Professional Documents
Culture Documents
Fig.4.1 (a) The schematic of a two-terminal MIS structure. (b) Band diagram of a two-terminal
MIS structure at zero gate voltage, showing accumulation of holes near the surface. VFB is the
flatband voltage, Xm is the metal work function, Xi is the electron affinity of the insulator, Xs is
the electron affinity of the semiconductor, and Eg is the band gap of the semiconductor.
Assumptions:
-Ideal MIS structure with no charges in the insulator layer and no surface states at the
semiconductor-insulator interface.
-The insulator layer has infinite resistivity, thus there is no current across the insulator
when a bias voltage is applied => Fermi level constant across the device.
Some definitions:
-Work function: energy required to remove an electron from the Fermi level to the
vacuum level (free space).
-Electron affinity: energy required to remove an electron from the conduction band to the
vacuum level.
At zero bias voltage, the band bending in the semiconductor layer is determined by the
work function difference between the metal and the semiconductor, and it can be
compensated by applying a voltage VFB to the gate
where VFB is called the flat-band voltage, Xm is the metal work function, and Xs is the
semiconductor electron affinity.
Note: this equation for VFB is applicable for an ideal MIS structure; however, if there are
charges in the insulator or at the insulator-semiconductor interface, then the gate voltage
required to obtain flatband condition would change.
Fig.4.2 The band diagram of the two-terminal MIS structure under the flatband condition. Vg is
the applied gate voltage.
Therefore, Si work function s = Xs + (Eg/2) + (Ei EF) = 4.05 + 0.56 + 0.35 = 4.96 eV
(a) For Al gate, VFB = 4.1 4.96 = 0.86 V
Note: all these numbers can be equivalently represented either in volts or in electron-volts,
depending on whether potential or energy is represented.
It is assumed here that the Fermi level of the n+-poly gate is coincident with the conduction
band.
Fig.4.3 The band diagram of a two-terminal MIS structure at (a) depletion and (b) inversion.
Note: at the onset of strong inversion Vs = , and also, that nsps = => consequence of
zero current in the semiconductor (perpendicular to the semiconductor-insulator
interface) => corresponds to constant (as a function of distance) EF in the semiconductor.
Note: deep into the bulk, from charge neutrality condition, NA = pp0 np0.
Thus,
Using the definition of the electric field F = dV(x)/dx, the above equation can be
rewritten as
Integrating this equation with respect to V, one gets
Thus,
where
EXAMPLE 4.3: Draw the low- and high-frequency C-V characteristics, clearly showing all the relevant points,
including the flatband capacitance, for a two-terminal MIS structure having 30 nm thick oxide and substrate doping
of 1015 cm 3 (p-type). Assume VFB = 1 V.
Fig.4.9 (a) The exact high-frequency equivalent circuit of a two-terminal MIS structure, and (b)
its simplified equivalent.
where
and
Note: both Ceq and Req are frequency dependent: in the limiting case of
+ Cdep, and in the other limiting case of
Fig.4.10 The C-V characteristics for a two-terminal MIS structure at different frequencies.
4.4.1 Extraction of Parameters from the C-V Characteristic
Fig.4.11 Parameter extraction from the C-V characteristic for a two-terminal MIS structure. The
parallel shift in the characteristic after the bias-temperature stress test (described later) is also
shown.
The maximum measured capacitance Cmax in the accumulation region gives the dielectric
thickness
The minimum measured capacitance Cmin at high frequency gives the doping
concentration (assumed uniform) in the substrate. Steps:
o First, determine the depletion capacitance Cdep in the strong inversion region from
1/Cdep = 1/Cmin 1/Cmax.
o Then, obtain the depletion region thickness from
o And, finally, calculate the doping concentration from the following two equations:
o These two equations need to be solved by iteration: first choose a suitable value
for (say, 0.3 V), obtain NA, recalculate , obtain another fine tuned value of
NA, and repeat the process until the desired accuracy is achieved.
It also gives the information about the flatband voltage VFB. Steps:
o The device capacitance CFB under flatband condition can be given by CFB = CiCs0/
(Ci + Cs0) =
o Thus,
o From a knowledge of di and NA, CFB/Cmax can be obtained, and the intercept can
be found on the C-V curve to yield VFB.
In most of the commercially available MOS capacitors and MOSFETs, silicon (Si) is used
as the semiconductor and silicon dioxide (SiO2) is used as the insulator.
Si being a crystalline material and SiO2 being an amorphous material, there is a sudden
discontinuity in the lattice structure at the Si-SiO2 interface.
Fig.4.12 Different types of charges in the Si-SiO2 interface and in the SiO2 layer.
This interface has attracted considerable interest over the last few decades, and
significant studies have been made on this structure, however, a detailed understanding of
many of its features is still lacking.
The interface and the oxide contains various types of charges, which can be broadly
categorized into the following:
o Charges due to fast surface states (or interface trapped charges) located at the
interface.
o Charges due to mobile impurity ions located in SiO2.
o Charges due to traps ionized by radiation within SiO2.
o Fixed surface state charges located at the interface.
4.5.1 Fast Surface States
These are also referred to as Tamm and Shockley states, after their inventors.
These are created at the interface due to the sudden termination of the crystal periodicity,
since all the bonds of the atoms at the surface are not fulfilled these unfulfilled bonds are
referred to as the dangling bonds.
Obviously, the density of these states is a function of the crystal orientation (since (100)
planes have lower atom density than (111) planes, MOSFETs are universally fabricated
on (100) oriented Si).
