2, MAY 2007
137
AbstractThis paper presents the contents of a lecture for enhancing the interest of electrical engineering students in semiconductor device modeling and providing a deep appreciation of the
conflicting requirements for an ideal model, of the sequence and interconnectivity of steps in the modeling procedure, and of the types
of models existing in modern times. The lecture can be included in
courses on semiconductor devices or device modeling. The lecture
uses a resistor with 3-D effects and much simpler physics than a
p-n junction for illustration, and coins mnemonics for key aspects.
Objective assessment data confirm the need and efficacy of the proposed lecture. The work of this paper shows how proper introduction of a topic with carefully chosen illustrations and mnemonics
can improve the teaching/learning process and how such improvements can be assessed.
Index TermsAnalytical modeling, compact modeling, device
modeling, numerical modeling, semiconductor devices, spreading
resistance.
I. INTRODUCTION
HE topic of modeling has acquired great importance in
engineering education because of astronomical increase
in computing power. As a result, a number of universities
across the globe are including a full-fledged modeling course
in their undergraduate or graduate curricula. For example, in
the electrical engineering curriculum of the authors institution,
following a mandatory semester-long course on semiconductor
device basics, an advanced one-semester course on semiconductor device modeling is taught both at senior undergraduate
and graduate levels to the students specializing in microelectronics and Very Large Scale Integration (VLSI) design. The
basic course follows the topics discussed in books such as
[1][5]. The topics in the advanced course are selected from
books, such as [6][11] and research papers; the topics covered
in such a course in the authors institution are provided in the
Appendix. A knowledge of the modeling procedure is required
at the beginning of the advanced course, and is also useful in
the second half of the basic course while deriving the device
characteristics from fundamentals covered in the first half.
Literature survey [1][11] shows that a good introduction to
the device- modeling procedure is still lacking at present. According to educational psychology, the method of introduction
of a topic significantly influences student learning [12]. Thus,
pedagogy attaches great importance to an effective and efficient
introduction to any topic. Here, effectiveness is related to stu-
138
C. Mathematical Equations
The previously mentioned concepts/principles 14 translate
to the following equations, respectively.
Flux creation
(1)
(2)
Flux conservation
(3)
(4)
Here, is the current density per unit area; is the electron
charge; is the electron concentration; is the electric field;
is the electric potential;
is the current across the resistor
cross-section at any ; and is the space-charge. Four physical
, and , associated with the four equations.
variables are
The popular names for the various equations have been indicated on the left of each equation. Equation (1) describes the
electron current created by the field, while (3) shows that the current exiting region equals the current entering since no charge
build-up is present. Equation (2) depicts the creation of electric
field from potential gradient, while (4) relates the change in field
over a distance to the space-charge within that distance.
D. Approximations
The following approximations are made about the behavior
, and , to simplify the solution of the
of the variables
equations.
0, resulting in = , eliminates
1) No space-charge, i.e.,
(4) from the set of equations.
2) The current is effectively restricted to an area of cross, which increases at a constant angle from the
section
smaller electrode area to the larger electrode area (Fig. 1).
are negligible com3) The lateral components of and
pared to their vertical or y-components which are uniform
but may vary with .
over the cross-sectional area
E. Solution
Using approximations 1) and the transport equation (1), at
any point,
, which, based on approximations 2) and
, at any cross-section. As
3) leads to
is also the terminal current,
per the continuity equation (3),
, so that
. As per the electric field equation
(2),
, where
Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY MADRAS. Downloaded on January 27, 2010 at 10:41 from IEEE Xplore. Restrictions apply.
139
(5)
F. Verification
The assumptions and the transport and continuity equavaries
tions (1) and (3) lead to the conclusion that
with
as per
. Thus,
, which is inconsistent with
the assumption of space-charge neutrality, as per the Gauss law
(4). To see the extent of inconsistency, compare the previously
, which is the left-hand side (LHS) of (4),
mentioned
on the right-hand side (RHS) of (4). If
with the term
, then
0 is justified, else it is not.
