• No results found

Structural and optical characterization of Si/Ge quantum dots

N/A
N/A
Protected

Academic year: 2021

Share "Structural and optical characterization of Si/Ge quantum dots"

Copied!
51
0
0

Loading.... (view fulltext now)

Full text

(1)

Department of Physics, Chemistry and Biology

Master’s Thesis

Structural and optical characterization

of Si/Ge quantum dots

Dan Wigblad

LITH-IFM-A-EX–08/1916–SE

Department of Physics, Chemistry and Biology Linköpings universitet

(2)
(3)

Master’s Thesis LITH-IFM-A-EX–08/1916–SE

Structural and optical characterization

of Si/Ge quantum dots

Dan Wigblad

Supervisor: Stanley Wissmar

Acreo AB

Phd. Henry H. Radamson

KTH

Examiner: Prof. Per-Olof Holtz

ifm, Linköpings universitet

(4)
(5)

Avdelning, Institution

Division, Department Material physics

Department of Physics, Chemistry and Biology Linköpings universitet

SE-581 83 Linköping, Sweden

Datum Date 2008-17-003 Språk Language  Svenska/Swedish  Engelska/English   Rapporttyp Report category  Licentiatavhandling  Examensarbete  C-uppsats  D-uppsats  Övrig rapport  

URL för elektronisk version

http://urn.kb.se/resolve?urn:nbn:se:liu:diva-11672

ISBN

ISRN

LITH-IFM-A-EX–08/1916–SE

Serietitel och serienummer

Title of series, numbering

ISSN

Titel

Title

Structural and optical characterization of Si/Ge quantum dots

Structural and optical characterization of Si/Ge quantum dots

Författare

Author

Dan Wigblad

Sammanfattning

Abstract

In this study silicon-germanium quantum dots grown on silicon have been in-vestigated. The aim of the work was to find quantum dots suitable for use as a thermistor material. The quantum dots were produced at KTH, Stockholm, using a RPCVD reactor that is designed for industrial production.

The techniques used to study the quantum dots were: HRSEM, AFM, HRXRD, FTPL, and Raman spectroscopy. Quantum dots have been produced in single and multilayer structures.

As a result of this work a multilayer structure with 5 layers of quantum dots was produced with a theoretical temperature coefficient of resistance of 4.1 %/K.

Nyckelord

(6)
(7)

Abstract

In this study silicon-germanium quantum dots grown on silicon have been investigated. The aim of the work was to find quantum dots suitable for use as a thermistor material. The quantum dots were produced at KTH, Stockholm, using a RPCVD reactor that is designed for industrial production.

The techniques used to study the quantum dots were: HRSEM, AFM, HRXRD, FTPL, and Raman spectroscopy. Quantum dots have been produced in single and multilayer structures.

As a result of this work a multilayer structure with 5 layers of quantum dots was produced with a theoretical temperature coefficient of resistance of 4.1 %/K.

(8)
(9)

Acknowledgements

I would like to thank everyone who has suported me and helped me during this work. My supervisors Stanley Wissmar at Acreo and Henry Radamson at KTH have been very supportive and I have enjoyed working with them.

A special thanks to Jonathan Askesjö and Sollentuna Budoklubb for arranging a place for me to live in Stockholm, and also for the good training after work.

A big thanks to Acreo for giving me the chance to do this thesis work in a very exciting area of research. The work has also given me the opportunity to work with several advanced characterization methods.

(10)
(11)

Contents

1 Abbreviations 1 2 Introduction 3 2.1 Infrared imaging . . . 3 2.2 Bolometers . . . 4 2.3 Computer simulations . . . 5

2.4 Aim and purpose . . . 5

3 Semiconductors 7 3.1 Strain . . . 7

3.2 Low dimensional structures . . . 8

3.3 Growth process . . . 9

3.4 Intermixing . . . 9

3.5 Multilayer and capping . . . 10

3.6 QD morphology . . . 11

3.7 Growth process . . . 11

3.7.1 Chemical vapor deposition . . . 11

3.7.2 Molecular beam epitaxy . . . 12

4 Characterization methods 13 4.1 High resolution scanning electron microscopy . . . 13

4.2 Atomic force microscopy . . . 14

4.3 High resolution X-ray diffraction . . . 15

4.4 Fourier transform photoluminescence . . . 15

4.5 Raman spectroscopy . . . 17 5 Experimental details 19 5.1 Single layer . . . 19 5.2 Multilayer structures . . . 21 6 Results 23 6.1 Single layers . . . 23 6.1.1 Time . . . 23 6.1.2 Gas-flow . . . 24 6.1.3 Temperature . . . 25 ix

(12)

x Contents

6.1.4 Summary single layers . . . 25

6.2 Anneal and standby . . . 27

6.3 Silicon Capping . . . 27

6.4 Multilayer structures . . . 28

6.4.1 X-ray diffraction . . . 28

6.4.2 Photoluminescence . . . 29

6.4.3 Raman spectroscopy . . . 30

6.4.4 Summary multilayer structures . . . 31

7 Conclusions 33

8 Further work 35

Bibliography 37

(13)

Chapter 1

Abbreviations

AFM Atomic force microscopy CCD Charge-coupled device

CMOS Complementary metal-oxide semiconductor CVD Chemical vapor deposition

FIR Far infrared

FTPL Fourier transform photoluminescence Ge Germanium

HRSEM High resolution scanning electron microscopy HRXRD High resolution x-ray diffraction

KTH Royal institute of technology LWIR Long wave infrared

MBE Molecular beam epitaxy

MEMS Microelectromechanical systems MWIR medium wave infrared

NIR Near infrared PL Photoluminescence ROIC Readout integrated circuit

RPCVD Reduced pressure chemical vapor deposition Si Silicon

TCR Temperature coefficient of resistance QD Quantum dot

QW Quantum well WL Wetting layer

(14)
(15)

Chapter 2

Introduction

Silicon-germanium structures are very appealing to produce due to their low man-ufacturing costs and simple integration with existing silicon-based electronics. By creating quantum dots (QDs) new areas of use are possible. Most of the research on silicon-germanium QDs today is focused on creating photonic detectors in the infrared spectrum. The photonic detectors generally need to be cooled down to about 70K. However the structures investigated in this report have a novel use:

as thermistor materials used in a room-temperature operated infrared imaging system.

Computer simulations [17] have shown that large quantum dots are preferential for the properties of QD thermistors. To be able to create thermistors of good quality as well as using QDs for other applications, the manufacturing process has to be thoroughly investigated. An understanding of how to control the quantum dots size, density and composition is essential. A general knowledge of how the QDs are formed and evolve can be achieved by reading articles published on the subject, but for the more precise knowledge of how manufacturing parameters influence the creation of QDs in the KTH lab, experiments are needed. The equipment used to manufacture germanium quantum dots is a RP-CVD reactor of a type that can be found at industries worldwide. The purpose of this work is to produce a QD structure suitable as a silicon-germanium thermistor material. Optical and structural properties of Si-Ge structures have been investigated in single layers of quantum dots (section 6.1) as well as multilayer structures of QDs (section 6.4).

2.1

Infrared imaging

IR detectors are of great interest since objects in our surroundings have thermal radiation in the infra-red spectra. An object’s thermal radiation depends on two things: its temperature and its emission properties. This will give objects of different temperatures a difference in thermal radiation. Also objects of the same temperature will radiate differently if their emission properties are different. The thermal radiation is independent of visible light, the detection of thermal radiation

(16)

4 Introduction

is therefore widely used in night vision systems. Numerous other uses have also been commercialized including thermal imaging to study fires. The infrared region is also used in telecommunication applications, the wavelengths used for telecom applications are 1.33 µm and 1.5 µm.

