Physical characterization of hfo2 hfo2 al2o3 alloy thin films

Thus, a higher capacitance can beachieved either by decreasing the dielectric thickness t, or by using a material of a higherdielectric constant, κ.. The equivalent oxide thickness, teq

Department of PhysicsNational University of Singapore

2005

Finally, I would like to thank my husband Tianfu for his support and continuousencouragement during the course of this work.

i

1.1 Preliminaries: The MOSFET 1

1.2 Scaling and Improved performance 2

1.3 High κ Gate Dielectric 4

1.3.1 Organization of thesis 7

2 Thin Film Deposition 9 2.1 Atomic Layer Chemical Vapour Deposition 9

2.2 Sample Preparation 10

2.2.1 HfO2 and HfO2-Al2O3 alloys 10

3 Composition of Thin Films 12 3.1 Rutherford Backscattering Spectrometry 12

3.1.1 Introduction 12

3.1.2 Experimental 13

ii

3.1.3 Results & Discussion 14

3.2 X-ray Photoelectron Spectroscopy 15

3.2.1 Introduction 15

3.2.2 Experimental 17

3.2.3 Results & Discussion 18

3.3 Conclusion 20

4 Thermal Stability of Thin Films 22 4.1 Introduction 22

4.2 Glancing Angle X-Ray Diffraction 24

4.2.1 Introduction 24

4.2.2 Experimental 25

4.3 Results & Discussion 26

4.3.1 Phase study of HfO2 26

4.3.2 HfO2-Al2O3 alloy samples 27

4.4 Conclusion 29

5 Diffusion Studies of HfO2-Al2O3 Alloy Thin Films 31 5.1 Introduction 31

5.2 Secondary Ion Mass Spectrometry 31

5.3 Bulk & Grain Boundary Diffusion 33

5.3.1 Bulk Diffusion 33

5.3.2 Grain Boundary Diffusion 34

5.4 Experimental 34

5.5 Results & Discussion 34

5.6 Conclusion 38

iii

A XPS spectra 45

iv

The continued scaling of complementary metal-oxide-semiconductor (CMOS) devicefeature sizes have allowed the semiconductor industry to achieve unprecedented gains inproductivity and performance This rapid scaling has caused the channel length and thethickness of the gate dielectric to decrease rapidly As a result, high gate leakage currentacross the thin gate dielectric now necessitate the introduction of alternative high κ gatedielectric materials in order that the stringent requirements for leakage current can bemet

Due to its high dielectric constant, HfO2 (κ ∼ 20) is a potential candidate for gateoxide replacement However, as HfO2 is thermally less stable and it crystallizes at about

400◦C, Al2O3 was co-deposited with HfO2 to obtain an alloy that is more resilient tocrystallization In this work, we will investigate the physical characteristics of HfO2 andHfO2-Al2O3 alloy thin films (∼ 20 nm) prepared by ALCVD The samples were deposited

on p-type Si(100) substrate at 300◦C Their composition was first determined using RBSand XPS The HfO2 thin film was found to be in good stoichiometry For the HfO2-Al2O3

alloy thin films, the Hf to Al cation ratios obtained were much higher than the expected.From the results, the films were also found to be of good compositional uniformity.Glancing incidence XRD was then employed to study the microstructure and ther-mal stability of HfO2 and HfO2-Al2O3 alloy thin films that were annealed using RTPfor temperature up to 1000◦C As-deposited HfO2 was found to have a small degree ofcrystallization in mixed tetragonal and monoclinic phase Upon annealing at 1000◦C, thetetragonal polymorphs transformed into monoclinic polymorphs and the thin film crys-tallizes fully in the monoclinic phase For the HfO2-Al2O3 alloy samples, crystallizationwas observed to be suppressed in the thin films up to an annealing temperature of 800◦C

At 1000◦C, the HfO2-Al2O3 thin films crystallize in the monoclinic phase in the [111]direction The degree of crystallization and the crystallite size was found to decrease forincreasing Al2O3 content in the samples

