Download NEGATIVE BIAS TEMPERATURE INSTABILITY

Survey
yes no Was this document useful for you?
   Thank you for your participation!

* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project

Document related concepts

Negative mass wikipedia , lookup

Anti-gravity wikipedia , lookup

Transcript
A Project Report on
NEGATIVE BIAS TEMPERATURE INSTABILITY
EE311 Term Paper
Prashant Khokhar
Shiv Prakash
Shouvik Ganguly
Rakesh Meena
Kundan Kanwaria
Asutosh Tiwari
Y9427
Y9551
Y9558
Y9474
Y9300
Y9152
1.Introduction
Negative Bias Temperature Instability (NBTI) is a key reliability issue in MOSFETs. It is of
immediate concern in p-channel MOs devices, since they almost always operate with negative gateto-source voltage; however, the very same mechanism affects also n-MOS transistors when biased in
the accumulation regime, i.e. with a negative bias applied to the gate too. NBTI manifests as an
increase in the threshold voltage, a degradation of the mobility, drain current and trans-conductance.
This instability in MOSFETs has been known since 1996. It is become a reliability issue in silicon
integrated circuits, because gate electric field have increased as a result of scaling, increased chip
operating temperature, surface p-channel MOSFETs have replaced buried channel devices, and
nitrogen is routinely added to thermally grown SiO2.
A brief history on NBTI shows that reliability issue in MOSFETs is a serious issue:
Experiments in late 1960s by Deal and Grove at Fairchild
Role of Si-H bonds and BTI vs. NBTI story (J. Electrochem Soc. 1973;114:266)
Came out naturally as PMOS was dominant
Important in FAMOS and p-MOS EEPROMS (Solid State Circuits 1971;6:301)
Theory in late 1970s by Jeppson (JAP, 1977; 48:2004)
Generalized Reaction-Diffusion Model
Discusses the role of relaxation, bulk traps
Comprehensive study of available experiments Early 1980s
Issue disappears with NMOS technology and buried channel PMOS
Early 1980s
Issue disappears with NMOS technology and buried channel PMOS
Late 1980s and Early 1990s
Begins to become an issue with dual poly gate, but HCI dominates device reliability
Late 1990s/Early 2000 (Kimizuka, IRPS97; 282. Yamamoto, TED99; 46:921. Mitani, IEDM02;
509)
Voltage scaling reduces HCI and TDDB, but increasing field & temperature reintroduce NBTI
concerns for both analogue and digital circuits.
Numerical solution is extensively used for theoretical modeling of NBTI
What is NBTI?
NBTI is an increase in the absolute threshold voltage, a degradation of the mobility, drain current, and
transconductance of p-channel MOSFETs. These typesof effects are generally found in pMOSFETs.
When a PMOS transistor is biased in inversion, the dissociation of Si-H bonds along the silicon-oxide
interface causes the generation of interface traps. The rate of generation of these traps is accelerated
by temperature, and the time of applied stress. These traps cause an increase in the threshold voltage
(Vth) of the PMOS transistors. An increase in Vth causes the circuit delay to degrade, and when this
degradation exceeds a certain magnitude, the circuit may fail to meet its timing specifications. This
effect is known as the Negative Bias Temperature Instability (NBTI). Mechanism of NBTI is the
degradation of Si-H bonds broken by the chemical reaction with high energy holes on SiO2/Si surface.
In PMOS voltage between gate and the source is negative (Vgs = -Vdd), causes the interface traps and
when (Vgs = 0) causes the reduction in the interface traps. Thus the effect of NBTI on the PMOS
depends on the time the device has been stressed, and relaxed.
1|Page
It is commonly accepted that two kinds of trap contribute to NBTI:
When Vgs is negative, interface traps are generated. Those traps cannot be recovered over a
reasonable time of operation i.e. Vgs= 0. It is believed that the electric field is able to break Si-H bonds
located at the Silicon-oxide interface. H is released in the substrate where it migrates. The remaining
dangling bonds (Si-Pb center) contribute to the threshold voltage degradation.
