Leakage Currents in Illuminated Junctions and Biased Transistors

1. Summary

Analysis of the frequency dependence of the junction photo-voltage (JPV) in illuminated

junctions (under weak forward bias) yields the amplitude of the ideal diode I-V
characteristic, Io, dominated by the carrier recombination effects in the depletion layer.
This carrier recombination leakage is strongly dependent on process conditions, which
determine the dopant activation and profile shape, junction location and residual damage
distribution in the junction.

Leakage in reverse biased transistors and diodes includes the effects of carrier generation,
related to residual damage density and location relative to the junction boundary, as well
as structure and bias dependent effects of gate oxide leakage, band-to-band tunneling at
the drain junction and thermionic emission from metal contacts. All of these effects
depend on process conditions, through dependence on dopant activation and profile
shape, junction location and local electric fields. However, only the
generation/recombination current is strongly sensitive to the residual defect density and
location relative to the junction boundaries, which are strongly dependent on implant and
annealing process conditions.

Comparisons of the dependence on process-sensitive effects (related to doping and

damage distributions) for leakage currents measured in forward (RsL) and reverse
(diodes & transistors) bias show that RsL measures have a much wider dynamic range
than diode and transistor measurements, which are limited by effects of structural, contact
and additional leakage factors, such as gate oxide conduction.

2. Sources of leakage currents in p-n junctions and transistors

The p-n junction (for example, a p+ top junction with n+ sub-junction doping) leakage
current under reverse bias includes the contributions of diffusion current, Jdiff, space
charge generation current, Jgen, band-to-band tunneling current, Jtun, and thermionic
emission current, Jthem, [1]:

J leak = J diff + J gen + J tun + J therm . (1)

Additional terms for MOS transistors include leakage from gate oxide conductance, as
sketched in Fig. 1.

Comparison of leakage currents in RsL measurements and transistors

January 4, 2006
Page 1 of 9
 Gate Ox Band-band
Leakage Tunneling
Metal
G
Ox
Metal Metal
XjSDE SDE
Halo S D
EOR damage ************************ ******* *******
Depletion layer

Well

Substrate

Carrier Thermionic
Recombination Emission
Figure 1. Sketch of leakage mechanisms in p-n junctions and transistors.

Carrier recombination/generation
Carrier recombination leakage, observed under forward bias, is determined primarily by
the sub-junction doping density (which determines the width of the depletion layer) and
the density and location of the residual damage after annealing from the accumulated
damage due to the various implantation cycles. Damage accumulation during
implantation depends on the ion type, target (usually Si), ion energy and dose as well as
process specifics such as the ion beam current and wafer temperature during
implantation. Under reverse bias conditions, defect centers are sources of carrier
generation, with similar amplitude to the recombination rate. The generation current is
described by:

qW qε S ε 0 ( Na + Nd )(Vbi − Vrb )
J gen = σ ⋅ vth ni N = σ ⋅ vth ni N t (2)
2 2N A N D

where Nt, σ are the concentration of recombination centers and their capture cross section
in a depletion region of p-n junction, W is depletion width, vth is thermal velocity of
carriers, q is the charge of the electron, ε0 is permittivity of vacuum, εS is dielectric
constant of silicon, Na, Nd are doping in top and bottom layers of p-n junction, ni is
intrinsic concentration of carriers in silicon, Vbi is build in voltage of p-n junction, , Vrb is
applied reverse bias [2].

Comparison of leakage currents in RsL measurements and transistors

January 4, 2006
Page 2 of 9
The key features of generation/recombination current is that it is (1) dependent on the
number of generation/recombination centers, usually identified with lattice imperfections
such as end-of-range (EOR) dislocation layers and various forms of metal and oxygen
precipitates and (2) dependent on the sub-junction doping, where the depletion layer
decreases in thickness with increasing sub-junction doping. The most effective
contributions to generation and recombination currents occur when the defect centers are
located in the middle of the depletion layer, where the populations of carrier types
available for recombination are close to equal. The principle process-sensitive
parameters are (1) the location and density of residual implant damage, (2) surface
junction depth and profile shape and (3) active sub-junction doping density.

