Implementation of complex nanosystems using a versatile ultrahigh vacuum nonlithographic technique

IOP PUBLISHING NANOTECHNOLOGY

Nanotechnology 18 (2007) 445202 (8pp) doi:10.1088/0957-4484/18/44/445202

Implementation of complex nanosystemsusing a versatile ultrahigh vacuumnonlithographic techniqueBiswajit Das and Arghya Banerjee

Nevada Nanotechnology Center, University of Nevada, Las Vegas, NV 89154, USA

E-mail: [email protected]

Received 18 February 2007, in final form 7 August 2007Published 9 October 2007Online at stacks.iop.org/Nano/18/445202

AbstractWe have developed an ultrahigh vacuum technique for the implementation ofcomplex nanosystems incorporating nonlithographic nanoparticles, ohmiccontact metals and isolation dielectrics. The technique is compatible with thesilicon integrated circuit manufacturing process and is versatile, allowing thedeposition of nanoparticles of any metal, semiconductor or insulator withdiameters as small as 2 nm with less than 5% size variation. In addition, thetechnique allows the creation of multi-layered structures of nanoparticles ofdifferent dimensions. The flexibility and the versatility of the technique havebeen demonstrated by depositing nanoparticles of various materials as well asfabricating multi-layered structures incorporating nanoparticles.

1. Introduction

Nanoparticles of semiconductors and metals are promising forthe implementation of a variety of photonic and electronicdevices with significantly improved performances as wellas with new functionalities. Examples of such devicesinclude quantum dot lasers and LEDs, single-electron devices,quantum computing and plasmonic devices. For theimplementation of such devices, some of the necessarycomponents are (i) nanoparticles, (ii) contact metals and(iii) tunneling and/or isolation dielectrics. While a numberof devices based on nanoparticles have been proposed, theirsuccessful implementations have been limited due to thelack of appropriate implementation systems and processes.For most electronic and photonic devices, it is typicallyrequired for the nanoparticle dimensions to be in the 1–20 nm range with size variations of 10% or less. Currentlithographic techniques are not suitable for the implementationof such nanoparticles, and nonlithographic techniques arebeing increasingly used for their fabrication [1, 2]. However,most nonlithographic techniques are based on natural self-organization processes and suffer from a lack of flexibilityor a lack of engineering control. Among the variousnonlithographic techniques, the predominant ones are solution-based, examples of which include sol–gel synthesis [3],chemical synthesis and electrochemical synthesis inside self-organized nanoporous templates [4]. While such solution-

based techniques are capable of producing nanoparticles withthe required dimensions and size distributions, they requirecomplex surface passivations involving organic cappingmolecules to prevent aggregation. These capping moleculesmodify the electrical surface properties of the nanoparticlesmaking charge injection/extraction difficult. In addition,the solution-based synthesis techniques are not compatiblewith solid-state device technology, the primary manufacturingprocess for electronic and photonic devices. The problemsassociated with solution-based methods can be addressed tosome extent by using nonlithographic fabrication techniquesbased on physical vapor deposition of nanoparticles. Examplesof such techniques include the Stransky–Krastanov growthby an epitaxial process, metal–organic vapor phase epitaxy(MOVPE), ion implantation in molecular beam epitaxy (MBE)grown films, inductively coupled plasma-enhanced chemicalvapor deposition (ICPECVD), etc [5–8]. However, mostof these techniques have severe constraints in terms ofnanoparticle size, material, location and the choice of substrate.A versatile nanofabrication technique that is capable ofproducing high purity nanoparticles with flexibility in terms ofparticle size, nanoparticle material and the choice of substrate,together with the capability for in situ deposition of ohmiccontact and isolation dielectric materials, will be an importantstep towards the realization of a variety of nanoparticle-basedelectronic and photonic devices.

0957-4484/07/445202+08$30.00 1 © 2007 IOP Publishing Ltd Printed in the UK

Nanotechnology 18 (2007) 445202 B Das and A Banerjee

We have developed an ultrahigh vacuum system for theimplementation of nanoparticle-based electronic and photonicdevices that address the above issues. With this system,nanoparticles of any metal, semiconductor or insulator canbe deposited with dimensions as low as 1 nm with less than5% size variation on any kind of substrate. The uniquenessof this system lies in the fact that, besides the nanoclustersource, it consists of an electron-beam evaporation systemas well as a pulsed DC sputtering unit installed in thesame system. Therefore, the deposited nanoparticles can beembedded within, or coated with, metallic, semiconducting orinsulating layers without breaking the vacuum. Thus multi-layer compound nanoscale structures can be created whichhave diverse applications in the fields of nanoscale detectors,nanooptics, nanosensors, field emitters, etc [9–12]. In addition,the system and the process are compatible with the siliconintegrated circuit (IC) fabrication technique and can be easilyintegrated into existing CMOS process lines, thus making thistechnique suitable for volume manufacturing.

