Chapter 4

Performance Analysis

4.1 Introduction

The past few years have seen significant progress in the development of SiGe heterojunction bipolar transistor (HBT) technology. Today, the use of SiGe-base HBTs is becoming increasingly popular in wireless and high-speed digital communications. The most significant material parameter to be specified in the simulation of SiGe HBTs is the bandgap narrowing induced by incorporation of a Ge fraction in the base. In addition to the Geinduced bandgap narrowing, the high doping in the base induces additional bandgap narrowing, similar to that observed in silicon. Although several bandgap narrowing model and its affect have been proposed for silicon, the new effects and nuances of operation of SiGe HBT are still being uncovered and as transistor scaling advances with different application targets steadily increasing, the comprehensive treatment of its working is still desired. While designing the SiGe HBT, doping is considered a critical issue as it affects bandgap narrowing. In lightly doped semiconductors the dopant atoms are sufficiently widely spaced in the semiconductor lattice that the wave functions associated with the dopant atoms electrons do not overlap. The energy levels of the dopant atoms are therefore discrete. Furthermore, it is reasonable to assume that the widely spaced dopant atoms have no effect on the perfect periodicity of the semiconductor lattice, and hence the edges of the conduction and valence bands are sharply defined. In heavily doped semiconductors the dopant atoms are close enough together that the wave functions of their associated electrons overlap. In addition, the large concentration of dopant atoms disrupts the perfect periodicity of the silicon lattice, giving rise to a band tail instead of a sharply defined band edge. At high doping 1

concentrations, the Fermi level approaches the band edge and can even move above the band edge. In these circumstances, the Boltzmann statistics used in are inaccurate and it is necessary to use Fermi-Dirac statistics to calculate the position of the Fermi level. To model heavy doping effects in the emitter of a bipolar transistor, it is necessary to combine the effects of bandgap narrowing and Fermi- Dirac statistics. For ease of modelling, these effects are rolled into a single parameter called the apparent bandgap narrowing or the dopinginduced bandgap narrowing. Although significant amount of work on Fermi Dirac Analysis to estimate the bandgap narrowing has been done but its effect on some other parameters like gain, cutoff frequency etc. needs some more explanation. A performance comparison of Fermi Dirac Statistics with Boltzman approach is reviewed in this paper with the help of physics based model and simulated results. These results are obtained using 2D SILVACO device simulator known for its authentication in the industry. An attempt has been done in this paper to study the impact of Ge fraction (in SiGe base) on bandgap narrowing considering the important design parameters and issues like conductance, current gain, cutoff frequency, maximum oscillation frequency and junction capacitances.

4.2 Bandgap Narrowing

In lightly doped semiconductors the dopant atoms are sufficiently widely spaced in the semiconductor lattice that the wave functions associated with the dopant atoms electrons do not overlap. The energy levels of the dopant atoms are therefore discrete. Furthermore, it is reasonable to assume that the widely spaced dopant atoms have no effect on the perfect periodicity of the semiconductor lattice, and hence the edges of the conduction and valence bands are sharply defined. In heavily doped semiconductors the dopant atoms are close enough together that the wave functions of their associated electrons overlap.

This causes the discrete impurity level to split and form an impurity band, as shown in figure 1. In addition, the large concentration of dopant atoms disrupts the perfect periodicity of the silicon lattice, giving rise to a band tail instead of a sharply defined band edge It can be seen that the overall effect of the high dopant concentration is to reduce the bandgap from Ego to Ege.

At high doping concentrations, the Fermi level approaches the band edge and can even move above the band edge. In these circumstances, the Boltzmann statistics used in are inaccurate and it is necessary to use Fermi-Dirac statistics to calculate the position of the Fermi level. To model heavy doping effects in the emitter of a bipolar transistor, it is necessary to combine the effects of bandgap narrowing and Fermi- Dirac statistics.

Fig-4.1 Effect of High doping on energy bands.

For ease of modelling, these effects are rolled into a single parameter called the apparent bandgap narrowing or the dopinginduced bandgap narrowing in the emitter Ege, which is defined by thefollowing equation:
2 2

Where Ege = Ego Ege, nie is the intrinsic carrier concentration in the emitter, and nio is the intrinsic carrier concentration for lightly doped silicon. As the name implies, the apparent bandgap narrowing is not the same as the bandgap narrowing obtained from optical measurements, because it includes the effects of Fermi-Dirac statistics.

4.3 Performance Parameter

Summarized below are some the key performance parameters of the Heterojunction Bipolar Transistor.