The rate of carrier recombination, U(z), in the depleted layer is given by the Shockley-
Read-Hall equation:

⎡ ⎛ qV ⎞ ⎤
σ nσ p vth N t ( z )ni ⎢exp⎜ ⎟ − 1⎥
⎣ ⎝ ⎠
kT⎦ (3)
U ( z) =
⎡ ⎛ ψ − ϕn ⎞ ⎛ E − E i ⎞⎤ ⎡ ⎛ ϕ − ψ ⎞ ⎛ E − Et ⎞⎤
σ n ⎢exp⎜ q ⎟ + exp⎜ t ⎟⎥ + σ p ⎢exp⎜⎜ q ⎟ + exp⎜ q i ⎟⎥
p
⎟
⎣ ⎝ kT ⎠ ⎝ kT ⎠⎦ ⎣⎢ ⎝ kT ⎠ ⎝ kT ⎠⎦⎥

where and V is the JPV, ψ=-Ei/q is the internal potential corresponding to the intrinsic
level, Ei, ϕn , ϕp are the potentials corresponding to quasi-Fermi levels for electrons and
holes, vth, σn ,σp are thermal velocity and the capture cross sections of electrons and
holes, Et, Nt is the energy level and concentration of recombination centers, ni is intrinsic
carrier concentration.

The leakage current is the integral of the recombination rate over the thickness of the
depletion layer:

W
J rec = ∫0
qU ( z ) dz (4)

The recombination profile and leakage current, Io, can be calculated for low photo-
voltage V=ϕn - ϕp << kT/q for an abrupt p-n junction, for the example with a junction
depth, Xj = 100 A, with the residual damage profile described by:

⎡ ( z − z max ) 2 ⎤
N t ( z ) = N t max exp ⎢− ⎥ + N tsub (5)
⎣ ∆2 ⎦

where the peak concentration of recombination sites Ntmax= 1018 traps/cm3 at a depth of
zmax= 200 A, with a trap profile width of ∆ = 150 A, σn ,σp = 10-14 cm2 , Et=Ei, and Ntsub=
1012 traps/cm3 is the concentration of recombination centers in the substrate, much deeper
Comparison of leakage currents in RsL measurements and transistors
January 4, 2006
Page 3 of 9
than zmax. These parameters are descriptive of conditions for a shallow junction formed
in an implanted halo profile. The recombination rate profiles, U(z), for substrate doping
of 1015, 1017 and 1019 dopants/cm3 are shown in Fig. 2. And the integral leakage currents
for three damage density values are shown in Fig. 3 [3].

Figure 2. Recombination rate profiles below a shallow junction (Xj = 100 A) for a model
end-of-range damage profile (dotted line) and substrate doping levels of 1015, 1017 and
1019 dopants/cm3.

Figure 3. Leakage currents as a function of substrate doping for three trap densities.

Comparison of leakage currents in RsL measurements and transistors

January 4, 2006
Page 4 of 9
Note that, according to the integral of Eq. 4, shown in Fig. 3, for lightly-doped substrates,
with n ~ 1015 dopants/cm3, the recombination leakage current is low and does not depend
strongly on the density of recombination centers. For doping levels characteristic of
shallow junctions in halo/well profiles, with n ~ 5x1018 dopants/cm3, the leakage
current levels are ~ 106 times higher and increase in proportion to the density of trap
centers. Since the trap centers are predominately provided by residual implant damage
centers, the leakage current for a SDE junction in a halo/well profile is strongly
dependent on the success of the implant/anneal process sequence to provide strongly
activated, shallow junctions and low defect density regions around the junction.

The increase in junction leakage with increased sub-junction doping and the presence of
residual damage has been clearly illustrated for the case of “diffusionless” anneals at 650
C after Ge pre-amorphization implants and shallow B doping [4].

Carrier diffusion
The carrier diffusion contribution to leakage is given by [1,2]:

D p ni2
J diff = q (6)
τ p NA

where τ is the minority carrier lifetime, which follows an Auger recombination

mechanism for heavily-doped junctions and Dp is the carrier diffusion coefficient.

The carrier lifetime follows:

τ p = 1 C p N A2 (7)

where Cp is the Auger recombination coefficient, so the diffusion contribution to leakage

is small for the case of highly-doped junctions and

D p ni2
J diff =q ~ q *ni2* (Dp*Cp)1/2. (8)
τ p NA

For biased transistors with metal junction contacts, the additional leakage mechanisms
include structure-dependent effects of band-to-band tunneling at the drain junction,
thermionic emission from the metal contact and gate-to-channel leakage through the gate
dielectric.