2. Fabrication technique

A schematic diagram of the nanofabrication system, calledthe NanoSys1, is shown in figure 1. NanoSys1 consists of(i) a nanoparticle unit that provides the capability to depositnanoparticles of any metal, semiconductor or insulator ofdiameters as low as 1 nm with less than 5% size variationon an arbitrary substrate, (ii) a four-pocket electron-beamevaporation unit that allows the in situ deposition of fourdifferent materials (metals or insulators) with less than 5%thickness uniformity and (iii) a UHV pulsed DC sputter sourcethat allows the in situ deposition of thick layers of insulators(semiconductors or metals) including isolation dielectrics. InNanoSys1, the nanoparticle deposition unit, the ohmic contactmetallization unit and the isolation dielectric deposition unitare all housed inside an ultrahigh vacuum (10−10 Torr)chamber to ensure purity and good surface properties ofthe nanoparticles. A loadlock is used for transferring thesubstrate to and from the ultrahigh vacuum chamber. Thesubstrate is mounted on a rotating substrate holder that canbe heated up to 800 ◦C. An important objective in thedevelopment of NanoSys1 was the commercial viability of thenanoparticle-based devices. Since it is widely believed that, fornanoparticle-based devices to be commercially viable, at leastin the near future, the fabrication process has to be compatiblewith the silicon IC process, special attention was given to makeall processes in NanoSys1 silicon IC compatible. Electron-beam evaporation and pulsed DC sputtering are standardtechniques used by the silicon IC industry: however, selectionof the nanoparticle source required serious considerations. Thenanoparticle unit in the NanoSys1 is based on a nanoclustersource developed by Oxford Applied Research Inc. [13] andprovides the desired flexibility for nanoparticle deposition, andis described below.

The nanoparticle unit in NanoSys1 consists of ananocluster source and a quadrupole mass filter. Thenanocluster source consists of a DC sputtering unit, whichis used to sputter particles into an aggregation region wherethese particles form clusters, which are then channeled througha quadruple mass filter (QMF) to allow a pre-selected size

Figure 1. Schematic diagram of NanoSys1 with the ability to depositnanoparticles of any metal, semiconductor or insulator of diametersas low as 2 nm with less than 5% size variation on an arbitrarysubstrate, as well as for the implementation of complex multi-layeredstructures.

distribution of the nanoparticles. The sputtering source has theadvantage over all other types of sources in terms of the widecluster size range, which varies from a fraction of a nanometerto a few tens of nanometers. The variation of cluster size isdependent on three main parameters: (i) the length in whichthe clusters aggregate, (ii) sputtering power and (iii) the flow-rate of the aggregation gas. Another important feature of thenanocluster source is the presence of ionized clusters in theaggregation region, which is suitable to form highly adherentand uniform coatings even on insulating substrates by the so-called technique of energetic cluster impact. These ionizedclusters are electrostatically manipulated and filtered by theQMF, intercepted between the aggregation region and the maindeposition chamber. The QMF has been designed specificallyfor the purpose of high-resolution measurement and filteringof nanoclusters between 50 and 3 × 106 amu rather than onlyto detect the presence of elemental or low-mass compoundmaterials by currently available quadrupoles. Positive andnegative AC voltages are applied to the opposite pairs of polesof the QMF, respectively, and the cluster ions are selectedaccording to their charge-to-mass ratio. The parameters whichcan be varied to allow clusters of particular mass (or diameter)to pass through are: (1) amplitude of the AC voltage (V ),(2) frequency ( f ) and (3) the DC component of the appliedvoltage (U ). The ratio U/V , called the resolution, determinesthe size distribution of the cluster diameter transmitted throughthe filter. The higher the value of U/V , the narrower is the sizedistribution transmitted through the filter. Theoretically, theresolution can be better than 0.01%, but in reality the optimumresolution is determined by a number of other parametersincluding the mechanical construction (diameter, toleranceand length of poles) and variations in the initial cluster ionenergy. Therefore, the typical usable cluster size resolution

