with the irradiated region Figure 1 shows the AFM images, for various positions (a–e) inside or outside the irradiated region as denoted in the upper-left inset, of an SPLA sample produced with a Ge layer thickness of 0.9 nm, a peak SPLA laser fluence of 115 mJ/cm2, and a scan speed of 1.0 µ m/s (corresponding to accumulation of 5000 laser shots at every position

in the irradiated region) It was observed that Ge/Si QDs formed throughout the entire irradiated region In contrast, there were no Ge/Si QDs in the non-irradiated region, as shown in Figure1(a) The non-irradiated region of the Ge thin film is flat with a surface roughness of 0.1 nm The absence

of any kind of surface structure in the non-irradiated region proved that the growth of QDs was in fact due to the laser annealing rather than any other growth mechanism In addition, it was found that QDs formed with very similar dot size and dot density over a 2 mm × 2 mm region, as shown

in Figure1(b)–1(e) This indicates that SPLA could be used to produce Ge/Si QDs of uniform char-acteristics on chip-size scale The vertical width of the region with uniform QD charchar-acteristics was determined by the uniformity of the laser beam in the vertical direction, which could be improved

by implementing beam homogenization technique or just expanding the beam further vertically The horizontal width of the uniform QD region was simply determined by the scan range, and thus it could be expanded as much as needed by applications

Figure2(a)–2(d)show the AFM images of SPLA samples produced using various scan speeds with a Ge layer thickness of 0.9 nm and a peak SPLA laser fluence of 115 mJ/cm2 Figure2(e)and

2(f)show the mean dot height, diameter, and density as functions of scan speed (and corresponding number of accumulated laser shots) As shown, the mean QD height decreased from 5 nm to 1.6 nm

FIG 2 AFM images of SPLA samples produced using various scan speeds: (a) 0.5 µm/s, (b) 1.0 µm/s, (c) 1.5 µm/s, and (d) 2.0 µm /s, with a Ge layer thickness of 0.9 nm and a peak SPLA laser fluence of 115 mJ/cm 2 (e) and (f): mean dot height, diameter, and density as functions of scan speed (and corresponding number of accumulated laser shots) The error bars indicate the standard deviation (size spread) The scale for all images is 1 µm×1 µm.

FIG 3 AFM images of SPLA samples produced using various peak SPLA laser fluences: (a) 92 mJ /cm 2 , (b) 100 mJ /cm 2 , (c) 107 mJ /cm 2 , and (d) 115 mJ /cm 2 , with a Ge layer thickness of 0.9 nm and a scan speed of 1.0 µm /s (e) and (f): mean dot height, diameter, and density as functions of peak SPLA laser fluence The error bars indicate the standard deviation (size spread) The scale for all images is 1 µm×1 µm.

and the mean diameter decreased from 46 nm to 28 nm with increasing scan speed from 0.5 µm/s to 2.0 µm/s Concomitantly, the QD density increased from 2.5×1010cm−2to 6.3×1010cm−2with the increase of scan speed

Figure3(a)–3(d)show the AFM images of SPLA samples produced using various peak SPLA laser fluences with a Ge layer thickness of 0.9 nm and a scan speed of 1.0 µm/s Figure3(e)and

3(f)show the mean dot height, diameter, and density as functions of peak SPLA laser fluence As shown, the mean QD height increased from 3.5 nm to 11.2 nm and the mean diameter increased from 24 nm to 39 nm with increasing peak SPLA laser fluence from 92 mJ/cm2to 115 mJ/cm2 Concomitantly, the QD density decreased from 5.1×1010cm−2to 1.5×1010cm−2with the increase

of peak SPLA laser fluence Formation of QDs was not observed with a peak SPLA laser fluence lower than 92 mJ/cm2, and ablation of sample surface occurred when the peak SPLA laser fluence was substantially higher than 115 mJ/cm2

Figure 4(a)–4(d) show the AFM images of SPLA samples produced using various Ge layer thicknesses with a peak SPLA laser fluence of 92 mJ/cm2and a scan speed of 1.0 µm/s Figure4(e)

and4(f) show the mean dot height, diameter, and density as functions of Ge layer thickness As shown, a flat, smooth film without QDs was obtained when the Ge layer thickness was 0.3 nm That is, the processing with SPLA did not change the film surface morphology Formation of QDs was observed when the Ge layer thickness exceeded 0.6 nm The mean QD height increased from 2.9 nm to 8.1 nm and the mean diameter increased from 25 nm to 45 nm with increasing Ge layer