The effect of thermal annealing on precursor contamination and interdiffusion of silicon

v

in the thin films were then investigated using SIMS SIMS results indicate the presence of

an interfacial layer between the thin film and the silicon substrate Chlorine contaminationfrom precursor by-products was also found in the thin films and the interfacial region inthe as-deposited films After the samples were annealed at 1000◦C, it was observed thatthe chlorine contamination persists at the interface only Silicon out-diffusion from thesubstrate into the thin film was also observed in HfO2-Al2O3 samples that were annealed

at 1000◦C This is likely to be due to enhanced diffusion along grain boundaries when thesamples crystallize during the annealing process

vi

List of Tables

1.1 Material properties of high κ candidate Table adapted from reference[8] 6

2.1 Precursors used in the deposition of the various types of films 11

2.2 Target composition and estimated thickness of thin films 11

3.1 Film composition in atomic fraction of the samples obtained by RBS 14

3.2 Film composition in atomic fraction of the samples obtained by XPS 20

4.1 Crystallite size for samples annealed at 1000◦C 28

A.1 Sample 1 peak parameters 46

A.2 Sample 2 peak parameters 48

A.3 Sample 3 peak parameters 50

A.4 Sample 4 peak parameters 52

A.5 Sample 5 peak parameters 54

vii

List of Figures

1.1 (a) Schematic diagram of a MOSFET (b) Energy band diagram of a

Metal-Insulator-Semiconductor (MIS) capacitor under inversion condition 2

2.1 Deposition of Al2O3 thin film 10

3.1 Setup used for RBS analysis 13

3.2 RBS spectra obtained for HfO2-Al2O3 samples Inset shows an enlarged view of the Hf peaks 14

3.3 RBS spectra for sample 4 (red line indicates simulation of spectra) 15

3.4 Sampling depth of XPS experiment 16

3.5 Schematic Diagram of an XPS spectrometer 17

3.6 Hf4f, Al2p and O1s spectra obtained for the various HfO2-Al2O3 samples 18 3.7 XPS wide scan and Hf4f and O1s fitted core level spectra for sample 1, pure HfO2 19

4.1 Diffraction of X-ray by a crystal 24

4.2 Effect of crystallite size on diffraction curves 25

4.3 Sample 1(HfO2): XRD diffractograms of as-deposited and annealed samples 26 4.4 XRD patterns of HfO2-Al2O3 alloy samples: (a) as-deposited and (b) an-nealed at 800◦C 27

4.5 XRD patterns of HfO2-Al2O3 alloy samples annealed at 1000◦C 28

5.1 Schematic diagram showing the initiation of the collision cascade and the sputter removal of material from the sample surface[34] 32

viii

5.2 Schematic representation of the diffusion profile showing contributions from

the bulk and grain boundary diffusion regimes.[35] 33

5.3 SIMS profiles obtained for sample 1, 2 and 4 before and after annealing at 1000◦C The first row shows as-deposited sample profiles while the second row shows profiles of annealed samples 35

5.4 ClCs+ and SiCs+ curves for sample 2 and sample 4 36

5.5 Bulk diffusion in sample 4 annealed at 1000◦C Linear fit to the data is indicated by red line 37

5.6 Grain boundary diffusion in sample 4 annealed at 1000◦C Linear fit to the data is indicated by red line 38

A.1 Sample 1 Wide Scan 45

A.2 Sample 1 high resolution scans for Hf4f, O1s and C1s 46

A.3 Sample 2 Wide Scan 47

A.4 Sample 2 high resolution scans for Hf4f, Al2p, O1s and C1s 47

A.5 Sample 3 Wide Scan 49

A.6 Sample 3 high resolution scans for Hf4f, Al2p, O1s and C1s 49

A.7 Sample 4 Wide Scan 51

A.8 Sample 4 high resolution scans for Hf4f, Al2p, O1s and C1s 51

A.9 Sample 5 Wide Scan 53

A.10 Sample 5 high resolution scans for Al2p, O1s and C1s 53

B.1 SIMS depth profiles for Sample 1 at various annealing temperature 55

B.2 SIMS depth profiles for Sample 2 at various annealing temperature 56

B.3 SIMS depth profiles for Sample 4 at various annealing temperature 56

ix

Chapter 1

Introduction

The idea of a field effect transistor (FET) was first proposed by Julius Lilienfeld in 1926