2

Nanotechnology 18 (2007) 445202 B Das and A Banerjee

Figure 2. Field emission scanning electron microscope (FESEM) images of silicon nanoparticles deposited on aluminum-coated siliconsubstrates with average diameters of (a) 10 nm, and (b) 15 nm. Aluminum grain boundaries are visible in the background. Inset of (b) showsthe size distribution of nanoparticles determined from FESEM micrographs.

is ∼2%–5%. Another important feature of this system is thepair of X , Y deflection plates present after the QMF. Neutralparticles present in the incoming flux, which are not affectedby the QMF, can be separated from the ionized clusters byapplying suitable voltages in the X , Y deflection plates, therebydeflecting the charged particles to the substrate mounted at anangle.

3. Experimental results

We have used the NanoSys1 to deposit nanoparticles of avariety of metals (Au, Cu and Al) and semiconductors (Si,CdS and CdSe) onto a variety of substrates including silicon,aluminum-coated silicon, SiO2-coated silicon, GaAs, glassand plastic substrates. Si nanoparticles are of particularinterest for a number of device applications includingnanoflash memory, vertical transistors and potentially silicon-based optical devices [14–17]. Nanoparticles of direct gapsemiconductors are of interest for optical applications andthat of metals for biomedical applications. We have alsosynthesized complex structures incorporating nanoparticles,such as stacked layers of nanoparticles separated by metalsor insulators, and nanoparticles embedded in metals orinsulators. Some of the results are presented below; thedepositions for the results shown below were carried outat room temperature. Figures 2(a) and (b) show the field-emission scanning electron microscope (FESEM) images ofSi nanoparticles with average diameters of 10 and 15 nmdeposited on an Al-coated Si substrate. From the figures,the size distributions of the nanoparticles are found to bevery uniform. The inset of figure 2(b) shows the particle-size distribution as determined from image analyses of FESEMmicrographs, assuming the particle area to be the projection ofa spherical particle. The column charts are the experimentaldata measured from figures 2(a) and (b). The data are wellapproximated by Gaussian distributions with peak diameters of9.99 nm (curve (a)) and 15.04 nm (curve (b)) for figures 2(a)and (b), respectively, which agree well with our pre-selecteddiameters during deposition. The pre-selected diameters ofthe nanoparticles were 10 and 15 nm, respectively; thus theexperimentally obtained values are in good agreement with the

pre-selected parameters with relative size variations as smallas 4.0% and 3.0%, respectively. This demonstrates the powerof the mass filter to obtain nanoparticles with highly accuratedimensions with precisely controlled size distributions. TheAr gas flow-rate during the deposition of the nanoparticlesshown in figure 2 were set to 70 sccm. As mentionedpreviously, the cluster size variation within the aggregationregion depends on three main parameters, such as effectivelength of the aggregation region, sputtering power and flow ofthe aggregation gas (Ar, in our case). In our experiments, wehave kept the sputtering power and aggregation length constantthroughout the process, but varied the Ar flow-rate from 70to 30 sccm to investigate the effect of variations of mass-flow-rate on the size distribution of deposited nanoparticles.Physically, as the sputtered particles agglomerate within theaggregation region to form clusters, it is expected that theshorter the time the nanoparticles spend within the aggregationregion, the lower will be the probability of the smaller particlesagglomerating to form larger clusters. Therefore, with differentmass flow-rates, the nanoparticles spend different times inthe aggregation region, leading to a potential variation inthe size distribution of the nanoparticles with variations inthe Ar flow-rate. These particles are then passed throughthe QMF, which allows only particles with a pre-selectedsize and size distribution to be deposited. To investigate theaccuracy and effectiveness of the QMF as a size selector, wehave deposited nanoparticles with the QMF in ON and OFFconditions, and compared the size distributions. Figure 3(a)shows the FESEM image of Si nanoparticles deposited onan Al-coated Si substrate at 70 sccm Ar flow-rate withthe QMF turned OFF. Figure 3(b) shows the cluster sizedistribution of the nanoclusters as obtained from figure 3(a).The size distribution is found to be quite large with 21.1% sizevariation. As this image is taken with the QMF OFF condition,therefore, physically, this size distribution is representativeof the nanocluster distribution present within the aggregationregion at those particular deposition parameters used for clusterdeposition.

