Formation of ultra shallow junctions in silicon germanium by pulsed laser annealing

amorphous layers to define the depth of electrically active junctions.5 For the current text, the process of laser irradiating a doped amorphous layer to achieve full melt and subsequent

CHAPTER 1 – INTRODUCTION

Laser thermal processing (LTP) is a paradigm shift in the formation of shallow junctions in silicon The sub-processes of LTP are amorphization, dopant implant, laser irradiation, liquid melt and liquid phase epitaxy During laser irradiation the amorphous region is liquefied allowing the dopant atoms to distribute homogeneously throughout the melt by diffusion Upon rapid solidification, the dopants largely retain the even distribution and produce an abrupt junction that is box shaped, with the depth of the junction defined by the position of the amorphizing implant Additionally, due to rapid conductive cooling, the liquid phase epitaxy is non-equilibrium, resulting in the substitutional incorporation of the dopant atoms into the crystalline lattice in excess of equilibrium solid solubility

Since the invention of light amplification by stimulated emission of radiation or LASER in 1960, lasers have been used in a variety of ways to process semiconductors From as early as 1968 lasers were used to modify the electrical resistivity of semiconductors.1 By 1976 lasers were used to remove lattice damage caused by ion-implantation and to electrically activate dopants in a process termed laser annealing.2

In 1978 re-crystallization of an amorphous silicon film was achieved by irradiation with a single laser pulse.3,4 More recently lasers have been used to melt doped

amorphous layers to define the depth of electrically active junctions.5 For the current text, the process of laser irradiating a doped amorphous layer to achieve full melt and subsequent epitaxial re-growth resulting in the depth of the electrical junction being defined by the original amorphous layer thickness is termed laser thermal processing

The following five sections further describe laser thermal processing in terms of amorphization, dopant implantation, laser irradiation, liquid melt and liquid phase epitaxy

1.1.1 PRE-AMORPHIZING IMPLANT

The depth of the amorphous layer controls the depth of the melt Implanting the substrate with a high dose of a low energy, high mass ion, forms a shallow amorphous layer Amorphization is a result of the damage introduced to the crystalline lattice during ion implantation.6 For crystalline silicon, amorphization requires displacement of approximately 10~12 % of the lattice atoms Formation of an amorphous layer progresses gradually, as damage to the crystal lattice is accumulated with each implanted ion

The number of ions per unit area required to amorphize a layer is referred to as the threshold dose Amorphizing implants must have an implant dose higher than the threshold dose The threshold dose for an ion with low mass is larger than that for ions with a high mass The depth of the amorphous layer is determined by the energy of the implant ion Ions with low implant energy do not travel as deep into the substrate as high-energy implants; consequently, low energy implants result in shallower amorphization depths

1.1.2 DOPANT IMPLANT

The dopant is implanted into the amorphous surface layer at a low energy and high dose Implantation of the dopant is consistent with conventional ion implantation7 The implant energy of the dopant ions needs to be sufficiently low to confine the dopant

to the amorphous layer, thus maintaining an electrical junction defined by the amorphous depth Additionally, the dose of the dopant needs to be high enough, in order to provide sufficient population of carriers to achieve a low sheet resistance

1.1.3 LASER IRRADIATION

Lasers provide a high power-density energy source that is both monochromatic and coherent, which can be controlled to achieve nanosecond pulse durations For laser thermal processing several types of lasers have been used, differing primarily in wavelength (e.g., XeCl 3088, 9 , frequency doubled Nd:YAG 53210 and Nd: YAG 1064 nm) The optical properties of the material; specifically reflectivity, absorption and absorbance are wavelength-dependant Reflectivity is the amount of incident energy that

is reflected from the surface of the material relative to the incident intensity Absorption determines the degree to which the amplitude of the radiation is damped as a function of distance into the sample

Output of the laser is directed through optics that homogenize, focus and define the cross-sectional area of the laser irradiation In the irradiated area of the sample,

absorption of laser energy by the silicon may result in emission of phonons that are expressed as thermal heating11 Absorption occurs when a photon hits an electron and the electron absorbs the energy and as a consequence moves up to a higher energy state The electron can transition between energy bands by a direct transition or indirect transition

A direct transition produces a photon that is reflected, while an indirect transition produces a photon and a phonon It is the phonon that produces the effect of heating For amorphous silicon, the transition mechanism is predominately indirect, thus resulting in phonon induced lattice heating

