INSTITUTE LECTURE

What Chemical
Engineers Can Learn
from Mother Nature
conomic pressures on the chemi- The problem

Biological
processes can serve
E cal process industries (CPI), and
particularly on research and de-
velopment, are currently quite
severe. Most specifically, the high cost of
innovation must be reduced if the pros-
The challenge facing the CPI has re-
cently been characterized by J. A. Miller,
chief technical officer of Du Pont, in
these terms:
“In the last 14 years, our net returns
as a model for perity of the CPI is to continue. Success have not been sufficient to allow industry
will require recognizing that the primary to attract the capital needed for sustained
improving the basic function of the engineer is neither analy- growth. We seem trapped in a box de-
sis nor design. It is creating new process- fined by sociopolitical forces and by the
means by which es, products, concepts, and organizations. way we develop and implement technolo-
The only means we have for conducting gy. Our future depends on finding a way
our profession such creative activities is to mimic, in out of this box.”
some sense, the evolutionary processes of The situations in academia and gov-
operates, nature, and even a cursory examination ernment are different, but comparable to
shows that such processes are inherently those in industry: all of us live in fast-
particularly for uncertain and “expensive.” changing environments dominated by
Fortunately, Mother Nature has de- forces that we cannot control. To prosper,
R&D, as this article vised some complex and effective mech- and perhaps even to survive, we must
derived from the anisms for increasing the economy of continually evolve, and we must do so on
evolutionary activity, and some of these many levels. We must react to changing
1996 Institute are already being used by chemical engi-
neers. Others remain to be applied, and
conditions on time scales from hours or
days for discharging immediate responsi-
Lecture suggests. there appear to be ways for us to im-
prove on natural evolution for restricted
bilities, to decades if we wish to have
successful careers.
ranges of creative activity. Moreover, re- Since sociopolitical forces are beyond
cent results of nonlinear dynamics and our direct control, we must all learn to in-
E. N. Lightfoot, complexity theory suggest that there are novate more quickly and economically.
University of Wisconsin powerful innate organizing forces in Moreover, as suggested in Figure 1, it is
the physical world of as yet unrealized the inventive aspect of engineering which
potential. must drive all others, and, at the moment,
The primary purpose of our analysis is it is also that which has received the least
to suggest examination of the rapidly systematic attention.
growing literature in biological evolution It will be suggested below that all evo-
and nonlinear dynamics as bases for im- lutionary processes exhibit similar dy-
proving innovation rates in the CPI. Sug- namics, and that the large literature on bi-
gestions are also made for developing al- ological evolution and dynamics of com-
ternative approaches that are more suit- plex processes should prove helpful for
able to our situation. increasing innovation rates in the CPI.

CHEMICAL ENGINEERING PROGRESS • JANUARY 1998 ©Copyright 1997 American Institute of Chemical Engineers. All rights reserved. Copying and downloading permitted with restrictions.
INSTITUTE LECTURE

Specific examples are given, but we • Even producing anything as im- One can, however, provide an or-
start here with the back side of evolu- pressive as a Shakespeare in what ganization which supports the activ-
tion: its essential wastefulness and turns out to be about 109.6 years by ities of creative people, and empha-
lack of predictability which lie at the simple Darwinian selection is so un- sis below will be on identifying de-
heart of the innovation problem. likely as to be effectively impossible; sirable characteristics of such a
and structure.
Back side of evolution • It follows that biological evolu-
Shown in Figure 2 is a variant of tion uses some tricks that are worth Nature of evolution and
the famous problem posed over two- investigating. We will initiate such an requirements for success
and-half centuries ago by Jonathan investigation below. We shall begin here by para-
Swift (“Travel Into Several Remote However, there is another implicit phrasing Manfred Eigen and define
Nations of the World, by Lemuel lesson which is just as important: evolution as an increase in function-
Gulliver” (1726)): could a Shake- evolution is inherently unpredictable, al efficiency manifesting itself as
speare play be produced by random and no simple random evolutionary the spontaneous generation of useful
typing on what would now be a com- process could produce a Will Shake- information. At least in principle
puter? It is easily shown that it would speare from a primitive life form. He then, evolution can be quantified in
take the order of 10300,000 years to was just one of an exceedingly large terms of information theory. More
type through all the combinations of a number of equally probable results of practically, biological evolution is
120,000 character manuscript or evolution dynamics. most commonly understood as the
10299,990 times the probable age of our This inherent unpredictability is generation of increasingly “fit” liv-
universe. It is immediately clear that the reason that we must always start ing organisms, with fitness defined
such a random process could not pro- with a desired solution, and not in terms of reproductive success.
duce the desired result, and it seems with the analytical tools later need- However, ecological and social
equally obvious that simple trial-and- ed to refine it. That is why there are communities can also be said to
error could hardly produce William asymmetric interactions in Figure 1, evolve, and, for our purposes, we
Shakespeare, either. and why many research programs can generalize this term to concepts,
We, thus, obtain our first useful fail. The leader of any innovative equipment, processes, and indus-
bits of information from evolution group has to have a sense of mis- tries. We need only replace the ad-
dynamics: sion and know what kind of infor- jective “reproductive” by “finan-
• Simple random processes are mation is needed from those work- cial” or other appropriate term.
enormously inefficient. It can be ing at more-detailed levels. No In spite of their obviously distinct
shown by extension that simple Dar- committee of analysts can innovate features, these various evolutionary
winian selection is not a whole lot effectively and one cannot legislate processes share the same small set
better; breakthroughs. of essential characteristics: variation,