Roughly, one fast surface state is assigned for every surface atom, resulting in a density
Proper cleaving of the surface and consequent heat treatment with H2 drastically reduces
the density of these states to or so, since H2 compensates some of these
dangling bond by the formation of SiH.
These states behave acceptor-like or donor-like, depending on the position of the Fermi
level at the surface and the amount of band bending, and these are referred to as fast
states, since they capture and release the carriers at a fast rate.
When the surface potential changes, the charges in the surface states change as well, and
leads to a shift in VT and a change in the C-V characteristics.
Fig.4.13 The experimental C-V characteristics showing the difference between them due to the
presence of fast surface states.
There is a shift of the C-V curve towards the left due to the fast surface states, which
changes the flatband voltage.
In the equivalent circuit of an MIS structure, the fast surface states can be represented by
an additional series combination of an equivalent capacitance Css of the surface states,
and an additional resistance Rss, with the time constant RssCss representing the time
response of the surface states.
Fig.4.14 The overall high-frequency equivalent circuit for a two-terminal MIS structure showing
the additional components Css-Rss to account for the effects of fast surface states.
A major difficulty with early MOS devices was the instability of the threshold voltage
VT, i.e., it used to vary with bias under elevated temperatures.
This happens due to the rearrangement of the mobile ions within the oxide,
which are introduced into the oxide from the furnace walls during oxidation.
Fig.4.15 Shift in the C-V characteristic after the bias-temperature stress test due to ionic
contamination in the oxide, and its partial recovery after annealing with gate-substrate shorted.
The initial C-V characteristic is marked by (1), while those observed after 30 minutes at
127 C with VG = +10 V applied is marked by (2), and after heating the device for 30
minutes at the same temperature with the gate shorted to the substrate yields
characteristic marked by (3)- this experimental procedure is known as the bias-temperature
stress test.
Fig.4.16 Charge distribution during the various stages of the bias-temperature stress test and post
annealing.
Initially, all the positive ionic charges are located at the metal-SiO2 interface, exerting no
influence on Si; after positive gate bias at high temperature, all these ionic charges cluster
near the Si-SiO2 interface and induce all the image charges in Si; finally after recovery,
the ions create an arbitrary distribution (x) within the oxide, inducing image charges in
both the gate and the semiconductor.
For any arbitrary distribution of the oxide charges (x), the shift in the flatband voltage
can be given by
The menace created by mobile ions is reduced to a large extent in today's technology due
to the improvements in the fabrication process.
EXAMPLE 4.4: In a two-terminal MIS structure having 40 nm thick oxide, the shift in the flatband
voltage after a bias-temperature stress test was found to be 10 mV. Determine the mobile ionic
contamination per unit area in the oxide in numbers per unit area.
SOLUTION: The oxide capacitance per unit area
The shift in the flatband voltage due to the mobile ionic contamination after bias-temperature
stress test is given by Thus, the mobile ionic contamination per unit area in the
oxide
A positive space charge is seen to build up in SiO2 films when it is irradiated by ionizing
radiation of various kinds, e.g., X-ray, gamma ray, low- and high-energy electron
irradiation, etc. (potential danger during ion implantation).
The physical origin of this charge is completely different from the ionic contamination.
Due to irradiation, EHPs will be generated within the SiO2.
In the absence of any electric field within the oxide, these carriers will immediately
recombine; however, under a positive applied gate bias, due to the electric field within
the SiO2, the generated electrons and holes would separate, with the electron moving
towards the metal-SiO2 interface, and the hole moving towards the SiO2-Si interface.
Thus, a space charge layer starts to build up within the oxide due to these charges, thus
creating an electric field within the oxide, which is opposite to that of the applied field =>
changes VFB, and, thus, VT.
These charges can be eliminated by thermal annealing.
A fixed charge is seen to exist within the oxide very near the Si-SiO2 interface, which
results in a parallel translation in the C-V characteristics along the voltage axis these
charges are called the surface state charges, and the density of these charges per unit area
is denoted by
These surface states have the following properties:
o It is fixed, i.e., its charge states cannot be changed over a wide variation in the
band bending.
o Unchanged under bias-temperature stress test and thermal annealing.
o It is located within 200 of the Si-SiO2 interface.
o Its density is not significantly altered by the oxide thickness, or by the type or
concentration of impurities in Si.
o Its density is a strong function of the oxidation and annealing conditions, and the
orientation of the Si crystal.
The ratio o f in (111), (110), and (100) Si are in the ratio 3:2:1, and is a strong
function of the oxidation condition.
Popular theory: originates from the excess ionic Si in the oxide, which moves into the
growing SiO2 layer during the oxidation process.
can be reduced by a large extent by H2 heat treatment
The general expression for the flatband voltage VFB can be given by
The standard charge control model (SCCM) postulates that the interface inversion charge
of electrons qns is proportional to the applied voltage swing VGT = VG -VT.
This model is an adequate description of the strong inversion region of the MIS capacitor,
but fails for applied voltages near and below VT (i.e., in the depletion and weak inversion
regions).
A new model has recently been proposed which has been shown to model the device
behavior adequately both in the weak and strong inversion regions, and is given as:
The ideality factor reflects the gate voltage division between the insulator layer
capacitance Ci and the depletion layer capacitance Cdep.
