Results from a recent reassessment
of the state of computational materials
science and engineering (CMSE)
education are reported. Surveys were
distributed to the chairs and heads of
materials programs, faculty members
engaged in computational research,
and employers of materials scientists
and engineers, mainly in the United
States. The data was compiled to assess
current course offerings related
to CMSE, the general climate for introducing
computational methods in
MSE curricula, and the requirements
from the employers’ viewpoint. Furthermore,
the available educational
resources and their utilization by the
community are examined. The surveys
show a general support for integrating
computational content into MSE education.
However, they also reflect remaining
issues with implementation, as well
as a gap between the tools being taught
in courses and those that are used by
employers. Overall, the results suggest
the necessity for a comprehensively developed
vision and plans to further the
integration of computational methods
into MSE curricula.
INTRODUCTION
Materials science and engineering
(MSE) encompasses metallurgy,
semiconductors, ceramic engineering,
and polymer science. It is a multidisciplinary
field that enables new technologies
required to address a wide variety
of critical challenges facing society,
such as clean energy production. While
traditionally viewed as an experimental
discipline, many researchers have
begun to take advantage of rapidly
growing computing resources and associated
algorithmic and theoretical
developments, and the capabilities of
integrated computational approaches
are increasingly being utilized to accelerate
materials design and development.
Recent National Research
Council (NRC) reports1,2 indicate that
successful integration of computational
tools has also begun to be demonstrated
in industrial settings, comparing its potential
impact to that of bioinfomatics.
The reports summarized recommendations
that include incorporation of computational
modules into a broad range
of materials science courses in order to
train the next generations of materials
engineers with the abilities required to
exploit these tools. However, the degree
to which such efforts are already
under way, and what steps must still be
taken to address these NRC recommendations
remain unclear. Therefore, we
have undertaken a survey of the field
to assess the current status of computational
materials science and engineering
(CMSE) education. A summary is
presented below, which serves as an
update to a previously published report3
based on similar surveys performed in
2003–2004. See the sidebars for a survey description, the list
of respondents, and CMSE resources.
Survey Procedures |
Following the procedure of the previous study,3 three sets of surveys were performed.
The first was a short survey sent to the chairs and heads of materials programs (hereafter
referred to simply as chairs) that participate in the University Materials Council. The
second survey was more detailed and sent to faculty members engaged in computational
research and education (hereafter referred to as computational faculty), which was initially
sent to the chairs and department heads to be forwarded to their faculty and was
followed up by the authors. A third survey was sent to employers of materials scientists
and engineers.
The surveys were mainly distributed to U.S. universities, national laboratories, and
materials-related industries. The chair survey was comprised of five questions, all of
which focused on undergraduate education. Information was sought to determine the
chairs’ perspective on how important CMSE education is, how this material should be
covered in undergraduate curriculum, and how much CMSE education is supported
by their faculty. Nineteen responses were returned for this survey. The second survey
targeted computational faculty. In addition to questions similar to those in the chair
survey, information describing the current and planned course offerings in CMSE,
as well as publicly available CMSE education resources, was requested. The survey
of computational faculty also included questions pertaining to graduate curriculum
in CMSE and the availability and preparedness of graduate students to perform
computational research. Twenty-three responses were received from computational
faculty. Finally, employers of MSE degree holders were asked to provide input. Four
responses from national laboratories and twelve responses from industry were received.
Respondents from large organizations generally answered for labs or groups they lead.
Some large corporations had two respondents answering for different labs or groups.
