dopant diffusion1

Upload: gurivireddy

Post on 05-Apr-2018

227 views

Category:

Documents


0 download

TRANSCRIPT

  • 7/31/2019 Dopant Diffusion1

    1/20

    Diffusion 1

    Dopant Diffusion (Jaeger Chapter 4 and Campbell Chapter 3)

    As indicated previously the main front-end processing in building a device or

    integrated circuit is to selectively introduction of dopant atoms into silicon wafer.

    Dopant introduction by high temperature diffusion is one of the two major processes

    for achieving this. Diffusion is carried out at high temperature ( 800 to 1000 C) in adopant-rich gaseous ambient.

    Diffusion also covers the redistribution in the wafer of dopant atoms introduced into

    the wafer by other methods such as ion implantation.

    For simplicity the diffusion of dopant in silicon is modelled by the simple diffusion

    theory described by Ficks law in which the flux of dopant atoms at any point isassumed to be proportional to the gradient of the dopant concentration at that point.

    The proportionality constant is the diffusion coefficient, D . This is assumed to be a

    constant for a given dopant species at a given temperature. Actually D is not constant

    by is dependent on dopant concentration.

    This section covers the diffusion techniques used and the simple 1-D analysis

    describing the dopant distribution into the wafer.

  • 7/31/2019 Dopant Diffusion1

    2/20

    Diffusion 2

    Diffusion System

    The most common practice of introducing dopantinto wafers by diffusion is the horizontal quartztube furnaces similar to that used for thermal

    oxidation of silicon wafers. A separate furnace,wafer holder etc are reserved for a particulardopant to avoid contamination by other dopants.

    In general the dopant atoms are introduced in twoseparate diffusion steps, each in its own tubefurnace. The first step is referred to as pre-deposition diffusion, it is aimed to introduce acontrolled amount of dopant atoms into the silicon

    surface (dopant atoms per unit surface area). Thesecond step known as the drive-in diffusion whichis aimed (at least theoretically) to redistribute thedopant atoms introduced during predepositionfurther into the silicon and reducing the dopantconcentration near the surface.

    Predeposition Diffusion

    In pre-deposition diffusion the wafer is exposed to

    excess amount of dopant atoms in a gaseousambient to ensure that the wafer surface has themaximum concentration of dopant atomsdetermined by the diffusion temperature (referredto as the solid solubility limit). Not all introduceddopant atoms are electrically active.

  • 7/31/2019 Dopant Diffusion1

    3/20

    Diffusion 3

    The introduction of dopant atoms into the silicon (particularly at high dopant

    concentration) coupled with the high thermal gradients experienced by the wafers when

    they are inserted or removed from the high temperature of the furnace can cause

    significant damage to the crystal structure of the single crystal (dislocations and slip)which would affect the electrical properties of the material (lifetime reduction for example).

    To alleviate this, wafers are pushed into and removed from the furnace in slow controlled

    manner using mechanical puller which has a slow control pull rate (~a few cm per

    minute). Mechanical puller is now used for all furnaces. In modern furnaces rather than

    keeping the furnace at the diffusion temperature when the wafers are inserted, it is

    possible to introduced the wafers into a cool furnace (~600 C) and then ramp up the

    temperature to the diffusion temperature in a controlled fashion (similarly for a ramp downon wafers removal).

    The dopant atoms can be introduced in solid, liquid or gaseous form. The following table

    gives the sources used for Boron and Phosphorus:

    Dopant Boron Phosphorus

    Form

    Solid B2O3,BN P2O5Liquid BBr3 POCl3

    Gaseous BCl3 PH3

    The BN solid source (in solid disc form) is actually converted to B2O3 in an oxidising

    ambient in a separate furnace and then loaded in the predeposition furnace with each BN

    disc facing the silicon wafers. The B2O3 is then transferred to the surface of the wafers.

  • 7/31/2019 Dopant Diffusion1

    4/20

    Diffusion 4

    These diagrams show how these are done.

  • 7/31/2019 Dopant Diffusion1

    5/20

    Diffusion 5

    Drive-in Diffusion

    The drive-in diffusion is typically carried out in an oxidising ambient but with no dopant

    atoms. This is carried out after in a furnace very similar to an oxidation furnace usually at

    a temperature higher than the predeposition temperature. Prior to the drive-in diffusion, avery thin layer of dopant rich silicon and/or silicon dioxide formed during predeposition is

    etched away (hydrofluoric acid dip).