In the subthreshold regime,
At the onset of strong inversion (VGT = 0), the surface potential Vs has the value
Below threshold, we have the following approximate relationship:
Note: in general, is dependent on VGT, and at low substrate doping levels, is close to
unity near threshold where the gate depletion width is large (corresponding to Cdep <<
Ci).
Usually, Cdep can be estimated as follows:
and the value for the unified capacitance per unit area at threshold becomes
From the experimentally determined gate-channel capacitance, the inversion carrier sheet
density can be calculated as
According to UCCM, this should agree with Eq.(4.29), which can be written as
The values of obtained from the slopes in Fig.4.18 agree very well with those
determined directly from the subthreshold I-V characteristics, and the value of di
calculated from a is in excellent agreement with that measured by ellipsometry.
In Fig.4.19, the value of VGS corresponding to the peak value of should
coincide with the value of VGS at which the gate-channel capacitance has dropped to one-
third of its maximum value.
In Fig.4.20, the agreement between the measured and the calculated data is excellent for
the entire range of gate bias.
Fig.4.20 Measured (solid lines) and calculated (UCCM, symbols) ns versus VGS characteristics
for different values of Vsub in (a) semilog scale and (b) linear scale. In (b), the results obtained
from the simple charge control model (SCCM) are also shown.
The deviation in the measured curves found in the deep subthreshold region is due to two
reasons: one is the C-V measurement error, and the other is the leakage current, which
dominates deep subthreshold operation.
At deep subthreshold, the channel offers a large series resistance compared with the
reactance of the capacitance.
Note: the UCCM does not have an exact analytical solution for the inversion charge in
terms of the applied voltage even though an accurate approximate solution can be
obtained.
Above threshold, the sheet density of carriers in the inversion layer can be given as
Below threshold, the electron sheet density in the channel can be written as
From Eq.(4.37), the following expression is obtained for the subthreshold differential
channel capacitance per unit area
In this case, the quantization of the energy levels in the potential well at the
semiconductor-insulator interface in the direction perpendicular to the interface must be
taken into account.
Once quantization of energy levels take place, then the dispersion (E-k) relation in the
direction parallel to the interface is given by:
where En is the electron energy, Ej is the energy level of the jth subband, and ky and kz are the
wave vector components parallel to the interface.
Fig.4.21 Schematic diagram of energy subbands at the semiconductor-insulator interface
(assuming constant effective field approximation).
For a relatively thick electron gas layer, the number of subbands is large and the energy
difference between the bottoms of the subbands is small (<< kT).
For a relatively thin electron gas layer, only the lowest few subbands are important for
electron occupation, and the energy difference between the bottoms of the subbands may
become large compared to the thermal energy kT.
In this case, the electron gas is often referred to as a two-dimensional electron gas
(2DEG).
The density of states D for each subband is given by which is a constant
and independent of the subband energy Ej => the overall density of states has a staircase
dependence on energy for a triangular quantum well, which is characteristic for the
semiconductor-insulator interface of an MIS structure.
The number of electrons occupying a given subband j can be found by multiplying the
density of states D for a single subband by the F-D distribution function, and integrating
from Ej to infinity:
Fig.4.22 Energy levels (bottoms of subbands) and density of states for a triangular quantum well
structure (j = 1, 2, …, correspond to the different subbands).
After evaluating this integral and adding the contribution from all subbands, one obtains
The quantized energy levels for the subbands can be found using a numerical self-
consistent solution of the dinger and Poisson's equations.
However, an excellent approximation for the exact solution can be found by assuming a
linear potential profile (i.e., constant effective field Feff) in the semiconductor and close to
the semiconductor-insulator interface.
In this case, the energy levels are given by
where is the effective mass for electron motion perpendicular to the (100) surface, and Ec(0)
is the minimum conduction band energy at the Si-SiO2 interface.
The effective field Feff is expressed through the surface field FS and the bulk field FB.
For electrons, the relationship linking Feff, FB, and FS, giving the best fit to the self-
consistent solution of dinger and Poisson's equation is given by
and Feff = (FS + FB)/2, where ns is the interface electron sheet
density, and qnB (= qNAddep(av)) is the sheet density of depletion charge.
Similarly, for holes, FS = q(ps + where ps is the
interface hole sheet density, and qpB (= qNDddep(av)) is the sheet density of depletion
charge.
In reality, it has been found that a slightly different form of the effective field Feff1 = (FS +
2FB)/3 gives a better fit to the measured data.
Solving these equations iteratively, one can obtain the relation between ns and the Fermi
level [EF Ec(0)].
Fig.4.23 Comparison of the interface carrier density versus EF Ec(0) characteristics for different
substrate doping densities in (a) semilog plot and (b) linear plot. Symbols: calculations based on
a 2DEG formulation, solid lines: charge sheet model, straight line in b): linear approximation to
2DEG formulation, the slope gives
The difference between the curves at high substrate doping levels is caused by the fact
that the large bulk field quantizes the energy levels even in the subthreshold region.
However, at strong inversion, the difference between the charge sheet model and the
2DEG formulation is large.
As can be seen from Fig.4.23, the dependence of ns on EF in the above threshold regime
can be approximated by a straight line: where EF0 is the intercept of
this linear approximation with ns = 0.
This approximation means that a fraction of the applied voltage, equal to is
accommodated by a shift in the Fermi level with respect to the bottom of the conduction
band.
The shift in the Fermi level with respect to the bottom of the conduction band changes the
above-threshold capacitance from to where the parameter can
be interpreted as a correction to the insulator thickness.