On top of the interface states generation some pre-existing traps located in the bulk of the dielectric
(and supposedly nitrogen related), are filled with holes coming from the channel of p-MOS. Those
traps can be emptied when the stress voltage is removed. This Vth degradation can be recovered over
time.
NBTI Models
Many models have been discussed on the NBTI. Here we discuss some model on which NBTI affects,
which include Reaction Diffusion model, Models of Source Drain bias.
Model of Source Drain Bias
In this experimental model following devices were used: surface channel P+poly-silicon gate pMOSFETs fabricated with silicon oxynitride (SiON) gate dielectric using a 90 nm nodeCMOS
technology, with channel width of 10µm, andvarious channel lengths. The gate dielectrics
withequivalent oxide thickness (EOT) = 1.9 nm were grownby thermal oxidation followed by plasma
nitridation and post-deposition thermal annealing.
In this experiment, negative voltage was applied to the gate and drain, source and substrate grounded.
The stress circuit diagram is given here:
Fig 1.1: Stress Circuit Diagram
The experimental result of source drain bias shows the effect of Vds on ∆Vth (Vth shift). The graph
given below shows the effect of source drain bias in threshold shift of pMOSFET for different
channel lengths.
2|Page
Fig 1.2: Vth vs Vds for different channel lengths
From the graph, we observe that for L = 0.1m, when |Vds| changes from 0 to 0.5V then ∆Vth does not
change much but further increasing |Vds| increases NBTI degradation. For L = 0.3m the NBTI
degradation is initially reduced but increases after |Vds| = 1.5V. We also observe that minimum
degradation occurs around |Vds| =1.5 V. This degradation leads to the generation of non-uniform
neutral hydrogen. Generation of more interface traps leads to the diffusion of hydrogen from the
source to the drain. The following figure shows the diffusion of hydrogen.
Fig 1.3: Hydrogen Diffusion
When there is S/D bias the interface traps density (Nit) at the drain canbe reduced and the Vth shift will
be less. In above figure it is also observed that devices are more stressed in the linear region since |Vgs|
> |Vth| and |Vds| < |Vgs|. The drain current, which is linearly dependent on the Vds, can be expressed as
Id =
C(Vgs –
Vth)Vds
where w is the channel length in cm, is the carrier mobility in cm2/(V-sec), and C is oxide
capacitance inF/cm2.
The drain current can also be expressed in terms of carrier velocity as
Id =qpvA
where p is carrier density in cm-3, vis carrier velocity incm/sec and A is the cross-section area in cm2.
A = W.H
where H is the inversion layer thickness in cm.
From the above equations we can find the carrier velocity as
3|Page
v= C (Vgs − Vth ) Vds
qH
L
In the linear region, the carrier density (p) can beexpressed as the oxide charge (Qin
Coulomb/cm2)divided by the inversion layer thickness (H) and the unitcharge (q):
p=
Q = C(Vgs – Vth)
Then
v = .
Vds
L
The average energy of hole is
E = mpv2 = mp µ 2
Vds
L2
2
So, the larger is Vds, larger will be the hole velocity and higher the hole energy. Generally, the energy
required to break the Si-H bond is around 0.3eV. So, energy higher than this enhances the NBTI and
this enhanced NBTI leads to more Si-H bond breakdowns. Fig. 1.4 shows the variation of average
hole energy with the S/D bias i.e. Vds.
Fig 1.4: Average Energy vs S/D bias
The main contribution to the Vth shift is the energetic hole which induces higher NBTI degradation
with higher S/D bias.