The questions addressed the importance of computational expertise in their organization,
what computational tools are utilized, what background (including degrees) those who
utilize them have, and future trends. We also solicited recommendations concerning
what aspects of CMSE education may require improvement and how. Percentages are
calculated out of the number of respondents to each specific question. |
CMSE Education Resources |
Table A lists examples of software packages that are being used in CMSE-related
courses. Most of the reported packages are not tailored for education. Of the software
listed in Table A, nine are commercial packages, while four are free software. Overall,
the number of packages used for instructional purposes is relatively low, compared
to the amount of research and development software and numerical libraries that are
commercially or freely available. A great fraction of these packages (including all of the
open source packages) focus on the atomistic aspects of materials simulation, revealing
that there is little emphasis on simulations at meso/microscales and macroscopic scales,
which may reflect the research expertise of many computational faculty members. The
contrast between the tools utilized in teaching and those employed in industry, many of
which focus on the continuum scales, was discussed previously and is shown in Table
I. Programming languages used in CMSE classes can be divided into two classes: the
low-level, yet more computationally efficient, languages like C, C++, and Fortran, versus
higher-level languages such as Mathematica, MATLAB, and Python.
A number of computational faculty believe there is a shortage of appropriate software
tools that could be integrated into their courses. There have been efforts to make CMSE
education modules or MSE education modules based on computational tools freely
available to interested educators. However, there has not been an assessment of how well
these resources are utilized at large. One of the questions to computational faculty asked
about their knowledge and use of four existing National Science Foundation sponsored
websites with CMSE education modules. Of the 22 who responded to this question, 45%
were aware of the resources at the Materials Technology@TMS ICME Community site
(now a part of the Education Community), and of these, half (23%) used the content
in their instruction. Forty-five percent were aware of the nanoHUB site although none
used the tools from this web site in their courses. Forty-one percent were aware of
the UIUC Materials Computation Center software archive, but only 9% used the tools
available there in their courses. Twenty-seven percent were aware of the MatDL Digital
Library Courseware, but only 5% used it in teaching. Overall, we find that many CMSE
researchers/educators are not aware of the available resources, and those who are aware
typically do not utilize them. Furthermore, the survey shows a lack of incorporation of
the latest developments such as parallel computing and cloud computing (dynamically
scalable internet-based computing) into CMSE education (see Figure 3). |
Undergraduate
Education in CMSE
The status of undergraduate CMSE
curriculum was assessed through five
survey questions directed to department
chairs, as well as corresponding
questions included in the survey targeted at computational faculty. Overall,
42% of the surveyed chairs consider
integration of CMSE into the required
undergraduate curriculum “very important,”
versus 53% that consider it
“somewhat important,” leaving only
5% (one respondent) stating “not important.”
As an elective, however, a
majority (56%) report that it is “very
important” to have CMSE-related undergraduate
courses, as compared to
31% that consider it “somewhat important.”
These responses thus indicate a
somewhat greater support for providing
undergraduates an option (rather than a
requirement) to study CMSE. Overall,
the chair surveys suggest a majority
view that CMSE plays a supporting
role in undergraduate materials education,
rather than being a focal point in
the preparation of new generations of
material scientists and engineers. In the
responses from computational faculty,
62% consider it “very important” to
integrate CMSE into the required curriculum,
while 33% consider it only
“somewhat important.” Perhaps not
surprisingly, integration of CMSE into
the required materials undergraduate
curriculum is supported by a larger
fraction of computational faculty than
chairs.
Another question addressed the types
of skills in CMSE that are considered
important, specifically programming
skills vs. skills to utilize tools (see
Figure 1). More than 63% of surveyed
chairs chose the ability to use computational
tools over programming skills.
Similarly, 62% of the surveyed faculty
selected the development of skills to
use computational tools, while only
14% considered programming skills a
critical priority in the education of materials
scientists and engineers. (The
remainder considered the two skills
equally important, and amongst these
faculty some stated that programming
was one of the skills required to effectively
use computational methods.)
Overall, the responses indicate a general
preference toward teaching students
to use computational tools in problem
solving, rather than the skills required
to develop computational tools.
Another important question addressed
the mechanism for including
CMSE in the undergraduate curriculum (see Figure 2). In response to the
question of whether CMSE should be
implemented into existing core courses
or as a stand-alone course, only
11% of the chairs chose a stand-alone
course, compared to 53% preferring
implementation into the existing core
courses; 36% chose “equal in usefulness.”