From the straight-line approximation in Fig.4.23b), is obtained, which is much
smaller than that of
This difference is caused by
o a much larger effective mass in the conduction band in Si, which makes quantum
effects much less pronounced, and
o the large difference in the dielectric constants between the insulator and the
semiconductor for the MOS system.
Practice Problems
4.1 Clearly draw the band diagrams for an ideal MOS structure and no oxide charge)
on n-type Si for i) accumulation, ii) depletion, and iii) inversion. If the oxide thickness tox = 40
nm and VG = 1 V, determine the magnitude and sign of the charge density in the
semiconductor. What is the status of the surface?
4.2 Show that for an MOS structure on p-type Si, the electron and hole concentrations as
functions of position are given by where n0 and p0
are the equilibrium electron and hole concentrations respectively, and is defined by =
[Ei(bulk) Ei(x)]/q.
4.3 Continuing with the derivation given in Section 4.2, show that the electric field E in the
semiconductor in an MIS capacitor can be given by where all the notations
carry their usual meanings.
4.4 Sketch the electric field and voltage distribution in an MOS structure at the threshold gate
voltage. Data: substrate voltage = 0, and VFB = 0. Compute the
threshold voltage VTH from the voltage distribution.
4.5 Calculate and plot the semiconductor surface charge per unit area for an MIS structure as
a function of the surface potential
4.6 Starting from Eqn.(4.16), show that at flatband (i.e., when Vs = 0), the flatband capacitance
per unit area Hence, compute its magnitudes for substrate
dopings of
4.8 (a) Find the voltage VFB required to reduce to zero the negative charge induced at the
semiconductor surface by a sheet of positive charge located below the metal.
(b) In the case of an arbitrary distribution of charge in the oxide, show that
4.12. Calculate and plot the maximum width of the depletion region for an ideal (i.e., VFB = 0)
MIS capacitor on p-type Si with as a function of the substrate bias Vsub for -2 V <
Vsub < 0.1 V. Assume that the voltage difference between the inversion layer at the interface and
the gate contact is maintained constant when the substrate potential is changed (charge
screening), so that the substrate voltage reverse biases the inversion layer/p-type substrate
junction. Also, calculate the threshold voltage VT, and the capacitance of the structure at low and
high frequencies for V >> VT for Vsub = 0. Data: ni =
4.13 Calculate and plot the surface potential as a function of the gate voltage VG in depletion
and inversion for a two-terminal MIS structure. Identify the weak inversion, moderate inversion,
and the strong inversion regions in the plot (as per Tsividis). Can the plot be really linearized in
subthreshold? Determine an effective value of in subthreshold from the plot.
4.14 Calculate and plot the gate-to-substrate capacitance Cmis as a function of the gate voltage VG
for a two-terminal MIS structure with area = The plot should show all the regions of
operation (i.e., accumulation, depletion, weak inversion, and strong inversion). Mark Cso in the
plot, with the magnitude shown. (Note: the externally measured capacitance includes the oxide
capacitance).
4.15 Calculate and plot the temperature dependence of the surface charge per unit area for the
surface potential i) in the temperature range between 150 K and 450 K.
Data: effective densities of states in conduction and valence bands and
respectively at 300 K (with both of them having a dependence), and the
energy gap Eg = 1.12 eV (the variation of the energy gap with temperature may be neglected).
4.16 From the equivalent circuit for an MIS structure, determine the expression for the
impedance across its two terminals as a function of frequency. Hence, calculate and plot the
effective capacitance of the structure as a function of the gate voltage VG (varying from -5 V to
+5 V) for frequencies of
4.17 As a practice problem, draw any arbitrary C-V curve of your choice, and following the
parameter extraction algorithm discussed in Section 4.4.1, obtain the i) oxide thickness, ii)
threshold voltage, iii) substrate doping, iv) flatband capacitance, v) flatband voltage, and vi)
fixed oxide charges.
4.21 (a) Compute and plot the surface electron concentration ns as a function of [EF - EC(0)]
under the 2DEG approximation for
(b) Repeat part (a) under the 3D approximation (i.e., the 3D charge sheet model as given by
Brews).
Data: (Note: the
constant energy surface for Si consists of six ellipsoids of revolution, and ml ( ) and mt (
) represents the lateral and transverse effective mass respectively. For {100} direction four
of these ellipsoids will lye on the surface and two ellipsoids will be perpendicular. Refer to
Problem 22 also.)
4.22 In the classical limit, the separation of the energy subbands in a 2D electron gas is small
compared to the thermal energy kT. In this case, the sheet density of the 2D electron gas is given
by the classical charge sheet model, given by Eqn.(4.46), which is derived using a conventional
3D electron gas approach. Show that in this limit (i.e., Ej Ej 1 << kT), the equation
reduces to Eqn.(4.46). In the above equation, mpi is the parallel effective mass for the valley i,
and Eji is the energy level of the jth subband in valley i. Note: the effective mass mpi is mt for two
valleys, and for four valleys, where mt and ml is the transverse and lateral effective mass
respectively.
METAL-OXIDE-SEMICONDUCTOR FIELD-EFFECT TRANSISTORS (MOSFETs)
Principle of Operation
Schematic diagram of an n-channel MOSFET, where two n+ source and drain regions are
diffused into a p-type substrate, making it a four terminal (drain, source, gate, and substrate)
device.
