0% found this document useful (0 votes)
104 views

CMD - Introduction To Computational Material Design

This document discusses computational materials design and engineering. It describes how computational materials design integrates materials processing, structure, and property models within systems frameworks to meet engineering needs. The goal is to apply scientific understanding to facilitate optimal tradeoffs that efficiently meet design requirements. Mechanistic models require sufficient fidelity to determine best design solutions over large parameter spaces, and quickly identify failures. Computational materials design aims to find optimal solutions while efficiently navigating tradeoffs among competing requirements.

Uploaded by

Raden Eka G.
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
104 views

CMD - Introduction To Computational Material Design

This document discusses computational materials design and engineering. It describes how computational materials design integrates materials processing, structure, and property models within systems frameworks to meet engineering needs. The goal is to apply scientific understanding to facilitate optimal tradeoffs that efficiently meet design requirements. Mechanistic models require sufficient fidelity to determine best design solutions over large parameter spaces, and quickly identify failures. Computational materials design aims to find optimal solutions while efficiently navigating tradeoffs among competing requirements.

Uploaded by

Raden Eka G.
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 7

Computational materials design and

engineering
C. J. Kuehmann* and G. B. Olson
Computational materials design integrates targeted materials process–structure and structure–
property models in systems frameworks to meet specific engineering needs. Design inherently
consists of many competing requirements that require judicious decisions regarding key
tradeoffs. The goal of computational materials design is to apply the best scientific understanding
to facilitate decisions regarding the optimal tradeoffs that meet desired needs in the most time
and resource efficient manner. Mechanistic materials design models require adequate fidelity to
determine the favourability of one design solution over another but also the ability to be
extrapolated over large parameter space to search for design optima in unexplored terrain.
Design processes must not only efficiently find optimal solutions, but quickly identify failures.
More broadly, materials design can only be as successful as the ability to identify the correct
requirements for an application, and those requirements must address not only performance but
also qualification hurdles including prediction of manufacturing variation.
Keywords: Computational modelling, Materials design, Accelerated qualification

Introduction The performance of the alloy is embodied in the


combination of properties outlined in the column on
The application of systems engineering principles, the right. The design process determines suitable
utilising computational materials science, allows the microstructural concepts to meet these property goals,
rapid and efficient development of design solutions for as indicated by the middle column. Available processing
materials with specific application requirements. The paths to access the microstructural objectives are
general application of the technique has been well quantified in the left column. The links between the
developed in other references.1–4 Expansion of compu- subsystem blocks in the flow block diagram represent
tational materials design for the qualification and process–structure and structure–property models
implementation of new materials into engineering required to quantitatively design an alloy to meet the
systems was established by the US Defence Advanced desired performance objectives.
Research Projects Agency (DARPA) Accelerated The DARPA AIM programme extended the use of
Insertion of Materials (AIM) programme.5 Following computational modelling into the challenges of scale up
up on these efforts, the US Office of Naval Research and materials qualification and certification. In the
has partnered with DARPA and established the Digital context of classical materials development, AIM applies
3-D (D-3D) initiative creating the next generation to materials discovered experimentally in the laboratory
computational and analytical tools to support compu- and addresses the significant failure rate of such
tational materials engineering incorporating the AIM materials during scale-up and qualification. Applying
methodology.6 advanced computational modelling to scale-up, AIM
The driver for an engineering approach to materials accelerates and reduces the risk associated with produ-
design is meeting a specific need for materials perfor- cing new materials reliably at scale. In addition, the
mance in the context of a system. Such a perspective programme addressed the issue of establishing quanti-
necessitates the focus of the design activity on the wider tative estimates of material property variation used for
context of the application. Combinations of properties component design. Currently, such variation is quanti-
must be considered within specified process, cost, fied by expensive and time consuming statistical
environmental and life cycle constraints. Tradeoffs are experimentation. AIM applies a predictive probabilistic
inherent in the approach and tools to effect the design approach that integrates limited experimental data with
must address the optimisation of such tradeoffs expli- computational materials modelling to efficiently estab-
citly. An example of the systems approach is contained lish variability. In the pilot programme, AIM methods
in Fig. 1 for ultrahigh strength corrosion resistant steel. quantified the strength variation experienced in the
production of 1000 turbine discs with as few as 15
individual measurements.7 In the context of computa-
QuesTek Innovations LLC, 1820 Ridge Ave., Evanston, IL, USA tionally designed materials, AIM becomes even more
*Corresponding author, email [email protected] effective since mechanistic modelling directly applicable