) [
, and
Here,
,
, and is given by
whose maximum occurs at
. This maximum value is zero only for
, increases as is increased, but depends on the aspect
rather than individual values of and .
ratio
m and
Consider a silicon block having
m,
cm
cm /V-s, and
F/cm. Such a resistor geometry forms a part of many modern
represents the concentration of dopants
devices. Note that,
which contribute electrons, and not silicon atoms, and the eftimes that of elecfect of holes whose concentration is
trons has been neglected. The formula (5) gives a resistance
k . Suppose a current
mA is passed through
V. For these condithe block causing a voltage drop
tions, maximum
which is not
.
is inconsistent with the asHence, the assumption of
sumption of uniform and over , meaning that the simple
model (5) is not accurate for the geometry considered.
The extent of inaccuracy in the model is seen by comparing
model results with accurate computer calculations. The 3-D device simulator named ATLAS supplied by Silvaco Inc. [15] was
used to calculate the resistance of the geometry shown in Fig. 1
(see [16], [17] for other
for various aspect ratios / and
simulators available). The results are plotted in Fig. 2(a) in a
compact form using normalized quantities, and compared with
, has been norresults of model (5). The resistance,
malized with respect to
, which corresponds to uniform area of cross-section. Two important qualitative trends of the numerical calculations, namelythe
aspect ratio dispersion, and the saturation at a nonzero value for
large , are not captured by the model (5), for which the normalized resistance is independent of and decreases monotonically to zero with increase in . This discrepancy is traced to
the fact that in reality the current spreads beyond the flow area
assumed, and the actual field distribution is nonuniform and
3-D (Fig. 3).
G. Modification
Fig. 3. (a) Actual electric flux lines along a vertical cross-section of the
spreading resistor geometry of Fig. 1. The dotted lines enclose the effective
current flow area assumed in the model. (b) The y-component of the electric
field along the cut-section A-A.
increasing
values intersect the results of (5) at increasing
values, and saturate beyond the point of intersection.
To discern some constancy in the pattern of intersection, one
at the point of intersection
should consider the ratio of
to the corresponding
, i.e., the ratio of two numbers
which increase with each other, for each of the curves. This
2.2 or
ratio is found to have the value
2.2
at the point of intersection. Hence, the formula (5) is
, and the resistance is assumed to saturate
used for
for
(6a)
(6b)
This adjusted model captures the resistance saturation at a
aspect ratio dispersion
nonzero value for large , and the
in saturation, and thus improves model accuracy significantly.
Along with accuracy and physical basis, continuity is an overriding concern for compact modeling. The derivatives of (6a),
. One way of eliminating
(6b) are discontinuous at
this discrepancy and transforming (6) into a compact model is as
and
follows. Interpolate the variation between
using a function which approaches the results of the segment
(6a) for small and segment (6b) for large , e.g.,
(7a)
140
(7b)
(7c)
An empirical adjustment of (7c) by multiplication with 1.7
(8)
leads to an accurate and continuous model as shown in Fig. 2(b).
IV. DISCUSSION
The contradictions or tradeoffs among the CAPS requirements for an ideal model should be noted. The comparison between (5) and numerical calculations shown in Fig. 2(a) demonstrates the tradeoff existing between simplicity and accuracy.
,a
Similarly, the discontinuity in derivatives of (6) at
discrepancy that was absent in the simple model (5), shows the
tradeoff between accuracy and continuity, and the direct relation
between simplicity and continuity. Finally, the lack of physical
basis and the purely empirical nature of the constant 2.2 in (6a),
(6b), and (7b) and the constant 1.7 in (8), introduced to achieve
accuracy, demonstrate the tradeoff existing between accuracy
and physical basis.
The choice of the variables on coordinate axes in Fig. 2 and
the related description in the last paragraph of section (g) brings
out the use of scaling or normalization for compact presentation
of a large amount of data, an important modeling skill.
The organization of the basic physical tenets and their mathematical translations into flux creation and flux conservation categories in Section III (B, C) enhances the connectivity among
diverse concepts and the physical insight, both of which should
improve student retention and facilitate generalization of the
procedure to all the devices by a process of inductive thinking.
For instance, in the context of semiconductor devices, where
holes also contribute to the current apart from electrons, an additional equation in the flux creation and conservation categories
each, for holes, raises the number of basic equations to six. Further, additional terms in charge flux creation equations exist to
represent transport mechanisms different from drift, such as diffusion and thermoelectric current. Another noteworthy aspect is
the verification of the consistency of the assumptions made in
the derivation of the closed-form spreading resistance formula.
This aspect has not been discussed in any of the books or research papers dealing with spreading resistance. The compact
model of the spreading resistance presented in this paper is also
new.