2. Reaction-Diffusion model
This analytical model captures the effect of stress and relaxation. These types of R-D model limited to
a single stress cycle and a single relaxation cycle.This analytical model includes stress and relaxation
phase as well as frequency independent property of NBTI.In this model interface traps are occurred at
the SiO2/Si interface and this reaction release the Hydrogen and this reaction also depend linearly on
4|Page
the stress time. In the next diffusion phase, hydrogen diffuses from the interface into oxide diffusion
with time dependence, tn where n = 0.25 for neutral hydrogen. This reaction occurs for a short time, so
we cannot observe it directly, but we can observe and verify the diffusion phase. The interface
generation equations are given below
∆Nit (t) =
2 At
1 + 1 + 4 ABt 3 / 2
2k F N 0 t
=
1 + 1 + k f N 0 k R t 3 / 2 / 0.5 DH
where kF and kR are the forward and backward reaction rates respectively, N0 is the initial defect
density, and DH, the hydrogen diffusion coefficient. During reaction phase i.e. for a small time
interval,
∆Nit (t) At = kFN0t
For diffusion phase
∆Nit (t) kF N0
A 1/4
t =
(DHt)1/4
B
2k R
The experimental data shows that the time exponent observed is less than 0.25. The generation of
interface traps is universally true, but the positive oxide charges generated during NBTI is not
confirmed yet.
The experimental plot of generation of interface traps verses stress time is given below.
Fig 2.1: ∆Nit vs Stress Time
Solution to the Reaction - Diffusion Model
According to the model, the rate of generation of interface traps initially depends on the rate of
dissociation of the Si - H bonds (controlled by the rate constant, ) and the local self-annealing
process (which is governed by the rate constant, ).
5|Page
Thus the reaction occurring in the R-D model can be quantified as
We now present the analytical solution of the R-D model assuming that alternate periods of stress and
relaxation, each of equal duration , are applied to the gate of a PMOS device(shown in the figure),
whose source and bulk are tied to while the drain is grounded.
Four different cases arise in this model, namely, the stress and relaxation phases, appearing alternately.
This analytical model is extended to the subsequent stress/relaxation phases. The final analytical
expression is given below. After the second recovery phase, we get
The analytical solution of the R-D model for the first two cycles is plotted. The black curve
corresponds to the case where the device is stressed continuously, while the blue curve corresponds to
the previous case, i.e., stress and relaxation applied alternately.
Fig 2.2: Analytical solution to the R-D model for the first two cycles and
Comparison with experimental data
3. Mechanism of interface traps and oxide charges
Silicon is tetrahedrally bonded to four other silicon atoms. When Si is oxidized, some Si atoms are
bonded to oxygen and some are bonded to hydrogen. The bonding configuration on the surface is
shown in the figure given below for (111) and (100) orientations.
6|Page
Fig 3.1: Bonding configuration for Si
An interface trap is an interface trivalent Si atom with one unpaired electron at the SiO2/Si surface.
This interface trap is denoted as
Si3≡Si•
Interface traps are also known as Pb centres. From the above figure we observe that in (111)
orientation, Pb centre orbital is perpendicular to the interface. In (100) orientation, the four
tetrahedralSi–Si directions intersect the interface plane at the sameangle. Two types of interface
centres are found in this orientation,Pb1 and Pb0, by electron spin resonance (ESR).The identification
of Pb0 and Pb1 is not perfectly calculated. The two types of traps in (100) orientation are
producedbecause ofthe strain relief in the (100) silicon. The defects result from the naturally occurring
mismatch-induced stress at the SiO2/Si interface during oxide growth.Pb0centres result when strain
relaxation occurs with adefect residing at (111) microfacets at the Si/SiO2 interface, while Pb1centres
result when strain relaxation occurswith a defect at (100)Si/SiO2 transition regions. These interface
traps are electrically active with an energy distribution throughout the Si band gap. The generation
and recombination processes of these traps causes the leakage current, low-frequency noise, and
reduced mobility, drain current, and trans-conductance. Electron and holes occupy the interface traps
and they become charge and contribute to threshold voltage shift. The potential voltage at the surface
is dependent on the occupied traps. The band diagram in Fig 2.4 illustrates the dependency of surface
voltage on occupied traps:
Fig 3.2: Band Diagrams
From Fig 3.2(a)we observe that interface traps are acceptor-like in upper half band and donor like in
lower half band gap. From Fig 3.2 (a) we also observe that flat-band with electrons occupy states
below the Fermi energy and the states in the lower half of the band gap are neutral (occupied donors
designated by ‘‘0’’). Those between mid-gapand the Fermi energy are negatively charged
(occupiedacceptors designated by ‘‘-‘‘), and those above EF are neutral(unoccupied acceptors). We
observe from Fig 3.2 (b) that the interface traps between mid-gap and the Femi level are now
unoccupied donors, leading to positively charged interface traps (designatedby ‘‘+’’). Hence, interface
traps in p-channel devices ininversion are positively charged, leading to negative thresholdvoltage
shifts. Negative bias stress generates donorstates in the lower half of the band gap.