ß 2009 Institute of Materials, Minerals and Mining


Published by Maney on behalf of the Institute
Received 12 February 2008; accepted 5 August 2008
472 DOI 10.1179/174328408X371967 Materials Science and Technology 2009 VOL 25 NO 4
Kuehmann and Olson Computational materials design and engineering

1 Flow block diagram for ultrahigh strength corrosion resistant steel indicates desired property objectives, microstruc-
tural subsystems and sequential processing steps needed for design: links between systems blocks indicate quantita-
tive models needed to affect design via science based computation

to the material is already available and scale issues can conflicting, requirements. A number of models will need
be initially incorporated into design considerations. to be employed and operated cooperatively to determine
The ONR/DARPA D-3D initiative is applying lessons the best combinations of process and composition to
learned in the DARPA AIM programme to establish the meet a diverse set of material objectives.
next generation of modelling and characterisation tools Figure 2 illustrates five stages of development in
enabling AIM adoption. The programme is further computational materials design and AIM development:
focusing on modelling approaches to incorporate three- conceptual, parametric, embodiment, detailed and quali-
dimensional considerations, including orientation and fication. The lower portion of the diagram depicts the
clustering effects, into key modelling components. New modelling activities that are associated with each of the
experimental techniques to quantify three-dimensional five stages as well as some typical modelling tools that may
characteristics are being developed to efficiently cali- be employed. Table 1 defines some of these modelling
brate and validate these new modelling tools.8 tools, identifies the methods underlying them and provides
information on specific software implementations that are
Computational materials modelling for available. The upper portion of Fig. 2 describes proto-
typing and production activities typical to the five stages
design of design and development, as well as examples of specific
Computational materials models for design purposes experimental techniques applied to them.
differ significantly from those used in traditional In the conceptual phase, general questions about
materials science research activities. While conventional mechanistic aspects of the alloy are addressed. What
computational materials science modelling strives to strengthening mechanisms should be deployed? And,
understand general phenomena, design models are used given the answer, what microstructures, phases and
to control a materials system and optimise it for a components are the most beneficial? Here, general
specific outcome. Computational materials design mod- materials science tools are very helpful. CALPHAD
els are used to make specific engineering decisions and models of thermodynamics and mobility can determine
tradeoffs, i.e. is option A more advantageous than relative phase stabilities, individual component solubi-
option B? Their fidelity depends on the type of question lities and rate constants given sufficient (often concur-
being asked. Models used in a conceptual design exercise rent) database development. If thermodynamic
need less absolute accuracy than those employed in the relationships are not available, or the literature lacks
design of a detailed embodiment, but they also must be sufficient experimental data to create them, first princi-
fast to allow a large design space to be explored. Design ples ab initio techniques can be used to calculate
models must also extrapolate well, requiring a strong formation energies, magnetic states and lattice
mechanistic foundation. By definition, a new design will parameters for candidate structures. This very basic
be outside the known parameter space on which a model level of ‘phase engineering’ can be quite productive
is based. Lastly, design models need to be highly when identifying new constituents to mimic particular
cooperative. A given design will have many, often mechanisms in a new alloy system that may have been