1.1.4 LIQUID MELT

Heating from the laser provides the thermal budget for melting, diffusion and electrical activation Amorphous silicon melts at approximately 200 ± 50 K lower than that of the crystalline silicon.12 Laser thermal processing utilizes this melting point difference to establish a process window As the laser energy density is increased, a threshold is reached at which the amorphous silicon begins to melt Further increase in the laser energy density increases the melt depth, until the entire amorphous layer is melted and the original amorphous/crystalline interface is reached The melt does not proceed into the crystalline material until sufficient thermal energy is added to raise the liquefied amorphous silicon temperature to the melting point of crystalline silicon The difference in laser energy density required to completely melt the amorphous layer and the laser energy density required to begin melting the crystalline silicon is termed the process window.10

Dopants distribute homogeneously throughout the liquid melt The diffusion coefficient of dopant atoms in liquid silicon is approximately eight orders of magnitude greater than that in crystalline silicon For example the diffusion coefficient of boron in liquid silicon is [3.3 ± 0.4]x10-4 cm2/s 13, while the diffusion coefficient in crystalline silicon is approximately 1.0x10-15 cm2/s The large diffusivity of the dopant atoms in the liquid region results in a rapid redistribution that tends to homogenize the dopant distribution throughout the melt In contrast, the amount of diffusion that occurs in the solid is negligible in comparison; thus in terms of dopant distribution, an abrupt box-like junction is formed at the maximum melt depth14

1.1.5 LIQUID PHASE EPITAXY

Liquid phase epitaxy (LPE) is the commensurate growth of a solid from a liquid, based on the crystal structure of the seed or substrate material The process of liquid phase epitaxy begins as one atom, in order to minimize free energy, attaches itself to a metastable lattice site on the surface of the seed material Subsequently, other atoms attach to the ledge or kink site formed by the original atom making it more stable Atoms continue to attach to the newly formed ledge sites until an atomic layer is formed This process continues until the liquid phase has been completely regrown

Liquid phase epitaxy results in the incorporation and re-distribution of dopant atoms, as the melt solidifies For a laser pulse with duration less than 18 ns, the melting and solidification times are less than 100 ns in silicon, resulting in rapid regrowth of the

melt Rapid regrowth is driven by the large thermal gradient that exists in the wafer due

to the surface being at the melting temperature of amorphous silicon (1480 ± 50 K)12, while the back side of the wafer is at ambient temperature (~298 K) This thermal gradient drives liquid phase epitaxy of the silicon with melt front velocities of 2 m/s to 4.5 m/s15 Large regrowth velocities can increase the segregation coefficient (k = Cs/Cl) of dopants by several orders of magnitude above equilibrium values The segregation coefficient increases as the regrowth velocity exceeds the diffusion velocity of the dopant impurities in the liquid The increase in the segregation coefficient results in a higher concentration of dopants incorporated into the solid Hence, high regrowth velocities achieve higher dopant concentrations Laser thermal processing utilizes high regrowth velocities to produce junctions that are highly doped with correspondingly low sheet resistance values

Excimer laser annealing (ELA) was thoroughly investigated in the 1970–1980s and is a long known approach for annealing of ion implanted Si Today there is a renewed interest from the semiconductor community for a possible application of ELA to sub-70 nm technology ELA offers considerable advantages over the conventional rapid thermal annealing (RTA) It provides the possibility to obtain ultra-shallow junctions, to remove completely the implantation defects and to form more abrupt dopant profiles with higher dopant activation efficiency The annealing of implantation induced damage during ELA occurs due to melting of the damaged region and subsequent liquid-phase re-

crystallization of the melt The process is characterized by extremely short times of phase transformation, with typical laser pulses of 10–30 ns and the melting time within the microsecond range, which results in a very localized effect of ELA on the material, where all the modifications take place mainly within the melting region Because of this highly non-equilibrium nature, a very high dopant incorporation efficiency can be obtained with the dopant concentration exceeding the solid solubility level.16

Pulsed excimer laser annealing has been proven to be effective for reducing the poly-Si gate depletion effect to less than 1 Å High active boron concentration and low resistivity can be achieved without boron penetration as can be seen from the work by Wong et al 17 Gate oxide quality is preserved, and ELA is found to be more compatible with HfO2 gate dielectric than RTA due to the ultrashort melt time Therefore, ELA is promising for application to sub-65-nm bulk-Si CMOS technologies.17

Future semiconductor technology nodes are not achievable with current processing technologies Shrinking of device dimensions is limited by the ability to fabricate doped junctions that are shallow, abrupt and highly activated In conventional processing, achieving high activation is at odds with achieving a shallow junction High activation is generally achieved by high thermal budgets, which cause the dopant to diffuse, thus increasing the final junction depth In contrast, a low thermal budget limits diffusion of the dopant but yields a low activation Consequently, new processing

techniques need to be developed which are capable of producing device features in compliance with the requirements of future technology nodes