A comparison of this data with those obtained withthe QMF turned ON (figure 2), under identical depositionconditions (similar sputtering power, aggregation length and

3

Nanotechnology 18 (2007) 445202 B Das and A Banerjee

Figure 3. (a) FESEM image of Si nanoparticles deposited on Al-coated Si substrate at 70 sccm Ar flow-rate, with QMF turned OFF, (b) sizedistribution of the deposited nanoclusters showing a large variation in the cluster size when the QMF is OFF.

Figure 4. FESEM images of Si nanoparticles deposited on Al-coated Si substrates at (a) 50 sccm and (b) 30 sccm Ar flow-rates. The QMFwas turned ON in both cases and set to 10.0 nm size selection. Insets of both the pictures show the corresponding size distribution obtainedfrom the respective images.

mass flow-rate of 70 sccm), shows that the size variationdrops down to as low as 3%–4% compared to the 21.1%obtained with the QMF turned OFF. This demonstrates theeffectiveness of the QMF in selecting nanoparticles with veryaccurate and narrow size distributions. We have also depositedSi nanoparticles with 50 and 30 sccm Ar flow-rates with theQMF in both OFF and ON conditions. Figures 4(a) and (b)show the FESEM images of as-deposited Si nanoparticles at50 and 30 sccm Ar flow-rates with the QMF ON and set to10.0 nm size selection. The insets show the correspondingsize distributions obtained from the respective images. Fromthe distribution curves, the experimentally obtained peak-size values are found to be 10.18 and 10.30 nm for 50and 30 sccm mass flow-rates with 7.0% and 3.2% relativesize variations, respectively. Corresponding distributions forQMF OFF conditions have also been determined. The sizevariations in these cases are found to be 18.6% and 24.1%for 50 and 30 sccm Ar flow-rates, respectively (images similarto figure 3(a)). These values again prove the effectivenessof the QMF to generate nanoparticles with very narrow sizedistributions. A relative comparison of nanoparticle sizedistributions for different Ar flow-rates and at QMF ON andOFF conditions are summarized in table 1.

To demonstrate the versatility of the Nanosys1, we havealso deposited nanoparticles of metals such as Cu, andnanoparticles of compound semiconductors such as CdS, atvarious gas-flow-rates with the QMF turned both ON andOFF. Figures 5(a) and (b) show the FESEM images of Cunanoparticles on Al-coated Si substrates at 70 and 30 sccmAr flow-rates. The QMF was turned ON in both cases with10.0 nm size selection. Figures 6(a) and (b) represent thecorresponding particle size distributions shown in figure 5.The shaded distributions are for depositions with the QMFON, whereas patterned curves are for depositions with theQMF OFF (images with QMF OFF are similar to the onein figure 3(a)). For an Ar flow-rate of 70 sccm, the sizevariation with the QMF-OFF is found to be 22.27% whereasthat with the QMF-ON is around 7.0%. The peak-size value inthis case is found to be 9.89 nm. Similarly, for an Ar flow-rate of 30 sccm, the size variations for QMF OFF and ONconditions are 27.29% and 5.0%, respectively. In addition,for 50 sccm gas flow-rate, these values are around 16.86% and6.0%, respectively.

Relative comparisons of the above data are provided intable 1. It may be observed from table 1 that, with decreasinggas flow-rate, the experimentally obtained peak cluster sizes

4

Nanotechnology 18 (2007) 445202 B Das and A Banerjee

Figure 5. FESEM images of Cu nanoparticles deposited on Al-coated Si substrates at (a) 70 sccm and (b) 30 sccm Ar flow-rates. QMF wasturned ON in both cases and set to 10.0 nm size selection.

Figure 6. Size distributions of Cu nanoparticles for (a) 70 sccm and (b) 30 sccm Ar flow-rates. The shaded charts are for the QMF ONcondition and patterned charts are for the QMF OFF condition.

Table 1. Average particle sizes and deviations of as-deposited Cu, Si and CdS nanoparticles for different QMF conditions and Ar flow-rates.