In 1960, the first working Metal Oxide Semiconductor(MOS) transistor, made using mally grown oxidized silicon, was demonstrated by Kahng and Atalla[1] However, due

ther-to surface trap states and ionic contamination at the silicon-silicon dioxide interface, vice characteristics fluctuate and were not reproducible Subsequently, intensive researchyielded improvements in the interface and today, the MOSFET is one of the most im-portant device in the electronic industry Figure 1.1(a) shows a schematic drawing of aMOSFET

de-In a MOSFET, free carriers are produced in the channel by the process of “inversion”where a gate bias pulls the bandedges to a point lower than the fermi level, EF and createsfree carriers near the oxide-semiconductor interface (Figure 1.1(b)) Then, by applying abias between the source and drain, a current will flow in the channel Using the gradualchannel approximation[2], the drain current is given by

CHAPTER 1 INTRODUCTION 2

Gate Oxide Gate

dia-gram of a Metal-Insulator-Semiconductor (MIS) capacitor der inversion condition

un-and VD are the voltages applied to the transistor gate and drain while VT is the thresholdvoltage

As VD increases, ID initially increases linearly with VD but eventually saturates when

The continued success of the semiconductor industry can be attributed to the large R&Dinvestments made to improve its products The main improvement trends include theincrease of number of components per chip, decreasing the cost per function, improvingthe speed, power, compactness and functionality of devices[3] These improvements areachieved mainly by scaling the minimum feature sizes at an approximate rate of 0.7x pertechnology node Historically, the semiconductor industry has been able to reduce thecost per function by an average 25 - 30% each year while doubling the functions per chipevery 1.5 - 2 years The continued market pressure to maintain this trend requires the

is the total gate capacitance (including parasitic gate overlap and fringing capacitance)per micron transistor width, VDD is the power supply voltage and ID,sat is the saturationdrain current per micron transistor width.

From equation 1.2, we see that an increased saturation drive current can be obtained

by increasing the Cinv or decreasing the channel length (The gate voltage, VG cannot beincreased much to avoid creating high electric field across the oxide while VT cannot be re-duced due to room temperature operating constraints) To increase the gate capacitance,

we consider a parallel plate capacitor where the capacitance is given by

C = κǫ0A

where κ is the dielectric constant, ǫ0 is the permittivity of free space, A is the area ofthe capacitor, and t is the thickness of the capacitor Thus, a higher capacitance can beachieved either by decreasing the dielectric thickness t, or by using a material of a higherdielectric constant, κ

The former solution was applied to the gate dielectric, SiO2, until the thickness of theoxide reaches below 1.5 nm In ultrathin SiO2, high gate leakage current was estimated

CHAPTER 1 INTRODUCTION 4

no longer meet strict performance requirements which is particularly important in lowstandby power logic applications Fundamental material limit thus precludes the furtheruse of SiO2 as the gate dielectric and hence introduction of an alternate gate dielectricmaterial into the gate stack is imminent

The development of silicon oxynitride (Si-O-N) has served as an interim solution for thescaling of the CMOS transistor Si-O-N, formed by substitution of nitrogen into SiO2[5],has a dielectric constant of about 6 and therefore, allows a thicker gate dielectric However,the scaling of Si-O-N is limited due to stringent gate leakage requirements of the variousapplications

These applications are divided into low power and high performance logic For lowpower logic which consists of low operating power (LOP) and low standby power (LSTP)logic applications, the allowable gate leakage limit is very low while the equivalent oxidethickness, teq,1

is relatively high High κ dielectric is a potential solution that is projectedfor 2006 and beyond when Si-O-N will no longer be able to meet the nominal gate leakagecurrent limit[3] The equivalent oxide thickness, teq of LSTP is expected to be 1.9 nmand the nominal gate leakage current density limit of 0.015 A/cm2

1 teq= t d ·3.9κd ; t d & κ d are the thickness and dielectric constant of the alternate dielectric respectively The equivalent oxide thickness is the theoretical thickness of SiO 2 (κ ∼ 3.9) that is needed to achieve the same capacitance density as that of an alternate dielectric.