  • 7/31/2019 Dopant Diffusion1

    6/20

    Diffusion 6

    Prior to the drive-in diffusion, a very thin layer of dopant rich silicon and/or silicon dioxide

    formed during predeposition is etched away (hydrofluoric acid dip). The drive-in oxidelayer formed on the wafer is to seal in the dopant atoms. In simple analysis. It isassumed that no dopant atoms are lost either to the ambient or to the oxide. Due to the

    conversion of silicon into silicon dioxide and the redistribution of dopant atoms at Si-SiO2interface during oxidation, there is always loss of dopant atoms, more so for born due to

    the suck-out effect.

    As mentioned before, during drive-in the dopant atoms diffuse further into the silicon and it

    also reduces the surface dopant concentration. The resistance (more correctly the sheet

    resistance) of the diffused layer increases and the junction depth increases.

  • 7/31/2019 Dopant Diffusion1

    7/20

    Diffusion 7

    Mathematical Modelling of Diffusion 1-dimensional

    2

    2

    x

    ND

    t

    N

    Dopant diffusion is governed by Ficks Laws(diffusion laws).

    Ficks 1st Law states that the diffusion flux F (i.e.dopant flow per unit cross-section area per unit

    time ) is dependent on the dopant concentrationgradient:

    The variation of dopant concentration with distance into the semiconductor is known

    as the dopant profile. Consider the 1-dimension case in which we consider only

    dopant concentration variation in x, the direction perpendicular to the semiconductor

    surface:

    where N(x) is the dopant concentration and D

    is the diffusion coefficient (or diffusion

    coefficient) at the diffusion temperature.

    Ficks 2nd Law, conservation of matter, givesthe rate of change of N with time:

    x

    NDF

  • 7/31/2019 Dopant Diffusion1

    8/20

    Diffusion 8

    Predeposition Diffusion

    For this process, the dopant concentration at the surface N(x=0) is limited to the solid

    solubility of the dopant at the predeposition temperature No for all t >0. Assuming thatthere are no dopant atoms of this element in the silicon prior to diffusion (N =0 for all x at

    t 0.Then it can

    be shown that the N(x) for t>0 is given by:

    .d)exp()y(erf

    asdefinedfunction,errorasknownfunctionalmathematicknownwelltheiserfwhere

    )y(erfd)exp()y(erfc

    asdefinedfunction,errorarycomplementtheiserfcwhere

    Dt

    xerfcNd)exp(N)t,x(N

    y

    y

    o

    )Dt(x

    o

    0

    2

    0

    2

    2

    0

    2

    2

    12

    1

    2

    21

    When x = 0, N(0,t) = No independent of t. This is expected because of the solid

    solubility assumption at the surface. For any x >0, N(x,t) increases with t. This gives

    dopant profile of the form:

  • 7/31/2019 Dopant Diffusion1

    9/20

    Diffusion 9

    This profile is often referred to as the complementary error function profile or the

    predeposition profile. The term 2(Dt) is known as the diffusion length (not to beconfused with the carrier diffusion length (Dt) when we consider carrier diffusion in theoperation of semiconductor devices). In some literature (Dt) is referred to as thecharacteristic length of the diffusion process. Increasing (Dt) extends the dopant atomsfurther into the bulk of the semiconductor wafer this simply means that by increasingthe predeposition diffusion time we have more dopant atoms moving further into the

    semiconductor while keeping the surface concentration constant at No.

  • 7/31/2019 Dopant Diffusion1

    10/20

    Diffusion 10

    The diffusion coefficient, D, is a function of diffusion

    temperature following the Arrhenius relationship:

    Typical Do and activation energy EA as well as D-

    temperature plots are given here. For first order

    approximation, D is taken to be independent of the

    actual dopant level.

    kT

    AE

    oeDD

  • 7/31/2019 Dopant Diffusion1

    11/20

    Diffusion 11

    For a given predeposition diffusion conditions (i.e. fixed No, D and t) the total amount of

    dopant atoms introduced into the silicon wafer, Q, measured in atoms per unit surface

    area of the diffusion window, is determined by predeposition diffusion. Q is given by:

    Concept of Junction Depth, xj

    When dopant of one type (p or n) is diffused into a wafer originally doped with dopant of

    the opposite type, a junction will be formed. The distance of the junction from the

    surface is known as the junction depth, xj. At xj the concentration of the diffused dopant

    = the concentration of the substrate dopant (assumed to have a uniform concentration).