Note: the C-V characteristic of this device shows low-frequency behavior (of the two-
terminal MIS structure) up to a fairly high frequency (of the order of the inverse transit
time of the carriers across the channel), since the heavily doped source/drain regions
provide an infinite reservoir, from which the carriers can move into the channel, or to
which they can escape from the channel.
The GCA, proposed by Shockley, is used in order to calculate the I-V characteristic of
the device.
This approximation states that the rate of variation of the lateral field within the channel
is much smaller than the rate of variation of the vertical field, i.e., ,
and the channel potential is assumed to be a gradually changing function of position.
Note: This approximation actually states that the channel potential varies very little along
the channel over a distance of the order of the insulator thickness , i.e., this requires
<< L, where L is the channel length.
The gradual channel approximation (GCA):
Fig (a) schematic comparison of the parallel and perpendicular electric fields in
the channel, and
Fig (b) qualitative potential profile in the channel.
However, modern MOSFETs have extremely short channel lengths, and this requirement
is not often met; thus, the GCA fails for most of modern MOSFETs, nevertheless its
discussion is important.
According to GCA, the charge induced at any position along the channel can be
determined from the formulas derived for the MIS structure, provided the constant
surface potential for an MIS structure is replaced by a variable channel potential
in the expression for the surface charge density per unit area in the semiconductor.
Assumption: The device is operating in the above threshold regime, i.e., the gate voltage
is sufficiently large to create strong inversion throughout the channel.
where is the insulator capacitance per unit area, and the term within the square
brackets is the voltage drop across the insulator.
Band diagrams at
Note: here we are considering an n-channel device, however, all the results are also
applicable for a p-channel device, provided appropriate sign changes are made.
however, elsewhere in the channel, the total band bending between the substrate and the
surface is , since the induced n-channel/p-substrate junction is reverse
biased by the .
The band bending increases in the channel as one moves from the source to the drain,
which leads to an increase in the width of the depletion layer and of the depletion charge
density, thus, the exact expression for can be given by
Since the drain current is carried entirely due to drift, its expression can be given by
In writing this equation, it is assumed that the electron drift velocity is proportional to
the component of the electric field parallel to the Si- interface, i.e., .
Note: for short channel devices, this electric field may be sufficiently high to cause
velocity saturation in the channel.
Note: is a function of ; thus, substituting the expression for in the above equation,
noting that is a constant throughout the channel, and integrating it from the source,
i.e.,x = 0 ( = 0) to the drain, i.e., x = L ,
Note: is the voltage difference between the inversion layer at the source end and the
substrate contact, and its sign should be such that it never forward biases the inversion
layer-substrate junction (a small forward bias, much less than may be allowed in
certain cases).
For both n- and p-channel MOSFETs, the magnitude of the threshold voltage VT
increases with an increase in |Vsub|.
A physical insight into the phenomenon of saturation may be obtained by analyzing the
electric field distribution under the gate.
Solving Eqs.(5.12) and (5.13) together, the field profiles can be calculated.
Fig.5.6 The variation of the electric field along the channel for drain voltage nearly equal
to the saturation voltage for gate voltages
Note: from the constancy of the drain current throughout the device, it can be seen that
as and the electric field F(L) diverges.
The differential drain conductance
tends to zero when and the I-V characteristics may be extrapolated in the
voltage region assuming a constant (independent of drain current
Note: this approach is only valid when the channel electrons do not suffer any velocity
saturation due to high electric fields.
Note: modern day MOSFETs have extremely small gate lengths, and the channel has
high electric fields (more than the critical electric field required for velocity saturation),
which creates the velocity saturation effects for the channel electrons.
Fig.5.7 The I-V characteristics of an n-channel MOSFET for different values of gate
voltage . The dashed line represents the drain-to-source saturation voltage.
Fig.5.8 The variation of the drain saturation current with gate voltage for three different
values of substrate doping.
For very small the terms under the curly brackets in Eq.(5.15) can be expanded in
Taylor series, leading to the following simplified expression for the I-V characteristics in
the linear region:
A physical justification of Eq.(5.16) can be given as follows:
At very small the charge induced in the channel is, to the first order, independent of
the channel potential, thus,
(5.17)
Now, for small the electric field F in the channel is nearly constant, and is
given by
The drain current is entirely due to drift, and is given by the electrons in transit model:
since
Compare Eq.(5.19) with Eq.(5.2): in Eq.(5.19), the variation of the depletion charge
density with the channel potential has been neglected.
The drain current can now be given by
Integrating Eq.(5.21) from x = 0 (source side) to x = L (drain side), which corresponds to
a change in from the following expressions for the I-V
characteristics are obtained:
Fig.5.9 The I-V characteristics of an n-channel MOSFET calculated using the charge
control model (solid curve) and the Shockley model (dashed curve).
Thus, in order to achieve a high value for the transconductance gm, the following steps
may be taken.
Higher value of low field electron mobility
Thinner gate dielectric layers, which in turn gives large values for the insulator
capacitance per unit area
Large widths (W) and short lengths (L).
Note: for short channel devices, where velocity saturation effects are important, the
dependence of transconductance on the low-field electron mobility and the gate length
gets strongly affected.
SOLUTION:
i)
Hence, the device is under linear mode of operation
Note the huge change in transconductance in saturation as compared to the linear region: this is
due to the square law dependence of current on the gate voltage in the saturation region (as
against the linear variation in the linear region).