A similar trend was observed
in the computational faculty survey.
Some respondents (including computational
faculty and chairs) pointed out
that CMSE has been used as a “virtual
laboratory” to deliver materials science
concepts that are difficult to demonstrate
in a laboratory or illustrate in a
traditional classroom setting. The role
of CMSE as a design tool was also discussed.
In addition, practical issues associated
with implementation of CMSE
were discussed in the respondents’
comments. A respondent questioned
whether elective CMSE courses would
be sufficiently subscribed. Others were
concerned whether there is space in
the required curriculum to incorporate
CMSE topics. Additional barriers to
implementing CMSE education more
broadly were discussed, including the
availability of funding and a lack of
computational faculty in the department.
Graduate Education in CMSE
Five questions were included in the
computational faculty survey to assess
the status of graduate CMSE education.
When asked about current offerings
of CMSE courses, 39% of the 18
responding departments indicated no
current offerings. Twenty-eight percent
of the departments offer multiple
computational courses, most of which
were offered sequentially in alternating
years. Seventeen percent described
a pair of courses consisting of a course
on quantum-mechanical, first-principles
methods and another focusing
on atomistic simulation (molecular dynamics
and Monte-Carlo) methods. In
one case, three different courses are offered
in a single department, including
a course on computational thermodynamics,
one on continuum and mesoscale
methods, and another on molecular
scale modeling. In one case, where
the department hosts both chemical
engineering and materials science programs,
two courses are offered: one
focused on molecular scale simulation
and another on computational thermodynamics
and Monte-Carlo simulation.
Eleven percent of the respondents
described single course offerings that
cover multiple methods and length
scales, while in 17% of the cases only
a single course is offered in the department
focusing exclusively on atomistic-
simulation techniques. In addition
to the graduate courses devoted exclusively
to computational methods, 36%
of the respondents reported integration
of computational approaches in courses
covering more traditional topics.
In two of these cases, computational
thermodynamics software was used
in a graduate-level thermodynamics
course. In another case, a Monte-Carlo
assignment is included in a course on
phase transformations.
The computational faculty surveys
also probed the availability of courses
that focus on computational methods
that are offered outside the department
but are nevertheless frequently subscribed
to by materials science graduate
students. Twenty-seven percent
mentioned related courses covering
aspects of numerical methods, atomistic
simulations, or electronic structure
methods in chemical engineering,
mechanical engineering, physics,
and chemistry departments. Another
survey question requested the number
of courses on topics related to CMSE
taken by a typical Ph.D. student pursuing
a computational thesis; the number
ranged from zero to six, with a typical
range between one and three.
When asked what aspects of CMSE
education require improvement, 36%
of the 11 who responded identified
the need for better software tools that
could be integrated into their courses.
In light of this, it is interesting that
most computational faculty were aware
of at least one web-based resource on
CMSE education, but the majority did
not utilize them.
One question addressed the level of
interest of MSE graduate students in
pursuing computational research. Of
the 15 who responded to this question,
40% mentioned difficulty in recruiting
students for computational research,
while 27% stated that they have no
such difficulty. Twenty percent mentioned
a strategy of recruiting students
from outside their departments (e.g.,
from physics or chemistry departments).
Thirteen percent mentioned
that students with an undergraduate
degree in MSE do not generally have
a background well suited for computational
research.
Employers’ Perspective
Employer responses represented organizations
ranging from small businesses
to global corporations. The
number of MSE graduates working at
these organizations ranged from 1 to
1,000, with some predominantly at the
BS/MS level and some predominantly
Ph.D. level. The number of employees
making significant use of computational
modeling and simulation ranged
from 1 to 50, with a greater representation
at the Ph.D. level than BS/MS level.