V. LECTURE EVALUATION
After introducing the lecture discussed in this paper in the
device modeling course, the author sensed higher levels of student interest, comfort, and class activity in terms of discussions
and questions during the entire course. When students (70 in
number) were asked to respond to the assignment, Write in
about 250 words what you learned in this course that was not
known to you earlier, as a part of the end semester examination,
Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY MADRAS. Downloaded on January 27, 2010 at 10:41 from IEEE Xplore. Restrictions apply.
141
Fig. 5. Student responses to Questionnaire-1. The numbers indicate percentages. See text for the method employed to estimate the mean score given in
the last column. Other columns indicate the percentage of students with the response given at the bottom of the column.
Fig. 4. The questionnaires used to get student feedback for measuring the impact of the introduction lecture. Questionnaire-1 is used in both prelecture and
postlecture surveys. Questionnaire-2 is used in postlecture survey.
in the table (see Fig. 5). Fourteen percent of students who agreed
with statement-2 (in the table) regarding the type of models gave
incorrect answers to the related question-1 in the descriptive section. These students (and another 52%) mistook the models for
various devices such as diodes, MOSFETs, and bipolar junction transistors (BJTs) for the types of models. The responses
of these 14% of students have been corrected from agree to
disagree in the score for statement-2 reported in Fig. 5. In the
postlecture survey, all the students filled up all the descriptive
responses, and all these responses were correct except for the
responses of 14% of students to the question on the types of
models. An overwhelming majority of the correct descriptive responses to the questions 1, 2, and 3 were written in the order of
the letters in the mnemonicsCAN, CAPS, and IQM SVM,
respectively, clearly establishing the positive role of mnemonics
in aiding retention and understanding. On the other hand, the
anomalies in the prelecture and postlecture responses related to
the types of models point to the need for extra care in clearing
the misconceptions about this topic.
The responses to the unguided Questionnaire-2 in the
postlecture survey should be considered. Students completed
this questionnaire before responding to the Questionnaire-1,
i.e., were not guided about the issues to which they should
respond. Therefore, the responses on Questionnaire-2 are valuable indicators of the impact of the lecture. Without exception,
the students expressed positive feelings about the material
covered in the lecture; 10% of students stated categorically that
the lecture inspired them to consider specializing in the area
of device modeling. Regarding the important points learned
in the lecture, 71% of students mentioned the properties of a
good model, 71% the types of models, and 76% the modeling
procedure; 57% of students mentioned all these three features.
These figures broadly agree with the percentages obtained in
the independent survey based on Questionnaire-1 (Fig. 5), confirming the fidelity of the responses and reliability of the study.
Seventy six per cent of students wrote the mnemonics CAPS,
SVM in their answers proving the worth
CAN and IQM
and appropriateness of the mnemonics coined in the lecture.
While 33% of students explicitly stated that the systematic
SVM modeling procedure
step-by-step nature of the IQM
was an extremely interesting aspect of the lecture, 29% mentioned the conflicting nature of the CAPS requirements as
an important point. In the words of the students, the lecture
brought out organized thinking, good intuition, ability
to make consistent approximations, deep understanding of
physical concepts, brains, and creativity as the qualities
required of an aspiring model developer. A notable omission
in the responses was that of normalization/scaling as an
Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY MADRAS. Downloaded on January 27, 2010 at 10:41 from IEEE Xplore. Restrictions apply.
142
APPENDIX A
SYLLABUS OF A COURSE ON Device Modeling
(Duration38 hours, prerequisite Basic course on Solidstate Devices) Introduction to device modeling.
Model of a modulation doped (grossly asymmetric) heterojunction under equilibriumpartial depletion, inversion, partial
ionization, and quantum-mechanical confinement effects in the
space-charge region.
MOSFET modeling: Level-1 SPICE modeldrain current;
junction leakage; junction, overlap, and intrinsic capacitances;
Meyer model; small-signal equivalent circuit; model parameter
input file; limitations of Level-1 SPICE model, field dependent mobility, small-geometry effects, subthreshold current, and
depletion charge variation along the channel; channel-length
modulation; source and drain resistances; nonuniform channel
doping; Level-3 SPICE model; and parameter extraction.