The infrared spectrum is divided into four regions of interest: • Far infrared (FIR)

• Long wave infrared (LWIR) • Medium wave infrared (MWIR) • Near infrared (NIR)

Objects in room temperature emit light in the MWIR region, specifically with a wavelength close to 10 µm. The emission of light based on temperature is referred to as thermionic radiation.

Infrared detectors can generally be divided into two groups: photonic and thermionic. The photonic detectors are often constructed to give a signal from photonic excitation of charge carriers, this is a very fast process and the response-time of the system is in the order of microseconds. The main draw-back of the photonic detectors is their need of a cooling system in order to reduce their dark-current. The operating temperature for photonic detectors is usually around 70K. Thermionic detectors are generally a lot slower than their photonic counterparts, but they can be produced much cheaper and in smaller sizes (no need for a cooling unit). The Si/Ge thermistor material studied in this work is intended for use in a bolometer. Bolometers use a thermionic principle of detecting IR radiation. Section 2.2 provides a more detailed description of bolometers and thermionic detection.

2.2

Bolometers

Bolometers use a two step process to detect thermal radiation. An absorbing film is heated by thermal radiation and the temperature change in the film is measured by a thermistor. The thermistor changes its resistance with temperature, the size of change in resistance is denoted as the temperature coefficient of resistance (TCR) in units of %/K.

The detectors are arranged in an array, where each bolometer represents a pixel. The ambient temperature of the operating environment is measured by us-ing a reference bolometer not subjected to the thermal radiation. The thermistor is contacted to the surrounding electronics by long and thin legs. The bolometers have a read-out integrated circuit (ROIC) typically placed underneath the ther-mistor material. The legs support the therther-mistor and create a cavity to thermally insulate the ROIC from the thermistor.

One of the disadvantages of bolometers compared to photonic detectors is the slow response time. The thermal response time is defined as the time it takes for the bolometer to return to 1

(17)

2.3 Computer simulations 5

temperature when the thermal radiation is removed. The thermal response time can be calculated as τt= CG where G is the sum of all thermal conductance from

all sources that remove heat and C is the heat capacity of the bolometer.

By designing a bolometer with thin and long legs the influence of heating from the ROIC is decreased, but the factor G is decreased and the thermal response time is increased. To calibrate the sensor array against self-heating and ambient temperature a reference bolometer without exposure to thermal radiation is used. The absorbing film and thermistor produced separately from the electronics and is bonded to a CMOS read-out integrated circuit (ROIC). The final bolometer structure is formed by the use of MEMS processing.

Figure 2.1: A bolometer with a sensing membrane and thin supporting legs.

2.3

Computer simulations

Quantum well thermistors have already been produced and the motivation for pro-ducing QDs is from previous computer simulations in a master thesis by Berntsen [17]. The result from the computer simulations are the energy levels of the valence band in Ge QDs with silicon barriers. The energy levels are used to calculate the materials temperature coefficient of resistance (TCR). An approximation of the TCR is given by the difference in energy between silicon valence band and QD valence band. The simulated energy levels are then used in the equation:

β= 1 kBT2  |EV(QD) − EV(Si)| + kBT 2 nsh2 8πm∗  (2.1) The result from the computer simulations show that large QDs give higher TCR (figure 2.2). The effects of silicon intermixing on the energy levels in the QDs are discussed in section 3.4. Simulated TCR for Ge QDs with diameter of 100 nm is close to 8.5%/K. Comparing the theoretical results from the simulations with measured TCR of current SiGe quantum well structures (TCR < 3.5%/K) show a potential for QDs to outperform QWs as a thermistor material.

2.4

Aim and purpose

The main purpose of this study is to produce a multilayer structure of QDs suitable for use as a thermistor material. The advantages of a thermistor material in silicon

(18)

6 Introduction

Figure 2.2: Computer simulations for Ge QDs [17]

and germanium are its low manufacturing cost and that it is easily integrated with silicon-based CMOS technology. Secondly this is a pre-study for future work with SiGe QDs at Acreo and KTH.

For ideal function as a thermistor material the following working aims were set for the QDs:

• High uniformity • High density

• High germanium content

• Diameter ≈ 100 nm (from simulations (figure 2.2) A multilayer structure of QDs should be produced.

(19)

Chapter 3

Semiconductors

In this work only quantum dots of SiGe are discussed. When a QD is mentioned it is considered a Si1−xGex self-assembled quantum dot on silicon substrate.

Semiconductors are materials which are characterized by their band structure, the separation and positions of the valence and conduction bands. The crystal structure of semiconductors is a periodic and ordered formation of the atoms within the crystal. Silicon and germanium are both diamond structured and their lattice constants are aSi= 5.43 Å and aGe= 5.64 Å. The lowest energy transition

over the band gap is phonon assisted for both silicon and germanium. Figure 3.1 shows a simplified image, without compensation of strain effects, of the valence and conduction band line-up for a heterojunction between silicon and germanium. Both silicon and germanium have indirect bandgaps.

Figure 3.1: Band structure for SiGe heterojunction

3.1

Strain

Between the bulk silicon lattice and the germanium lattice there is a 4.2 % lattice mismatch (Ge has the larger distance between its atoms). The germanium lattice will be compressed in the growth plane and elongated in the growth direction (figure 3.2). The strain is defined as:

= astrained− arelaxed arelaxed

(3.1) 7

(20)

8 Semiconductors

For germanium on silicon substrate the strain in the growth plane is negative (compressive). In thin films the layers can be strained without defects but for thicker films defects will relax the strained lattice. Examples of defects are vacan-cies (missing atoms) and threading dislocations which will spread in the growth direction through the lattice. To avoid defects in the lattice is important to the optical and electrical properties of the semiconductor. In the SiGe system strain will raise the energy levels in the valence band.

Figure 3.2: Compressive strain

3.2

Low dimensional structures

By creating low dimensional structures charge carriers within the semiconductor can be confined within the structure. A structure that confines carriers movement to two directions is the quantum well (figure 3.3a). Quantum wires confine the charge carriers movement to one dimension (figure 3.3b). Quantum dots (figure 3.3c) are considered a zero dimensional system and the charge carriers are trapped within the dot. The density of states within a quantum dot as a function of energy can be described by a Dirac function.

Figure 3.3: Low dimensional structures. (a) quantum well. (b) quantum wire. (c) quantum dot.

An advantage of creating a low dimensional structure is that it is possible to create wells, wires and dots without dislocations. The bandstructure of the semiconductor can be altered by modifications of the size of the QW or QD. One of the advantages of using QWs or QDs as thermistor material is that they can

(21)

3.3 Growth process 9

be designed to be dislocation free and do not have the grain boundaries found in amorphous silicon, this will reduce the thermistors noise.

3.3

Growth process

Self assembled quantum dots (QDs) are also referred to as islands or clusters. The nucleation of dislocation-free quantum dots following the Stranski-Krastanov growth mode can be divided into three steps:

• First a germanium layer is formed on top of the silicon surface. This layer is referred to as the wetting layer. When the wetting layer has reached a critical thickness (a few mono-layers) the initial nucleation of quantum dots takes place. The dots grow in size by capturing deposited ad-atoms. The wetting layer is highly strained and without dislocations. The time of nucleation is denoted tN.