A potential candidate for reaching future technology nodes is laser annealing Laser annealing, in which the laser melts the surface layer of silicon and causes the dopants to be distributed uniformly within the melted region, produces abrupt, highly activated and ultra-shallow junctions The degree of melting is determined by the extent

of laser absorption and rate of heat dissipation, which are dependent on the substrate properties Laser annealing with extremely low energy is required to guarantee enough process windows Laser annealing offers several advantages over conventional processing techniques:

1) the junction depth is defined by the amorphous/crystalline interface Laser annealing, combined with pre-amorphization implant (PAI) has been reported to control ultra-shallow junction depth Since an amorphous silicon layer has approximately 300°C lower melting temperature than that of crystalline silicon, laser annealing energy can be reduced

by PAI

2) the amorphous region is regrown by liquid phase epitaxy providing an even distribution of the dopant across that junction, resulting in a box-like junction The amorphous silicon thickness and melted thickness are precisely controlled by PAI and the dopant diffusion occurs within the melted layer during very short laser annealing period,

so that a box-like profile is formed

Fig 1.1: Sheet resistance versus junction depth for p-type dopants activated by rapid thermal annealing, Levitor, Flash, SPER and LTP The solid solubility line represents an impurity concentration of 2.0x1020/cm3, which is obtained at a temperature of 1050 oC for boron.18

Fig 1.2: Vertical junction abruptness for p-type dopant gradients achieved by rapid thermal annealing, Levitor, Flash, SPER and LTP 18

According to the International Technology Roadmap for Semiconductors (2007)19, the source/drain extensions shallower than 14 nm will be required In addition, the sheet resistance must be lower than 900 and 400 ohms/sq for PMOS and NMOS, respectively Rapid-thermal-annealing (RTA)-based technology with low-energy ion implantation is the standard for Ultra-Shallow Junction (USJ) formation However, the obtainable junction depth and sheet resistance with these technologies will not meet the requirements for upcoming technology nodes For RTA-based annealing methods, wafer heating and cooling times are on the order of seconds Consequently, it is difficult to restrain the thermal diffusion of dopants, particularly the diffusion of boron In addition, the sheet resistance is limited by the thermal equilibrium solid solubility of dopants Various methods to form USJs have been reported as alternatives, including atomic layer doping, plasma doping, flash lamp annealing and laser annealing

These methods have their own advantages and disadvantages For example, the junction depth obtainable by atomic layer doping satisfies the demands of some of the next generation devices; however, its sheet resistance is too high for practical applications Laser annealing combined with low-energy ion implantation has been reported to be an effective method for simultaneously achieving a shallow junction depth and a low sheet resistance However, for actual production, the following problems still remain; the leakage current caused by residual defects and the difficulties of integration into the device fabrication processes.20

1.4 ADVANCED FRONT-END PROCESSES FOR THE 45NM CMOS TECHNOLOGY NODE

For the PMOS transistor, co-implantation or cocktail implants of combinations of germanium (Ge), boron (B), fluorine (F) and carbon (C) have been shown to hold great promise in improving the abruptness, increasing the activation and limiting the diffusion

of boron For the 45 nm technology, diffusion-less anneal will be paramount One potential annealing technology is Laser Thermal Processing (LTP) The specific implant conditions will also need to be examined in order to achieve the maximum improvement

as RTP-based solutions are not necessarily transferable

The results of a recent experiment for a millisecond laser anneal are shown in figures 1.3 (a) and (b) for 1c1015

cm 2 and 2c1015

cm 2, 500eV B implants The sheet resistance decreases with higher Ge+ PAI implant energy, from 790 Ω/sq at 2 keV Ge+ to

548 Ω/sq at 10 keV Ge+ for a boron dose of 1c1015

cm 2 The Ge+ dose has little effect

on the measured sheet resistance, as the data for 5keV Ge+ in Fig b demonstrate This is because for the Ge energies used here, the amorphous layer thickness does not increase significantly above a dose of 1c1015

cm 2 The weak sensitivity to dose and the fact that the two B doses have different optimum pre-amorphization conditions (see Fig b) point

to the importance of the relative positioning of both implants

The lowest Rs is now produced using a 10 keV Ge+ pre-amorphization implant, different from RTP anneal This is because the damage evolution is very different between a regular RTP and a millisecond anneal For the latter, all the implant damage beyond the amorphous/crystalline interface remains It is well known that this damage can prevent activation through clustering or even de-activate B For a deeper amorphous layer, the damage overlap with the B profile will be reduced, thus leading to better activation and lower Rs