CHAPTER 1 INTRODUCTION 5densities and mid-bandgap interface state densities of the order of 1010

cm−2 are routinelyobtained on transistor devices Finding an alternative material that can match SiO2 interms of its superior electrical isolation properties has proved to be a great challenge.Criteria of Alternative Gate Dielectric The search for an alternative gate dielectricrequires the consideration of many properties Broadly classifying them, we arrive atissues in two main areas that need to be considered First, fundamental properties ofthe material such as dielectric constant, barrier height, stability in contact with siliconand film morphology needs to be understood since this has a direct consequence on theelectrical properties of the material

The alternative gate dielectric clearly ought to have a dielectric constant that is greaterthan SiO2 From equation 1.4, we see that the leakage current associated with directtunnelling increases exponentially with decreasing barrier height This is also true for trap-assisted electron transport such as Frenkel-Poole or hopping conduction[6] For electronstunnelling from the silicon substrate to the gate, this barrier is the conduction bandoffset, ∆Ec, while for electrons tunnelling from the gate to the Si substrate, this barrier

is ΦB (figure 1.2) The dielectric constant, band gap and conduction band offset of somepotential high κ materials are listed in Table 1.1

Metal

Semiconductor Insulator

With the close proximity of the gate dielectric to the channel, it is crucial that thematerial is stable when in contact with silicon Reaction with silicon to form uncontrolledinterfacial layers such as SiO2, silicide or silicate compounds will have adverse effects onthe carrier channel mobility The overall capacitance of the gate stack will also be affected

CHAPTER 1 INTRODUCTION 6since the minimum achievable teq will then be limited by the lower κ material.

teq = tlow−κ+ κlow−κ

κhigh−κ thigh−κ (1.5)Using thermodynamic considerations, materials that are stable in contact with siliconunder equilibrium conditions have been identified[7] However, as deposition conditionscan depart from equilibrium, the resulting film quality may vary Thus, factors such assurface preparation and reaction kinetics need to be considered when characterizing thesefilms

From Table 1.1 we see that most of the materials, with the exception of Al2O3, areeither polycrystalline or crystalline even under moderate thermal treatment Polycrys-talline gate dielectric are not preferred as the presence of grain boundaries in these filmscan act as high leakage or impurity diffusion paths Impurity or dopant diffusion can

in turn result in threshold and flatband voltage shifts Problems with grain boundariescan in principle be avoided if single crystal oxides are used However, this would requiredevelopment of large throughput manufacturing tools that can be integrated into existingprocesses Consequently, it is understandable that an amorphous material is still the idealfilm structure for the alternative gate dielectric

Material Dielectric Band Gap ∆Ec to Si Crystal

Constant (κ) EG (eV) (eV) structureSiO2 3.9 8.9 3.2 Amorphous

Si3N4 7 5.1 2.0 Amorphous

Al2O3 9 8.7 2.1 Amorphous

Y2O3 15 5.6 2.3 Cubic

ZrO2 25 5.8 1.2 Mono., tetrag., cubic

HfO2 25 5.7 1.5 Mono., tetrag., cubic

La2O3 30 4.3 2.3 Hexagonal, cubic

Ta2O5 26 4.5 1.5 Orthorhombic

TiO2 80 3.5 1.2 Tetrag

∆E c = Conduction Band Offset, Mono = Monoclinic, Tetrag = Tetragonal, cubic

The second area concerns device processing, integration and performance issues In