7|Page
The traps are positively charged when occupied by the holes and neutral when occupied by hydrogen.
Oxide charges located near the oxide/substrate interface lead to higher threshold shift than the charge
near the gate/oxide interface. We have already discussed the change of threshold voltage in the
section ‘S/D bias’. The change in the threshold voltage is determined by changing the channel
holedensity and measuring the change in threshold voltage.Thehole density can be changed by
changing Vth through fabrication or substrate bias. The p-MOSFET threshold voltage can be given as
Vth = φ MS −
Qox Qit ( 2φ F )
Q
−
− 2φ F − S
Cox
Cox
Cox
where φ MS is the work function difference between the gate and substrate, φ F the Fermi potential, Qox
the positive oxide charge density, QS the semiconductor charge density,and Cox the oxide
capacitance/unit area. Qit depends on the surface potential, according to the equation
Qit= qDit∆E = qNit
where∆E is the energy range over which interface traps areactive.
The threshold voltage change is given as:
∆Vth= −
∆Qox + ∆Qit
q ( ∆N ox + ∆N it )
t ox
=−
Cox
K ox ε 0
= -4.6 × 10 −7 ( ∆N ox + ∆N it )t ox
We have already discussed the effect of stress time on the change in threshold voltage and change in
the interface trap density. From the experimental database the change in threshold voltage (∆Vth) and
change in the generation of interface traps density (∆Nit) verses stress time data are shown in the
graphs given below:
8|Page
Fig 3.3: (a) Relative ∆VT vs Stress Time
(b) Relative ∆Nit vs Stress Time
4. Frequency Independence
An important requirement of any physical model for NBTIis that it should be able to explain the
phenomenon of frequencyindependence, observed over a wide range of frequencies. Frequency
independence implies that for any two periodicwaveforms with frequencies and, and the same
duty cycle, thenumber of interface traps generated at any time t, such thatt>> andt>> is the same.
Experiments have consistently shown that forfrequencies less than 10MHz, this is true.Mathematical
proof for square waveforms is as given below.
Fig 4.1: Two different square waveforms with different time periods ( and ) but equal
duty cycles (50%)
Let the number of interface traps after time for the two cases be denoted as ( , ,) and ( , ,) respectively. By “frequency independence”, we imply
9|Page
We can prove this equation mathematically by using expressions for "#
Using the above equations, we have
5. Signal Probability and Activity Factor based Solution to the R-D
Model (SPAF Method)
While the NBTI impact has been analyzed fora square wave, such waveforms are rarely seen in digital
circuits,due to the random distribution of node probabilities. We describea methodology to estimate
the NBTI-induced $% degradation of aPMOS device. This method, called the SPAF (Signal
Probability andActivity Factor) method, is based on signal probability (SP) andactivity factor (AF),
i.e., with respect toa reference clock.
Fig 5.1: (a) Signal probability dependency of trap generation shown for fourwaveforms of equal
frequency but varying duty cycles.
(b) Simulations for25% and 75% duty cycle waveforms showing the SPAF method
SPAF Method
The SPAF method helps to convert a random aperiodic signal to anequivalent deterministic waveform,
based on its statistical information(SP). Since the interface trap density is independent of theAF of a
signal, the crux of the SPAF method is to use the signal probabilities of any given node,
computed using logic simulators or otherwise. The SP information can now be used to compute
anequivalent rectangular periodic waveform with the same SP value. Such a transformation is
10 | P a g e
equivalent to representing a 100Hz square waveform by a 1Hz square waveform, both of which have
the same SP.
Let the total time period of this new waveform be k cycles, such that it is low for m cycles and high
for k - m cycles. Let n indicate the number of such periods, each of duration k. We can determine the
number of interface traps generated. Using the "# notation, we have:
The SPAF method of equivalent waveform construction is verifiedfor two random waveforms of SP =
0.25 and SP = 0.75. The samplesare accumulated over 10000 cycles.