Fig 1.3: (a) Sheet resistance as a function of Ge pre-amorphisation implant energy (b) Sheet resistance as

a function of Ge implant dose for 5 keV Ge with 500 eV B implants at 1

1015 cm 

2 and 2

1015 cm 

2 21

Fig 1.4 shows the B and Ge SIMS profiles for three different conditions:

1015 cm 

2

B dose, and (C) with 10 keV Ge at 2

1015 cm 

2 and 2

1015 B dose 21

1.5 MOTIVATION BEHIND THESIS

With the continuous scaling of the device dimensions, there is a need for shallow, and highly active junctions in order to avoid short channel effects and to decrease the resistance of the S/D junctions The ideal profile is box-like with a high activation, high abruptness and showing almost no diffusion with respect to the as implanted profile Therefore, in recent years, a huge effort has been done to drastically reduce the thermal budget for dopant activation A decrease of the anneal temperature can

ultra-be combined with longer anneal time (i.e the concept of solid phase epitaxial re-growth, SPER, where activation is limited to a shallow amorphized top layer) On the other hand, the dwell time can be drastically reduced and this is done using techniques such as flash and laser annealing where typical dwell times are in the order of msec for flash and < 1 msec for laser annealing

In the recent work by K.K Ong et al,22 it is shown that laser annealing has a strong dependence on the thermal conductivity of the substrate used which causes a great challenge in maintaining the integrity of the strained-Si layer during shallow-melt laser annealing They have shown that non-melt laser annealing produces dopant profiles of negligible diffusion and improved activation in the strained-Si/SiGe substrate From their work, it can be concluded that non-melt laser annealing is an attractive process in the dopant activation of thermally less conductive semiconductor substrates

In the work by Susan Earles, Mark Law et al,23 it is shown that for a shallower implant, Non-melt Laser Annealing (NLA) shows reduced junction depth and sheet resistance The NLA also repairs crystalline damage from the implant resulting in

improved mobilities and lower leakage currents, making NLA a suitable method for satisfactory layers for shallower implants

In the work by G Fortunato and L Mariucci24, it is shown that for the formation

of ultra-shallow junctions it is essential to combine ELA with ultralow energy ion implantation to avoid producing a deep tail of active boron, preventing the possibility to form abrupt and ultra-shallow junctions

The impact of sub-melt laser annealing on embedded Si1-x,Gex and source/drain defectivity was studied by E Rosseel et al25, by means of transistor data and blanket layer studies In their work, it is concluded that if the Si1-xGex source/drain and laser anneal modules in the process flow are not optimized, defects can be introduced into the Si as a result of the laser induced enhanced thermal stress These defects give rise to enhanced junction leakage By means of blanket Si1-x,Gex stacks, it was demonstrated that the laser peak temperature onset for defect formation depends on the dwell time and Si1-xGex stack parameters They have shown that the possible peak temperature budget increases for shorter dwell time, lower Ge content or smaller layer thickness Hence it is better to have

a smaller layer thickness, shorter laser annealing dwell time and lower Ge content for a better temperature budget for less defect formation

From the work done by Poon, Tan et al,27 it was found that despite the many advantages of laser annealing using a moderate laser fluence, subsequent thermal treatment results in junction deepening due to Transient Enhanced Diffusion (TED) as a result of the End-Of-Range (EOR) defects which cannot be efficiently removed, since the

melt depth for moderate laser fluence is not beyond the amorphous layer To make the multiple-pulse laser annealing technique more attractive, it is thus essential to remove the EOR defects without melting beyond the amorphous layer Melting beyond the amorphous layer would compromise the multiple-pulse LA technique since junction depths would then be controlled by the laser fluence instead of the pre-amorphized depth.26,27 Some work has been carried out to understand and reduce the formation of extended defects that influence boron diffusion in silicon during conventional annealing techniques like RTA However, none of these techniques have been applied to laser annealing.28

Based on the above results by previous research work, the objective of this work

is to study the activation of p-type ultra-shallow implants on blanket Si (100) by submelt laser annealing while making use of the fact that the melting point of the SiGe alloy is lower than that of c-Si, thus giving a process window within which the melt depth and hence junction depth is controlled by the thickness of the SiGe layer instead of the laser fluence

The application of sub-melt or non-melt laser annealing in sub-30nm MOSFET fabrication technology is proposed by M Narihiro, T Iwamoto, et al.29 It is shown that the two major features of the laser annealing, i.e diffusion-less and higher dopant activation enable us to apply more elaborate channel engineering, involving multiple halo implantations and optimized gate-predoping, that contributes further scaling of a functional gate-length (Lg) and effective gate-insulator thickness (Tin,), maintaining

sufficient current drivability prior to any local stress engineering applied Hence this method is promising for future technology nodes