• When the initial quantum dots reach a critical radius, early stage growth takes place. Nucleation of new dots has ceased and the dots grow by captur-ing adatoms and by Ostwald ripencaptur-ing (large dots grow by capturcaptur-ing atoms from smaller dots). Dots with a radius above the critical radius are favored by the growth process, resulting in a bimodal size distribution. The time when a critical radius is reached is further on referred to as tC and the QDs

with a radius less or equal to the critical radius as type 1 QDs. QDs formed at during early stage growth are in this work referred to as type 2 QDs. • The late stage growth is characterized by Ostwald ripening and coalescence

(Dots growing into each other). Coalescence is the dominating process. At this stage, the dot density has started to decrease as a result of coalescence and Ostwald ripening.

During the growth process two stages may at times co-exist. Figure 3.4 shows the evolution of the QDs during growth. Figure 3.4b shows the initially nucleated QDs and the transition to early stage growth, one QD in the image has started to grow beyond the critical size. Figure 3.4c shows a state further into early stage growth. Some coalescence and Ostwald ripening can also be seen in Figure 3.4c. Figure 3.4d is an example of where late stage growth is dominating, the number of QDs are rapidly decreasing due to Ostwald ripening and coalescence.

3.4

Intermixing

Intermixing occurs at two stages when creating multilayer structures, at QD growth and at silicon capping. When germanium is deposited on the surface some silicon from the substrate is mixed into the QDs. During capping the intermixing occur at the top of the QDs. The intermixing during growth and capping will create a composition gradient within the QDs, typically the QDs have a germa-nium rich core and the germagerma-nium content decreases toward the edges of the QDs.

(22)

10 Semiconductors

Figure 3.4: SEM micrographs of the structural evolution of QDs grown at 600C.

The temperatures of growth and capping are very important for the germanium content, higher temperatures increase the silicon intermixing.

The lattice constant of a SiGe alloy can be approximated by Vegard’s law: aSiGe= XaGe(1 − X)aSiGe (3.2)

The intermixing will affect the strain in the structure since the lattice constant of SiGe alloys is closer to Si than for bulk Ge. Silicon intermixing will lower the valence band energy in the SiGe alloy.

3.5

Multilayer and capping

The capping of QDs results in increased silicon intermixing and restructuring of the QDs. Capped QDs have a lowered height to width ratio. Their height is decreased and their diameter has been increased. Under some conditions dome shaped QDs may undergo transformation to pyramid shape after capping.

When creating multilayer structures of QDs, the thickness of the silicon spacing layer is important. When thin capping layers are used there will be some layer to layer influence of the QDs. The layer to layer influence results in a vertical alignment of the QDs. If no layer to layer influence exists, the QDs will nucleate randomly on the surface of the silicon capping layer. Another effect of the layer to layer influence is a reduction of the critical thickness of the wetting layer. The reduced thickness results in faster nucleation of QDs and with the same growth conditions the QDs will change in size between each layer. The thickness of the wetting layer decreases for each consecutive layer and stabilizes at a constant thickness. For QDs with a diameter of 100 nm (before capping). Than et. al. [7] show a decreasing wetting layer for the first 5 QD periods and a stabilization for the consecutive layers. They also show that the vertical alignment ceases and that the thickness of the wetting layer of the second layer is the same as for the first layer when the QDs are capped by 160 nm silicon.

(23)

3.6 QD morphology 11

3.6

QD morphology

As the structure of quantum dots evolves during growth, four configurations can be distinguished, hut clusters, pyramids, domes and superdomes. Hut clusters are small quantum dots, typically around 30-50 nm in diameter and they have a rectangular base and pyramidal shape with a height of about 1-2 nm. Nguyen-Duc et al.[13] consider these hut clusters to be a meta-stable phase between layer by layer growth and three-dimensional growth. Nguyen-Duc et al. also report hut cluster growth to take place at temperatures lower than 550C in a chemical

vapor deposition system (CVD). Pyramids are distinguished by their square base and pyramidal shape and the height and diameter vary depending on growth conditions. The height is typically higher than that for hut clusters but lower than domes. A typical pyramid height is about 10 nm. Domes have an increased volume and are higher than pyramids. They are also multifaceted. The increased volume compared to pyramids can be explained by the dome’s height and higher angles between sides and the base plane. Superdomes are formed during late stage growth by Ostwald ripening and coalescence. The superdomes are strain-relaxed by misfit dislocations at the base.

The formation of domes and pyramids depend on the growth conditions. Pyra-mids and domes often co-exist, resulting in a bimodal dot distribution (and some-times bimodal size distribution). A bimodal dot distribution is not the same as a bimodal size distribution, the former indicates two types of quantum dot struc-tures and the latter the existence of two distinguishable sizes of quantum dots. As the germanium coverage increases, the pyramids undergo transition into domes.

3.7

Growth process

Mainly two methods are used to produce self-assembled quantum dots, chemical vapor deposition (CVD) and molecular beam epitaxy (MBE). When capping QDs using CVD, the surface tends to be rougher and follow the shape of the QDs compared to the same capping thickness using MBE [15]. Another difference between the resulting QDs from CVD and MBE is that at low temperatures the silicon intermixing is slightly lower for CVD than MBE [15]. In this work a reduced pressure CVD equipment was used.

3.7.1

Chemical vapor deposition

The material source in the CVD system is provided by gases. For deposition of germanium, germane (GeH4) gas is used and for the silicon coverage silane (SiH4)

is a common source. A chemical reaction takes place at the heated substrate. When depositing silicon the reaction is SiH4 → Si+ 2H2 and for germanium,

GeH4→ Ge+ 2H2. By using a mixture of the silane and germane, Si1−xGexcan

be deposited. The chemical reaction is temperature dependent and the growth rate increases with temperature. The growth rate in the CVD system is controlled by substrate temperature and gas-flow.

(24)

12 Semiconductors

3.7.2

Molecular beam epitaxy

The deposition atoms are evaporated from a solid-source. By heating the substrate the evaporated atoms condensate and form a thin film. The growth is controlled by atomic flux and substrate temperature. The growth rate using MBE is nearly constant at all temperatures since no chemical reactions take place on the surface of the substrate. The system is operated under ultra high vacuum to avoid collisions of atoms.

(25)

Chapter 4

Characterization methods

4.1

High resolution scanning electron microscopy

A high resolution scanning electron microscope (HRSEM) of model Zeiss Ultra 55 was used to create micrographs of single layer QDs. With the HRSEM equipment an image is generated by scanning an electron beam over the sample’s surface and using a detector to measure the number of scattered electrons. The electrons originate from a filament with a fine tip. The filament is situated in a chamber with ultra high vacuum. The energy of the emitted electron is controlled by an applied electric field (acceleration voltage). The Zeiss Ultra 55 can use accelerating voltage in the range 0.1 to 30 kV. In this work a typical accelerating voltage of 2 kV has been used. The electron beam is focused on the sample using objective and condenser lenses. An aperture is used to reduce aberrations.

Two types of scattered electrons are detected, backscattered electrons and sec-ondary electrons. Backscattered electrons are inelastically scattered by the atoms of the sample beneath the electron beam. The number of backscattered electrons is dependent on the atomic number of the sample and the number of incident electrons. Secondary electrons originate from the sample and are generated by the incident electrons within an interaction volume of the sample. The number of secondary electrons depend on the size and shape of the interaction volume. The interaction volume is related to the energy of incident electrons, material prop-erties of the sample and the angle of the incident electrons. The contrast of the image is due to differences in yield of detected electrons as the electron beam is scanned over the surface. The electrons detected consist of both secondary and backscattered electrons. An electrical field can be applied in front of the detec-tor to deflect low energy electrons, this is used to remove the secondary electron component of the signal.