4.3.1 Current Gain

The current gain of the HBT for an abrupt emitter-base junction was given previously by (2.14). It is clear from the above that for Ev > kBT, the current gain is large and an exponential function of Ev. This factor is sufficiently large in many cases so that the base doping can be made higher than the emitter doping while achieving a reasonable current gain.

Analysis of Current Gain A simple way of modelling bandgap narrowing in the emitter is through an effective doping concentration in the emitter N deff ( N E ) due to Bandgap Narrowing[9].
2 Ege nio Ndeff = Nde 2 = Nde exp kT nie

[2]

For heavily doped, n-type silicon, the model developed by del Alamo [10] gives a reasonably accurate description of the apparent bandgap narrowing. In this model, the apparent bandgap narrowing in the emitter E ge is described by the following empirical equation

Nde cm3 Ege = 18.7 ln meV 7 1017

concentration in the base N aeff (N B ) due to Bandgap Narrowing is given as[11]
2 Egb nio Naeff = Nab 2 = Nab exp kT nib

[3]

The effects of bandgap narrowing in the base can be simply modeled using an effective doping

[4]

The structure under consideration is npn SiGe HBT. In a SiGe HBT the ratio of the base and collector currents is given by [7]

 E I C DnBWE N E = exp v k T I B DpEWB N B B

[1]

where E v is the valence band discontinuity at the emitter-base heterojunction. For Si-SiGe heterojunctions, the consensus from the literature is that E v = E g so E c 0 [8]. ATLAS has a parameter called ALIGN that allows the user to incorporate the percentage of the energy gap difference at a heterojunction to the conduction band.

So putting the value of Effective doping effective

BGN = 0 exp{(E gb E ge )/ kT }

[5]

This equation clearly indicates that bandgap narrowing has the effect of reducing the effective doping concentration in the emitter, and hence also the gain of the bipolar transistor. The gain can be manipulated if the doping concentration in the emitter Nde is replaced by the effective doping concentration Ndeff.

In base region there are two source of Bandgap Narrowing (a) due to the strained Si1 x Ge x (b) due to Ge. The Ge dependent energy bandgap of the SiGe is given by E g (Ge ) = 700 x meV where x is the Ge content in mole fraction. The bandgap reduction due to Ge content has been incorporated by using the above function,

( )BGN

( = exp { 700 x E ge ) / kT }

[6]

where, 0 is the gain calculated considering Boltzman Statistics. E ge is constant as Emitter doping is constant.

Considering the Fermi Dirac and Boltzman Statistics and results obtained on Atlas we came to conclude that Bandgap Narrowing has great affect on highly doped SiGe HBT (1*1020 cm3). The most important parameter, the current gain, is reduced from 180 to 145. There is a corresponding reduction in Collector current from 3.12 mA to 1.25 mA. Hence Fermi Dirac Stats is essential for accurate modeling of highly doped SiGe HBT. So increasing the Ge fraction we can compensate the error occurring by bandgap narrowing.

4.3.2 Cutoff Frequency Usually the gain of the device is independent of the frequency at low frequencies. However, at high frequencies the current gain decreases due to capacitances within the device and transit time effects. The frequency at which the transistor incremental gain (ac) drops to unity is called the cutoff frequency (fT ). For HBTs, the base transit time is significantly reduced by the reduction in base width possible due to the heavy doping in the base, which reduces the overall ec of the HBT. Also, the emitter delay time e is reduced since Cje is proportional to N1/2 DE and the NDE is reduced.

4.3.3 Maximum Frequency of Oscillation In SiGe HBTs, the base resistance of a HBT is reduced due to the higher base doping and the fT is improved due to the thinner base. Together, these improve the maximum frequency of oscillation (fmax). 8 Analysis of Cutoff frequency Cutoff frequency is given by,

fT =

where = e+

b+ c .

e is the emitter charging time and is defined as the time required to change the base potential by
charging up the device capacitances through the differential base-emitter junction resistance.

e =

(Cje+Cjc)

where Cje and Cjc denote the junction capacitances for the base-emitter and base collector junctions, respectively, I denotes the collector currentand n denotes the ideality factor of the device.

c ,The collector charging time is given by, c= (RE + RC )CjC

where RE is the emitter resistance and RC is the collector resistance. The value of this charging time depends greatly on the parasitic emitter resistance RE and collector resistance RC.

is the base transit time and is defined as the time required to discharge the excess minority

carriers in the base through the collector current. It is given as, (Wb2)/2Dn where Wb is base width and Dn is diffusivity. Dn is a linear function of mobility and mobility is given by,
nSiGe(x)=

(1+3x)

nSi(x)

Hence, while increasing the Germanium fraction mobility is increased, consequently base transit time is decreased. In this study we have taken CONMOB, FLDMOB ie Concentration Dependent and Field Dependent mobility [14]

(VSATN and BETAN are constants.)