1.6 EXPERIMENTAL PROPOSAL

Fig 1.5: Proposed Experimental Structure

Figure 1.5 shows the structure for the proposed scheme where a 30nm thick SiGe epitaxial layer is grown on the bulk Si substrate The process flow outlining the various stages of the experiment, is shown in Fig 1.6

In this experiment, an undoped Si0.8Ge0.2 epitaxial layer, 30 nm thick, was grown

on a 300 mm n-type (100) Si wafer A 1 keV, 1x1015 cm2 Ge implant was carried out, followed by a 0.5keV, 5x1014 cm2 B implant The purpose of the Ge implant was to create a thin amorphized region near the surface of the SiGe epitaxial layer to reduce the effect of channeling during the subsequent B implant

After the implantation into the thin SiGe epitaxial layer, laser annealing was carried out using a 248 nm KrF excimer laser with a pulse duration of 23 ns and a

repetition rate of 1 Hz, at laser fluence ranging from 0.55 to 0.95 J/cm2 to activate the dopants and create ultra-shallow junctions with low resistance

Fig 1.6: Process flow showing various stages of fabrication

In this proposed scheme, the melt depth and hence the junction depth are independent (within limits) of the laser fluence This makes use of the fact that the melting point of the SiGe alloy is lower than that of c-Si, thus giving a process window within which the melt depth is controlled by the thickness of the SiGe layer instead of the laser fluence The main objective is to ensure that the laser annealing should not lead to

overmelt of the crystalline region, and to maintain the dopants within the SiGe epitaxial layer for ultra-shallow junction formation

1.7 THESIS OUTLINE AND ORIGINAL CONTRIBUTIONS

The objective of this document is to describe the electrical and physical characteristics of junctions formed by annealing of silicon-germanium layers doped with boron Chapter 1 presents the motivation for the work by reviewing the trends in scaling and highlights the associated advantages of laser annealing Chapter 2 presents the simulation work done to determine process variables prior to the actual fabrication Chapter 3 describes the experimental matrices and explains the basic principles of the analytical techniques used for characterization Chapter 4 describes the initial investigation work using sub-melt laser annealing Finally, Chapter 5 presents the experimental results and discussion, summarizing the experimental findings of this work

CHAPTER 2 – SIMULATION WORK

Simulation is a fast method to estimate process fabrication variables such as ion implantation energy and dopant dosage before carrying out actual fabrication Furthermore, simulation is an essential tool to investigate the effect that changes on the substrate have, on the substrate material and electrical properties It allows room to explore more choices for laser annealing enhancement such as the reduction of defect density and surface roughness of the surface after ion implantation Three process simulators were used – SRIM, TRIM and TAURUS The objective is to simulate the fabrication processes used in the making of semiconductor devices In this work, the focus is on ion-implantation

SRIM 30 (Stopping and Range of Ions in Matter) is a group of programs which calculate the stopping and range of ions into matter using a quantum mechanical treatment of ion-atom collisions This calculation is made very efficient by the use of statistical algorithms which allow the ion to make jumps between calculated collisions and then averageing the collision results over the intervening gap TRIM (Transport of Ions in Matter) is the most comprehensive program included TRIM will accept complex targets made of compound materials with up to eight layers, each of different materials TRIM uses the Monte Carlo method to calculate both the final 3D distribution of the ions

sputtering, ionization, and phonon production All the target atom cascades in the targets are followed

TAURUS31 is a process simulator that simulates each of the process steps carried out with the help of a variety of mathematical models Numerical solutions of the model equations are computed over a triangular grid or mesh structure consisting of triangular elements and nodes Taurus-Visual lets us visualize data from physical simulation software tool or other sources in one, two, and three dimensions It is an interactive visualization tool to analyze physical simulation results and manipulate the resulting plots

to gain a new perspective Taurus-Visual allows us to visualize fields, geometries and regions, visualize the properties in a structure, e.g P/N junctions and depletion layers, view I-V curves and doping profiles, annotate plots: modify axis and grid, change placement and style for labels and legends, zoom, pan and rotate images, animate images

to show plot sequences in movie mode and extract data using rulers and probes

From a review of the literature, researchers make use of a dose level range of 1x1015cm-2 to 5x1015cm-2 and an energy range from 10keV to 30keV for initial ion implantation If germanium is involved in initial implantation, a second ion implantation

is at a dose level range of 1x1014cm-2 to 5x1015cm-2 and an energy range from 0.5keV to 20keV In ‘Variation of end of range defects density with ion beam energy and dose: Experiments and simulations’ by L.Laânab et al,32 they had experimented with different doses and energies for germanium ion implantation to determine the density of End of Range defects (EOR) The disappearance of dislocation loops was observed when the

energy of the implantation fell to 15keV Further observation was made that when the dose level decreased, the density of EOR decreased as well Since Ge was implanted for pre-amorphising purposes only, 1 keV should be sufficient energy as confirmed from simulation results