Comparison of leakage currents in RsL measurements and transistors

January 4, 2006
Page 5 of 9
Band-to-band tunneling
The band-to-band tunneling current, Jtun, is approximated by:

πq 2 mrVrb E g E m
J tun = exp(− E 0 / E m ) (9)
h 3 E0

2ε S ε 0 (Vbi − Vrb ) ⎛ N A + N D ⎞
W (Vrb ) = ⎜⎜ ⎟⎟ (10)
q ⎝ NAND ⎠

Vbi − Vrb
Em ≈ (11)
W

where Em is maximum electric field calculated using W, h is a Plank constant, mr is the

effective mass value for the tunneling electron in an indirect band gap material, Eo is a
band gap and Vrb is applied reverse bias [5]. This form of leakage current is centered on
the high-field region of the channel/drain junction and represents a fundamental limit for
semiconductor materials, scaled by the band gap energy, Eo.

Thermionic emission
Thermionic emission of carriers from metal contacts through shallow junctions depends
on the junction doping level, depth and profile shape, with lower carrier flows for
junctions with abrupt, or box-like, profiles [6]. For practical transistors, the effects of
thermionc emission are minimized by the deeper junctions used for the source/drain
contacts, however this can be a significant issue for measurements with direct contact
probes onto shallow extension junctions.

Gate leakage and other effects

Gate-to-channel leakage is, at present, a dominant leakage issue for transistors and the
driver for replacement of SiO2 with high-k dielectrics with EOT values of ~1 nm and
less. The thermal stability of these new dielectric materials effects transistor doping by
limiting the thermal cycles that can be used for damage annealing and dopant activation.

Additional leakage effects includes surface recombination currents, which are small for
heavily-doped, well-activated surface junctions (but could become more important for
poorly activated junctions, such as those resulting from unsuccessful ms-timescale
anneals), as well as a variety sub-threshold leakage current effects in short-channel
CMOS transistors.

Comparison of leakage currents in RsL measurements and transistors

January 4, 2006
Page 6 of 9
3. Leakage currents in RsL measurements
RsL methods measure the junction leakage through analysis of the dependence of the
JPV signal on the modulation frequency of the incident LED light beam [3]. The
extracted parameter is the amplitude of the diode equation, Io, where the diode current is
given by the Shockley equation:

Jdiode = Io(eqV/kT -1) (A/cm2). (12)

The recombination current amplitude, Io, is measured under forward bias in RsL but is
also equal to the reverse bias leakage of an “ideal” diode dominated by carrier generation.

In the case of RsL, leakage is measured under very small forward bias, V<<kT/q, created
by photo carriers. In this case leakage current includes only contribution of diffusion
current Jdiff and space charge generation current Jgen (operating in this case as a
recombination current):
D p ni2 qε S ε 0 ( Na + Nd )Vbi
J leakRsL = J diff + J gen = q + σ ⋅ v th n i N t . (13)
τ p NA 2N A N D

The effect of trap density, Nt, on leakage current for forward (RsL) and reverse
(transistors) bias is shown in Fig. 4. In this calculation, the surface junction is heavily
doped (so Jdiff is small), the sub-junction doping levels are high (~5x1018 dopants/cm3)
approximating the levels for a halo profile near a SDE junction, and the capture cross-
section is σ ~10-16 cm2. For the reverse bias (assumed to be -1V) case, the main
additional leakage contribution is assumed to be band-to-band tunneling at a level of Jtun
~5x10-4 A/cm2 [5], which sets the leakage current limit for the case of low trap densities.

Comparison of leakage currents in RsL measurements and transistors

January 4, 2006
Page 7 of 9
 0.1

Leakage Current Density, A/cm2

0.01

1 .10
3 Reverse bias
4
1 .10

5
1 .10

6
1 .10

1 .10
7 Forward bias
8
1 .10
14 15 16 17 18 19 20
1 .10 1 .10 1 .10 1 .10 1 .10 1 .10 1 .10
.
Trap concentration per cm3
Fig. 4. Leakage current versus trap concentration for reverse bias (-1 V, red curve) and
forward bias (+50 mV, blue). The reverse bias curve corresponds to leakage current
measurements with diodes and transistors and the weak forward bias case occurs in RsL
measurements.