Drain Conductance
This is due to the independence of the saturation drain current on the drain voltage. In reality,
channel length modulation creates a change in drain current with respect to the drain voltage in
saturation, and finite drain conductance
The analysis so far neglects the effects of the source/drain series resistance, and the entire
voltage is assumed to drop along the channel.
However, for modern day MOSFETs, this effect cannot be ignored, due to smaller
diffusion cross-sections and smaller drain currents.
The extrinsic (measured) voltages can be related to the intrinsic (device)
voltages by the following equations:
where are the source and drain resistances respectively.
Fig.5.11 The drain current drain-to-source voltage characteristics for different values of
The series source resistance reduces the drain current, and the series drain resistance
increases the drain-to-source saturation voltage.
Both series source resistance and series drain resistance reduce the drain conductance at
low drain-to-source voltages.
In modern day MOSFETs, the channel length is very small, the electric field in the
channel is very high, and the velocity saturation effects are very important.
The measured electron and hole mobilities in the inversion layer may be quite different
than those measured in the bulk.
Note: the channel, in reality, is under a two-dimensional electric field, one directed
longitudinally from the gate to the substrate, and the other directed laterally along
the length of the channel.
The effective inversion layer thickness is approximately given by thus, a
large vertical field creates a narrow inversion layer, and vice versa.
Fig.5.12 The random path of electrons in the channel, undergoing surface scattering,
which is more intense in narrow channels.
Electrons in the channel move in random directions, undergoing surface scattering, which
increases for narrow channels thus their mobility drops.
Fig.5.13 The variation of the electron and hole mobilities in the channel as a function of
the gate electric field.
The dependence of the electron and hole mobilities on the gate field can be crudely
approximated by
For this derivation, a simple two-piece linear approximation for the electron velocity is
used:
where is the electric field required for velocity saturation, and is the saturation
limited thermal velocity.
Recall: in the linear region, the I-V characteristic can be given by:
where
The saturation current can now be found by assuming that the current saturation
occurs when the electric field at the drain side of the channel exceeds the critical field
required for velocity saturation.
This is a much more realistic assumption than the Shockley model, which assumes
saturation occurs when
The constant mobility model is still used for drain voltages below the saturation voltage.
The absolute value of the electric field in the channel at drain
voltages below the saturation voltage can be obtained from Eq.(5.21):
Integrating Eq.(5.35) from 0 to x, the following equation for the channel potential is
obtained for drain voltages below the saturation voltage:
Substituting Eq.(5.37) into Eq.(5.35), the following expression for the electric field as a
function of distance is obtained:
and the electric field F(L) at the drain side of the channel (where it is the largest),
From the condition the drain saturation current can now be found as
Fig.5.14 The variation of the drain saturation current as a function of the gate length for
three different values of the gate voltage (3 V, 5 V, and 7 V). The drain saturation current
predicted by the constant mobility model (shown by the dashed line) is also shown for
comparison.
The effects of source/drain series resistance, for these cases, can be accounted for (as
done earlier for long channel devices), and the following expressions for the drain
saturation current and the drain saturation voltage are obtained:
Interpolated Relation
The following interpolation formula for the MOSFET I-V characteristic has been
proposed by Shur, which describes both limiting cases
correctly:
This was one of the earlier formulas, and a huge amount of work has been done in this
area for the last ten years or so, in order to further refine the description of the behavior
of short-channel MOSFETs.
In practical devices, the I-V characteristics do not completely saturate at large drain-to-
source voltages, and this is related to the short channel and other nonideal effects in
MOSFETs.
In order to account for the finite slope of the output characteristics in saturation, the
following modification to the drain current expression has been proposed:
Fig.5.15 I-V characteristics of two n-channel MOSFETs: (i) L = 0.5 (dashed lines),
and (ii) L = 0.75 (solid lines).
Fig.5.16 The variation of the threshold voltage with the effective channel length.
Another interesting feature seen in short channel devices is that the saturation current
increases as the device length is reduced.
Now, based on the existing model for the threshold voltage, which states that it is
independent of the device length this behavior cannot be explained.
In reality, it has been shown that the threshold voltage is a strong function of the channel
length (for short channel devices), and it actually decreases with a decrease in the channel
length, which explains the reason behind the larger saturation current.
The reduction of the threshold voltage with a reduction in the channel length can be
explained by the charge sharing model.
Fig.5.17 The depletion charge profiles for (a) a long channel device, and (b) a short
channel device.
For a long channel device, the depletion layer thickness at the source end of the
channel and at the drain end of the channel are much less than the channel length L,
and, thus, the depletion charge enclosed by these sections are much smaller than the total
depletion charge under the gate.
However, for a short channel device, the widths of these depletion regions are a non-
negligible fraction of the total depletion charge under the gate.
Note: essentially, the depletion regions near the source and the drain are contributed by
the source-substrate and the drain-substrate bias, and gate has no role to play.
Under an applied drain-source bias, the depletion region thickness near the drain will
obviously be larger than that at the source side.
The net effect is that the gate now has to compensate for a lower depletion charge density
than that for a long channel device, which qualitatively explains the reduction of the
threshold voltage with a reduction in the channel length.
The exact analysis of the charge sharing effects requires a two-dimensional analysis,
however, to the first order, it is assumed that the effect of the depletion width at the
drain side of the channel is to reduce the effective channel length in the saturation region
from L to where
Here, is the effective channel length, and the voltage dropped along this section is
assumed to be equal to the drain saturation voltage , and is length of the pinched-
off portion of the channel (related to the drain depletion width), where the excess drain
voltage beyond , i.e., is dropped, where is the applied drain voltage.