In this work, simulation was carried out to determine the optimal Ge concentration, B and Ge implant doses, oxide and Si-Ge epitaxial layer thicknesses that would provide us with high quality, highly activated and ultra-shallow junctions

2.2 SIMULATION RESULTS USING SRIM2006 SOFTWARE

2.2.1 GE PRE-AMORPHISATION IMPLANT (1 KEV)

Fig 2.1: SRIM Simulation Results showing Ion Range, Straggle, and Recoil Energy when Ge Amorphising Implant of 1keV is implemented into a 30nm thick SiGe epitaxial layer

Pre-From the above simulation results, it is shown that Ge pre-amorphisation implant at 1keV would confine the Ge atoms implanted, well within the 30nm epitaxial layer The ion range is found to be 3.6nm with a straggle of 1.7nm This implant is very shallow as only

a thin amorphous layer near the surface is needed to prevent boron ion channelling during the subsequent B implant

2.2.2 GE PRE-AMORPHISATION IMPLANT (2 KEV)

Fig 2.2: SRIM Simulation Results showing Ion Range, Straggle, and Recoil Energy when Ge Amorphising Implant of 2keV is implemented into a 30nm thick SiGe epitaxial layer

Pre-From the above simulation results, it is observed that the Ge implant ion range is found to be 5nm with a straggle of 2.4nm Although the Ge pre-amorphisation implant at 2keV would confine the Ge atoms implanted within the 30nm epitaxial layer, the higher energy of the implant would cause more implant damage Hence, it is more prudent to implant the Ge at 1keV since the Ge implant is purely for pre-amorphisation purposes

2.2.3 BORON IMPLANT (0.5KEV)

Fig 2.3: SRIM Simulation Results showing Ion Range, Straggle, and Recoil Energy when 0.5keV boron is implanted into a 30nm thick SiGe epitaxial layer

It is important to confine the implanted boron atoms well within the SiGe epitaxial layer TRIM software also takes into account the effects of channeling, hence increasing the reliability of this simulation results An ultra-shallow boron implant at 0.5keV is selected

as it fulfills the purpose of this research work to obtain highly-activated and ultra-shallow junctions

2.3 SIMULATION RESULTS USING TAURUS SOFTWARE

2.3.1 30 NM THICK Si0.8Ge0.2 EPITAXIAL LAYER

Fig 2.4: TAURUS Simulation Results showing boron dopant profile when 5 x 1014cm-2 boron is implanted into a 30nm thick Si 0.8 Ge0.2 epitaxial layer

Fig 2.5: TAURUS Simulation Results showing germanium concentration profile when 1 x 1015cm-2germanium is implanted into a 30nm thick Si 0.8 Ge0.2 epitaxial layer

2.3.2 30 NM THICK Si0.7Ge0.3 EPITAXIAL LAYER

Fig 2.6: TAURUS Simulation Results showing boron dopant profile when 5 x 1014cm-2 boron is implanted into a 30nm thick Si0.7Ge0.3 epitaxial layer

Fig 2.7: TAURUS Simulation Results showing germanium concentration profile when 1 x 1015cm-2germanium is implanted into a 30nm thick Si 0.7 Ge0.3 epitaxial layer

From the above comparison of the TAURUS simulation results, it is found that the optimal value of x in Si1-xGex epitaxial layer is x=0.20 as 20% Ge is sufficient for the containment of the dopants within the 30nm SiGe epitaxial layer, at the proposed implant conditions

In addition, it is important to note that E Rosseel et al,25 have shown that the possible peak temperature budget increases for shorter dwell time, lower Ge content or smaller layer thickness Hence it is better to have a smaller layer thickness, shorter laser annealing dwell time and lower Ge content for a better temperature budget for less defect formation

Hence, from the simulation results, the optimal boron dose and energy were found

to be 5 x 1014cm-2at an energy of 0.5keV and Ge pre-amorphization implant dose at 1 x

1015cm-2at an energy of 1keV for ultra-shallow junctions The Si-Ge epitaxial layer was selected to be 30nm thick to provide high quality, highly activated and ultra-shallow junctions