    Consider a p-type substrate with uniform dopant concentration NA subjected to the n-

    type predeposition diffusion with the profile:

    .dy)y(erfc

    :thatfacttheofusemadehavewewhere

    DtN

    dxDt

    xerfcNdx)x(NQ

    o

    o

    0

    1

    12

    2

    0

    0

    .NDt

    xerfcN

    :bygiventhenisxdepth,junctionthe

    ,DtxerfcN)x(N

    Aj

    Do

    j

    DoD

    02

    2

  • 7/31/2019 Dopant Diffusion1

    12/20

    Diffusion 12

    For x xj it remains p-type. The net dopant concentration

    on n-side is ND(x) NA and on p-side it is NA-ND(x). See Left plot below. Because thedopant concentration level varies several order of magnitude, the dopant concentration

    axis is usually plotted on log scale. See npn bjt dopant concentration plot below right.

  • 7/31/2019 Dopant Diffusion1

    13/20

    Diffusion 13

    Drive-in Diffusion

    Dopant diffusion during drive-in is governed by the same diffusion equation i.e. Ficks 2ndLaw for the diffusion of dopant atoms inside the silicon, i.e.

    The only difference is the boundary conditions for this equation. It is usually assumed that

    a layer of SiO2 is formed immediately at the surface and this serves as a mask preventing

    the inward (from ambient to silicon) or outward (silicon to ambient) diffusion of dopant

    atoms. Mathematically, the boundary condition is that the dopant atom flux at x=0 is zero

    for all t >0 i.e.

    where D1 and t1 are the diffusion coefficient and diffusion time for predeposition diffusion.

    In order to obtain analytical solution for the final profile, we have to make simplifying

    assumptions based on our knowledge about the diffusion sequence.

    s.x'allfor0tat,)tD

    x

    (erfcNN(x)

    i.e.condition,initialtheasprofile

    ionpredepositthehavewethenion,predepositafteroutcarriedisdiffusionthatAssuming

    0tafor0xatx

    N

    o

    112

    0

    2

    2

    x

    ND

    t

    N

  • 7/31/2019 Dopant Diffusion1

    14/20

    Diffusion 14

    In general, predeposition diffusion is carried out at a lower temperature than drive-in

    diffusion and also for a shorter time. If D2 and t2 and the diffusion coefficient and

    diffusion time for drive-in, then

    D2t2 >> D1t1.Under this condition, it is usual to treat the predeposition diffusion profile as a delta

    function at x=0 with the area of the delta function = Q, the dopant atoms per unitsurface area introduced by predeposition i.e.

    N(x,t=0)= Q.d(x),

    where,

    In this case, the solution of the drive-in diffusion is:

    The right hand side is the well known Gaussian function and the profile is referred to as

    the Gaussian diffusion profile. The peak of the profile occurs at x =0, i.e. at the surfaceas we would expect it. However notice that the magnitude of this peak decreases as the

    diffusion time t increases. This is the consequence that as diffusion time increases, the

    fixed amount of dopant atoms, Q, is spread out further into the silicon wafer. The plot of

    Gaussian dopant profile is shown below:

    1

    2 DtNQ o

    .Dt

    xexp

    Dt

    Qt)N(x,

    4

    2

  • 7/31/2019 Dopant Diffusion1

    15/20

    Diffusion 15

    Erfc and Gaussian profiles are only

    approximate description of the actual

    profiles but they are used extensively for

    estimation of junction depth, resistance

    (known as sheet resistance) etc of diffusedlayers.

    Some reasons for deviations from erfc and

    Gaussian profiles:

    1.Dopant concentration dependent

    diffusion coefficient2.Lateral diffusion of dopant atoms

    3.Boron suck-out and phosphorus pile-up

    during drive-in

    4.Shallow junctions where delta function

    approximation of pre-deposition profile is

    not accurate.