To achieve good resolution, a high grade filament is required, the tip of the filament is manufactured to create as narrow an electron beam as possible.

HRSEM uses a high vacuum in the sample chamber to reduce absorption of electrons, and can use very short working distance (around 2 mm was used in this work). The short working distance and the high vacuum reduce the need for a

(26)

14 Characterization methods

Figure 4.1: AFM. Detection of cantilever deflection.

high excitation energy, which will be beneficial if the sample is not conductive. The requirement of the sample to be studied with SEM is that it is solid. To avoid charging of the surface gold is often deposited on isolators and semiconductors, but with the relatively low excitation energy used, this was not necessary for for the samples studied in this work.

The resolution of the Zeiss Ultra 55 with an accelerating voltage of 2 kV is 1.5 nm. The HRSEM was used to obtain information about QD density, diameter and size distribution.

4.2

Atomic force microscopy

Atomic force microscopy (AFM) operates by using a small tip attached to a can-tilever. As the tip approaches the surface of the sample it will experience short-range atomic forces. The position of the tip is determined by reflecting a laser beam off the cantilever on to a CCD detector. This setup will allow for a very precise detection of the tip’s position, figure 4.1. The relative heights of structures on the sample’s surface are detected as the tip scans the surface, and by using software 3D images of the surface can be created. When the tip is approaching the surface it will first experience attracting atomic forces and as it is moved closer to the atoms of the surface it experiencez repulsing forces. Piezo-electric crystals are used to move the tip over the scanning area.

One technique of scanning the surface of the sample is to use the contact mode, the tip is close enough to the surface to experience repulsing atomic forces and is moved over the scanning area to give a topographic image of the surface. This technique provides very high resolution but may damage the sample. The tip-to-surface forces change the resonant frequency of the cantilever, the oscillation amplitude and the phase of the oscillations. Using AFM in tapping mode is a non-invasive technique and does not damage the sample. No sample preparation is required for the atomic force microscopy.

AFM in tapping mode has been used to create topographic images of the samples. Single layers have been investigated by AFM to obtain height information and the surface of capped layers has been studied.

(27)

4.3 High resolution X-ray diffraction 15

Figure 4.2: Incident and diffracted angle for the HRXRD setup

4.3

High resolution X-ray diffraction

A high resolution x-ray diffraction (HRXRD) equipment was used to find informa-tion about the periodicity of the QD structure of multilayers and the germanium content in the QDs.

Bragg’s law states that constructive interference occur when 2dsinθ = nλ. In HRXRD x-rays with λ = 0.1504 nm are used and reflections occur at lattice planes. The constructive interference can be measured by scanning the diffraction angle using an ω/2Θ scan. ω is the incident angle of the x-ray beam and θ the diffracted angle (figure 4.2). The result will be intensity peaks at angles corresponding to the lattice distance and from diffraction between layers of different materials.

Differentiation of Bragg’s law gives: ∆d2sinθ + 2dcosθ∆θ = 0 ⇒ ∆dk1(θ) +

∆θk2(θ) = 0 where k1(θ) and k2(θ) in this setup are positive for all θ.

Compres-sively strained SiGe (QDs) will have an elongated lattice in the growth direction compared to bulk silicon (section 3.1) which means that ∆d > 0 ⇒ ∆θ < 0. This positions the peak from the QDs at lower diffraction angle (θ) than the silicon substrate peak.

Ideally the measured peaks are very narrow and correspond to a single distance in the lattice. The width of the peaks is an indication of the quality of the inter-face between two materials, an intermixing gradient will broaden the peak for that layer. Some troubles appear when studying QDs, intermixing and strain affect the distance between crystal planes is the lattice and the QDs are typically not uni-formly strained or composed. These effects for QDs will contribute to broadening of the QD peak. If the multilayer structure contains several layers of QDs with varying compositions, this will also contribute to a broadening of the peak. The measured composition based on the QD peak position should be considered an average composition of the QDs.

4.4

Fourier transform photoluminescence

To determine the bandgap of the QD structure Fourier transform photolumines-cence (FTPL) was used on samples from batch F. The samples were placed in a cryostat and excited by a laser with an energy higher than the semiconductors bandgap. The luminescence originating from the recombination over the bandgap

(28)

16 Characterization methods

Figure 4.3: Recombinations of electrons for Si/Ge QD structure.

Figure 4.4: Schematic drawing of the FTPL setup.

was studied using a Michelson interferometer. The two detected recombinations are illustrated in figure 4.3 and correspond to spatially direct and indirect transi-tions [9]. An argon laser was used to excite the charge carriers of the samples. A germanium detector and a CaF2 beam splitter were used. With the CaF2 beam

splitter the spectrum from 680 to 1050 mEv was measured by 100 scans.

In the Michelson interferometer the incident beam is divided in two parts by using a semi-transparent beam splitter. The transmitted part of the beam is reflected on a stationary mirror and the reflected part is reflected by a moving mirror which creates an optical path difference between the two beams. The two beams will give rise to constructive and destructive interference due to the optical path difference. This will create a modulated intensity variation of the beam. The modulated beam is recognized as the Fourier transform of the spectral distribution. In figure 4.4 a schematic illustration of the FTPL setup is shown.

To study both the recombinations EA and EB the samples were placed in a

cryostat and the photoluminescence was measured at temperatures down to 36K. At low temperatures the spatially indirect transition EBis dominant and the peak

(29)

4.5 Raman spectroscopy 17

4.5

Raman spectroscopy

Raman spectroscopy is used on silicon-germanium alloys and structures to deter-mine their germanium content and strain. The basic principle of Raman spec-troscopy is to measure the change in energy of photons that are scattered by the sample. Photons are provided by a laser with known wavelength, the incident photons are then inelastically scattered by the sample. The change in energy of the scattered photons is due to either creation or absorption of vibrational energy. When part of the incident photons energy is absorbed and vibrational energy is created it is referred to as stokes scattering. If the photons are scattered with higher energy than the incident photons it is called anti-stokes scattering. The energy shift of incident and scattered photons is referred to as the Raman shift. The size of the Raman shift is given in units of wavenumber (cm−1).

The vibrational energy of the atoms is described by phonon energies. Materi-als with crystal structure have characteristic phonon energy levels related to the bonds between the atoms in the crystal. The Raman shifts can correspond to first order or higher order energy levels of the phonons. Typically only Raman shifts corresponding to first or second order phonon energies are observed with Raman spectroscopy. As distances between the atoms in a crystal lattice change, the phonon energies are shifted.

Three characteristic Raman shifts are observed for silicon [3]. The dominant shift is at νSiSi= 520cm−1, a first order phonon energy mode corresponding to the

bond between silicon atoms. Two second order transverse acoustic and transverse optical phonon energies can be detected, νSiT A= 435cm−1and νSiT O≈300cm−1.

In fact the νSiT Ois a doublet with peaks at 299 ± 1 and 303 ± 1cm−1. Depending

on experimental conditions one or both peaks can be seen (merged to one peak or separated). For Si-Ge alloys two first order Raman shifts occur. One is originating from bonds between germanium atoms νGeGe = 296.5cm−1 and the other from

bonds between silicon and germanium atoms νSiGe = 520cm−1. The positions

of the peaks are shifted when the germanium composition changes and with the strain of the material. The resulting position of the Raman shifts [14] are described by equations 4.1 and 4.2:

νGeGe= 282.5 + 16XGe−575⊥[cm−1] (4.1)

νSiGe= 400.5 + 14.2XGe−384⊥[cm−1] (4.2)

Where XGeis the fraction of germanium in the alloy and ⊥the strain

perpendic-ular to the growth direction.