It is important to note that the strong effect of trap (residual damage) density on leakage
current shown in Fig. 4 is observed for heavily-doped sub-junction regions similar to the
case of SDE junctions formed in halo/well profiles, where the depletion layer thickness,
W, is small (~20 nm). For SDE junctions formed in lightly-doped wafers (~ 10 Ohm-
cm), the depletion layer is large (~1 um) so the main carrier recombination and
generation occurs deep (~ 0.5 um) below the SDE junction and far from the EOR damage
layer [3]. In the lightly-doped substrate case, the recombination leakage current will be
much lower and show little or no dependence on trap density.

4. Comparison of leakage current measurements for RsL and transistors/diodes

The direct linkage of leakage current measurements between RsL probes of p-n junctions
and 3-D structures, such diodes and transistors, is complicated by the number of
additional leakage mechanisms in the 3-D structures. The key factor is that RsL
measures the components of device leakage which are dependent on process variables,
implant and anneal sequences. These are the components of the device leakage which are
accessible to optimization without wholesale changes to the transistor structure and
materials. Since transistor structures contain additional leakage mechanisms, the
recombination leakage measured by RsL is the minimum leakage current levels that
characterize a given junction-substrate doping and damage condition.

Comparison of leakage currents in RsL measurements and transistors

January 4, 2006
Page 8 of 9
The effect of the additional leakage current factors, beyond the carrier generation and
recombination effects, that are active in diodes and transistors is to limit the dynamic
range of leakage current dependence on carrier trap distribution density and location. For
the case shown in Fig. 4, where the minimum transistor leakage is limited by a tunneling
current of ~5x10-4 A/cm2 [6], the dynamic range of sensitivity to defect density is <102.
At the high end of the leakage current measurement, the leakage current is limited by
factors related to limited thermal processing of shallow junctions; poor dopant activation,
thin active and depletion layers and incomplete removal of implant damage.
The RsL method, which is not limited by the materials and structural leakage
mechanisms in transistors, has a measurement dynamic range of ~105; from ~10-7 A/cm2
for low defect and low sub-junction doping levels to ~5x10-2 A/cm2 for high defect levels
in heavily-doped profiles. The wide dynamic range of RsL methods (~104 for Rs, ~105
for leakage) allows for testing across the full range of CMOS doping processes with a
single, well-known parameter for Rs and provides a means of assessing the effects of
process variables on defect-driven leakage over a range well beyond the capability of
diode testing and with significantly shorter turn around time.
The types of process and recipe-related variables that the RsL can monitor include the
effects of dopant dose and depth, damage accumulation, dopant activation and diffusion
and local effectiveness of damage reduction during annealing. Damage accumulation
during implant depends on a combination of recipe-driven effects (ion type, energy and
dose) and process related variables (ion beam current, scan rates and wafer temperature
during implant). The recipe related effects can be explored during process development
and the process-drive effects are checked for during routine process control sampling.

Doping processing with ms-scale annealing

With the use of ms-scale anneal cycles to reduce dopant diffusion to the low nm level,
proportional limitations on Si diffusion create a much higher sensitivity to initial damage
distributions following implant and to variations in the complex, multi-component
thermal cycles for laser and flash-type annealers. The advantage of the RsL tool is to
provide leakage current data on the same time scale and with the same monitor
processing costs as sheet resistance monitoring for implant dose, energy and anneal
activation. The rapid turn around time for testing, with high-resolution mapping data for
Rs and Io available within minutes of the completion of the anneal cycle, is compatible
with in-process control activities for critical doping processes.

References:
[1] S.M. Sze, Physics of Semiconductor Devices.
[2] D. K. Schroder, Semiconductor Material and Device Characterization.
[3] V. N. Faifer et al., Proc. of USJ05, to appear in JVST-B, Jan06.
[4] R. Lindsay et al., JVST B22(1), (2004) 306.
[5] P. M. Solomon, et al., IEDM03.
[6] E. C. Jones, N. W. Cheung, JVST B14(1), (1996) 236.

Comparison of leakage currents in RsL measurements and transistors

January 4, 2006
Page 9 of 9