  • 7/31/2019 Dopant Diffusion1

    16/20

    Diffusion 16

    Sheet Resistance and Irvin Curves

    With varying dopant concentration

    N(x), the conductivity and resistivity ofa diffused layer also change with

    distance: (x)= q(mn.n+mp.p) =1/(x).However the layer can carry current

    parallel to the wafer surface and will

    obey Ohm Law. For the purpose of

    characterising this layer and for the

    use of the layer to form resistors, thelayer is characterised by a quantity

    known as sheet resistance, Rs (ors).This is the resistance between the two

    opposite sides of a unit square (side

    length L) of the diffused layer as

    shown below. It is assumed thatcurrent flows parallel to the surface.

    Breaking this layer into elemental

    layers each of thickness dx. They are

    conductors in parallel.

  • 7/31/2019 Dopant Diffusion1

    17/20

    Diffusion 17

    The elemental layer between x and x+dx has net dopant concentration [N(x) NB]. Ithas a conductance:

    We have assumed here that only the majority carriers contribute to the current flow.We have also ignored the effect of the space charge region of the junction.

    Since all elemental layers are in parallel, the total conductance of the diffused layer,GS, the sheet conductance, is:

    Or, the sheet resistance, RS, is:

    RS has the unit of and it is independent of the actual value of L i.e. so long it is asquare piece of the diffused layer. RS is more commonly referred to in per square.

    Also notice that if we have the diffused layer of width L (direction perpendicular tocurrent flow) and length (direction parallel to current flow) 2L, its resistance is that oftwo square units connected in series i.e. R = 2RS.

    dx]N)x(N[q

    L

    Ldx]N)x(N[q)x(g

    B

    B

    m

    m

    mjx

    Bs dx]N)x(N[qG

    0

    m

    jx

    Bs

    s

    dx]N)x(N[qG

    R

    0

    11

  • 7/31/2019 Dopant Diffusion1

    18/20

    Diffusion 18

    Diffused layers are used to produce resistors in bipolar technology, typically

    the base diffusion is used for this purpose because it has higher sheet

    resistance. The value of a diffused resistor is given by:

    R = Rs. (L/W)Where Rs is the sheet resistance of the diffused layer and L and W are the

    length and width of the diffused area as seen from the top.

    L

    W

  • 7/31/2019 Dopant Diffusion1

    19/20

    Diffusion 19

    Sheet Resistance Measurement

    Sheet resistance of a diffused layer ( or an

    implanted layer) can be measured fairly

    easily using what is referred to as the

    (linear four-point probe measurement on atest wafer (of dopant type opposite to that

    of the diffusion) on which the diffusion is

    carried out on the whole wafer. The

    measurement is setup is shown in

    diagram. The four probes are in line equal

    distance a apart (typically 1mm). Constantcurrent I is flowing into wafer at probe 1

    and out of wafer at probe 4. Potential

    difference between probes 2 and 3, V is

    measured by a voltmeter.

    Consider first the situation that there is

    only current I into probe 1 (no current in or

    out of probe 4). We expect this current toflow uniformly radially outward in the

    diffused layer. At a radius r from probe 1,

    the current density (sheet current density)

    is:

    r

    IJs

    2

  • 7/31/2019 Dopant Diffusion1

    20/20

    Diffusion 20

    The sheet current density is related to the radial electric field E through Ohms Law:

    E is related to the radial change in potential dV/dr by:

    By integrating the above expression from probe 2 to probe 3, we gat the voltage drop

    from probe 2 to probe 3 due to current I flowing into probe 1:

    By similar consideration of a current I flowing out at probe 4 (with no current in or out of

    probe 1), we get the voltage across probe 2 and 3 as:

    The combination of the two above situations is equivalent to the case of the four point

    probe measurement. By superposition theorem, the measured voltage between probe 2

    and 3 for the four point probe measurement is:

    If the current is set at 4.532 mA, then the voltage measured is the sheet resistance in

    k per square.

    r

    IRJRE sss

    2

    r

    IREdr

    dVs

    2

    )ln(I

    R)a

    2aln(

    IRV ss 2

    22

    123

    )ln(I

    RV s 22

    423

    IV.

    IV

    )ln(Ror)ln(IRV ss

    2323

    235324

    22