Care has to be taken when evaluating the Raman spectra for Si/Ge QDs. The νSiT O may coincide with νgeGe for strained Si/Ge QDs [3]. The position of the

νSiGeRaman shift is often [3] considered a feature resulting from the wetting layer,

the νSiGecontribution from the wetting layer dominate the contribution from the

QDs. The yield from Raman shifts of single layer QDs is expected to be weak and multilayer structures are often necessary for Raman spectroscopy.

Techniques such as polarized Raman spectroscopy can be used to eliminate the contribution from νSiT Oto the νGeGeRaman shift [11]. Several differences can be

(30)

18 Characterization methods

found in the way the Raman spectra is analyzed. Xu et. al. [12] use the integrated intensity ratios between the νSiGeand νGeGepeaks. The contribution from silicon

peaks is removed by subtraction of a silicon reference spectrum normalized by the ratio νSiSi peak. The resulting intensity of the sample’s spectrum is:

Iresult(ν) = Isample(ν) − (

Isample(SiSi)

Iref erence(SiSi)

)Iref erence(ν) (4.3)

The connection between intensity ratios of the two SiGe peaks and germanium composition is an established method for analysis of relaxed SiGe alloys [1]. Bara-nov et.al. [14] use a combination of subtracting a silicon spectrum and relations between integrated intensities to obtain information about the composition and equation (4.1) to determine the strain.

The equipment used for Raman measurements was an Argon laser with a wave-length of 514 nm. The PL/Raman equipment HR800 was equipped with a white notch filter of 514 nm, a grating of 1800 l/mm and a CCD detector array was used. The laser was focused on the sample by an objective lens. With this setup a resolution of 1 cm−1 was obtained. The chosen method for analyzing the data is

described in section 6.4.3 and is based on the separation of the νGeGe and νSiT O

(31)

Chapter 5

Experimental details

The work to develop a Si-Ge thermistor was divided into two parts, single layer and multilayer structures. The process parameters that were changed were:

• Temperature (TG).

• Time (tG).

• Flow (f).

It is known that growth temperature has a strong influence on the QDs, espe-cially on their size and on the amount of intermixing that occurs. The temperature will affect the critical thickness of the wetting layer and thereby influence when the initial nucleation occurs. At higher temperatures the atoms will be more mo-bile due to thermal energy, which will result in a faster growth of QDs but also increase the amount of silicon that intermixes into the QDs. A temperature range of 500C to 700C has been investigated. The growth time has been varied from

20s to a maximum of 320s. The growth times have been doubled for each consec-utive sample, i.e. the growth times are all part of the distribution t∈[20 40 80 160 320] s. Two values for the gas-flow were used to investigate if a high or low gas flow will affect the QDs. Gas-flow of either 24 sccm or 48 sccm was used. The quantum dots were produced in a RPCVD rector, ASM Epsilon 2000, with a base pressure of 10−3 Torr. The substrate used was p-type 4 inch silicon(100) wafers.

The production of the QDs was handled by KTH and Acreo.

5.1

Single layer

Characterization of single layer QDs was made by HRSEM and AFM. Diameter and density was obtained from the HRSEM micrographs and height information from the AFM micrographs. Raman spectroscopy was performed on single layer QDs as a reference to the multilayer structures. Investigation of the single layer structures and their process parameters was divided into three steps. Each step has a corresponding production batch:

(32)

20 Experimental details

A. Temperature constant, time and flow varied. Table 5.1 B. Temperature constant, time and flow varied. Table 5.2 C. Temperature and time varied, flow constant. Table 5.3

The difference between batches A and B is that the growth temperature has been changed from 500C to 600C. Each production batch was thoroughly

in-vestigated and used as a base for how the following batch was processed. Process parameters for the samples can be found in table A.

Sample Temperature [C] Flow [sccm] Time [s]

A.1 500 24 20 A.2 500 24 40 A.3 500 24 80 A.4 500 24 160 A.5 500 24 320 A.6 500 48 20 A.7 500 48 40 A.8 500 48 80 A.9 500 48 160 A.10 500 48 320

Table 5.1: Process parameters for batch A. Sample Temperature [C] Flow [sccm] Time [s]

B.1 600 24 40 B.2 600 24 80 B.3 600 24 160 B.4 600 48 40 B.5 600 48 80 B.6 600 48 160

Table 5.2: Process parameters for for batch B. Sample Temperature [C] Flow [sccm] Time [s]

C.1 650 24 20

C.2 650 24 40

C.3 700 24 20

C.4 700 24 40

(33)

5.2 Multilayer structures 21

5.2

Multilayer structures

The initial step towards creating multilayered structures is to cap the QDs with silicon. Due to the very low growth rate of silicon using silane gas (SiH4) at

temperatures lower than 600C, the QDs grown at such temperatures will have

to be capped at a higher temperature. To investigate how the QDs are affected by the annealing step required for low temperature QDs, a short investigation of annealing and standby effects was made. Standby is of interest since the epitaxial growth process involves two gases (silane and germane), and the switch between gas sources (QD growth and silicon capping) can be achieved with or without a short standby step.

Batch E was produced to investigate how the thickness of the silicon capping layer could reduce surface roughness and minimize layer to layer influence.

The investigation of multilayer structures has been divided into three steps: D. Anneal and standby effects.

E. Capping of single layers. F. Multilayer structures.

Batch D was studied using AFM and HRSEM. The surface roughness of batch E was investigated using AFM. Batch F was investigated by AFM to determine surface roughness, HRXRD to determine composition and FTPL to determine the bandgap. Batches E and F (table 5.5) were also investigated using Raman spectroscopy. The process parameters can be found in tables 5.4 and 5.5. The annealing procedure used was to increase temperature from TG to the annealing

temperature and remove the sample after tA minutes. Standby denotes that the

temperature has been kept constant at TG for for tS minutes.

Sample Growth Flow Time Anneal/Standby Anneal/Standby temp. [C] [sccm] [s] temp. [C] time [min]

D.1 500 48 40 650 tA= 3

D.2 500 48 80 650 tA= 3

D.3 650 24 20 650 tS= 3

D.4 650 24 20 650 tS= 6

(34)

22 Experimental details

Sample Growth Flow Time Capping Capping Layers temp. [C] [sccm] [s] temp. [C] thickness [nm]

E.1 650 24 20 650 25 1 E.2 650 24 20 650 50 1 E.3 650 24 20 650 100 1 E.4 650 24 20 650 200 1 F.1 650 24 20 650 50 5 F.2 650 24 20 650 100 5 F.3 650 24 20 650 200 5

(35)

Chapter 6

Results

6.1

Single layers

6.1.1

Time

The structural evolution of Ge QDs over time is described in section 3.3. To create uniformly sized and large QDs is a challenge as large quantum dots are typically associated with a bimodal size distribution. However, the initially nucleated QDs have a narrow size distribution and their size can be maximized with retained distribution, if the growth is stopped when larger dots start to appear (early stage growth, figure 3.4c). Figure 6.1 shows when the tC can be found to maximize the

size of type 1 QDs. Figure 6.1 also includes a graph of the size evolution with time of type 2 QDs, the dotted and dashed lines indicate the size of type 1 QDs from batch A and B.