Invention
Could Swift's apes type a
Shakespeare play?

No!
Design Production Use

120,000 Characters
Analysis and Technology 10301,008 Combinations
Experiment
10301,000 Years
■ Figure 1. The business of engineering.
Invention is the key element in all
engineering processes, but information
flow is bidirectional.

■ Figure 2.
Magnitude of the
combinatorial problem.

CHEMICAL ENGINEERING PROGRESS • JANUARY 1998

selection, and heritability (being ca-
pable of passing from one generation The Genes: Information Storage and Replication
to the next). Moreover, since evolu-
tionary change increases the degree of Shorthand for organic bases that
make up the genetic code
order in the system under considera-
tion, change can occur only at the ex- Single Chains:
pense of entropy increase in the sur- U C U U G A G C C
roundings. However, entropic forces
are always at work, and they can
overwhelm weak evolutionary strate- (Independent in RNA viruses)
gies, producing what is known as an
“error catastrophe.” Double Chains:
The nature of variation, selec- Negative
tion, and heredity differ greatly
among various evolutionary pro-
cesses, and these differences will
have a major impact on our discus- Positive
sion. If we are to benefit from bio-
logical examples, we must be able Negative
to identify generic characteristics of
evolution dynamics — and to
evolve specific strategies effective
for our purposes. Driving energies Positive
also vary, and for most engineers ■ Figure 3.
the immediate source is money The chemical
or its equivalent. However, one nature of genes.
should not underestimate nonfinan-
cial motivation.
A particularly important charac- Looking ahead sent the genotype of the organism.
teristic of evolution is the large pa- Now, we will review known The genes, in turn, produce the phe-
rameter space in which it normally salient characteristics of biological notype, or the organism we see,
occurs and its basic unpredictability, evolution, followed in each case by a largely by the production of en-
both already illustrated above. The projection to engineering systems. zymes via the elegant and complex
immediately obvious conclusion is We then will look at some nonbiolog- processes of developmental biology.
that the blind forces of classic Dar- ical dynamic models and conclude The phenotypes then produce genes
winian dynamics (“chance and ne- with some specific suggestions for to form the next generation, thus
cessity” as formulated by Jacques chemical engineers. These sugges- completing a cycle used by all living
Monod) could not by themselves tions will necessarily reflect the systems. Evolutionary change takes
produce the often surprisingly rapid limited experience of the author, but, place via sequence changes at the
pace of evolution which is due to a hopefully, they will imply wider genetic level and differential sur-
variety of factors, some of which applications. vival rates of the phenotypes. The
will be considered below. The fre- reproductive cycle is, in many ways,
quency of convergent evolution, in Fundamental mechanisms well understood, but, in spite of
which very similar species are Biological evolution is firmly much speculation, the origin of this
formed from very different routes, based on the information content of dualism is still obscure and is likely
in particular, suggests more pre- long, double chains of desoxy ribose to remain so for a long time.
dictability than can be explained to nucleic acids (DNAs), or, for very The situation in engineering shows
date. There is much to be learned simple organisms such as viruses, intriguing analogs to the genotype-
here, and we have hardly started. single chains of the closely related phenotype cycle, but also strong con-
However, we are already beginning ribose nucleic acids (RNAs). This trasts. At the purely descriptive level,
to find that even the qualitative na- information is coded in linear ar- we can see incremental evolution of
ture and feasibility of biological rangements of elements, usually products and processes, and even
evolution, and by extension success- known as base sequences. These close analogs of Linnaean “family
ful engineering developments, are chains, illustrated schematically in trees,” the familiar organization of
subjects for serious discussion. Figure 3, form the genes and repre- plant species. We can also see how