The two curves practicallycoincide, thereby validating the SPAF method of equivalent
waveformconstruction. The SPAF method can reduce a random aperiodicsignal to an equivalent
periodic rectangular waveform, and hence, candecrease the amount of computation, without loss of
accuracy.
6. Impact of NBTI on Circuit Delays
In this section, we present the simulation methodology to computethe delay degradation of ISCAS85
benchmarks, and the simulationresults. We pre-computethe statistical information (SP and AF) of the
various nodesin the digital circuit. The SPAF method is used to convert these probabilisticwaveforms
into deterministic periodic signals. We choose the equivalent waveform tobe of frequency 1Hz in
order to reduce the amount of computation.
We note here that for higher frequencies, thedependency of trap generation on the frequency is rather
weak.
The results are presented in Table above. The percentage increase in delays is plotted in thelast
column and an average degradation of 8.33%is seen for thesecombinational benchmarks.
Expectedly, the rise and fall arrival times show asimilar extent of increase in delay due to NBTI.
11 | P a g e
References
[1] Negative bias temperature instability: What do we understand?
Dieter K. Schroder, Department of Electrical Engineering and Centre for Solid State Electronics
Research, Arizona State University, Tempe, AZ 85287-5706, USA
[2] Negative Bias Temperature Instability Basics/Modelling
Muhammad A. Alam Purdue University West Lafayette, IN
[3] An Analytical Model for Negative Bias Temperature Instability
Sanjay V. Kumar, Chris H. Kim, and Sachin S. Sapatnekar, Department of Electrical and Computer
Engineering, University of Minnesota, Minneapolis, MN 55455
[4] Negative Bias Temperature Instability of Deep Sub-Micronp-MOSFETs
Under Pulsed Bias Stress
B. Zhu, J. S. Suehle, Y. Chen, and J. B. Bemstein
Centre for Reliability Engineering, University of Maryland, College Park, MD 20742
Semiconductor Electronics Division, National Institute of Standards and Technology,
Gaithersburg, MD 208993, Agere Systems, Orlando, FL 32837
[5] Investigation and Modelling of Interface and Bulk Trap Generation during Negative Bias
Temperature Instability of p-MOSFETs
Souvik Mahapatra, P. Bharath Kumar, Student Member, IEEE, and M. A. Alam
[6] Plasma Damage-Enhanced Negative Bias Temperature Instability in Low-Temperature
Polycrystalline Silicon Thin-Film Transistors
Chih-Yang Chen, Student Member, IEEE, Jam-Wem Lee, Wei-Cheng Chen, Hsiao-Yi Lin,
Kuan-Lin Yeh, Po-Hao Lee, Shen-De Wang, and Tan-Fu Lei
[7] Negative Bias Temperature Instability in Low-Temperature Polycrystalline
Silicon Thin-Film Transistors
[8] Models of Source/Drain Bias on Negative Bias Temperature Instability
Z. H. Gan, C. C. Liao, M. Liao, J. P.Wang, W.Wong, B.G. Yan, J. F. Kang, Y. Y. Wong,
Semiconductor Manufacturing International Corporation
[9] Influence of Nitrogen on Negative Bias Temperature Instability in Ultrathin SiON
Yuichiro Mitani, Hideki Satake, and Akira Toriumi, Member, IEEE
[10] Anomalous Negative Bias Temperature Instability Degradation Induced by Source/Drain
Bias in Nanoscale PMOS Devices
Baoguang Yan, Student Member, IEEE, Jingfeng Yang, Zhiliang Xia, Xiaoyan Liu, Gang Du, Ruqi
Han, Jinfeng Kang, C. C. Liao, Zhenghao Gan, Miao Liao, J. P. Wang, and Waisum Wong
[11] A Rigorous Study of Measurement Techniques for Negative Bias Temperature Instability
Tibor Grasser, Senior Member, IEEE, Paul-Jürgen Wagner, Philipp Hehenberger, Wolfgang Goes,
and Ben Kaczer
12 | P a g e