Materials Science and Technology 2009 VOL 25 NO 4 473


Kuehmann and Olson Computational materials design and engineering

2 Five stages of development in computational materials design include conceptual, parametric, embodiment and
detailed design phases as well as materials qualification: details above each stage represent experimental efforts and
specific analytical tools involved in validating design; details below each stage indicate computational modelling
approach and specific tools employed in creating design

well studied and characterised in another. General efficient precipitation strengthening. CALPHAD models
analytical models for mechanistic process–structure of equilibrium or metastable thermodynamics can then
relationships are very applicable in the conceptual phase determine if such phase fractions and particles sizes are
where rough estimates of required microstructural possible. Model alloys can be produced and charac-
features are needed to establish feasibility of particular terised to determine the validity of the mechanistic
concepts. For example, analytical relations for Orowan assumptions on which a given concept is based in order
strengthening or particle shearing can be used to to establish overall feasibility and enhance fundamental
estimate phase fractions and particle sizes needed for parameters.

Table 1 Listing of available methods and models used in computational materials design and engineering including
available software platforms

Method/modelDescription Tools

TECD Topological electron charge density (TECD) TECD is available from the Molecular Theory
analysis uses charge density input from Group at the Colorado School of Mines, Golden, CO, USA.
pseudopotential or full electron DFT methods
to visualise gradients, critical
points and other features in the charge density
of atomic polyhedra
FLAPW The fully linearised augmented plane wave The code QMD-FLAPW is maintained and supplied
(FLAPW) method is an all electron implementation by Quantum Materials Design Inc. The code WEIN2k
of density functional theory (DFT) that allows is available from the Vienna University of Technology.
the calculation of total energy and electronic The Flair Consortium and the FLEUR project are other variants
structure for given crystal structures
CALPHAD CALculation of PHase and Alloy Diagrams ThermoCalc is available from ThermoCalc Software,
(CALPHAD) is a technique that uses various Stockholm, SE. Pandat is available from CompuTherm
models of the composition and temperature LLC, Madison, WI USA. FactSage is available from CRCT,
dependence of the free energy of phases Montreal, CA and GTT-Technologies, Aachen, DE
to calculate equilibrium and metastable states
of multicomponent systems
DICTRA DICTRA is an extension of CALPHAD specific to DICTRA is available from ThermoCalc Software, Stockholm, SE
the ThermoCalc platform that incorporates
atomic mobility and uses finite difference methods
to solve 1-D diffusion and moving boundary problems