In device processing, the deposition process of the dielectric needs to be compatiblewith current CMOS processing, cost and throughput The techniques available for gatedielectric deposition includes, physical vapor deposition, chemical vapor deposition(CVD)and molecular beam epitaxy Of these, the atomic layer CVD method that is based on self-limiting chemistries for layer by layer deposition of the thin film appears to be favourable

as films of good conformity can be obtained

With the introduction of alternative gate dielectrics, reliability characteristics is likely

to deviate from that of SiO2 Dielectric breakdown characteristics and transistor failuremodes such as hot carrier effects and negative bias temperature instability of the materialwill need to be re-established Hence, to meet reliability requirements, an understanding

of the physics of each failure mode and the development of practical engineering tools will

be necessary

In this project, we focus mainly on the understanding of the first area mentioned in theprevious section, i.e materials properties The high κ material that we are studying isHfO2-Al2O3 alloy deposited using atomic layer CVD We will first look at the microstruc-ture of the as-deposited thin films followed by the effect of thermal treatment on the films.Following which, we will look at the effect of thermal annealing on the interdiffusion ofimpurities between the dielectric and silicon substrate

The thesis is organized as follows: chapter 1 consists of a general introduction to thesubject of high κ dielectric, chapter 2 on the method of thin film deposition and chapter

3 discusses the composition of the thin films obtained by Rutherford Backscattering and

CHAPTER 1 INTRODUCTION 8

X-ray Photoelectron Spectroscopy In chapter 4, we will look at the effect of thermaltreatment on the microstructure of the sample while in chapter 5 the diffusion of impuritiesinto the thin film will be covered Chapter 6 is an summary of results obtained in thiswork

Chapter 2

Thin Film Deposition

The thin films used in this study were deposited using atomic layer chemical vapourdeposition (ALCVD) technique In this chapter, we will look briefly at the principle ofthe technique and the deposition conditions used for our samples

Atomic layer chemical vapour deposition (ALCVD) is a self-limiting film growth methodcharacterized by the alternate exposure of gaseous precursors to the substrate surface in arepeated cycle In between cycles, an inert gas is used to purge volatile by-products of thechemical reactions The self-limiting mechanism ensures that film growth stops automat-ically at one or two monolayers ALCVD relies on surface exchange reactions betweenchemisorbed metal precursors fragments and adsorbed nucleophilic reactant molecules.Figure 2.1 illustrates how an Al2O3 thin film is obtained

ALCVD is a deposition technique whereby highly conformal films with precise control

of layer thickness can be obtained at low deposition temperature It offers potential forlarge area growth as it is integrable into existing manufacturing methods to obtain goodcomposition control and high uniformity films[9, 10]

As ALCVD of thin films is a relatively slow process, it’s primarily used only in areas ofapplications where thin films are required Presently, ALCVD is slated to be useful for gatedielectric applications and metal barrier films for Cu metallization in the semiconductorindustry

9

CHAPTER 2 THIN FILM DEPOSITION 10

Si

OSiOAlOH

Si

OSiOAlOH

(c)

Si

OSiOAlCH

Si

OSiOAlCH

(b)

Si

OSi

O

Si

OSiO

saturation followed by gas purge Figure adapted from ence [11]

Samples used in this study were prepared by Genus Inc (USA) using ALCVD Depositionwas carried out on p-type(100) silicon substrate using Genus Lynx2TM

dielectric module.The wafers were HF vapour cleaned prior to deposition at 300◦C

Precursors used for deposition are trimethyl aluminium (TMA, Al(CH3)), hafnium chloride (HfCl4) and water (H2O) Table 2.1 shows the precursors used for deposition ofHfO , Al O and HfO -Al O alloy films respectively Nitrogen was used as the carrier

tetra-CHAPTER 2 THIN FILM DEPOSITION 11and purge gas.

Thin Film Deposited Precursors Used

HfO2 HfCl4, H2O

Al2O3 TMA, H2OHfO2-Al2O3 HfCl4, TMA, H2O

A series of HfO2-Al2O3 thin films of different compositions were prepared as shown inTable 2.2 The recipes prescribed to obtain the targeted composition for the HfO2-Al2O3

films were based on the HfCl4:TMA pulse ratio in a cycle As such, the HfO2-Al2O3 alloyfilms obtained using such method are often not well characterized for its composition.The thickness of films were monitored by spectroscopic ellipsometry In the next chapter,

we will proceed to characterize the composition of the material

Sample No Target Composition Hf/Al Estimated thickness/nm

Chapter 3

Composition of Thin Films

In this chapter, we set forth to determine the composition of the deposited thin films Twomethods were employed: namely Rutherford Backscattering Spectrometry (RBS) and X-ray Photoelectron Spectroscopy (XPS) Results from the former will first be discussed,followed by a comparison with the latter