The typical number of MSE graduates
hired per year at each organization was
roughly 1 to 3 at the level of Ph.D., and
1 to 5 for the BS/MS levels. In addition
to MSE graduates, these organizations
also employ graduates from mechanical
engineering, physics, and chemical
engineering for the type of work performed
by computationally inclined
MSE graduates. More than half of
the respondents (57%) stated that the
number of MSE graduates hired, or the
desired skill set, had changed over the
past decade. The current percentage of
hires with a CMSE background is now
37% on average, with an expressed desired
future percentage of 50% on average.
While there was a large range in
the responses (less than 5% to 90%),
this change in the future percentage of
CMSE-trained hires appears to be significant.
![Table 1](table1.jpg)
Categories of CMSE software tools
and specific codes are listed in Table
I, ranked by the frequency they were
cited by employers as used in their
organizations. Finite element analysis
(FEA) tools such as ABAQUS
and DEFORM were by far the most
frequently cited tools. Together with
casting simulators such as PROCAST and MAGMA, these tools predict
macroscopic material behavior during
processing and service and are
closely linked to activities of other
engineering disciplines that need to be
integrated. Control of material structures
is represented predominantly by
CALPHAD computational thermodynamics
tools such as ThermoCalc and
Pandat. Atomistic-level tools based on
density functional theory (e.g., VASP)
were found to be widely utilized, often
to support the thermodynamic tools,
while other atomistic models such as
molecular dynamics (e.g., LAMMPS)
and Monte-Carlo methods were cited
much less frequently. Microstructure
evolution simulators such as DICTRA,
PrecipiCalc, or JMatPro were cited
less frequently. Half of the respondents
also cited significant use of in-house
developed computational tools. These
organizations also develop software, in
addition to using available tools. Other
specific tools cited by more than one
respondent include MATLAB, Materials
Studio, and process integration and
design optimization software such as
iSIGHT.
![Table A](table2.jpg)
Several suggestions were offered
for improvements in CMSE education,
citing a need at the BS/MS level to
evaluate problems quantitatively and
understand what tools exist and how
they can be meaningfully applied to
practical engineering problems such
as design. Experience in applying the
tools in an interdisciplinary collaborative
setting was also recommended.
There were comments on early experimental
validation, the need for ability
to utilize visualization techniques, the
relative utility of different numerical
methods in different situations, and a
desire for students’ familiarity with the
tools most commonly used in industry.
All these needs were also cited at the
Ph.D. level, with the addition of more
extensive knowledge of various methods
to guide appropriate application
choices. There was general consensus
for a strong base in thermodynamics
and mechanics tools, with an expressed
need for “a much better view
of real-world engineering” as well as
“a strong additional bank of knowledge
in statistical mechanics, quantum
mechanics and computer science, and
software development.” Some recommended
teaching the tools in the same
courses where the underlying science
is taught.
The majority of respondents felt that
CMSE would play an even greater role
in their organizations in the future. It
was broadly expressed that moving
away from slow and costly “trial and error”
empirical development was vitally
important for the future competitiveness
of the MSE profession. About one
third of the respondents emphasized
the future importance of these tools
in enabling computer-aided design of
new application-specific materials.
THE CLIMATE
The surveys summarized above indicate
that the majority of chairs and
computational faculty view that CMSE
is an important aspect of MSE that
should be implemented into undergraduate
curricula. However, to modify
the undergraduate curriculum, an endorsement
from faculty is critical. We
inquired of the chairs whether there is
a general support amongst the faculty
within their departments for such modifications.
Only 13% of the respondents
answered “no,” and an additional 19%
responded “not sure.” The remainder,
69%, reported that there is support to
integrate CMSE into the undergraduate
curriculum; of these, 28% responded
that integration into the required curriculum
is supported. Integration of
CMSE as elective courses appears to
be most supported, with 41% of the
respondents choosing this answer.
DISCUSSIONS AND CONCLUSIONS
The results of the surveys can be
summarized as follows:
- A general consensus exists in academia,
national laboratories, and
industry that CMSE is of growing
importance and requires attention
in the curricula of MSE.