Figure 6.1: Time evolution of QDs type 1 and 2. The dotted and dashed lines indicate the critical sizes of type 1 QDs from batch A and B

Samples C.1 and C.3 are are still in the nucleation stage and the critical size of QDs have not yet been reached. The consecutive samples C.2 and C.4 are both

(36)

24 Results

in early stage growth which gives that 20 ≤ tC <40 s for TG = 650, 700◦C and

f= 24 sccm. Sample B.1 has a few type 2 QDs (density 1 · 108cm−2). Figure 6.2

shows the relative number of type 2 QDs of the total QD density, from the figure tC is expected to be around 30-35 s for TG= 600 and f = 24 sccm (batch B 1-3).

Figure 6.2: Relative density of type 2 QDs of the total QD density.

6.1.2

Gas-flow

The difference between gas-flows of f=24 sccm and f=48 sccm was investigated at 500C and 600C (batches A and B). It is obvious that the growth is faster for 48

sccm than 24 sccm, QDs A.4 (24 sccm, 500C and 160 s) and A.8(48 sccm, 500C

and 80 s) are in the same stage of growth. Both QDs A.4 and A.8 have similar size distribution but the density of dots differ significantly. Low gas-flow has reduced the QD density as seen in fig.(6.3).

The maximum density of QDs is found to be lower for f=24 sccm than for f=48 sccm at TG = 500C and T=600C. The samples with maximum density

at TG = 500C are in approximately the same growth stage and therefore is the

comparison relevant. Type 2 QDs are of the sizes 51 nm (sample A.4) and 58 nm (sample A.8). For TG= 600C the QDs are in different stages of growth, the sizes

of type 2 QDs are 91 nm (sample B.1) respectively 133 nm (sample B.4). It is possible that sample B.4’s higher density of QDs is somewhat contributed by being in later growth stage than sample B.1 and a few new QDs may have nucleated. An opposing effect of later growth stage is that the effect of coalescence is expected to be higher for sample B.4 than B.1. The nucleation of new QDs is not enough to explain the higher density of QDs in sample B.4, the increased gas-flow has contributed to the increase in density at TG = 600C.

The growth time is lower for f=24 sccm than f=48 sccm, at a growth time of t = 80s and TG = 500C no QDs have appeared for f=24 sccm (sample A.3),

however nucleation has occured for QDs grown at same conditions but with f=48 sccm (sample A.8).

At t=40 s and TG = 600C (sample B.1) early stage growth has just started

(37)

6.1 Single layers 25

Figure 6.3: Maximum density of QDs as a function of gas-flow (f). type 1 QDs is just below 40 s.

The gas-flow affects the speed of growth, more so at low temperatures, about half the growth time at 500C is needed when the gas-flow is increased from 24

sccm to 48 sccm. Size differences of type 2 QDs with equal growth time can be explained by the QD’s growth stage. The size of type 2 QDs is the same or slightly lower for increased gas-flow, the change in size is very small (3-4 nm at TG = 600C). The decreased size and increased density of type 1 QDs follows

results from Cappelini et. al. [8], showing that increased growth rate increase the probability of nucleation and decrease the size of QDs.

6.1.3

Temperature

A raise of the growth temperature results in increased silicon intermixing. An additional effect of high temperatures is that the critical size of Qds increase, i.e the size of Qds before the size distribution becomes bimodal has increased. The QD’s increased size can be explained by relaxation due to silicon intermixing and also by increased diffusion length of ad-atoms [4].

The critical size of quantum dots can effectively be controlled by changing the growth temperature, fig.(6.4) shows the average critical size of QDs as a function of temperature. The critical size has increased from 24 nm at TG = 500C to 96

nm at TG = 700C.

6.1.4

Summary single layers

The QDs have a uniform size distribution until early stage growth takes place. To optimize uniformity and size, using growth time tG = tC is favorable. The

critical size of QDs when early stage growth starts can be tuned by adjusting growth temperature. Increased gas-flow results in higher growth rate, increased QD density, and a decrease of QD size. The negative effect of increasing the growth temperature is that silicon intermixing is increased.

(38)

26 Results

Figure 6.4: Average critical size of QDs

After evaluation of batches A and B, batch C was produced to find tc and

critical size of QDs at TG = 650C and TG = 700C. The gas-flow f = 24 sccm

was used to maximize the QD’s size. The reason for not decreasing f lower than 24 sccm was to maintain a high density and correlation to previous batches A and B.

Sample C.1 was chosen to be used in multilayer structures. The QDs have the following characteristics:

• Uniform size distribution • Density: 9.5 · 109cm−2

• Diameter: 57 nm • Height: 8 nm.

The aim for the QD’s diameter was 100 nm, and the sample C.1 only has an average diameter of 57 nm. As illustrated in figure 6.4 57 nm is achieved at TG = 650C

and creating QDs with diameter of 100 nm would require TG 700C. The size

reduction is motivated by decreasing silicon intermixing (due to lowered TG)and

that the simulated TCR for 57 nm QDs is close to that of 100 nm QDs, illustrated in figure 2.2. Sample C1 (figure 6.5) was used as a reference to standby, capping and multilayer structures.

(39)

6.2 Anneal and standby 27

6.2

Anneal and standby

Standby and anneal procedures give the same results, the QDs have radically changed their distribution on the surface as well as their size. Height and diameter have increased and density has decreased for all samples in batch D, the increase of diameter is illustrated in figure 6.6. Sample D.1 had no QDs before anneal (only a thin wettinglayer), and no QDs could be found by SEM after anneal. The conclusion from standby and annealing experiments is that to maintain the QDs original properties, standby and annealing should be avoided. Growth and capping should be at the same temperature and standby times when switching from QD growth to silicon capping should be minimized.

Figure 6.6: Size evolution during anneal and standby.

6.3

Silicon Capping

Silicon capping of the QDs was made without standby time after growth and the temperature was kept constant at TG= 650C. Figure 6.7 shows the surface

rough-ness after silicon capping. The surface roughrough-ness is found in table 6.1. Samples E.3 and E.4 have the same height between lowest and highest point, but sample E.4 is more uniformly covered by silicon. The roughness at the surface indicates that there will be some influence between QD layers, but the layer to layer influ-ence decrease with silicon thickness. Le Thanh and V.Yam [7] report that layer to layer influence ceases for silicon thickness greater than 160 nm (the following QD layers are randomly distributed on the surface). The designed and real silicon thickness differ slightly, the real value is based on HR-XRD scans (section 6.4.1) on multilayer structures (batch F).

(40)

28 Results

Sample Silicon cap [nm] Surface Real Silicon cap [nm] Sample (designed) roughness [nm] (measured)

E.1 25 4.5 20

E.2 50 3.5 40

E.3 100 2.5 80

E.4 200 2.5 160

Table 6.1: Surface after silicon capping. The corrected silicon cap is based on HR-XRD measurements on corresponding multilayer structures(table(6.2))

.

Figure 6.7: AFM micrograph of surface roughness after capping. 20, 40, 80 and 160 nm silicon capping layers.

6.4

Multilayer structures

6.4.1

X-ray diffraction

HR-XRD scans show that significant intermixing has occured, average Ge content is measured to be 36% in sample F.3. The QD peak position in the Ω/2Θ scan has shifted closer to the substrate peak for increasing capping thickness (less ger-manium content). The time required to cap sample F.3 with silicon is 4 times the time required to cap sample F.1 (constant growth rate of silicon for all samples). The increased intermixing in sample F.3 can possibly be related to the increased time exposed to high tempereatures. Another feature of the QD peak is that it is broadened for low silicon cap. As discussed in sect 6.3, some layer to layer influence is expected, which will result in a thinner wetting layer and QDs that increase in size for each layer. The spread of QD’s size is a possible explanation to the broadening of the QD peak in samples F.1 and F.2. The distance between two repeated QD layers is measured to be 42, 82 and 158 nm respectively for samples F.1,F.2 and F.3.