Gas (Ar) flowrate (sccm) 70 (sccm) 50 (sccm) 30 (sccm)

MaterialQMFcondition

Pre-selectedparticle sizeset in QMF(nm)

Experimentalpeak (average)size (nm)

Sizevariation(%)

Pre-selectedparticlesize set inQMF (nm)

Experimentalpeak (average)size (nm)

Sizevariation(%)

Pre-selectedparticle sizeset in QMF(nm)

Experimentalpeak (average)size (nm)

Sizevariation(%)

Cu QMFOFF

— 12.82 22.27 — 18.15 16.86 — 22.06 27.29

QMFON

10.0 9.89 7.0 10.0 9.86 6.0 10.0 9.98 5.0

Si QMFOFF

— 38.55 21.1 — 43.98 18.6 — 54.96 24.1

QMFON

10.0 9.99 4.0 10.0 10.18 7.0 10.0 10.30 3.215.0 15.04 3.0

CdS QMFON

15.0 17.14 12.0 15.0 17.62 14.0 15.0 18.15 6.3

increase with the QMF OFF, which is expected due to thelonger time the particles spend within the aggregation region,as explained earlier.

Figures 7(a) and (b) represent the deposition of CdSnanoparticles at 50 and 30 sccm Ar flow-rates with QMFturned ON and set to 15.0 nm size selection. In figure 7(a),

the substrate used is an Al-coated Si substrate, whereas infigure 7(b) it was bare Si substrate. Inset of figure 7(b)shows the size distributions of the as-deposited particles shownin figures 7(a) and (b). For 50 sccm gas flow-rates, theexperimental peak value is found to be 17.6 nm with a relativesize variation 14.0%. On the other hand, for 30 sccm Ar

5

Nanotechnology 18 (2007) 445202 B Das and A Banerjee

Figure 7. FESEM images of CdS nanoparticles at (a) 50 sccm Ar flow-rate, on Al-coated Si substrate and (b) 30 sccm Ar flow-rate on bare Sisubstrate. In both cases the QMF is set to 15 nm size selection.

flow-rates these values are 18.15 nm and 6.3%, respectively.Similarly, for CdS nanoparticles deposited at 70 sccm, thesevalues are 17.14 nm and 12.0%, respectively (images notshown). It should be noted here that a relative comparisonof size variations of Si, Cu and CdS nanoparticles with theQMF ON shows that the size variations for CdS nanoparticlesare larger than that for Cu and Si nanoparticles for similargas flow-rates. For example, at 50 sccm gas flow-rate, therelative size variations for Cu and Si nanoparticles are 6.0%and 7.0%, respectively, whereas for CdS nanoparticles thisvalue is 14.0%. This deviation is believed to be due to thepresence of a relatively large percentage of neutral particlesfor compound CdS material in the aggregation region incomparison to elemental Cu and Si nanoparticles; these neutralparticles are usually not affected by the QMF and are difficultto filter out. We are currently modifying NanoSys1 to mountthe substrate at an angle where the charged particles can bedeflected to, thereby eliminating the deposition of the neutralparticles. This is expected to significantly improve the particlesize distribution for compound semiconductor nanoparticles aswell.

We also carried out experiments for the calibration ofthe sputtering-gas-aggregation chamber of NanoSys1 underdifferent deposition conditions. In this section we report thesyntheses of Cu and Si nanoclusters in the gas-aggregationregion, with the size of the nanoclusters controlled byvarying the different sputtering conditions, such as sputteringpressure, gas-flow-rate, etc, and compare the results withtheoretical modeling [18–20]. The sputtering-aggregationprocess typically involves the vaporization of target materialsby magnetron sputtering followed by an inert gas condensationto form clusters of varying sizes. As mentioned earlier, thesize distributions of the clusters typically follow a normaldistribution and the peak cluster sizes of the distributionsdepend on several factors, which include gas-flow-rate, lengthof the growth region, deposition pressure, etc. To investigatethe dependence of the peak cluster size on the gas flow-rates,we deposited Cu and Si nanoparticles at different Ar flow-rates with the QMF turned OFF. The aggregation length andthe sputtering power were kept constant for these experiments.The size distributions of the deposited nanoclusters were then

Table 2. Theoretical and experimental values of mean cluster sizefor Si and Cu clusters at different Ar flow-rates.

Mean clustersize (nm)

MaterialAr flow-rate(sccm) Theoretical Experimental

Cu 30 26.0 22.0650 20.4 18.1570 11.8 12.82

Si 30 60.5 54.9650 47.0 43.9870 42.0 38.55

analyzed from the FESEM micrographs in a similar wayas described earlier. Physically, these size distributions arerepresentatives of the cluster distribution present within theaggregation region under the applied deposition conditions.These data are then matched with the existing model usedby Hihara and coauthors [18] and the system was calibratedaccordingly. Table 2 shows the theoretical mean cluster sizeobtained from the existing model and compared with theexperimentally determined peak cluster size from FESEMmicrographs (one of the images and corresponding distributionare shown in figure 3). The values are found to be ingood agreement with the theoretical values. The results aresignificant since they demonstrated that proper optimizationof operational conditions can lead to the desired cluster sizesas well as the desired cluster-size distributions. A detaileddiscussion on this process is described elsewhere [21].