JANUARY 1998 • CHEMICAL ENGINEERING PROGRESS

INSTITUTE LECTURE

these evolutionary changes are driven ment of which is subject to random ganisms, including mammals, and
by market competition and other replication errors. Moreover, both has important consequences. First,
politicoeconomic forces. However, the RNA sequences and replication there are no well-defined species,
the interplay of concepts and imple- error rates can be determined exper- only groups of closely related indi-
mentation, the analog of the biologi- imentally. These viruses can then be viduals now normally known as
cal reproductive cycle, has not been used to test Darwinian evolution “quasi-species.” Second, diversity
studied in sufficient detail to give a dynamics quantitatively by subject- is a source of robustness for all or-
real understanding of the process. ing a stable culture to an abrupt ganisms and even ecosystems. This
In many cases, for example, chro- change in conditions, measuring its point has relevance for all human
matography, we can actually pinpoint response rate, and characterizing organizations: hiring in one’s own
the initial discovery and the circum- the new dominant sequence. Such image is a dangerous procedure
stances leading to it, and there are tests, by Manfred Eigen and others, leading to inflexibility and limited
many detailed discussions of creativi- produced some major surprises — capacity for dealing with changing
ty. Moreover, it is widely recognized and the first quantitative characteri- circumstances.
that some individuals are more cre- zation of evolution dynamics. By extension, generalists, such
ative than others. However, we seem The first surprise was that re- as rats and coyotes, tend to fare
far from optimum fostering of cre- sponse to change was very much much better in fast-changing envi-
ativity, and this is clearly an impor- faster than predicted, and it was ronments than specialists like the
tant area for future research. found that this speed results from famous dodo bird. The current per-
Further exploration of these heterogeneity in the original cul- ilous conditions of Hawaii’s highly
ideas must wait on a more-detailed ture: there is always a multidimen- specialized birds and plants are cur-
study of what is known about evo- sional distribution of base se- rent examples, but some academic
lution dynamics, and we now begin quences about the mean, or a domi- departments and industrial organi-
this process. nant sequence known as the “wild zations may soon illustrate this
type.” Some of these variants are same principle.
Darwinian dynamics better adapted to the new conditions The third surprise is in the rela-
and the error catastrophe than the previously dominant strain tion of evolutionary speed to error
The simplest of organisms are and, thus, need little or no evolu- rates, suggested in Figure 4 for a
small viruses whose genes consist tionary sequence modification. reasonable virus model. Shown here
of single strands of RNA, each ele- Such diversity occurs for all or- are the frequencies of variants dis-
tributed about a wild type as a func-
tion of error rate in RNA replica-
tion. Here d is the number of indi-
1.0
vidual RNA elements differing from
those in the wild type. It may be
seen that the fraction of individuals
corresponding to the mean, or wild
d = 0 (Master)
type, steadily decreases as the error
rate increases, but the mean remains
stable up to the error threshold. Up
Error Threshold
to this point then, information is re-
Relative Population Number, xd

tained, and the speed of response to

0.5 environmental change increases.
However, there is then an abrupt
d=1
change to a situation in which all
information is lost, and the organ-
d = 25
d = 24 and d = 26 ism dies out: entropic forces have
d=2
d = 23 and d = 27 taken over.
d=4 d=3 d = 22 and d = 28 This type of behavior also seems
d = 21 and d = 29 characteristic of many larger sys-
d = 20 and d = 30
0.0 tems, and it has consequences out-
side of biology, as well. In both re-
Error Rate search and business it is found that
up to a certain point riskier strate-
■ Figure 4. Error rate and information loss. (Adapted from Ref. 1.) gies tend to produce greater re-