So when we consider the bandgap Narrowing due to high doping effect high electric field can cause decrement in mobility, hence base transit time is decreased after high Ge fraction( as Ge fraction is responsible for band gap narrowing) by following analysis.

Eg effective band gap reduction in the base due to the presence of Ge (EgGe) and due to heavy doping effects (EgDOP) Eg(x) = EgGe(x) + EgDOP(x) EgGe(x) The band gap narrowing due to the presence of Ge is assumed to have a linear dependence on Ge concentration Eg,Ge = 700x(meV) and EgDOP(x) given by equation 3. Considering all the terms (ie. Mobility, Effective doping, Electron Saturation Velocity) in base transit time depends on Bandgap Narrowing which has high effect on high Ge concentration. Further if we consider the Bandgap narrowing there is drastic change in Base transit Time which consequently effect Cutoff frequency. 7

ConductanceThe SiGe layer with a constant Germanium mole fraction results in a Valence band discontinuity at both the emitter base and the collector base junctions. The Valence band discontinuity prevents back injection of holes from the base to the emitter. The performance of the SiGe HBT greatly depends on the Ge profile in the base. In the case of constant Ge profile, similar grading is assumed on the emitter side as well the device with constant Ge profile is much greater than that with graded Ge profile. The reason for the improved performance with increase in Ge fraction can be attributed to an increased hole mobility and improved current gain. The hole mobility in the SiGe increases with increasing Ge content as given by [13], 400 29 4737

ph

, thereby reducing the base resistance.

Fig.-4.3 Dark color line shows the variation of Conductance of Emitter Base junction in (Siemens/mm).It clearly shows that upto some extent it increases (correspondingly base resistance decreases) but for higher Ge fraction in SiGe it decreases. This analysis is for considering Fermi Dirac Statistics. Light color line does not show the actual variation in

conductance as it considers the Boltazman Statistics which does not count Bandgap Narrowing. It shows the decrement in Conductance which signifies the decrease of mobility for high Ge fraction. It shows the Base resistance is increasing and hence Cutoff frequency is decreasing. Mobility of holes also depends on Doping and Electric field (CONMOB, FLDMOB ie Concentration Dependent and field dependent mobility) in our Simulator ATLAS. So increasing Ge fraction by a certain limit (0.25) emitter resistance decreases rapidly as shown in graph.

Fig.4 Graph shows that there is a decrement in Transconductance(used in formulae) with higher Ge fraction.Which is the main cause to decrease the Peak Cutoff frequency. This barrier also increases the base transit time and consequently degrades the cutoff frequency (fT) and maximum frequency of oscillation (fmax). It has been found that parasitic barriers at both junctions are also formed due to nonalignment of the p-n junction and heterojunction.

Capacitive Effect-

Cutoff frequency is increased so the maximum frequency of oscillation increases with Ge content. In case the of a constant Ge profile, the current gain and hence cutoff frequency are improved due to the presence of the emitter-base heterojunction at the valence band, which significantly reduces back injection of holes from the base into the emitter. The valence band discontinuity increases with Ge fraction at the emitter base junction, thereby improving the performance of the device improving the current gain. It also reduces the base transit time, which gives a larger cutoff frequency. The falloff at higher current density (thereby reducing cutoff frequency) is much steeper when the Ge mole fraction is higher (>24% as from Fig.-2) at the collector junction due to the formation of the parasitic barrier. The displacement of the heterojunction away from the p-n junction results in a parasitic barrier at the base-collector junction that degrades the performance of the SiGe.

Fig.3 It shows the formation of Junction Capacitance at Peak Cutoff Frequency. Concluding all the theories on the formation of Parasitic Barrier this graph justify that there is an abrupt change in capacitance at high value of Ge concentration.

10

The formation of the Parasitic barrier at the emitter-base and base-collector junctions due to high current effects and base dopant outdiffusion at high current densities, hole accumulation at the base end of the collector-base junction induces electron pile up at the collector end of the basecollector junction. This leads to the formation of a parasitic field which acts as a potential barrier to the electron flow in the conduction band as shown. This barrier increases the recombination in the base and produces a saturation tendency in the collector current, both of which degrade the current gain.