FIG 4 AFM images of SPLA samples produced using various Ge layer thicknesses: (a) 0.3 nm, (b) 0.6 nm, (c) 0.9 nm, and (d) 1.2 nm, with a peak SPLA laser fluence of 92 mJ /cm 2 and a scan speed of 1.0 µm /s (e) and (f): mean dot height, diameter, and density as functions of Ge layer thickness The error bars indicate the standard deviation (size spread) The scale for all images is 1 µm×1 µm.

thickness from 0.6 nm to 1.2 nm Concomitantly, the QD density decreased from 6.2×1010cm−2to 1.2×1010cm−2with the increase of Ge layer thickness

The induction of formation of QDs by SPLA could be explained with the sequence of physical processes illustrated in Figure 5 For each position in the irradiated region, before the laser beam moves to cover this position, although the 4.2% lattice mismatch between the Ge layer and the underlying Si layer results in a compressive strain, formation of QDs does not occur This is because the total strain energy is not high enough to produce a thermodynamic tendency for forming QDs when the Ge layer thickness is below the critical thickness In addition, the process of QD formation

is prohibited kinetically since the temperature of the film surface is not high enough for effective surface diffusion of Ge atoms When the laser beam moves in to cover this position, the absorption

of the laser energy by the Ge/Si film results a transient rise and fall of surface temperature upon the incidence of each laser pulse, driving the formation of QDs The rise in surface temperature results from the energy deposition by the laser pulses, while the heat conduction into the Si substrate leads to the fall of surface temperature after the end of each laser pulse During the period of high temperature generated by each laser pulse (laser shot), the strain energy increases to the level that can thermodynamically drive formation of QDs to relieve the compressive strain even with such a small Ge layer thickness, and the surface diffusion mobility is high enough for enabling the move-ment of Ge atoms to preferable positions The increase in the strain energy could come from two factors Firstly, the absorption coefficient of Ge is higher than that of Si by two orders of magnitude

at 532-nm wavelength,19resulting in a higher temperature in the Ge layer than in the underlying

FIG 5 Proposed model for the induction of formation of QDs by SPLA: (a) illustration of the induction of formation of QDs by the scanning laser beam and (b) time sequence of the underlying physical processes.

Si layer Secondly, the thermal expansion coefficient of Ge (6.0×10−6/◦C) is significantly higher than that of Si (2.5×10−6/◦C).19Both effects lead to an increase in the lattice mismatch, thus raising the strain energy Once QDs have formed, Ostwald ripening occurs.20This thermodynamic process leads to escape of Ge atoms from smaller QDs to join larger QDs, in order to reduce free energy As

a result, the mean QD size increases while the spread of QD size distribution gets smaller.20

Since the number of accumulated laser shots at each position in the irradiated region is propor-tional to the width of the laser beam in the scan direction divided by the scan speed, the dependence

of QD morphology on scan speed is equivalent to its dependence on the total number of laser shots irradiating that position Therefore, Figure2actually reveals the evolution of QD morphology with increasing number of accumulated laser shots Because each laser shot produces only a short time window (a few hundred nanoseconds) with sufficiently high surface temperature, the processes of

QD formation and coarsening take place in a stepwise manner with each laser shot Therefore, the observation in Figure2that QD size (both height and diameter) increases and QD density decreases with increasing number of accumulated laser shots (decreasing scan speed) can be expected from the progression of the process of Ostwald ripening with accumulation of the action from successive laser shots It is also expected from this model that a higher laser fluence should result in a higher surface temperature due to a higher density of energy deposition, leading to a higher surface di ffu-sion mobility for Ge atoms and thus faster progresffu-sion of Ostwald ripening Therefore, the mean

QD size should get larger and the QD density should get lower when a higher laser fluence is used for the same number of laser shots, consistent with what was observed and shown in Figure 3 Figure6shows the QD diameter distributions of SPLA samples produced using various peak SPLA laser fluences with a Ge layer thickness of 0.9 nm and a scan speed of 1.0 µm/s As can be seen, the mean QD diameter becomes larger and the relative spread gets smaller when a higher laser fluence

is used These characteristics support the explanation based on Ostwald ripening

The dependence of QD size and density on Ge layer thickness shown in Figure 4 can be ascribed to two reasons Firstly, there are more Ge atoms involved in the processes when a thicker

Ge layer is used A larger mean QD size is expected because more Ge atoms are transported upon the action of each laser shots Secondly, the areal density of energy deposition from the laser increases when a thicker Ge layer is used This results in a higher peak surface temperature and a longer duration of high-temperature window upon the action of each laser shot, since the heat dissi-pation into the underlying Si substrate is at roughly a constant rate Therefore, Ostwald ripening can progress at a higher rate, leading to a larger mean QD size for the same number of laser shots

FIG 6 QD diameter distributions of SPLA samples produced using various peak SPLA laser fluences: (a) 92 mJ /cm 2 , (b)

100 mJ /cm 2 , (c) 107 mJ /cm 2 , and (d) 115 mJ /cm 2 , with a Ge layer thickness of 0.9 nm and a scan speed of 1.0 µm /s.