CHEMICAL ENGINEERING PROGRESS • JANUARY 1998

 dP/dt = kP N Indicators of Industrial Strength
1E5

+ - 1E4 British Coal

French Coal

Kilotons/yr
1E3
dN /dt = kN P 2
P N
kP , kN = constants British Pig Iron
1E2

■ Figure 5. A simple hypercycle. 1E1

(Redrawn from Ref. 2.) ■ Figure 6. 1700 1750 1800 1850 1900
A nonbiological
hypercycle. Year

wards, but that this happy situation proportional to the population of P2. past decades provide a classic ex-
reverses abruptly at a finite level of This leads to the very fast hyperbol- ample of trying to reach too far
risk. As a result, there is a danger- ic kinetics and makes up in speed with existing resources and suffer-
ous tendency to become increasing- for a high error rate. Such a pro- ing repeated error catastrophes as a
ly bold — and to encounter disaster cess, called a hypercycle, is known result. Synthetic fuels programs
just as success seems most spec- to occur, and there are in fact many provide another example.
tacular. The newspapers are full of more complex and powerful vari-
examples. ants known. It is one of nature’s Roles of DNA and sex
ways of getting around replication Single RNA strands proved too
Hypercycles errors, and there are indeed nonbio- error-prone, even with the aid of hy-
It is found that most small virus- logical hypercycles. percycles, to support the evolution
es operate just below their error One is the Industrial Revolution of bacteria, let alone the macroscop-
threshold, and that each variant in England during the eighteenth ic life forms we see around us. As a
grows at its own exponential rate. and early nineteenth centuries. result, the genes of almost all inde-
However, the number of errors per Thus, it is shown in Figure 6 that pendently existing life forms be-
generation increases with base se- British iron production rose at a came organized into very long dou-
quence length, and above a certain faster-than-exponential rate. A com- ble strands of DNA, which permit
size, error rates of RNA replication bination of circumstances, including very low copying error rates. How-
exceed the expected error threshold. Justus Liebig’s demonstration that ever, these low error rates also
This limitation would prevent the North European agriculture could be strongly inhibit adaptation, especial-
development of all higher organ- made more productive by use of ly if changes can only be passed
isms, and one consequence is that phosphate fertilizer, caused an enor- down along single lines of descent.
simple Darwinian dynamics could mously vigorous increase of popula- This problem was resolved by the
not produce the myriad of species tion, well being, and productivity in development of sexually based re-
we see around us. The first biologic the north of England. It is probable production which permits recombi-
response to this problem was devel- that American biotechnology and nation of large segments of DNA
opment of autocatalytic replication the computer industry are currently between pairs of individuals without
which permits faster than exponen- enjoying mutually supportive hyper- frequent catastrophic errors. Such
tial growth rates and, thus, raises cycles, and there have been many homologous recombination, sug-
the error threshold. other historical examples. gested in Figure 7, proved a very
A very simple example is shown The lesson for engineers is that difficult process, with both pheno-
in Figure 5 for a virus whose geno- research should begin paying its typic aspects (mating rituals) and
type is denoted as P and the pheno- way as early as possible, providing genotypic (meiosis). Its develop-
type as N. Here, the P form acts as income streams before ultimate ment was very slow: some two-and-
its own catalyst and to a first ap- goals have been met. The repeated a-half billion years, or more than
proximation rate of N formation is failures of fuel cell technology over twice as long as required for the de-

JANUARY 1998 • CHEMICAL ENGINEERING PROGRESS

INSTITUTE LECTURE

the evolution of our planetary sys-

tem and its life forms, to the organi-
+ zation of human societies (Figure
8). Successive developments are
treated via the techniques of nonlin-
ear dynamics as a series of relative-
ly abrupt bifurcations or paradigm
shifts. As one moves through time,
complexity steadily increases, while
energy levels decrease. Specific ex-
■ Figure 7. Homologous gene recombination. amples such as the Industrial Revo-
lution, our information age, and the
velopment of asexual single-celled levels, and it is only very recently rise of the biological sciences ap-
organisms. In the process, a wide that management of research and pear to fit well into this framework,
variety of intermediate processes design has begun to receive serious and it may well prove useful to in-
developed such as bacterial conju- attention. However, sociobiological dividual engineers to try placing
gation and virus incorporation. factors are already recognized as their activities within Laszlo’s
The operation of research organi- important. framework.
zations is, in many ways, like a Laszlo also suggests that evolu-
multisexual extension of homolo- Nonbiological approaches tionary processes do not depend en-
gous recombination. Corresponding To date, deeper discussions of tirely upon purely random events,
to the phenotypic mating ritual are evolution than those sketched out as once postulated by Monod, but
hiring, and forming trust and coop- above have been dominated by that they are favored by the self-or-
eration between individuals, while speculation, but many have provid- ganizing nature of these systems.
the genetic level is represented by ed useful insight. Among these is a He is joined here by many others,
the exchange of specific informa- coherent generalized theory of including Stuart A. Kauffman and
tion, which appears to be the basis Ervin Laszlo which covers all John H. Holland.
of most creative activity. Success known dynamic processes from the Kauffman has studied the dynamic
has proved elusive at both of these “big bang” of cosmologists, through response of model systems, typically
banks of light bulbs connected by
Boolean switches to perturbations.
These may be viewed as crude ap-
Successive Levels of Organization proximations to the metabolic net-
works shown schematically in Figure
Low High 9: the central energy transduction
pathways common to most organisms
Social Systems
as visualized by V. Hatzimanikatis.
Kauffman finds that sparsely connect-
Ecosystems ed systems, with each bulb controlled
Bonding Energy