Fig.4 This graph justify the above discussed theory on parasitic effect on Cutoff frequency. Using a gradually reduced bandgap in the quasi-neutral region an accelerating electric field may be introduced for decreasing the base transit time. However, a large electric field in the base can be counter-productive as a result of mobility degradation in the high-field regions. In this study it has shown that Ge fraction about 24% can causes decrement in Cutoff frequency 11

E Eg

1.155

2.010

0.43

1.47

0.206 for 0.85<

for >0.25 0.25

The use of SiGe in the base of Si/SiGe/Si NPN HBT causes the presence of a heterojunction at the collector-base junction as well as the emitter-base junction. Since, the energy gap difference between the two materials is primarily in the valence band, the conduction band difference is almost negligible. This is desirable since the presence of a Ec at the collector-base heterojunction forms a potential energy barrier impeding electron injection from the base into the collector. For SiGe HBTs, the valence band discontinuity due to the heterojunction at the basecollector junction prevents holes from spilling into the collector from the base, at the onset of basepushout. Since the Ev barrier prevents holes from moving into the collector, a net positive charge accumulates at the collector end of the base, a corresponding negative charge forms nearby in the collector and a parasitic barrier forms in the conduction band at the collector-base junction. The presence of this parasitic barrier limits the collector current, causes the base current to increase and the current gain of the device drops drastically. Further, the base transit time increases which decreases the cutoff frequency (ft) and the maximum frequency of oscillation (fmax).

4.4 Conclusion
It is to conclude that there is trade off between current gain and cutoff frequency (unity gain bandwidth). So for high speed application (LAN and Mobile Communication), where Current Gain is not the major concerned, low Germanium fraction (<0.25) in SiGe Base region must be used. At this fraction Band gap narrowing has lower effect on Cutoff frequency and current gain also. Beyond this fraction, Junction Capacitive effects significantly come into existence which causes a drastically decrement in Cutoff frequency. Considering our analysis Boltzman Statistics (which we generally use), does not give an accurate estimation of Device parameters. Error caused by this approximation is very high at low fraction of Ge. Thus we conclude that compensating the bandgap narrowing effect due to high doping can be compensated by adequate Germanium fractionin SiGe base region. So Femi Dirac statistics must be considered for Gain and Cutoff frequency estimation.

12

Chapter 5

Simulation Software and Technique

5.1 Introduction
Computer Aided Design(CAD) are software packages that are powerful tools to economically simulate semiconductor devices and obtain information which cannot be experimentally observed. This information can help to understand the device physics and characteristics and the factors limiting device performance. In addition, CAD simulation also provides a means to predict device behavior that can reduce device prototyping and development costs. Semiconductor devices can be modeled in two ways; one is to determine the terminal electrical properties of a device based on a fit to empirical results and the other is to study the carrier transport processes taking place within the device. The method used for this work is a physical model, which can be used to predict both terminal characteristics and examine the transport phenomena within the device. Physical device models are based on a description of carrier transport physics, so that they calculate electron and hole concentrations within the device, their spatial variation, and their response to material properties and biasing. They can be used to characterize the DC and AC operation of the device and to provide detailed insight into the physical aspects of device operation, provided they include all the important device physics. An 13

important advantage of this type of model is that it can not only be used to predict the terminal current-voltage characteristics of the devices for comparison with device measurements, but also to investigate the origins of the observed behavior.

5.2 SILVACO-TCAD Tool

Silvaco International produced the device modeling and simulation software utilized in this work. Silvacos ATLASTM is a versatile and modular program designed for one, two, and threedimensional device simulation. BLAZETM and GIGATM, ATLASTM sub-modules (Figure 8), perform specialized functions required for advanced materials, heterojunctions, and temperaturedependent conditions. To control, modify, and display the modeling and simulation, the Virtual Wafer Fabrication (VFW) Interactive Tools, namely DECKBUILDTM and TONYPLOTTM, were utilized (Figure 4.1).

Unlike some other modeling software, Silvaco uses physics-based simulation rather than empirical modeling. In truth, empirical modeling produces reliable formulas that will match existing data but physics-based simulation predicts device performance based upon physical structure and bias conditions. To perform the modeling, the Silvaco software graphically represents a device on a two-dimensional grid with designated electronic meshing parameters. At every mesh intersection, the program simulates carrier transport by means of differential equations derived from Maxwells laws. To achieve accuracy, the program incorporates the appropriate physics via numerical procedures. To accurately model the III-V semiconductors, ATLAS must employ the BLAZE program extension to modify calculations that involve energy bands at heterojunctions. The heterojunctions require changes in calculating current densities, thermionic emissions, velocity saturation, and recombination-generation.