With an increase in the length of the pinch-off region also increases, leading to a
reduction in the effective channel length .
This effect is called the channel length modulation effect, and this effect leads to a higher
drain saturation current, and finite output conductance in the saturation region.
A very crude estimate of the pinch-off length (also referred to as the drain region
length) can be obtained from the solution of the one-dimensional Poisson's equation:
A more accurate and realistic expression for may be obtained by assuming that the
electrons are injected from the inversion layer into the drain depletion region, and they
spread uniformly, leading to the current density
Here, is the diffusion depth of the drain region, and is the thickness of the
inversion layer
It is also assumed that the velocity of electrons in this region is saturated, thus their
volume density can be given by
Now, the one-dimensional Poisson's equation can be rewritten as:
The solution of this equation leads to the following complicated expression for :
For gate lengths larger than or about 1 , and drain-to-source voltages smaller than or
about 10 V, this expression may be simplified to give
In short channel devices, the depletion charge under the channel [dependent on the
channel potential and has been represented by the second term within the brackets in the
right-hand side of Eq.(5.7)], which has been neglected in the charge control model [Eq.
(5.19)], has to be accounted for.
This effect may be taken into account by introducing an additional parameter a into the
equations of the charge control model, with the resulting equations given by
Linear Region
Saturation Region
For Si, the (empirical and fitting) parameter a describes the influence of the bulk
substrate depletion layer on the device characteristics, and can be approximated by the
following expression
The threshold voltage and the parameter K can be determined from the experimentally
measured data for a given device.
In addition, the dependence of electron mobility on the longitudinal and transverse
electric field in the channel should be included for a more realistic device modeling,
however, this simple empirical model gives adequately good fit with the measured data.
Fig.5.18 The measured and calculated I-V characteristics for a Si n-channel MOSFET.
Similar to the short channel device, the threshold voltage of a narrow channel (along the
width) device increases with a reduction in the effective device width Weff due to the
fringing fields outside the gate region, and the change in the threshold voltage as a
function of Weff can be given by
where is a constant.
Another non-ideal effect that may be especially important for short-channel devices is the
injection of electrons from the channel directly to the gate dielectric, where these
electrons get trapped => hot electron effect.
This phenomenon takes place because the carriers gain sufficient energy while traversing
the drain depletion region, which contains a high electric field, and has been used to
advantage in the FAMOS (Floating gate avalanche MOS) structures used in memories.
Avalanche breakdown of the drain-substrate junction can cause a sharp increase in the
drain current, and can damage the device unless it is controlled by some external means.
Typically, avalanche breakdown for a heavily doped drain-moderately doped substrate
junction takes place at approximately 8 to 10 V.
Another very important nonideal and potentially hazardous situation may arise due to
punchthrough, where the drain and source depletion regions touch each other and cause
abnormally large current to flow through the device: this effect is particularly severe for
short channel devices.
Punchthrough effect creates a superlinear increase in the drain current with the drain
voltage, even at gate voltages below the threshold voltage.
Subthreshold Conduction
So far, we have considered current flow in a MOSFET only when the gate voltage
exceeds the threshold voltage.
However, in reality, a finite (nonzero) current does flow in a MOSFET even for gate
voltages below the threshold voltage, and this effect is more marked for short channel
length devices than their long channel counterparts.
This current is referred to as the subthreshold current, and it flows for
when the surface potential lies between the ranges of the onset of weak inversion and the
onset of strong inversion.
The mechanism responsible for subthreshold current is quite different for long-channel
and short-channel devices. 5.6.1 Subthreshold Current in a Long Channel Device
In a long channel device, the situation is similar to a BJT, where the source plays the role
of the emitter, the drain is equivalent to the collector, and the substrate is the base.
The drain voltage drops almost entirely across the drain-substrate depletion region.
Thus, the component of the electric field parallel to the interface is small, and the
subthreshold current is contributed primarily by diffusion, just as the case for BJTs.
Fig.5.20 The depletion regions associated with a (a) long channel and (b) short channel
device.
Thus, the subthreshold current can be evaluated as
where is the region where most electrons are located) is the effective
cross-sectional area.
Thus, the effective depth where most of the electrons are concentrated, can be
estimated as where y = 0 corresponds to the interface.
If the diffusion length of electrons in the substrate is much greater than the channel
length L, then the electron density n should be a linear function of x, decreasing from the
source towards the drain (just like the linear distribution of minority carriers in the base
of a BJT):
where the volume concentrations for electrons at the source and the drain sides
of the channel are given by
For long channel devices, it is assumed that the depletion widths at the source and the
drain sides of the channel are small compared to the channel length L, and
Also, note that since
Using all the relations given above, the subthreshold current for a long channel MOSFET
can be given by
The surface potential at the source can be expressed as a function of the gate voltage
by noting that thus,
where
Note: For the subthreshold current becomes independent of the drain voltage.
This is expected since in a long channel device, most of the applied drain voltage drops at
the drain-substrate depletion region, and since the current is diffusive in nature, there is
no change in the current with the drain voltage.
Also, for large since the gradient of n is not affected by the drain voltage:
a situation similar to BJTs, where the collector current in the forward active mode is
independent of the collector-to-emitter voltage.
Note: the subthreshold current is almost independent of the drain voltage
The substrate bias shifts the threshold voltage to a more positive value, affects the surface
potential, and thus the subthreshold current changes.