474 Materials Science and Technology 2009 VOL 25 NO 4


Kuehmann and Olson Computational materials design and engineering

In the parametric design phase, individual design The qualification phase demonstrates the ability of the
models, representing the links in Fig. 1, are first material to meet requirements in a component with a high
combined to attempt an optimisation of the materials level of confidence. Here AIM methods integrate compu-
system as a whole. Again simpler models for phase tational modelling and targeted experiments to rapidly
stability, solubility and mechanistic analytical process– determine minimum material performance from produc-
structure relationships are employed, in this case to tion processes. Figure 3 demonstrates the method con-
establish if the conflicting requirements of the design can ceptually in which Monte Carlo simulations of material
be accommodated with a judiciously chosen set of design performance are performed using the expected distribu-
parameters. Here the model predictions may be para- tion of processing parameters to produce a modelled
meterised and normalised to the design objective set by dataset that contains the intrinsic shape of the property
the design requirements. This exercise tests the design distribution due to known sources of variation. A limited
tradeoffs of conflicting requirements, and determines if experimental dataset is then used to translate the modelled
suitable flexibility exists to meet all requirements distribution to match the mean and width of the
simultaneously. Because these parameterised (or para- experimental measurements. The resulting modified dis-
metric) models typically predict trends well but may tribution is then used to estimate the minimum material
have limited accuracy, a carefully chosen set, usually less properties. This method allows minimum properties to be
than six, of design solutions, is prototyped to maximise predicted well before adequate experimental data would
the likelihood of spanning an acceptable design. The exist to do so.
characterisation of these prototypes is completed to
directly validate design assumptions and model accuracy Examples of computation in design
as much as to strictly demonstrate the achievement of
The best example of a commercial alloy created by this
design goals.
methodology is the Ferrium S53 (AMS 5922) corrosion
Using the results from the parametric designs,
resistant landing gear steel allowing a drop in replacement
computational models and fundamental parameters are for current non-stainless landing gear steels, eliminating
updated and expanded to increase accuracy and the need for cadmium plating.11 As depicted in the flow
incorporate additional complexity in the embodiment block diagram in Fig. 1, the S53 alloy is a secondary
design phase. Where a parametric design may use hardening steel strengthened by efficient M2C carbide
driving force and precipitation rate constants to define precipitates and contains sufficient Cr content to provide
precipitate microstructures, the embodiment design passivation against general corrosion. Grain boundary
phase typically employs more rigorous precipitation chemistry is controlled to maximise cohesion, leading to
simulations to define precipitate distributions, using excellent SCC resistance. The sequential processing steps
numerical implementations of nucleation, growth and experienced by the alloy are depicted on the left most
coarsening.9 Strength modelling may move from column of the flow block diagram and are constrained to
mechanistic analytical models to explicit simulations of existing processes for steels employed in current structural
dislocation dynamics.10 Simple phase stability consid- aircraft applications to maximise manufacturability. The
erations may move to multicomponent diffusional subcomponents of this system are connected by process–
simulations of moving boundary problems with tools structure and structure–property relationships, represent-
like DICTRA. The first considerations of manufacturing ing the mechanistic modeling tools necessary to employ
scale are addressed utilizing full scale processing quantitative computational design.
conditions and rates in these simulation based process– The application of the system flow-block diagram in
structure and structure–property calculations. The goal conjunction with the computational models begins with
of the embodiment design is to determine a unique identification of key design trade offs. Key to achieving
design optimum with as much accuracy as possible. the stated design goals is the development of an efficient
Prototypes at this stage are again used to validate the strengthening dispersion. M2C carbide is an efficient
mechanistic underpinnings of the design but are also strengthener in steels due to its high modulus misfit with
intended to be a very accurate representation of the BCC Fe and its ability to precipitate coherently at the
design when produced at scale. nanoscale. It is important to note that the coherency
Detailed design considerations complete the compu- leads to significant complications in predicting the
tational design approach for a material and transition precipitation kinetics as the coherency strain energy
into qualification aided by AIM methods. Here must be accommodated in the precipitation model. This
computational modelling is used to further specify has been well studied and reported in the literature4,12,13
the full range of conditions that may be required for an and such models were utilised in Fig. 4 where the overall
engineering material. Intermediate states and proces- precipitation driving force and the normalised coarsen-
sing considerations are further developed and may ing rate constant are reported as a function of Mo and V
include modelling annealed states of the material to content in the S53 alloy.
address issues such as machinability or formability. This design calculation maximises the resulting
Forging process parameters and windows may be strength through the precipitation driving force, while
determined from incipient melting considerations or assuring this strength can be achieved with reasonable
phase stability. Compatibility with secondary proces- tempering parameters. Maintaining adequate martensite
sing such as surface treatments may also be modelled kinetics to ensure a martensitic alloy while achieving
and optimised. The detailed design phase fully high strengthening is another key design tradeoff.
completes the design and considers the total life cycle Quantitative martensite kinetic models14 are used to
of the material from the first processing step involved predict the martensite start temperature along with the
in its manufacture to the last step in its disposal or M2C precipitation driving force in Fig. 5 as a function
recycling. of Co and Ni contents.