14

ATLAS attempts to find solutions to carrier parameters such as current through electrodes, carrier concentrations, and electric fields throughout the device. ATLAS sets up the equations with an initial guess for parameter values then iterates through parameters to resolve discrepancies. ATLAS will alternatively use a decoupled (Gummel) approach or a coupled (Newton) approach to achieve an acceptable correspondence of values. When convergence on acceptable values does not occur, the program automatically reduces the iteration step size. ATLAS generates the initial guess for parameter values by solving a zero-bias condition based on doping profiles in the device.

5.2.1

ATLAS/BLAZE

The software employed in this study is a commercial simulator called ATLAS produced by Silvaco and configured to run on LINUX Operating System. This software package is a comprehensive and flexible tool for simulating the electronic performance of arbitrary two dimensional device structures, which can be composed of semiconductors, insulators, conductors and terminals. The simulation can be conducted to get not only DC and small signal AC analysis of the device terminal behavior, but also analysis of internal device parameter distributions, such as the conduction and valence band energies, electrostatic potential, electric fields, electron and hole concentrations and other current components. It also includes the interface physics models, which are provided for all topology combinations, such as heterojunctions, metal contacts, and surfaces. These capabilities enable accurate simulation of parasitics, leakage currents, trapped charges and contacts. ATLAS predicts the electrical characteristics of the device by solving the key semiconductor equations including the Poissons equations, the current continuity equations for electron and holes and the current density equations that describe the physics of the device operation. The simulator consists of a device simulation sub framework and a modular set of application oriented tools like BLAZE, which is a 2-D device simulator for III-V material devices and devices with position dependent band structure (heterojunctions), as well as a Graphic User Interface called TONYPLOT for graphic presentation.

15

5.2.2 ATHENA It also has a two dimensional fabrication process simulator called ATHENA. The ATHENA TwoDimensional Process Simulation Framework is a comprehensive software tool for modeling semiconductor fabrication processes. ATHENA provides facilities to perform efficient simulation analysis that substitutes for costly real world experimentation. ATHENA combines high temperature process modeling such as impurity diffusion and oxidation, topography simulation, and lithography simulation in a single, easy to use framework.

Simulation of a device for its characteristics involves the following methodology. The physical structure of the device, along with material parameters in the different regions and the doping profiles are initially specified by the user. A mesh is defined for the initial structure which generates the nodes at which the device equations are solved. In order to avoid overloading the computer, the mesh is defined very dense only near the junctions where quantities change rapidly. Physical models are incorporated that specify the dependence of one parameter on the other, such as carrier mobility on doping levels. Appropriate physics-based models are then selected by the user including recombination models (Auger, ConSRH), mobility models(Klaassen, Fldmob, Conmob), and bandgap narrowing models (BGN). Then, the bias conditions are specified for performing the AC, DC and transient simulation, if any. Appropriate numerical techniques for solving can also be specified.

16

Fig5.1 Flow Diagram of Simulation.

5.3 Models and Material

The default material parameters can be overridden using the MATERIAL, IMPACT, MODEL, and MOBILITY statements.In this project following Models are used-

5.3.1 Parallel Electric Field-Dependent Mobility (FLDMOB)

As carriers are accelerated in an electric field, their velocity will begin to saturate at high enough electric fields. This effect has to be accounted for by a reduction of the effective mobility since

17

the magnitude of the drift velocity is the product of the mobility and the electric field component in the direction of the current flow. The model used in ATLAS to represent this reduction is FLDMOB. The following Caughey and Thomas(Equation 1-2) expression is used to implement a field dependent mobility that provides a smooth transition between low field and high field behaviour.

[1]

[2]

where E the parallel electric field and no and po are the low field electron and hole mobilities, respectively. The model parameters BETAN and BETAP are curvature parameters that are set equal to unity for these simulations in the absence of adequate data for the Ge dependence in SiGe alloys.

5.3.2 Concentration Dependent Mobility (CONMOB)

The model used here is based on empirical data for the doping dependent low-field mobilities of electrons and holes in silicon at T= 300K. The mobility of holes and electrons is found to decrease with increase in doping concentration (Equation 3) due to ionized impurity scattering. The dependence of the hole mobility on doping can be given by the following equation-

[3]

where min is the carrier mobility at its minimum value and N is the doping level.