Another important feature of the NanoSys1 is that thedensity of nanoparticles can be controlled by accuratelymonitoring the deposition time, as demonstrated in figure 8.The micrograph on the left shows silicon nanoparticles of highdensity while the one on the right shows the deposition of asingle nanoparticle on the substrate (indicated by the arrow).To investigate the quality of the deposited nanoparticles,high resolution transmission electron microscopy (HRTEM,TECNAI G2 S-TWIN) analyses were performed on the siliconnanoparticles. From the HRTEM data, it was observed thatapproximately half of the nanoparticles were single crystallinein nature, while the other half were polycrystalline.

6

Nanotechnology 18 (2007) 445202 B Das and A Banerjee

Figure 8. Control of nanoparticle density in NaniSys1 by controlling deposition time. The micrograph on the left shows a high density ofsilicon nanoparticles while the micrograph on right shows only one nanoparticle.

Figure 9. High resolution transmission electron micrographs of silicon nanoparticles deposited in NanoSys1: (a) single-crystalline siliconnanoparticle, the inset showing the corresponding Fourier transform; (b) polycrystalline nanoparticle. Approximately half of the nanoparticleswere found to be single crystalline in nature.

Considering that the nanoparticles in NanoSys1 areformed by the aggregation of smaller nanoparticles, it wassurprising to find such a large percentage of the nanoparticlesto be single crystalline. Figure 9(a) shows the HRTEMimage of one of the single-crystalline Si nanoparticle withan average diameter of 15 nm, with corresponding Fouriertransform in the inset. The image demonstrates the single-crystalline nature of the nanoparticle, confirming the highquality of the nanocrystals that can be deposited by NanoSys1.The highly oriented atomic planes are clearly visible and theanalysis of the micrograph depicts the (111) lattice orientationof the nanoparticles. A closer look into the HRTEM imagereveals the presence of some point defects which may beattributed to the sample preparation procedure by the ion-milling process for HRTEM imaging. Figure 9(b) shows theHRTEM image of a polycrystalline silicon nanoparticle withan average diameter of about 15 nm. The polycrystalline natureof the silicon nanoparticle is clearly evident in the image. Itmay be pointed out that the nanoparticles shown above wereall deposited at room temperature, and it is expected thatdeposition at elevated temperatures or a post-annealing stepcan significantly increase the percentage of single-crystalline

nanoparticles over polycrystalline ones. It is interesting to notehere the effects of some of the deposition parameters, such asthe sputtering power, deposition pressure, gas flow-rates, etc,on the single-crystalline nature of the deposited nanoparticles.The Nanosys1 is a sputtering-gas-aggregation type depositionsystem, where most of the cluster formations occur within theaggregation region. As a result, the effect of the sputteringpower on the crystallinity of the nanoparticles is not expectedto be significant. On the other hand, the mass flow-rateand the length of the aggregation region are parameters thatcan potentially affect the crystallinity of the nanoparticles.Qualitatively, the longer the nanoparticles stay within theaggregation region, the higher is the probability of the smallerparticles agglomerating into clusters, thus increasing theprobability of the nanoparticles to be polycrystalline. Thus,if the length of the aggregation region is increased, or thegas flow-rate during deposition is decreased, the particlesstay longer in the aggregation region, leading to a potentialincrease in the percentage of polycrystalline particles. In ourexperiments, we have kept the aggregation length constant inall the experiments and varied the gas flow-rates. The HRTEMimages shown in figures 9(a) and (b) are for Ar flow-rate

7

Nanotechnology 18 (2007) 445202 B Das and A Banerjee

Figure 10. Cross-sectional FESEM image of a multi-layeredstructure incorporating Si nanoparticles of diameters 11, 8 and 5 nmseparated by 100 nm layers of Ni. Inset shows the correspondingschematic diagram of the structure.

of 70 sccm, in which case as mentioned earlier, 50% of thenanoparticles were found to be single crystalline in nature.We have also performed HRTEM imaging of nanoparticlesdeposited at 50 and 30 sccm gas flow-rates and found that,on average, 35% and 20% of the deposited nanoparticlesare respectively single crystalline in nature, which agreesqualitatively with the argument stated above.

