Tip:
Highlight text to annotate it
X
>> Thank you very much Paul and it's a pleasure
to have the opportunity to speak to you today.
This talk will be about the role of computing,
particularly how performance computing
in developing models not just in chemistry but in other areas
because the ability of computers to develop accuracy
and increasing a reliable models,
the complex systems has been one of the major developments
in science in the last 30, 40 years.
It's recent development but it's a recent development
that relates to one of the oldest scientific activities
and that is model building because models--
model building is as old as scientific thought.
What I'm showing you here is the Ptolemaic model of the universe,
a very beautiful model.
This is a real physical model just an illustration here
developed in the ancient world and a successful model.
This model helped to rationalize many
of the astronomical observations that were available
in the ancient world and of course that model was swept away
by the heliocentric Copernican model in the 16th century
but illustrates the fact that scientist have need models--
needed models since the very beginning of scientific thought.
So, these are areas in which we use computers to build models.
It's also an exhaustive list
but it illustrates how broadly computational modeling is now
used in science.
Cosmological modeling absolutely key,
the area of contemporary cosmology,
not an area that I've expertise in.
Atmosphere, ocean, and climate modeling a great deal
of what we are predicting about the evolution
of the earth climate derives
from sophisticated computer models.
Geophysical models, I'll you an example in a few minutes.
Aerodynamic modeling, absolutely key to the aviation industry.
Epidemiological model, again, absolutely vital in the field
of public health and then my own specialty that I'll speak
about later on, the development of models
from molecules and materials.
So what do we do in computer modeling?
Well, it's-- anyway it's an interesting kind
of conceptual area that we used very basic scientific knowledge,
for example, gravitation, the knowledge of gravitation,
quantum mechanics which is the basic knowledge
of how electrons behave and atom molecules.
Hydrodynamics, we'll give examples later on.
So we use this basic knowledge to construct models,
convert complex real systems and you can think of it
as the converse of reductionism.
Reductionism was absolutely key discovery as it were
in scientific thought.
The idea there is you strip nature then its very essentials
and pick up the fundamental underlying principles.
[Inaudible] has been very successful in that respect
so we can now use these fundamental underlying
principles with the immense processing power of computers
as I said to construct models of complex real systems.
And so, it's an integral part of contemporary physical,
biological, medical sciences, and engineering
as my previous slide showed.
So, what's the motivation?
Well, the first motivation,
model building I've already discussed.
Scientist need models to have helped them understand the
complex systems which they are attempting to, who's behavior
that they are attempting to follow.
So model building, very, very broad general need.
Bur here are some most specific ones
that computer models can be very valuable sources now
of numerical data.
For example, a great deal of what we know about impurities,
defects, and semiconductors that control their behavior,
key materials in the contemporary world that comes
from computer modeling.
Inaccessible system, I'm gonna give you an example
in a few moments of beautiful applications here from UCL
with computer modeling helping us
to understand the earth's core
which most certainly is an inaccessible system.
And then most ambitiously, we use computer modeling
to predict, to predict new systems and phenomena and,
again I hope to give you an example later on.
So, we're going to have some grand challenges.
UCL is a great place for grand challenges.
Here are four not the UCL grand challenges
but the grand challenges that I'm going
to discuss in a few moments.
We're going to look very briefly at how computers have been used
to model star formation then I gonna come down in length
and time scales then we'll look modeling the earth's core
that I already referred to, then modeling turbulent flow,
key area in engineering and then it begins my own specialty
of modeling materials at the molecular level.
Well, let's just look pictorially
at the first challenge modeling star formation
and here I'm going to highlight fantastic work from Kinwah Wu
from our Department of Space and Climate Physics at UCL.
And essentially what we're doing here is applying hydrodynamics
at the galactic scale and I'll just show you a couple of images
but this was supposed to be movies
but in some way they're more effective
as single image, as single images.
And was these are showing are the early stages of condensation
of a gas cloud of galactic dimensions into stars.