When a beam of α particles is incident on a target, some α particles are scattered at largeangles (> 90◦) This occurs when an α particle suffers an elastic collision as it comes intoclose encounter with the nucleus of a target atom The energy of the scattered particle ischaracteristic of the mass of the target atom

The ratio of the energy of the projectile after collision to that before the collision isgiven by the kinematic factor

12

CHAPTER 3 COMPOSITION OF THIN FILMS 13The areal density of the ith element, (N t)i, in atoms per unit area, can be found from

(N t)i = Aicos θ1

Q Ω σi(E, θ) (3.2)when Ω the detector solid angle, Ai the integrated peak count, Q the number of incidentparticles and σi the scattering cross section are known θ1 refers to the angle between thesample normal and the incident beam direction while Ni is the atomic density of the ithelement and t is the physical film thickness From equation 3.2, the stoichiometric ratiofor a compound film, AmBn, is given by

If the density of the film is known, it is then possible to determine the physical thickness

t of the film from equation 3.2

RBS is a useful technique routinely used to obtain information on the compositionand depth profile of thin film samples Further details on the technique can be found inreference [12]

Figure 3.1 shows the set-up used in the experiment A 2.0 MeV He+

ion beam with

a beam spot of approximately 5µm2

was incident on the sample at a glancing angle of

70◦ to the sample normal to improve the depth resolution The backscattered ions weredetected at a scattering angle of 160◦ with the vacuum chamber maintained at 2 × 10−6

torr

CHAPTER 3 COMPOSITION OF THIN FILMS 14

Figure 3.2 shows the RBS spectra obtained for the various samples The positions ofthe Hf, Al, O peaks were identified from their kinematic factors The Si high energyedge overlaps with that of the Al peak In the inset, the Hf peaks were observed to beflat-topped which suggests that the films are uniform

Hf

Al O

Si

Hf

an enlarged view of the Hf peaks

To obtained the stoichiometry, a simulation was carried out using the RUMP program[13].Figure 3.3 shows one of the fitted RBS spectra The composition of the samples obtained

From the table, we see that sample 1, pure HfO2, exhibits the expected stoichiometry

of HfO Decreasing Hf atomic fraction and increasing Al atomic fraction was observed

CHAPTER 3 COMPOSITION OF THIN FILMS 15

No 4

for sample 2, 3, and 4 respectively However, for sample 5 with expected composition ofpure Al2O3, the result obtained suggests that the film might be non-stoichiometric

In figure 3.2, the Al peak is observed to overlap with the high energy edge of the Sipeak This is due to the nearly identical atomic masses of the two elements The difficulty

in resolving the two contributions is made worse because of the weak intensity of the Alpeak The low intensity is due to a relatively smaller scattering cross-section for Al andthe low concentration of Al, especially in sample 2 and 3 XPS was therefore employed

as an alternative method to check the accuracy of the results obtained by RBS

XPS is a surface analysis technique based on the principle of the photoelectric effect.When a sample is irradiated with X-rays, photoelectrons are emitted from the core elec-tronic levels of the atoms Some of these photoelectrons escape from the surface withoutany energy loss and give rise to distinct peaks while others suffer inelastic collisions andcontribute to the background of the spectrum

The average distance travelled by an electron in a given matrix between two successive

CHAPTER 3 COMPOSITION OF THIN FILMS 16

collisions is known as the inelastic mean free path, λ 95% of the signal intensity iscontributed by electrons travelling 3λ or less Defining the sampling depth as

where θ is the photoelectron take-off angle, it can be shown that most of the detectedelectrons usually come from a thickness d, usually a few nanometers, for normal XPSoperation

The kinetic energy of the emitted electrons corresponding to a photoelectron peak isgiven by

KE = hν − BE − φs (3.5)where hν is the energy of the incident photon, BE is the binding energy of the electron,and φs is the spectrometer work function