18

5.3.3 Bandgap Narrowing (BGN)

In the presence of heavy doping, i.e. greater than 1018, the pn product in silicon becomes doping dependent. As the doping increases, a decrease in Eg occurs, where the conduction band is lowered by approximately the same amount as the valence band is raised. These bandgap narrowing effects are enabled by specifying the BGN parameter of the MODELS statement. The Ge dependence of the band gap narrowing has been analysed in this project.

5.3.4 Fermi-Dirac Statistics (FERMI)

Electrons in thermal equilibrium at temperature TL within a semiconductor lattice obey FermiDirac statistics. That is, f(E) is the probability that an available electron state with energy E is occupied by an electron. The expression for this relation is given by

f E

[4]

where EF is a spatially independent reference energy known as the Fermi level and k is the Boltzmanns constant. By utilizing the Fermi-Dirac statistics instead of Boltzmann statistics, more accurate simulation results are obtained.

In the limit that - EF >> kTL Equation 3-25 can be approximated as

exp

[5]

The use of Boltzmann statistics instead of Fermi-Dirac statistics makes subsequent calculations much simpler. The use of Boltzmann statistics is normally justified in semiconductor device theory, but Fermi-Dirac statistics are necessary to account for certain properties of very highly doped (degenerate) materials.

19

The remainder of this section outlines derivations and results for the simpler case of Boltzmann statistics which are the default in ATLAS. Users can specify that ATLAS is to use Fermi-Dirac statistics by specifying the parameter FERMIDIRAC on the MODEL statement.

5.3 Simulation Technique

The simulation method described in this chapter is the basis employed in a commercial device simulator, ATLAS, by Silvaco International [3]. It is a fully numerical model, which is based on using partial differential equations that describe the important device physics and include the different regions of a device in one unified manner. The semiconductor devices are governed by the basic semiconductor equations, such as Poissons equation and the carrier continuity equations, and are described as follows.

5.3.1 Poissons Equation

where is the electrostatic potential, + is the local permitivity, and is the local charge density, which can be a function of position. The reference potential can be defined in various ways for ATLAS. For the simulator used in this work, is always the intrinsic fermi level.

5.3.2 Carrier Continuity Equation

The carrier continuity equations for electrons and holes are

20

where n and p are the electron and hole concentrations, Jn and Jp are the electron and hole current densities, Gn(Rn) and Gp(Rp) are the generation (recombination) rates for the electrons and holes, and q is the magnitude of the charge on the electron. All of these (except q) are, in general, a function of position. The electron and hole current densities (Jn and Jp) are calculated based on the drift-diffusion model and are described as follows

Jn Jp

qnnE qnpE

qDnn qDpp

where n and p are the electron and hole mobilities, Dn and Dp are the electron and hole diffusion constants, Eis the local electric field, and is the three dimensional spatial gradient

The simulation principle incorporated in the ATLAS model is to numerically solve the above Poissons equation and the carrier continuity equation self-consistently to get electron and hole concentrations and electrostatic potential () at all points within the device structure. In general, the doping level as well as the material parameters in the above equation vary with location in the device structure. The solution is obtained by solving the above described differential equations subject to boundary conditions imposed by the devices contacts and biasing. Other models incorporated in the simulation include ones for the carrier generationrecombination mechanism, mobility models that depend on the doping level, band gap narrowing models, and density of states for different materials. Also included are velocity saturation effects at high electric fields.

21

Chapter 6

Conclusion

22

Appendices (A)
SILVACO Source

Code

go athena # SiGe HBT simulation # Establish initial grid and substrate material line x location=0.0 spacing=0.08 line x location=0.5 spacing=0.05 line x location=0.7 spacing=0.05 line x location=1.2 spacing=0.08 line x location=2.2 spacing=0.18 # line y location=0.0 spacing=0.01 line y location=0.1 spacing=0.02 line y location=0.5 spacing=0.05 line y location=0.8 spacing=0.15 # init silicon c.phos=2e16 structure outf=tmpr1.str init inf=tmpr1.str flip.y implant phos energy=60 dose=3e15 diffuse time=5 temp=1000 struct outf=temp init inf=temp flip.y