So, the scales here in length and time are enormous
with this model that gives--
gives us insight into the early stages
of this condensation process.
I can show you lots of examples here, these beautiful images.
This is another one where after which essentially a jet
of gas has come from the center of the galaxy
and again were beginning
to see this condensation into star formation.
Talk to Kinwah if you want to learn more
about this fantastic area
but it showing you modeling here is getting to grips with this
as it were cosmological,
astronomical event with no accurate.
Let's go on to my second example and that's we come
down a little bit in the length scales here
to modeling the earth's core.
And again here, I'll be highlighting fantastic work here
at UCL Dio Alfe from Mike Gillan, David Price,
and Lidunka Vocadlo in the Department of Earth Sciences.
Now, what they're doing here is they're applying quantum
mechanics to allow us to understand key aspect
of the behavior of the earth's core.
Quantum mechanics, as I said, is the basic theory
of how electrons behave and atoms, molecules, and solids
and they're applying this knowledge with the help
of enormous computer processing power to help us
as answer key questions about the earth's core.
This is part of a larger program led by Dario Alfe joint
between the Department of Earth Sciences
and Physics applying quantum mechanical techniques
to study materials under immense conditions
of temperature and pressure.
So let me remind you a kind of kids' model for the structure
of the earth which I think probably most of us know.
The earth in the center has the core
and that's got two components;
the inner core is solid metal mainly ion,
the outer core is liquid.
Then surrounding the core we have mantle
which is predominantly silicate minerals and then right
on the top as it were end is the crust,
a rather thin crust in which we all live.
The other point that this slide is making is the way
in which seismological data has been used
to probe the structure of the earth.
Now, we've known that basic idea, basic features
of the structure of the earth for a long time.
We've also known from a number of pieces
of evidence the composition of the core.
The evidence comes from cosmochemistry, distribution
of elements in the cosmos,
analysis of meteors, equations of state.
I'm just learning from seismology the properties
of the materials at various parts in the earth.
And we know, as I said a few minutes ago,
that the core is ion, alloy
with a small fraction of lighter elements.
Now, why is the core important?
It is important for a number of really very significant reasons.
It's the seat of major global process.
If the core, as we'll see in a few minutes, is extremely hot
and the heat coming out from the core drives the convection--
drives the convection in the mantle which is responsible
for plate tectonics which certainly influences
as we know dramatically what takes place on the surface
of the earth, but the other or one other important feature
of the core is that convection in the outer core,
the liquid part, generates the earth's magnetic field
and the earth's magnetic field is certainly more
than a curiosity, I love showing this slide.
The earth's magnetic field has an absolutely vital role
in making life on earth actually possible.
What we're seeing here is a solar wind.
Again, this is a picture that was provided for me by NSSL.
A peak of solar wind coming out [inaudible] it's a graph,
it's an illustration but the point we're making here is
that the earth's magnetic field protects the earth
from the consequences of that solar wind.
So, the earth's core is--
knowledge of the earth's core really is very important.
Now, the problem which Alpha
and Collins [phonetic] addressed was the temperatures.
So an absolute key quantity that we need to know the temperatures
at the inner outer core boundary
and then the outer core mantle boundary.
>> That was as an absolutely vital knowledge
for contemporary geophysics.
We may use this very sophisticated quantum mechanics
to work how the melting points of iron under the conditions
of huge pressure that we know are present in the earth's core.
They measure the melting point by finding the regions
at which liquid, which is on the right there
of this graphics, and solid coexist.
By this very sophisticated procedure, they're able
to determine the melting point of iron
under these huge conditions of pressure in the earth's core.
And to cut the long story short, this was the model
that they came that they derived for these key temperatures
at the inner outer core boundary and the boundary
of the outer core and the mantle.
So that is key knowledge for geophysics.
And as I said earlier, this is an inaccessible system
so this really is an absolutely vital method
for deriving this essential knowledge.