The binding energy is the energy difference between the initial and final states after theelectron has left the atom From Koopman’s theorem, the observed binding energy is ap-proximately the negative of the orbital energy for the ejected electron if no rearrangement

of other electrons in the atom occurs When chemical bonding occurs, the initial state ofthe atom is altered and this results in a shift in the binding energy known as chemicalshift Chemical shifts usually assume that only initial state effects are responsible

A schematic diagram of a XPS setup is shown in Figure 3.5 Photoelectrons ted from the sample enters the collection lens system before entering the hemisphericalelectron energy analyzer

emit-CHAPTER 3 COMPOSITION OF THIN FILMS 17

ultrahigh vacuum pump

Ultrahigh vacuum pump Ultrahigh vacuum pump

10 torr

Vacuum pump

10 torr

Sample Introduction Chamber

Sample Crystal

X-ray Anode

Lens Sytem

Electrostatic hemispherical electron energy analyzer

Large Area Detector

In this experiment, the system used to obtain the spectra is VG ESCALab MKII A

Mg Kα X-ray source of 1253.6 eV was used to irradiate the sample Vacuum pressurewas maintained at 1 × 10−9 mbar during data acquisition The system was operated inconstant pass energy mode where the pass energy was set at 50 eV for wide scan and 20

eV for high resolution scans The photoelectron take-off angle is 75◦ to the surface plane.Charging effect was corrected by using the binding energy of adventitious carbon (284.8eV) as an internal standard The spectra were collected and peak fitting was carriedout using XPSpeak software [14] Shirley background subtraction was used to correct fornoise due to inelastic scattering [15] The atom fraction of a sample was then determinedusing

Cx = Ix/Sx

P Ix/Sx

(3.6)where Ix is the number of photoelectrons per second in a specific spectra and Sx is thesemi-empirical atomic sensitivity factor of the element

CHAPTER 3 COMPOSITION OF THIN FILMS 18

Figure 3.6 shows the Hf4f, Al2p, and O1s spectra obtained for the various HfO2-Al2O3

samples A shift to higher binding energy with increasing Al2O3 concentration was served for all three spectra This indicates some extent of charge transfer from the Hfand Al cations to the oxygen anions

In the experiment, the HfO2-Al2O3 samples were first subjected to wide scans (0-1100eV) to determine the elements present and their positions before high resolution scans wereperformed Figure 3.7 shows a wide scan obtained for sample 1 The photoelectron peaksfor Hf, O and C were indexed with reference to data from known databases [16, 17] The

CHAPTER 3 COMPOSITION OF THIN FILMS 19oxygen KLL and carbon KVV auger peaks were also identified and indexed accordingly.

Hf Sample 1

12.3 14.3 16.3 18.3 20.3 22.3

B.E/eV 24.3

4f 7/2 4f 5/2

526.3 528.3 530.3 532.3 534.3 536.3

B.E/eV

O-Hf

O-C

1s

Using XPSpeak, the Hf4f spin doublet, 4f5 /2 and 4f7 /2 were fitted with constraints

of the same FWHM and a spin-orbit area ratio of 3:4 For sample 1, pure HfO2, thebinding energy of the Hf 4f7 /2 peak is 16.7 eV and the spin orbit splitting is 1.66 eV,

in good agreement with known data [16] Two peaks were fitted for the O1s core levelspectra The peak at ∼529.9 eV is attributed to Hf-O bond, while the peak at ∼531.5 eV

is attributed to O-C bond resulting from surface contaminants For samples 2 to 5, anadditional peak due to Al-O bond (∼530.8 eV) is also fitted for the O1s core level spectra

CHAPTER 3 COMPOSITION OF THIN FILMS 20

These peak fitting results can be found in Appendix A Table 3.2 shows the stoichiometryobtained for the samples

in the RBS technique In addition, low concentration of Al in the alloy samples furtheraggravated the situation Nonetheless, results for sample 1 and 5, pure Al2O3, were wellcorrelated using both techniques