Fig.5.21 The subthreshold characteristics for a long channel device as a function of the
gate voltage for different values of drain and substrate voltages.
where is the built-in voltage of the source/drain-substrate junction, and the surface
potential is now found from the solution of the following equation:
where
The curves clearly show shifts in the subthreshold current for different values of drain
voltages, a characteristic typical of short channel devices.
The subthreshold current is a strong function of temperature as well
Fig.5.22 The subthreshold characteristics for a short channel device as a function of gate
voltage for different values of drain and substrate voltages.
Fig.5.23 The subthreshold characteristics as a function of gate voltage for two different
temperatures (77 K and 300 K).
Note: in a MOSFET, the charges in the depletion region and the inversion layer depend
on the gate, source, drain, and substrate potentials; and the derivatives of these charges
with respect to the terminal voltages give rise to MOSFET capacitances.
The small signal equivalent circuit shown in Fig.5.24 is the one used by the popular
circuit simulation package called SPICE, and it contains:
the drain-to-source current source IDS,
two resistances (due to the quasi-neutral region resistances of the source and
drain respectively)
The coefficient
Therefore, and
respectively. Thus, significant
degradation in the transconductances and drain conductance may take place for large values of
source/drain series resistances.
Note: at low frequencies, when the effects of the capacitances can be neglected, the
voltage gain can be given by as expected.
Another simplified equivalent circuit, suitable for the calculation of the current gain, is
shown in Fig.5.26(c).
Fig.5.26(c) The alternate simplified equivalent circuit for a MOSFET suitable for the
calculation of the short circuit current gain.
From Fig.5.26(c), the short circuit current gain can be easily found to be:
Thus, the unity gain cutoff frequency (i.e., the frequency at which the absolute value of
the short circuit current gain is equal to unity) can be given by
where
EXAMPLE 5.4: Calculate the unity-gain cutoff frequency for the MOSFET considered in
Example 5.2. Compare this value with theoretical maximum value for , assuming
An actual device would show a cutoff frequency, which is smaller of the two, thus, the actual
unity-gain cutoff for the device considered in Example 5.2 would be 2.82 GHz.
Types of MOSFETs
Broadly, MOSFETs can be categorized into two types: enhancement and depletion.
Enhancement type devices are normally off, i.e., channel does not exist for and
the applied must be greater than for the device to turn on.
On the other hand, depletion type devices are normally on, i.e., channel does exist even
for and the applied must be reduced below for the device to turn off.
To put it simply, an n-channel enhancement type device has a positive , whereas an n-
channel depletion type device has a negative .
Similarly, a p-channel enhancement type device has a negative , whereas a p-channel
depletion type device has a positive .
The threshold voltage can be changed either by doping or by ion implantation, where
high energy ions are made to bombard the surface and get embedded into it: since these
are charged, they can change the charge state of the surface, and, hence, the threshold
voltage.
The shift in the threshold voltage is related to the ion density by the relation:
eg., negative ions (like Boron) implanted in a p-channel (n-substrate)
device will compensate some of the positive depletion charges and make the threshold
voltage less negative, however, note the same ions would shift the threshold voltage to
more positive for n-channel (p-substrate) device.
Since the threshold voltage is shifting towards negative value, hence, obviously, the type of
implant required is positive ions (e.g., P, As, Sb, etc.), which would compensate the negative
depletion charge of the substrate and push the threshold voltage towards negative direction.
For MOSFETs, the UCCM equation for MIS capacitors [Eq.(4.24)] has to be modified to
account for the channel potential, thus, the inversion charge is related to the gate-source
and channel potential as follows:
In order to get a better understanding of the term first consider the simplified version of
the charge control model, given by
Now, in reality, the threshold voltage depends on the depletion charge.
Taking into account the dependence of this charge on the channel potential, one can write
the corresponding position dependent threshold voltage as
This makes the charge control equation nonlinear and difficult to use in device modeling.
However, if Eq.(5.90) is linearized with respect to V, one can write
where now is the value of the threshold voltage at the source side of the channel.
Thus, one obtains
A generalized solution for ns is used in UCCM, given by
This equation allows the direct determination of the carrier distribution along the channel
as a function of
Of late, area of considerable interest, since an accurate modeling of the pinch-off region
is essential in order to obtain an exact drain current model in saturation.
Important to find a solution for the longitudinal field in the channel.
The model relies on the fundamental assumption that the carrier velocity in the saturated
part of the channel is constant and equal to the saturation velocity, which implies that the
carrier sheet density in the saturated part of the channel is also constant.
Another assumption made is that the substrate is lowly doped: this assumption
oversimplifies the true physics of the saturation region, however, it also leads to a
manageable theory with qualitatively correct features, which gives a fairly good fit to
experimental data with a judicious choice of parameters such as the saturation velocity
and the effective channel thickness.
The intrinsic saturation voltage can be defined as the intrinsic drain-source voltage
for which the longitudinal electric field at the drain end of the channel just becomes
equal to the saturation field
For the location in the channel where marks the boundary between
the saturated and the non-saturated regions.
The boundary point moves towards the source with increasing drain-source
voltage: this effect is called the channel length modulation.
Another important parameter is the channel potential at the boundary point
The two parameters and on the intrinsic gate-source voltage
and have to be determined self-consistently using the models for the two regions with the
requirement that the potential, the electric field, and the velocity be continuous at
For a description of the saturated region, it is necessary to consider a two-dimensional
Poisson's equation of the form