Materials Science and Technology 2009 VOL 25 NO 4 475


Kuehmann and Olson Computational materials design and engineering

3 AIM method integrates modelling with targeted experimental data to produce design allowable minimum properties
using anticipated supplier process variation estimates

This chart allows achieving maximum strengthening Design for corrosion utilised the results of Campbell
while requiring a sufficiently high MS temperature to and Olson4 modelling the stability of passivating films
maintain a predominantly martensitic alloy. in secondary hardening alloys. The analysis utilises

4 Maximizing M2C driving force while maintaining normal- 5 Maintaining sufficiently high MS temperature while max-
ised coarsening rate for adequate tempering kinetics imising carbide driving force allows efficient strength-
provides a secondary hardening alloy with highly effi- ening in fully martensitic alloy: contours of M2C
cient strengthening: S53 alloy achieves greater than precipitation driving force overlaid with alloy MS tem-
1930 MPa tensile strength with only 0?2 wt-% carbon perature determine optimal Co and Ni contents

476 Materials Science and Technology 2009 VOL 25 NO 4


Kuehmann and Olson Computational materials design and engineering

6 One-dimensional multicomponent diffusion simulations of solidification of S53 alloy indicate extent of segregation that
can be expected in final ingot: this segregation is verified qualitatively from SEM observations of ingot material

multicomponent thermodynamic effects to maximise Extending use of computation to


Cr partitioning in the spinel oxide primarily driven by
Co content. In addition, the nanoscale M2C carbides
qualification and implementation
are designed to be sufficiently small compared to the AIM methods were applied to the prediction of the
oxide scale to enable their oxidation during passiva- ultimate tensile strength of S53, the primary design
tion, freeing the Cr contained in these carbides to objective of the alloy. The goal of this exercise was to
incorporate into the passive film. predict property design minimums, which normally
Additional constraints on ductile fracture, grain require 10 heats and 300 individual observations, using
boundary chemistry and grain pinning dispersions are only three heats and 30 individual observations. A
used to complete the design optimisation and uniquely computational model employed predictions of the M2C
identify the alloy composition that represents the best precipitate size as a function of composition and
compromise of the diverse design goals and constraints. processing, as well as the resulting strength when
Grain boundary chemistry considerations are applied accounting for additional contributions from solid
based on first principles calculations of impurity and solution, martensite substructure and dislocation den-
alloying effects of the interfacial cohesion of BCC high sity effects. Using expected process variation, a large
angle grain boundaries.15 The grain pinning dispersion dataset with over 300 simulations was produced. This
design was based on calculations of the TiC/Fe distribution was then scaled to fit 20 experimental data
interfacial adhesion that demonstrated excellent resis- points spanning two heats using a linear transforma-
tance to microvoid formation.16 tion function determined by the best fit. This analysis
Full scale processing concerns were also integrated indicated that the 1% minimum ultimate tensile
into the design. Candidate design compositions were strength was predicted to be 1920 MPa, below the
used in solidification simulations using one-dimensional 1930 MPa design goal. This early indication allowed
multicomponent diffusion simulations employing additional process optimisation to be completed,
DICTRA.17 The results of these simulations in Fig. 6 increasing the ultimate tensile strength by an additional
show the segregation profiles across secondary dendrite 10 MPa. The analysis was repeated with 30 data
arms expected during large scale production of ingots produced from three heats of the alloy and is shown
using typical VAR processing conditions. These findings in Fig. 8a (and showing a prediction of 1930 MPa).
were later validated from actual production ingots. The resulting full dataset of 10 heats and over 600
Of greater importance is the ability to homogenise individual observations is shown in Fig. 8b. The AIM
expected solidification segregation with commercially prediction is within 7 MPa of the 1% minimum tensile
acceptable thermal treatments. Figure 7 shows strength statistically defined by the 10 heats of data. In
DICTRA predictions of homogenisation of Mo in the this AIM example, if data development had proceeded
as cast ingot of the S53 alloy at 1350uC as a function of and the property deficit was discovered when the full
time. The simulation guides processing recommenda- dataset was complete, over a year of development and
tions and indicates that sufficient homogenization can be in excess of $500 000 would have to be repeated to
reached in reasonable processing windows. meet the property goals.