The effect of SPLA is equivalent to lowering the critical thickness of self-assembly of QDs,

so that QDs can form with a Ge layer thickness that does not allow for it without SPLA, such

as the transition from figure 1(a)to figure1(b)–1(e) However, even with SPLA there still exists

a critical thickness below which the deposited Ge tends to form flat film instead of QDs due to insufficient laser-induced strain Therefore, the absence of QDs in the case of 0.3-nm Ge film shown

in figure4(a)can be ascribed to an equivalent critical thickness falling in the range of 0.3 – 0.6 nm under that SPLA condition

The evolution of Ge QDs in result of SPLA of a pre-deposited Ge layer is different from SK growth Unlike SK growth process, with SPLA no QD was observed in the form of pre-pyramid

or pyramid All QDs grown using SPLA were of dome-like structures with varying height and diameter depending on the laser parameters The circumferences of the QDs varied from round to more irregular These shapes conform to the transient nature of pulsed laser annealing Formation of pyramids or pre-pyramids could occur only in the continuous, thermodynamic equilibrium process

of SK growth, which allows for preferential growth in specific crystallographic directions

Figure 7(a)shows the Raman spectra of SPLA samples produced using various scan speeds (various numbers of accumulated laser shots) with a Ge layer thickness of 0.9 nm and a peak SPLA laser fluence of 115 mJ/cm2 The Raman spectrum of an as-deposited sample without SPLA treatment is also shown for comparison Figure7(b)shows the ratio of the intensities of the Ge-Ge and Si-Ge peaks (IGe-Ge/ISi-Ge) and the width of the Ge-Ge peak as functions of scan speed (number

of accumulated laser shots) The Raman spectra of all of these samples contained two distinct phonon peaks, which could be attributed to a Ge-Ge phonon mode (≈300 cm−1) and Si-Ge phonon mode (≈415 cm−1) respectively.21The Raman spectrum of the as-deposited sample also exhibited

a broad peak at ≈270 cm−1, which could be attributed to amorphous Ge.22 As can be seen, the amorphous Ge peak diminished and the Ge-Ge peak got narrower with increasing number of accu-mulated laser shots (decreasing scan speed) This revealed that the crystallinity of the Ge dot or layer got better with increasing number of accumulated laser shots This is consistent with what

FIG 7 (a) Raman spectra of SPLA samples produced using various scan speeds (various numbers of accumulated laser shots) with a Ge layer thickness of 0.9 nm and a peak SPLA laser fluence of 115 mJ /cm 2 (b) Ratio of the intensities of the Ge-Ge peak and Si-Ge peak I Ge-Ge /I Ge-Si ) and width of the Ge-Ge peak as functions of scan speed (number of accumulated laser shots).

is expected for a process of high-temperature annealing.23On the other hand, it was observed that

IGe-Ge/ISi-Gedecreased with increasing number of accumulated laser shots This indicated that the level of intermixing between Ge and Si layers got higher with increasing number of laser shots Since the improvement in the crystallinity of Ge dots and the degree of inter-diffusion between Ge and Si (dominated by the diffusion of Si atoms into Ge dots24) are kinetically controlled by bulk

diffusion of Ge and Si atoms, and the time window of high temperature produced by each laser shot is limited, the two processes are expected to occur progressively with increasing number of accumulated laser shots, leading to the dependence shown in Figure7(b) The values of IGe-Ge/ISi-Ge for the QD samples produced at different conditions are comparable to that for self-assembled QDs grown in SK mode.21This means that the production of Ge/Si QDs by using SPLA does not result in a stronger intermixing than self-assembled QDs produced with SK growth, thus equally promising for applications in this respect

The Ge-Ge Raman peak should shift to the red side or the blue side when there is compressive

or tensile strain in the Ge layer The observation that the Ge-Ge Raman peaks for the samples processed with SPLA at the three different scan speeds were all at 300 cm−1, the same as that for bulk Ge, indicated almost fully relaxed strain in the QDs for all these cases, in contrast to the compressively strained, non-irradiated film, which exhibits a slightly larger Raman shift Following our model that relaxation of transiently enhanced compressive strain induced by laser heating drives the formation of QDs, it seems from the Raman data that at these three different scan speeds the SPLA had driven the QDs to a state of almost no residual strain after completion of the annealing