Sample QD period [nm] Ge content [%]

F.1 42 38

F.2 81 37

F.3 158 36

(41)

6.4 Multilayer structures 29

Figure 6.8: HR-XRD Ω/2Θ scan for multilayer structures, batch F

6.4.2

Photoluminescence

Photoluminescence measurements give the spatially direct (EA) and indirect (EB)

transitions (figure 4.3). The difference in valence band offset between the silicon barrier and the Ge QD can be calculated from FTPL results and by using equation 2.1 the theoretical TCR can be calculated.

The luminescence peak is shifted towards higher energies for thicker silicon capping layers, this corresponds well with the increased silicon intermixing seen in HR-XRD Ω/2Θ scans (table 6.2). At low temperatures (around 40 K) the spatially indirect EB transition is dominant and for higher temperatures, the probability of

the spatially direct transition EA increases. The difference in separation between

the EA and EB differs for the samples, an explanation for this is that the peaks

are not fully separated for samples F.1 and F.2 (figure 6.9).

Figure 6.9: Normalized FTPL spectrum for batch F at 110K (EA)and 37K(EB)

(42)

30 Results

Sample EA [meV] EB [meV] Theoretical TCR [%/K]

F.1 830 805 4.2

F.2 840 810 4.15

F.3 865 815 4.1

Table 6.3: Multilayer data from HR-XRD Ω/2Θ scan.

6.4.3

Raman spectroscopy

This section contains the combined results from Raman spectroscopy on single and multilayer structures. Raman shifts can be detected for all samples, including single layer QDs. An explanation how even single layers of QDs can be detected is that all samples have high density of QDs and they cover a large part of the substrate surface. The signal for uncapped QDs and batch F (5 layers of QDs) is strong, but for the single layers of QDs capped with silicon the signal is weaker (figure 6.10).

Figure 6.10: Normalized Raman spectrum of samples C.1, E.1 and batch F. The spectrum was obtained using 200 mW laser intensity and integration time of 80 s. In this work the measured νSiGe is considered to be dominated by the

contri-bution from QDs and not the wetting layer. This assumption is based on the shift of νSiGe for increasingly larger QDs as seen in figure 6.11. The large size, high

density, and high surface coverage of QDs support that the νSiGe shift originates

from the QDs. Figure 6.11 shows a shift to lower wavenumber for the larger dots C.2 and C.4 over dots C.1 and C.3. The shift to lower wavenumber corresponds to a relaxation of the strain. This result is expected for large dots as they are not symmetrically strained and the higher height to diameter ratio of the type 2 dots in sample C.2 and sample C.4 results in relaxation of the strained lattice at the top of the dots (they are not capped with silicon).

In fig 6.11 a separation between two peaks at 295 and 301.5 is seen, but the separation is not big enough to clearly establish if the peaks are from the doublet of silicon peaks or a separation of the νGeGe and the νSiT O. Because of the problem

(43)

6.4 Multilayer structures 31

Figure 6.11: Normalized Raman spectrum of batch C. The spectrum was obtained using 200 mW laser intensity and integration time of 80 s.

to define the position of the νGeGe peak from the νGeGe, it is not possible to

evaluate the QDs with equation 4.1. As previously stated the νSiGepeak for these

samples originate from the QDs and not from the wetting layer, thereby it is valid to use equation 4.2 to evaluate the structure. Using the position of the νSiGe peak

also eliminates the need to subtract a silicon spectrum, which in this case would be reduce to reliability of the calculations since the νGeGepeak is very close to 300

cm−1. In order to use an equation with two unknown properties, composition and

strain, the assumption that the QDs are strained to fit the lattice of the substrate has been made. Combining Vegard’s law (equation 3.2) with the definition of strain and equation 4.2 gives the equation:

XGe = − 400.5−νSiGe+575 14.2 + aSi aGe−aSi 2 + v u u t 400.5−νSiGe+575 14.2 + aSi aGe−aSi 2 !2 −(400.5 − νSiGe)aSi 14.2(aGe− aSi) (6.1)

The results from measurements on multilayer structures are summarized in table 6.4. Since the intermixing is not entirely uniform through the dots (a germa-nium rich core is expected), the composition should be interpreted as an average composition of the QDs. The approximation of uniformly and fully strained QDs is not sufficient for uncapped QDs. Sample C.1’s calculated germanium content is 33%, which is lower than for the capped dots which shows that the approximation gives a very large error for uncapped QDs.

6.4.4

Summary multilayer structures

The QD multilayer structures show high silicon intermixing, and the structure changes with increased silicon capping. XRD measurements as well as AFM mi-crographs of the surface indicate that some layer to layer influence exists and is increased as silicon thickness is reduced. The composition of the QDs is affected by the thickness of the silicon capping layer. The high and increasing intermixing

(44)

32 Results

Sample νSiGe Ge content [%] C.1 413 33 E.1 416 41 E.2 416 41 E.3 416 41 E.4 414 36 F.1 416 41 F.2 415 38 F.3 414 36

Table 6.4: Multilayer data from Raman spectroscopy.

leads to the conclusion that decreasing growth and capping temperature would be beneficial to the composition and reduce the intermixing effect of capping thick-ness. The theoretical TCR for sample F.3 is 4.1%, calculated based on bandgap information from PL measurements.

Due to the challenges of Raman spectroscopy and the many different ways the results are interpreted, further investigation of the measurement technique on strained Si/Ge structures and the data analysis is recommended before final con-clusions are drawn from the measurements. The Raman results are close to what is measured with HRXRD, a possible explanation to why the result differs more for thinner capping thickness is that the layer to layer influence has resulted in size and composition variation between layers (broadening of the QD peak in HRXRD scan figure 6.8). The variation between layers may increase the uncertainty of the measurements. The reported composition of the QDs in this work is therefore based on results from HRXRD measurements.

(45)

Chapter 7

Conclusions

The resulting QDs used in a multilayer structure have the following characteristics: • Uniform QDs

• High Density: 9.5 · 109cm−2

• Diameter:60 nm (before capping) • Low germanium content: 36% • Theoretical TCR = 4.1%/K.

The TCR of the samples from batch F is mainly limited by the high degree of sili-con intermixing. To increase the TCR, Ge sili-content in the QDs should be increased, this can be done by decreasing the growth temperature. Decreasing growth tem-perature will result in formation of smaller QDs. QD size relative to intermixing need to be optimized with regard to TCR.

The time component of the growth should be optimized to be close to tC, which

will maximize the type 1 QDs before the size distribution becomes bimodal. The growth rate will affect the size of QDs and their density. It is possible to increase density and decrease QD size by increasing the growth rate.

The TCR calculated from the QD’s bandgap is low when compared to simulated values for pure Ge QDs. However the TCR of 4.1% is higher than current Si/Ge QW structures. Even though the QWs may be optimized further, this leaves good expectations for further development of Si/Ge QDs as thermistor material.

(46)
(47)

Chapter 8

Further work

The TCR for the current multilayer structure (batch F) has been theoretically cal-culated from the measured bandgap. To measure directly the TCR and the noise a structure with contacts needs to be produced. To achieve this for the current structure would provide a reference to compare with QW thermistors and future optimized components based on Si/Ge QDs. Further work with the characteriza-tion of the QDs also include using transmission electron microscopy to determine the composition profile of the QDs and their size after capping with silicon.