BJT Modeling: EbersMoll, charge control and
GummelPoon models, Early effect; quasi-saturation, base
widening, and Kirk effect; SPICE model; series resistances
and capacitances; transit time and cutoff frequency; beta and
cutoff frequency variation with collector current; small-signal
equivalent circuit; 2-D considerations, current crowding, lateral
base widening, sidewall injection; scaling consideration and
limitations; parameter extraction.
ACKNOWLEDGMENT
The author would like to thank the reviewers for adding value
to this paper with their critical comments.
REFERENCES
[1] M. S. Tyagi, Introduction to Semicondcutor Materials and Devices.
New York: Wiley, 1991.
[2] B. G. Streetman, Solid-State Electronic Devices. New Delhi, India:
Prentice-Hall of India, 1991.
[3] R. Muller and T. I. Kamins, Device Electronics for Integrated Circuits. New York: Wiley, 1977.
[4] E. S. Yang, Microelectronic Devices. New York: McGraw-Hill, 1988.
[5] M. Zambuto, Semiconductor Devices. New York: McGraw-Hill,
1989.
[6] T. A. Fjeldly, T. Ytterdal, and M. Shur, Introduction to Device Modeling and Circuit Simulation. New York: Wiley, 1998.
[7] Y. Tsividis, Operation and Modeling of the MOS Transistor. New
York: McGraw-Hill, 1999.
[8] T. Ytterdal, Y. Cheng, and T. A. Fjeldly, Device Modeling for Analog
and RF CMOS Circuit Design. West Sussex, U.K.: Wiley, 2003.
[9] N. D. Arora, MOSFET Models for VLSI Circuit Simulation. New
York: Springer-Verlag, 1993.
[10] D. J. Roulston, Bipolar Semiconductor Devices. New York: McGraw-Hill, 1990.
[11] C. Snowden, Introduction to Semiconductor Device Modeling. Singapore: World Scientific, 1986.
[12] K. L. Kumar, Educational Technology. New Delhi, India: New Age
International, 1996.
[13] S. Karmalkar, Simple unified elucidations of some semiconductor device phenomena, IEEE Trans. Educ., vol. 42, no. 4, pp. 323327, Nov.
1999.
[14] S. Karmalkar, Appealing analogies for aiding students assimilation of
some key physical concepts related to semiconductor devices, IEEE
Trans. Educ., vol. 42, no. 4, pp. 328330, Nov. 1999.
[15] ATLAS3D Device Simulation Software, SILVACO International. Santa Clara, CA.
[16] TAURUS3D Device Simulation Software, SYNOPSIS, Inc. Mountain View, CA.
[17] DESSIS Device Simulation Software, SYNOPSIS, Inc. Mountain
View, CA.
Shreepad Karmalkar received the B.Tech. and accelerated Ph.D. degrees in
electrical engineering from the Indian Institute of Technology, Madras (IITM),
India, in 1983 and 1989, respectively. His Ph.D. work involved analytical modeling of field-effect transistors for circuit simulation.
Currently, he is a Professor of electrical engineering with IITM, which he
joined as a Scientific Officer in 1983. From November 1991 to November
1993, he was on leave from the institute to work in industry on semiconductor
product development. From January 2000 to April 2001, he held visiting
research/teaching assignments at the University of California, Santa Barbara,
and Rensselaer Polytechnic Institute, Troy, NY. Since then, he has made
summer visits to Rensselaer Polytechnic Institute, Troy, NY, and the University
of South Carolina, Columbia. His areas of interest include modeling and
technology of semiconductor devices. In recent years, his research has focused
on heterojunction device modeling, electrochemical metallization, and microelectromechanical systems (MEMS). He has authored or coauthored numerous
technical papers and reports in reputed journals, conferences and books, and
has been granted several patents. He has served as a Consultant to industries
and as a Reviewer for several reputed journals. He has authored a one-semester
video course on physics and modeling of solid-state devices.
Dr. Karmalkar is the recipient of the 2005 Hari Om Ashram Prerit Vikram
Sarabhai Research Award in the field of electronics, informatics, telematics, and
automation for his work on device modeling and process development. He is
currently a member of the national working group on nanotechnology set up by
the Department of Information Technology, Government of India.
Authorized licensed use limited to: INDIAN INSTITUTE OF TECHNOLOGY MADRAS. Downloaded on January 27, 2010 at 10:41 from IEEE Xplore. Restrictions apply.