23

# Deposit Silicon germanium with composition fraction 0.2 for base deposit sige thick=.1 divis=12 ydy=0.05 dy=0.02 c.frac=0.2 c.boron=1e15 implant boron energy=10 dose=1.0e13 # Deposit silicon for the emitter #implanatation deposit silicon thick=0.2 divis=10 ydy=0.08 dy=0.04 c.phos=1.e15 implant boron energy=12 dose=3e14 diffuse time=0.5 temp=920 # Mask and implant the emitter deposit photo thick=.5 divis=5 etch photo left p1.x=0.5 #implant phos energy=40 dose=6e15 implant phos energy=38 dose=6e15 diffuse time=5 temp=920 strip # Deposit and pattern the contact metal deposit aluminum thick=0.05 div=1 etch aluminum start x=0.5 y=10. etch etch etch cont x=0.5 y=-10. cont x=1.2 y=-10. done x=1.2 y=10.

# Define the electrodes electrode name=emitter x=0.0 electrode name=base x=2.0 electrode name=collector backside

24

# Define impurity characteristics in each material impurity i.boron acceptor sige impurity i.phos donor sige structure outfile=hbtay03_0.str go atlas # Material parameter and model specification material material=Si taun0=1e-7 taup0=1e-7 material material=SiGe taun0=1.e-8 taup0=1.e-8 model bgn consrh auger fldmob conmob fermi print temp 300 output con.band val.band solve init # # Calculate Gummel plot and AC parameters versus Vbe (Vce) at 1 MHz solve prev # 1 - emitter 2 - base 3 - collector log outf=hbtay03_1.log solve v2=0.01 v3=0.01 ac freq=10 direct solve v2=0.025 v3=0.025 vstep=0.025 electr=23 nstep=2 ac freq=1e6 direct solve v2=0.1 v3=0.1 vstep=0.1 electr=23 nstep=5 ac freq=1e6 direct solve v2=0.65 v3=0.65 vstep=0.05 electr=23 nstep=6 ac freq=1e6 direct save outf=hbtay03_2.str solve v2=0.975 v3=0.975 vstep=0.025 electr=23 nstep=3 ac freq=1e6 # # Frequency domain AC analysis up to 100 GHz # log outf=hbtay03_3.log s.param gains inport=base outport=collector width=50 load inf=hbtay03_2.str master.in solve previous ac freq=1 direct solve ac freq=10 fstep=10 mult.f nfstep=8 direct solve ac freq=2e9 direct solve ac freq=5e9 direct 25

solve ac freq=1e10 direct solve ac freq=2e10 fstep=2e10 nfstep=5 direct # Extraction of parameters extract init inf="hbtay03_1.log" # Maximum cutoff frequency extract name="Ft_max" max(g."collector""base"/(6.28*c."base""base")) # # Base bias at maximum cutoff frequency Extract name="Vbe@Ft_max" x.val from curve g."collector""base"/(6.28*c."base""base" )) where y.val=$"Ft_max" # Input (base) capacitance at maximum cutoff frequency extract name="Cbb@Ft_max)" y.val from curve (v."base", abs(c."base""base" )) where x.val=$"Vbe@Ft_max" # # Transconductance at maximum cutoff frequency extract name="Gm@Ft_max)" y.val from curve (v."base", abs(g."collector""base" )) where x.val=$"Vbe@Ft_max" # Gummel plot tonyplot hbtay03_1.log -set hbtay03_1_log.set # AC current gain versus frequency tonyplot hbtay03_3.log -set hbtay03_4_log.set # S12 & S21 polar coordinates tonyplot hbtay03_3.log -set hbteay03_4_s12.set # # S11& S22 Smith chart tonyplot hbtay03_3.log -set hbtay03_4_s11.set # quit (v."base",