One of the major strengths of the NanoSys1 is its ability tocreate complex layered structures incorporating nanoparticles.To demonstrate this capability, we have created layeredstructures of nanoparticles of different dimensions separatedby both metals and insulators. In one case, we fabricated threelayers of Si nanoparticles separated by 100 nm layers of Ni.The nanoparticles in the layers have average diameters of 11,8 and 5 nm, respectively; a thin layer of Ni was also depositedon the top to prevent oxidation. Cross-sectional FESEMimage of the multi-layered structure is shown in figure 10and the corresponding schematic diagram of the structure isshown in the inset. The three layers are clearly visible inthe image separated by Ni films. In order to observe thenanoparticles in the cross-sectional image, we had depositeda large density of nanoparticles, some of which have formedbigger clusters as seen in the image. In another case, we havecreated a layered structure containing stacked layers of CdSnanoparticles separated by thin layers of aluminum oxide. Fivelayers of CdS nanoparticles with diameters ranging from 5 to15 nm were deposited separated by 100 nm of aluminum oxide,which was deposited by electron beam evaporation. Such astructure has important applications for light emission devices,

the layered structure providing higher light emission intensitywithout the risk of agglomeration of the nanoparticles. Inaddition, we have recently initiated some experiments wherethe nanoparticles and an insulator layer are co-deposited tocreate coated layers of nanoparticles.

4. Summary

In summary, we have developed a versatile technique forthe fabrication of electronic and photonic devices andsystems incorporating nonlithographic nanoparticles in anultrahigh vacuum environment. The technique providesexcellent engineering control over nanoparticle depositionas well as allowing the in situ fabrication of additionalcomponents needed for device/system implementation. Wehave demonstrated the versatility and the power of the systemby depositing nanoparticles of a variety of materials anddimensions on a variety of substrates, as well as the depositionof multi-layered structures incorporating nanoparticles. Weare currently utilizing this technique to implement specificelectronic and photonic devices and systems, the results ofwhich will be published in the future.

References

[1] Hoeneisen B and Mead C 1972 Solid State Electron. 15 819[2] Chou S Y, Krauss P R and Long L 1996 J. Appl. Phys. 79 6101[3] Prodan E, Radloff C, Halas N J and Nordlander P 2003 Science

302 419[4] Arai T, Gritschneder S, Troger L and Reichling M 2004

Nanotechnology 15 1302[5] Park N M, Kim T S and Park S-Ju 2001 Appl. Phys. Lett.

78 2575[6] Cui Y, Wei Q, Park H and Lieber C M 2001 Science 293 1289[7] Lu C H, Lin Y and Wang H-C 2003 J. Mater. Sci. Lett. 22 615[8] Das B and McGinnis S P 2003 Appl. Phys. Lett. 83 2904[9] Krause B, Durr A C, Ritley K, Schreiber F, Dosch H and

Smilgies D 2002 Phys. Rev. B 66 235404[10] Ruffenach S, Maleyre B, Briot O and Gil B 2005 Phys. Status

Solidi c 2 826[11] Ishikawa Y, Shibata N and Fukatsu S 1997 J. Cryst. Growth

175/176 493[12] Bapat A, Perrey C R, Campbell S A, Carter C B and

Kortshagen U 2003 J. Appl. Phys. 94 1969[13] http://www.oaresearch.co.uk/[14] Tiwary S, Rana F, Hanafi H, Hartstein A and Crabbe E F 1996

Appl. Phys. Lett. 68 1377[15] Canham L 2000 Nature 408 411[16] Nishiguchi K and Oda S 2000 J. Appl. Phys. 88 4186[17] Wang Y G, Wu H Y and Dravid V P 2005 J. Mater. Sci.

40 1725[18] Hihara T and Sumiyama K 1998 J. Appl. Phys. 84 5270–6[19] Bromann K, Felix C, Brune H, Harbich W, Monot R,

Buttet J and Kern K 1996 Science 274 956–8[20] Pratontep S, Carroll S J, Xirouchaki C, Streun M and

Palmer R E 2005 Rev. Sci. Instrum. 76 045103[21] Banerjee A, Krishna R and Das B 2007 Appl. Phys. A at press

8