by only one or two others, exhibit a

Complexity

Metazoa high degree of order, cycling through

only a very small number of states.
Moreover, most elements will quickly
Protozoa reach a steady state, leaving relatively
small islands of blinking lights, and
Molecules Macromolecules the qualitative features of the response
are insensitive to the specific nature of
Atoms
the connections. However, when the
Protons, Neutrons, Electrons number of connectors per bulb K ap-
Quarks proaches the number of bulbs N, the
High Low Compressed Time
system becomes effectively chaotic
and shows no apparent periodicity.
Now, it is the steady-state regions
which form small islands in a continu-
■ Figure 8. A global view of evolution. ally changing continuous “phase.”

CHEMICAL ENGINEERING PROGRESS • JANUARY 1998

This situation is analogous to passing
the error threshold of Eigen, and all
useful information is lost.
The most interesting feature of
Kauffman’s systems is that there is
a very abrupt shift from order to-
ward chaos when K increases from
1 toward 2, always assuming that N
is large. Orderly change to a new
pattern occurs only on the orderly
side of this shift and is fastest just
before the shift. In Kauffman’s
terms, evolutionary response to
change is fastest just on the deter-
ministic side of chaos, a situation
curiously analogous to Eigen’s find-
ing for simple viruses.
These results suggest that in a re-
search organization there is an opti-
mum degree of interaction between
individuals: too much isolation —
or too much hierarchical control
from the top — leads to stagnation,
and too much interaction to chaos.
This suggestion presents an inter-
esting challenge.
Another suggestion, by Holland,
among others, is that most creativi-
ty consists of rearranging known
components in new ways. This is a
generalization of the unit operations
concept which dates back to George
E. Davis and the very origins of our
profession at the University of
Manchester toward the end of the
last century.
The structure of research organi-
zations has also been an object of
much study by Peter Drucker,
among others, and some of Druck-
er’s findings are consistent with
Kauffman’s picture. Of particular ■ Figure 9. Biological energy transduction.
interest here is the suggestion of in-
creased lateral cooperation and less-
ened hierarchical control, a practice banyan tree with its complex of al- to develop improved understanding.
that seems to be gaining industrial most indistinguishable roots, trunks, It is, thus, desirable to examine se-
acceptance. Nature has also provid- and branches, rather than the tradi- lected existing applications to see
ed alternatives to the simple branch- tional tree with its simple sequential what we can learn from them. Bio-
ing structure familiar in our culture, branching structure. logical examples are attractive for
illustrated by the contrast between their growing practical importance,
an oak and a banyan as models for Suggestions as well as the massive literature that
organizational “trees.” The interrela- for chemical engineers has built up around them. However,
tions of chemical engineering with Evolution is such a pervasive ac- nonbiological applications will al-
its basic underlying sciences is bet- tivity that chemical engineers can ways be the primary concern of our
ter understood in terms of the hardly avoid it, and it is important profession, and these will require