However, comparing the Hf:Al cation fraction with Table 2.2, it is clear that thetargeted composition was not achieved A much higher fraction of Hf was deposited thanexpected This suggests that the HfCl4:TMA pulse ratio needs to be adjusted in order toobtain the desired composition

Using both XPS and RBS, the composition of sample 1 to 5 are obtained Hf:O ratio

in sample 1, pure HfO2, is approximately 1:2 Hence, the as-deposited thin film hasgood stoichiometry Decreasing Hf and increasing Al atom fractions were obtained forsample 2 to 5, HfO2-Al2O3 alloy samples Al:O ratio for sample 5, pure Al2O3, is approx-imately 0.33:0.67 in both techniques This suggests that the film is non-stoichiometric

CHAPTER 3 COMPOSITION OF THIN FILMS 21

and deposition conditions needs to be further optimized From the RBS spectra obtained,flat-topped Hf peaks were also observed for samples 1 to 4 This suggests that the sampleshave uniform compositions throughout the films In addition, we note that the resultsusing XPS, which has a shallower sampling depth, were well correlated and quite close tothat obtained by RBS This further affirms the compositional uniformity of the samples

Chapter 4

Thin Films under Thermal

Processing

Silicon dioxide, in its amorphous form has been used by the semiconductor industry sinceits inception as the gate oxide in the FET structure Its robustness when subjected tothermal treatment and the formation of a stable interface with the silicon substrate makes

it the material of choice for many years Hence, it is desirable that the replacement gateoxide exhibits physical characteristics akin to that of SiO2 in these respects This will bebeneficial as it enables the industry to integrate the new material into the present processflow

MicrostructureThe microstructure of thin films for gate oxide applications falls mainlyinto three categories, namely: amorphous, polycrystalline and epitaxial For gate oxideapplications, most oxides with the exception of Al2O3crystallize at rather low temperature(Table 1.1) to form a polycrystalline thin film A material that is polycrystalline isnot ideal as grain boundaries can facilitate both mass and electrical transport throughthe material thus affecting device reliability Increased interface roughness due to grainboundaries will also have an adverse effect on channel mobility

In principle, problems with grain boundaries can be avoided by using epitaxial

single-22

CHAPTER 4 THERMAL STABILITY OF THIN FILMS 23

crystal oxides However, epitaxial oxides may have anisotropic properties and an epitaxialmatch with the substrate will limit the number of available oxides Furthermore, commer-cial growth techniques will have to be developed for integration into the manufacturingprocess Nevertheless, research in epitaxial oxides is still important as it might be of use

in later generations or in alternative device structures

Due to the problems that are likely to arise with the use of polycrystalline or ial oxides, potential replacement high κ material that can maintain an amorphous filmstructure when subjected to thermal cycling is still preferred

epitax-Thermal Stability The thermal stability of an alternative gate dielectric is crucial indetermining the overall electrical properties of the MOSFET It is important that thematerial remains stable throughout all of the thermal cycling required for CMOS process-ing Temperatures of up to ∼1000◦C are typically used to activate implanted dopants

in MOSFET It is desirable that the potential candidate remains amorphous up to thistemperature and not form any undesirable interface with the silicon substrate The for-mation of an interfacial layer can result in overall reduced capacitance of the gate stack

if the layer is of a lower κ or affect the mobility of the carriers in the channel Similarly,thermal stability is required for junction annealing cycles used after gate formation.During thermal processing, alternative gate materials should be resistant to interdifffu-sion between gate, dielectric and silicon substrate Dopant diffusion from the gate into thedielectric or channel can result in undesired shifts in threshold and flatband voltages Inthe case of metal oxides, metal outdiffusion from the dielectric into the channel will result

in mobility degradation of carriers Other impurities may include reaction by-products orprecursors constituents during deposition Hence, thermal stability of the material is animportant consideration for integration of the alternative dielectric

In this chapter, we will study the microstructure of the HfO2-Al2O3 alloy thin filmsusing glancing angle X-ray Diffraction (GIXRD) and examine its stability when subject

to thermal treatment Interdiffusion of impurities in the thin film as a result of thermaltreatment will be treated in the following chapter