CHAPTER 3 – EXPERIMENTAL METHODS

The excimer laser used in the experiment is the Novaline 100 Excimer laser at the Singapore Institute of Manufacturing Technology (SIMTECH), which is a 248 nm KrF laser with pulse duration of 23 ns This laser system has a homogenized, uniform and flat-top beam profile and is equipped with a CCD camera for sample alignment and monitoring, and a high resolution (1 µm) xyz moving stage

This system operating on the projection image machining mode, has a mask stage (250x250mm) scannable synchronized with work stage (200x200mm) motion and is also equipped with an energy attenuator providing very subtle energy control without distorting or deflecting the laser beam

The laser wavelength determines the coupling of the irradiated material with the incident light, i.e., the absorbed energy and its depth distribution The energy density involved in the annealing of the silicon substrates range from 0.55 to 0.95 J/cm2 The controlling factors are the attenuation and the laser energy of the system to provide the laser fluence needed The chamber is purged with N2 gas to provide an inert ambience

As the laser beam size is small, in order to anneal an entire wafer, the laser beam would scan the whole wafer, annealing a small portion each time To achieve this, the

sample is placed on a motorized stage controlled using a computer, where after every laser shot on the wafer, the stage automatically moves by a preset value (corresponding to the size of the beam) to the next area for annealing

The four-point probe technique is the most common method for the measurement

of semiconductor sheet resistance The four-point probe consists of four probe tips arranged in a linear fashion, where the two outer probes are connected to a current supply and the inner probes to a voltage meter Typical probe spacing is in the range of 1 mm

Figure 3.1 shows the schematic diagram of a four-point probe set-up where S is the uniform probe spacing For an accurate measurement, the substrate must be placed such that the probe contacts are in the center of the wafer The sheet resistance Rs

(measured in ohms per square) characterizes thin semiconductor sheets or layers, such as diffused or ion-implanted layers For a layer of thickness ≤ S/2,

Fig 3.1: Schematic drawing of a four-point probe set-up 33

The Hall Effect is used to measure the conductivity type, active carrier concentration and mobility of a semiconductor sample at a given temperature Figure 3.2 illustrates the typical Hall measurement setup where, by forcing a constant current, I, through a sample within an orthogonal magnetic field, B, free charge carriers (namely electrons and holes) experience a Lorentz force FL= q(v × B), where v is the drift velocity

of the carriers and q the electronic charge When either carrier type is dominant, the accumulation of internal charge induces a steady state Hall voltage, VH, whose sign determines the carrier type Balancing the magnetic and electrostatic forces on a mobile charge, the following equation can be written:

W

VqF

FL= e= H

Fig 3.2: Schematic drawing illustrating the Hall Effect system

Expressing the current in terms of the drift velocity,

vtWqn

I = × × • ×

where n is the density of charge carriers and W• t the cross-section area through which the current passes Defining the Hall coefficient RH and combining the two equations above,

BI

tVqn

From RH, the sheet Hall coefficient RHSis given by:

BI

Vt

S

HS S

R

ρ

µ =

where the Hall sheet resistivity ρs

can be obtained using van der Pauw’s method In this method, a current source is applied to contacts 1 and 2 and the voltage measured across 4 and 3 to obtain Rx, after which, the current source is applied to contacts 1 and 4 and the voltage across 2 and 3 are measured to obtain Ry

The figure below illustrates the resistivity measurement, where contacts 1 to 4 are located

on the corners on the square sample

2532

3.4 SECONDARY ION MASS SPECTROMETRY

Secondary Ion Mass Spectrometry (SIMS) is predominately used for general surface analysis and for depth profiling It is based upon the bombardment of the sample surface with a beam of ions such as Ar+, Cs+or O2+, accelerated to an energy between 0.5

to 20 keV by applying a high dc potential

The impact of these primary ions causes the surface layer of atoms of the sample

to be sputtered off A small fraction of these atoms is ejected as positive (or negative) secondary ions that are drawn into a spectrometer for mass analysis The mass/charge ratio of the sputtered ions is analyzed and by monitoring the peaks of one or a few isotopes, as a function of time, the concentration profiles of the dopant can be obtained with a depth resolution of 5 to 10 nm Detection limits for some of the elements are in the range of 1014to 1015cm-3because of the small background interference signal