Appendices (B)
26

References
[1] Hashim, Md., R.F. Lever and P. Ashburn, 2D simulation of the effect of transient enhanced boron out-diffusion from base of SiGe HBT due to an extrinsic base implant, Solid State Elect., 43, pp. 131-140. 1999. H. Kroemer, Heterojunction bipolar transistors and integrated circuits, Proc. IEEE, 70, pp. 1325. 1982 F. Capasso, Band-gap engineering: from physics and materials to new semiconductor devices, Science, 235, 1726, 1987 N Jiang and Z. Ma, Current gain of SiGe HBTs under high base doping Concentration, Semicond. Sci. Technol. 22, pp.168-172, 2007. V. Palankovski and S. Selberherr, Critical modeling issues of SiGe semicondutor devices", J. Telecommun. Inform. Technol., no. 1, pp. 15{25}, 2004. M.K. Das, N.R. Das and P.K. Basu, Performance Analysis of a SiGe/Si Heterojunction Bipolar Transistor for Different Ge-composition , Proce. URSI, p. D03. 2005 K Suzuki and N Nakayama, Base transit time of shallow base bipolar transistors considering velocity saturation at base-collector junction, IEEE Trans. Electron Devices, 39, pp. 623-628, 1992. J. D. Cressler, SiGe HBT technology: a new contender for Si-based RF and microwave circuit applications, IEEE Trans. Microw. Theory Tech., vol. 46, issue 5, pp. 572589, 1998. Ankit Kashyap and R.K. Chauhan, A New Profile Design for Silicon Germanium based Hetero-Junction Bipolar Transistors, Journal of Computational and Theoretical Nanosciences, Vol.5 (11), pp. 2238-2242, Nov 2008. D. L. Harame, J. H. Comfort, J. D. Cressler, E. F. Crabbe, J. Y. C. Sun, B. S. Meyerson and T. Tice, Si/SiGe epitaxial-base transistors-part I: materials, physics and circuits, IEEE Trans. Electron Devices, 42, 455468, 1995. F. Capasso, Band-gap engineering: from physics and materials to new semiconductor devices, Science, 235, 1726, 1987 [2] B. L. Tron, M. D. R. Hashim, P. Ashburn, M. Mouis, A. Chantre, and G. Vincent, Determination of Bandgap Narrowing and Parasitic Energy Barriers in SiGe HBTs Integrated in a Bipolar Technology, IEEE Trans.on Electron Devices, vol. 44, NO. 5, May 1997. P. Ashburn, H. Boussetta, M. D. R. Hashim, A. Chantre, M. Mouis, G. J. Parker, G. Vincent, Electrical Determination of Bandgap Narrowing in Bipolar Transistor with 27

[2]

[3]

[4]

[5]

[6]

[7]

[8]

[9]

[10]

[11] [12]

[13]

Epitaxial Si, Epitaxial Si1-xGex, and ion Implanted Base, IEEE Trans. on Electron Devices, vol. 43, NO. 5, May 1996. [14] R R. King, and R M. Swanson, Studies of Diffused Boron Emitters: Saturation Current, Bandgap Narrowing, and Surface Recombination Velocity, IEEE Trans. on Electron Devices, vol. 38, NO. 6, June 1991. J. D. Cressler, Re-Engineering Silicon: SiGe Heterojunction Bipolar Transistor, IEEE Spectrum, vol. 32, p.49 (1995). Z. Matutinovic-Krystelj, V. Vennkataraman, E. J. Prinz, J. C. Sturm, and C. W. Magee, Base Resistance and Effective Bandgap Reduction in NPN Si/SiGe/Si HBTs with Heavy Base Doping, IEEE Trans. Electron Devices, vol. 43, p. 457 (1996). B. Mazhari, and M. Morkoc, Effect of Collector-Base Valence band Discontinuity on Kirk Effect in Double Heterojunction Bipolar Transistors Appl. Phys. Lett., vol. 59 p. 2162 (1991). A. Neugroschel, G. Li, and C. T. Sah, Low Frequency Conductance Voltage Analysis of Si/SiGe/Si Heterojunction Bipolar Transistors, IEEE Trans. Electron Devices, vol. 47, p. 187 (2000). P. Ashburn, C. Bull, K.H. Nicholas and G.R. Booker, Effects of dislocations in silicon transistors with implanted bases, Solid State Electronics, 20, 731 (1977).

[15]

[16]

[17]

[18]

[19]

[20]

J. del Alamo, S. Swirhun and R.M. Swanson, Simultaneous measurement of hole lifetime, hole mobility, and bandgap narrowing in heavily doped n-type silicon, IEDM Technical Digest, 290 (1985). J. Popp, T.F. Meister, J. Weng and H. Klose, Heavy doping transport parameter set describing consistently the AC and DC behaviour of bipolar transistors, IEDM Technical Digest, 361 (1990).

[21]

[22]

T. K. Carns, S. K. Chun, M. O. Tanner, K. L. Wang, T. I. Kamins, J. E. Turner D. Y. C. Lie, M. A. Nicolet, and R. G. Wilson, Hole Mobility Measurements in Heavily Doped SiGe Strained Layers, IEEE Trans. Electron Devices, vol. 41, p. 1273 (1994). Silvaco International Atlas Manual, version 1998.

[23]

28

29

30