- The academic surveys reveal a
majority opinion that it is more
important to teach the skills to use
computational tools rather than
the programming skills required to
develop such tools, although a significant
fraction feel that both are
equally important.
- A majority of academic respondents
favor integration of CMSE
into core courses (e.g., as modules),
although a significant fraction
support also a stand-alone
course.
- Faculty support appears to exist in
many departments, especially for
introducing CMSE as an elective
portion of the undergraduate curricula.
- Awareness of web-based resources
and software repositories is moderate,
but use of these resources for
the purpose of teaching is limited
so far. This underutilization may
unnecessarily raise the barrier to
incorporating CMSE into courses.
- A gap exists between what is typically
taught at universities (atomistic
methods) and the tools most
commonly used in industry (continuum
methods such as FEA).
- The employers responding to the
survey indicated their preference
in CMSE education for an overall
emphasis on use of commercial
tools to gain experience in solving
practical engineering problems.
- Employers commonly suggested
that undergraduate MSE educators
give more attention to the computational
tools used in industry. In
line with some faculty responses,
application in design courses was
also suggested in order to provide
context closer to employer needs.
The survey revealed critical questions
that need to be answered in order
to determine the best steps forward. For
example, despite the fact that there is a
majority consensus that CMSE topics
should be included in undergraduate
MSE curriculum, the best way to do this
remains uncertain. Even though more
respondents (both chairs and faculty)
preferred implementation of CMSE
into required core courses, only 28% of
the chairs felt that there was a general
support within the department for necessary
curriculum changes. Clearly this
approach requires signifi cant efforts by
multiple faculty members teaching the
core courses and cannot be achieved
without department-wide support.
Another important question is what
aspects of CMSE should be taught
in undergraduate courses. While the
survey results refl ect a majority view
that the emphasis should be placed on
developing skills to utilize tools, it is
unclear exactly what tools should be
introduced and how. While continuumlevel
tools may be more appropriate
for engineering, atomistic tools may be
more appropriate for teaching materials
science concepts.
All of these are complex issues that
require discussions within the materials
science and engineering community,
including CMSE and non-CMSE
researchers, educators, practitioners,
and managers. We hope that this article
serves as a starting point for active
discussions. Case studies from institutions
adopting various practices will be
a valuable resource that would provide
data points in such discussion. We plan
to collect such case studies and compile
them into a future JOM article. Other
possible information could be obtained
from recent graduates, who may provide
insights into the gaps that may
exist between MSE education and the
skills required in order to gain employment
and to succeed in the profession.
ACKNOWLEDGEMENTS
This report was made possible by
numerous responses from universities,
national laboratories, and industry.
The authors acknowledge funding from
the National Science Foundation DMR-
0502737 (K.T. & M.A.), DMR-0746424
(K.T.), and CMMI-0507053 (M.A.), as
well as the ONR/DARPA D3D Digital
Structure Consortium (G.B.O.), and the
Materials Design Institute, a Los Alamos–
UC Davis Educational Collaboration
(M.A.).
REFERENCES
1. Committee on Accelerating Technology Transition,
National Materials Advisory Board, “Accelerating
Technology Transition: Bridging the Valley of Death for
Materials and Processes in Defense Systems” (Washington,
D.C.: The National Academy Press, 2004).
2. Committee on Integrated Computational Materials
Engineering, National Research Council, “Integrated
Computational Materials Engineering: A Transformational
Discipline for Improved Competitiveness and
National Security” (Washington, D.C.: The National
Academy Press, 2008).
3. K. Thornton and M. Asta, Modelling and Simulation
in Materials Science and Engineering, 13 (2005), pp.
R53-R69.
K. Thornton and Samanthule Nola are with the
University of Michigan, Ann Arbor, MI 48109; R.
Edwin Garcia is with Purdue University, West
Lafayette, IN 47907; Mark Asta is with the
University of California, Davis, CA, 95616; G.B.
Olson is with Northwestern University, Evanston,
IL, 60208. Dr. Thornton can be reached at kthorn@umich.edu. |