For our experiments, the raw data obtained from SIMS is in the format of secondary ion yield in counts versus sputtering time To convert such a profile to one of dopant concentration versus depth requires the use of standards with composition and matrices identical or similar to that of the unknown Conventionally, such standards are prepared via ion implantation where implantation dose can be accurately controlled to an accuracy of 5% or better When such a standard is measured, the SIMS system can be calibrated by integrating the secondary ion yield signal over the entire profile Calibrated standards are, therefore, very important for accurate SIMS measurements The standard

sample used in our study was prepared via ion implantation with a known dose The to-depth conversion on the other hand, is done by measuring the sputter crater depth using a depth profiler after the analysis is completed

time-3.4.1 TIME-OF-FLIGHT SIMS (TOF-SIMS)

Fig 3.4: Particle beam interaction using ToF-SIMS Incident particles bombard the surface liberating single

ions (+/-) and molecular compounds.34

ToF-SIMS uses a focused and pulsed particle beam (typically Cs or Ga) to dislodge chemical species on a materials surface Particles produced closer to the site of impact tend to be dissociated ions (positive or negative) Secondary particles generated farther from the impact site tend to be molecular compounds, typically fragments of much larger organic macromolecules The particles are then accelerated into a flight path on their way towards a detector Because it is possible to measure the "time-of-flight" of the particles from the time of impact to the detector on a scale of nano-seconds, it is possible to produce a mass resolution as fine as 0.00X atomic mass units (i.e one part in a thousand

of the mass of a proton) Under typical operating conditions, the ToF-SIMS results analysis includes:

1 a mass spectrum that surveys all atomic masses over a range of 0-10,000 amu,

2 the rastered beam produces maps of any mass of interest on a sub-micron scale, and

3 depth profiles are produced by removal of surface layers by sputtering under the ion beam

ToF-SIMS is also referred to as "static" SIMS because a low primary ion current

is used to "tickle" the sample surface to liberate ions, molecules and molecular clusters for analysis In contrast, "dynamic" SIMS is the method of choice for quantitative analysis because a higher primary ion current results in a faster sputtering rate and produces a much higher ion yield Thus, dynamic SIMS creates better counting statistics for trace elements

The Transmission Electron Microscope (TEM) allows the users to determine the internal structure of materials, either of biological or non-biological origin Materials for TEM are specially prepared to thicknesses which allow electrons to transmit through the sample, much like light is transmitted through materials in conventional optical microscopy Because the wavelength of electrons is much smaller than that of light, the

optimal resolution attainable for TEM images is many orders of magnitude better than that from a light microscope Thus, TEMs can reveal the finest details of internal structure - in some cases as small as individual atoms

Fig: 3.5: Schematic drawing of the TEM equipment35

In this work, a relatively new technique called the focused ion beam (FIB) technique is used to prepare thin samples for TEM examination from larger specimens Because the FIB can be used to micro-machine samples very precisely, it is possible to mill very thin membranes from a specific area of a sample Materials that have dimensions small enough to be electron transparent, can be quickly produced by the deposition of a dilute sample containing the specimen onto support grids The suspension

is normally a volatile solvent, such as ethanol, ensuring that the solvent rapidly evaporates allowing a sample that can be rapidly analysed

Imaging techniques such as the HR-TEM are particularly important in the study

of materials Faults in crystals affect both the mechanical and the electronic properties of materials, so understanding how they behave gives a powerful insight By carefully selecting the orientation of the sample, it is possible not only to determine the position of defects but also to determine the type of defect present

The HR-TEM technique allows the direct observation of crystal structure and there is no displacement between the location of a defect and the contrast variation caused in the image However, it is not possible to interpret the lattice images directly in terms of sample structure or composition This is because the image is sensitive to a number of factors (specimen thickness and orientation, objective lens defocus, spherical and chromatic aberration), and although quantitative interpretation of the contrast shown

in lattice images is possible, it is inherently complicated

The AFM is a scanned-proximity probe microscope Such a microscope works by measuring a local property - such as height, optical absorption, or magnetism - with a probe or "tip" placed very close to the sample The small probe-sample separation (on the order of the instrument's resolution) makes it possible to take measurements over a small

area To acquire an image the microscope raster-scans the probe over the sample while measuring the local property in question In this case, the AFM measures the surface roughness of the samples

Fig: 3.6: Schematic drawing of the AFM equipment36

The AFM consists of a microscale cantilever with a sharp tip (probe) at its end that is used to scan the specimen surface The cantilever is typically silicon or silicon nitride with a tip radius of curvature on the order of nanometers When the tip is brought into proximity of a sample surface, forces between the tip and the sample lead to a deflection of the cantilever according to Hooke's law The AFM can be operated in a number of modes, depending on the application In general, possible imaging modes are divided into static (also called Contact) modes and a variety of dynamic (or non-contact) modes

The AFM has several advantages over the scanning electron microscope (SEM) -

It does not require vacuum conditions and provides a true 3-D surface profile However, a