By optimizing the current QD structure with TG = 650C it is possible to

increase growth time slightly since tG is close to but lower than tC. The design of

the multilayer structure therefore has room for optimization, only 5 layers of QDs have been produced, and it is possible to create multilayer structures with a high number of repeats. The structure can most likely be optimized to maintain size distribution between layers even for low silicon thickness. It has been proposed by Than et. al. [7] to adjust tG for each individual layer to compensate for the layer

to layer reduction of the critical thickness of the wetting layer.

As determined by HRXRD (section 6.4.4) the QDs grown at TG = 650C have

a low germanium composition. The QD’s size and composition can be altered to achieve a higher TCR. As a first step, reducing growth and capping temperature to TG= 600C would result in higher germanium content and still be suitable for

silicon capping at TG. The drawback of reducing the growth temperature is the

size reduction. Using the parameters for sample B.1 (table 5.2) but with a small reduction of growth time (to tG = 30-35 s figure 6.1) would result in QDs with

diameter of 40nm (size before capping).

(48)
(49)

Bibliography

[1] P. M. Mooney, F. H. Dacol, J. C. Tsang, and J. 0. Chu Raman scattering analysis of relaxed GexSi(1 − x) alloy layers Applied Physics Letters, v 62, n

17, 26 April 1993, p 2069-71

[2] J. H. Davies. 1998. The Physics of Low-Dimensional Semiconductors. Cam-bridge University Press.

[3] Kolobov, A.V. 2000. Raman scattering from Ge nanostructures grown on Si substrates: Power and limitations Journal of Applied Physics, v 87, n 6, 15 March 2000, p 2926-30

[4] Chaparro, S.A., Zhang, Y., Drucker J., Chandrasekhar D., Smith D. J. Evo-lution of Ge/Si(100) islands: Island size and temperature dependenceJournal of Applied Physics, v 87, n 5, Mar, 2000, p 2245

[5] P. W. Kruse. 2002. Uncooled Thermal Imaging. SPIE Press.

[6] P. Boucaud, V. Le Thanh, V. Yam, S. Sauvage, N. Meneceur, M. Elkurdi, D. Débarre and D. Bouchier. 2002. Aspects of Ge/Si self-assembled quantum dotsMaterials Science and Engineering B, Volume 89, Issues 1-3, 14 February 2002, Pages 36-44

[7] Le Thanh, Vinh , Yam V. Superlattices of self-assembled Ge/Si(0 0 1) quan-tum dots Applied Surface Science, v 212-213, n SPEC., May 15, 2003, p 296-304

[8] Capellini, G.; De Seta, M.; Evangelisti, F. Ge/Si(100) islands: Growth dy-namics versus growth rateJournal of Applied Physics, v 93, n 1, Jan 1, 2003, p 291-295

[9] M. Larsson, A. Elfving, P. O. Holtz, G. V. Hansson and W. -X. Ni. 2003. Photoluminescence study of Si/Ge quantum dots Surface Science, Volumes 532-535, 10 June 2003, Pages 832-836

[10] V. Yam, V. Le Thanh, D. Débarre, Y. Zheng and D. Bouchier Kinetics of Si capping process of Ge/Si(0 0 1) quantum dots Applied Surface Science, Volume 224, Issues 1-4, 15 March 2004, Pages 143-147

(50)

38 Bibliography

[11] Baranov, A.V. Fedorov, A.V.; Perova, T.S.; Moore, R.A.; Solosin, S.; Yam, V.; Bouchier, D.; Le Thanh, V. 2004. Polarized Raman spectroscopy of mul-tilayer Ge/Si(001) quantum dot heterostructures Journal of Applied Physics, v 96, n 5, Sep 1, 2004, p 2857-2863

[12] Xu, L.; McNally, P.J.; Dilliway, G.D.M.; Cowern, N.E.B.; Jeynes, C.; Men-doza, E.; Ashburn, P.; Bagnall, D.M. Raman study of the strain and H2 preconditioning effect on self-assembled Ge island on Si (001)Journal of Ma-terials Science: MaMa-terials in Electronics, v 16, n 7, July 2005, p 469-74 [13] T. K. Nguyen-Duc, V. Le Thanh, L.H. Nguyen, F.A. d’Avitaya and J. Derrien.

2005. Kinetic formation and optical properties of self-assembled Ge/Si hut clusters. Materials Science in Semiconductor Processing, Volume 8, Issues 1-3, February-June 2005, Pages 41-46

[14] Baranov, A.V.; Fedorov, A.V.; Perova, T.S.; Moore, R.A.; Yam, V.; Bouch-ier, D.; Le Thanh, V.; Berwick, K. Analysis of strain and intermixing in single-layer Ge/Si quantum dots using polarized Raman spectroscopy Physi-cal Review B (Condensed Matter and Materials Physics), v 73, n 7, 15 Feb. 2006, p 75322-1-6

[15] Baribeau, J.-M.; Wu, X.; Rowell, N.L.; Lockwood, D.J. Ge dots and nanos-tructures grown epitaxially on Si Journal of Physics: Condensed Matter, v 18, n 8, 1 March 2006, p R139-74

[16] A. Elfving. 2006. Near-infrared photodetectors based on Si/SiGe nanostruc-tures.IFM, Linköpings universitet.

[17] M. H Berntsen. 2007. Simulation of valence band structure in strained Ge/Si quantum dots. KTH

(51)

Appendix A

Appendix A

Sample Diameter Height Density Diameter Height Density 1 [nm] 1 [nm] 1 [cm−1] 2 [nm] 2 [nm] 2 [cm−1] A.1 0 0 0 0 0 0 A.2 0 0 0 0 0 0 A.3 0 0 0 0 0 0 A.4 22 4 1.7 · 1010 51 13 1.7 · 109 A.5 25 6 8.6 · 109 142 17 1.9 · 109 A.6 0 0 0 0 0 0 A.7 0 0 0 0 0 0 A.8 23 4 1.84 · 1010 58 13 2.7 · 109 A.9 27 4 1.2 · 1010 129 18 3.1 · 1010 A.10 36 7 3.6 · 109 243 7 1.8 · 109 B.1 43 7 1.7 · 1010 91 7 1 · 108 B.2 42 7 1.4 · 1010 160 20 5.6 · 108 B.3 47 1 · 1010 215 7.2 · 108 B.4 40 6 1.8 · 1010 133 25 9.3 · 108 B.5 42 5 9.6 · 109 175 30 1.42 · 109 B.6 43 12 3.4 · 109 294 45 9 · 108

Table A.1: Data for single layer quantum dots. QDs type 1 are the initially nucleated dots, type 2 are dots formed in early stage and late stage growth.

References

Related documents

1795, 2016 Department of Physics, Chemistry and Biology. Linköping University SE-581 83

The recommended recipe will be chosen on the basis of three different characteristics that the user can manipulate: user groups, the category of food (Italian, Thai, etc.), and

The two groups with no long-term expectations of health-related ability to stay in the same profession differed in age, and with regard to work conditions, older age was

Optical and Structural Characterization of GaN Based Hybrid Structur es and Nanor ods Linköping 2015 Mathias Forsber g.. Linköping Studies in Science and Technology

Mitochondria membrane potential measured as TMRM-uptake before and after GDP, CAT or the combination of CAT and GDP during baseline and inhibition of the ATP-synthase in control

Samtidigt som man redan idag skickar mindre försändelser direkt till kund skulle även denna verksamhet kunna behållas för att täcka in leveranser som

Accordingly, an even higher polarization degree (~ 73 %) is measured for the positively charged exciton. In a different study, pyramidal QD structures were employed. In contrast to

Department of physics, chemistry and biology