Materials Science and Technology 2009 VOL 25 NO 4 477


Kuehmann and Olson Computational materials design and engineering

7 Homogenisation of Mo segregation in S53 alloy as


function of time at 1350uC: two curves represent com-
position of Mo at dendrite centre and edge, and initial
profiles are provided by solidification simulations from
Fig. 6

Conclusion
8 a AIM analysis of ultimate tensile strength of S53 using
In summary, the example of the S53 steel (AMS 5922)
simulation results and 30 data from three individual
demonstrates how computational materials design
heats agrees well with completed dataset from 10 heats
enabled the development of a complex multicompo-
and b over 600 observations in determining 1% mini-
nent alloy that targeted properties enabling a drop-in
mum design allowable
replacement for Cd plated ultrahigh strength steels by
a corrosion resistant alternative. The S53 design
employs efficient M2C strengthening in a fine ductile References
lath martensite matrix. The matrix contains sufficient 1. G. B. Olson: Science, 1997, 277, 1237.
Cr for passivation that is further enhanced by matrix 2. G. B. Olson: Science, 2000, 288, 993–998.
Co that increases the Cr activity to levels greater than 3. A. Saha and G. B. Olson: J. Comput. Aided Mater. Des., 2007, 14,
common 12 wt-%Cr steels. Grain boundary chemistry 177.
4. C. E. Campbell and G. B. Olson: J. Comput. Aided Mater. Des.,
is controlled to maximise grain boundary cohesion
2000, 7, 145–170.
and to avoid SCC and H embrittlement. Additionally, 5. Committee on Accelerating Technology Transition, National
grain pinning particles were designed to be an effective Research Council: ‘Accelerating technology transition: bridging
pinning dispersion while minimising the impact on the valley of death for materials and processes in defense systems’;
microvoid formation during ductile fracture. The 2004, Washington, DC, The National Academies Press.
computational techniques employed here accom- 6. F. Vernerey et al.: JOM, 2006, 58, 45–51.
7. D. G. Harlow: Eng. Fract. Mech., 2007, 74, 2943–2951.
plished the ambitious S53 design in three design 8. A. C. Lewis et al.: JOM, 58, 52–56.
prototype test iterations and represented a significant 9. H.-J. Jou, P. W. Voorhees and G. B. Olson: in ‘Superalloys 2004’,
cost and time savings over traditional empirical 887–886; 2004, Warrendale, PA, TMS.
development techniques. 10. V. Mohles: Philso. Mag. A, 2002, 81A, 971–990.
11. C. J. Kuehmann, B. E. Tufts and P. Trester: Adv. Mater. Process.,
2008, 166, 37–40.
Acknowledgements 12. K. C. King, P. W. Voorhees and G. B. Olson: Metall. Trans. A,
1991, 22A, 2199–2210.
The authors gratefully acknowledge the support of US
13. G. Ghosh, G. B. Olson and C. E. Campbell: Metall. Trans. A, 1999,
Defense Advanced Research Projects Agency (DARPA) 30A, 501–512.
and the Office of Naval Research (ONR) under the 14. G. Ghosh and G. B. Olson: Acta Metall. Mater., 1994, 42, 3391.
current Digital 3-D consortium. Additionally, the authors 15. W. T. Geng, A. J. Freeman and G. B. Olson: Phys. Rev. B, 2001,
would like to thank the US Department of Defence’s 63B, 165415.
Strategic Environmental Research and Development 16. J.-H. Lee, T. Shishidou, Y.-J. Zhao, A. J. Freeman and G. B.
Olson: Philos. Mag., 2005, 85, 3683–3697.
Program (SERDP) and the Environmental Security 17. H. Lippard, C. E. Campbell, U. P. David, G. B. Olson,
Technology Certification Program (ESTCP) for support T. Björklind, U. Borggren and P. Kellgren: Metall. Trans. B,
in the development of the S53 alloy. 1998, 29B, 205–210.

478 Materials Science and Technology 2009 VOL 25 NO 4

You might also like