JANUARY 1998 • CHEMICAL ENGINEERING PROGRESS

INSTITUTE LECTURE

new approaches. For example, the narrow range of solutions that can acting with as wide a variety of other
use of “evolutionary algorithms” be handled by these familiar tech- professionals as possible.
modeled on Darwinian evolution, niques, and we make the claim here An important early step is estab-
which is to say complete ignorance, that this is best done by a hierarchi- lishing an approximate structure for
is unlikely to be efficient. cal sequence of carefully selected the problem at hand which identifies
approximations. the aspects of primary importance
Biological evolution The first step is the fundamental and their interrelations with each
problems creative one of establishing a new other. This requires a good grasp of
There is a wide variety of at- concept, and here biology has little to orders of magnitude and must in-
tractive biological problems for offer: in spite of many emphatic clude a practical familiarity with ex-
chemical engineers, and rapidly claims to the contrary, no one knows isting technology. Miller’s text is
mutating viruses are a prime exam- how or in what circumstances the first particularly useful here in emphasiz-
ple. John Yin at Dartmouth has al- living species arose. What is needed ing the importance of visualization,
ready begun such an investigation, here is a “creative” generalist who usually through rough sketches. It is
and there is plenty of room for more. has a wide experience and knowledge essential to avoid excess detail at
The problems of antibiotic resistance of available technology. The works of this early stage.
of bacteria is a pressing problem, J. H. Holland can be useful here and The next step is the use of geome-
and gene therapy is clearly in need also analytical biographies describing try-insensitive correlations to obtain
of attention. Bernhard Palsson of the the psychology and personalities of improved descriptions, and then one
University of California at San creative individuals. A. I. Miller’s goes on to successively more-refined
Diego has already shown that heuris- “Insights of Genius” is one of the calculations and experiments.
tic application of classic mass-trans- more recent, but there are many. The most important point is that
fer theory can be helpful in this lat- There is no substitute for studying the we are normally looking for anoma-
ter problem. lives of creative individuals and inter- lies: solutions that are better than
Frances Arnold of Cal Tech is brute-force ones based on some gen-
using modifications of genetic algo- eral formula. Such approaches can
rithms which she calls “directed evo- only be taught or discussed intelli-
lution” for development of improved E. N. LIGHTFOOT is Emeritus Hilldale Professor, gently by example. CEP

industrial enzymes, and Venkat Dept. of Chemical Engineering, University of

Venkatsubramanian of Purdue is tak- Wisconsin (608/262-6934; Fax: 608/262-5434;
ing a somewhat similar approach in e-mail: [email protected]). His long-term
interests are in mass-transport phenomena,
combinatorial chemistry for drug with emphasis on biological applications.
Literature Cited
discovery. These are well-defined A native of Wisconsin, he returned to that 1. Eigen, Manfred, “Steps Toward Life,”
problems in which biology plays a state in 1953, after receiving a BChE and Oxford, New York (1992) (translated
secondary role, and there is nothing PhD in chemical engineering from Cornell from, “Stufen zum Leben,” Piper, Mu-
University and spending three years in
inherently advantageous in using bi- process development at Chas. Pfizer and Co.
nich (1987)).
ological models for the algorithms. He is the author or co-author of six books
2. Lazlo, Ervin, “Evolution: The Grand
It is only important to have an eco- on chemical engineering on mass-transfer Synthesis,” New Science Library,
and bioengineering, the author of about Shambhala Publications, Boston
nomic means of seeking and evaluat- (1987).
200 research papers and patents, and has
ing new alternatives, and the blind presented about 250 lectures. He won
approach of Darwinian dynamics AIChE’s William H. Walker, Warren K. Lewis,
has, in fact, already been abandoned and the Food, Pharmaceutical & Further Reading
by all of these researchers. Bioengineering awards, and was an
Institute Lecturer for AIChE. Lightfoot has Holland, John H., “Hidden Order: How
been elected to the American Academy Adaptation Builds Complexity,” Helix
Nonbiological applications of Arts & Sciences, the National Academy Books, Addison-Wesley, Reading, MA
We must begin here with a clear of Sciences, and the National Academy (1995).
definition of invention which we of Engineering. He has won ACS’s E. V.
Kauffmann, Stuart A., “At Home in the
Murphree award and the American Society
differentiate from design as empha- for Engineering Education’s Corcoran
Universe,” Oxford, New York (1995).
sizing qualitative discovery and award. He has been a lecturer at numerous Kauffmann, Stuart A., “The Origins of
heuristic approaches to a “reason- universities, including the Technical Order,” Oxford, New York (1993).
University of Norway, California Institute Miller, A. I., “Insights of Genius,” Coper-
able” solution which can be further
of Technology, University of Michigan, nicus, Springer-Verlag, New York
refined, perhaps even optimized, by and the University of California, (1996).
the established techniques of pro- Berkeley.
cess design. The problem is now to
home in as rapidly as possible on a

CHEMICAL ENGINEERING PROGRESS • JANUARY 1998