Now, this work has been going on for quite a long time
but what I'm gonna show you now is a headline
from the Daily Mirror after one
of the earlier papers was published in the Journal Nature
in the late 1990's and they were greatly taken
by these predictions as to the temperatures of the rock,
5000 degrees which are the same temperatures, I think,
as the surface of the sun, and so they came
up with this wonderful headline, "Core Water Scorcher ".
And [inaudible] Mirror, October 1919 and that factor, I think,
perhaps the first time where results
of the computer modeling exercise and certainly
in science have got a full page spread in the Daily Mirror.
Anyway, let me go on to my next theme
and I should say here I'm a long way on this next one
from my own expertise but I'm highlighting some word
that was provided me by Ian [phonetic]
and Ian's many other colleagues involved but from the Department
of Mechanical Engineering really just to emphasize
that computational modelling is a really technique
in contemporary engineering and that concern
with fundamental fluid mechanics.
So we need to understand for instance how fluids flow
around complex shapes and this can be,
I'm sure if you all know, highly complex.
Slow it down, we have an example of turbulent flow,
extremely important phenomenon and one
that is actually very difficult in many years
to get grips with quantitatively.
Now, they are doing fantastic computational work
but the best computational work whenever it can be is tied
in with experiment
and so they're also doing experimental studies using this
new fluid mechanics, laboratory again of the way
in which complex flows occur around objects.
And I'm just going to show you [inaudible] some area
of expertise but three examples of work of this group.
This is I thought I found to be wonderful.
Well, this was a modeling of the breaking of tsunami.
You can see the graphics on the right hand side,
obviously a highly topically view of the tragedy
of the world a number of years ago
when the tsunami in the Indian Ocean.
And they're able to model this really rather accurately.
They can model the breaking of the elevation wave
and very importantly,
their models predict this very extensive backflow
that you see here, which is one of the major problems observed
in the Indian Ocean, the tsunami.
So then, please talk to Ian if you want
to learn more about this.
But then, just illustrating the diversity
of their work here is a summary of, again,
a very extensive study applying modeling
to airborne transmission in hospitals, as you all know,
a very important phenomenon.
They're able to model this really quite accurately
and the results are providing guidance
to reduce infection spread.
And if we want to go to an even more medical application,
here is work that they've done in collaboration
with the Moorfields Eye Hospital concernd
with retinal detachment, which is treated
by replacing the vitreous medium in the eye
by this gas liquid tamponade and you need
to know again the fluid flow properties of that fluid.
And again, they have modeled that successfully
and made a real serious input
into this important medical problem.
So, absolutely terrific work, again, here at UCL.
So, the remainder of this talk, I'll kinda come back to home
for me and discuss modeling materials,
which is my own specialty up in molecular level.
On here I'm going to highlight the work of a number
of colleagues in the chemistry department at UCL.
And what we're doing here, we're providing content--
applying quantum mechanics that I've already alluded to
and also-- and have you recall molecular mechanics, I'll try
and explain in a minute, but to predict--
to understanding or predicting structures and properties
of complex molecules and materials.
So, what we want to know
about materials are the atomic and molecular level.
Well, we want to know the structures the way
in which the atoms are arranged in them.
They can be arranged in a regular manner as in a crystal
or in a disordered manner as in an amorphous material.
We want to know what happens on the surfaces,
many very important phenomena take place on the surfaces
of materials, so we want to be able to have models for surfaces
and for interfaces between two types of material.
I've mentioned very briefly the area of defect,
solids contain impurities, imperfections
and they can often control many
of their most important properties.
We want to know how molecules can absorb,
can thick inside materials or occupy sides on their surface.
And very ambitiously we may want to use modeling to guide us
in the synthesis of new materials, understand the way
in which they grow, not a chemistry of course,
no talk on material science can mix something
on the normal theme but a really fascinating area
of contemporary chemistry.
We need models for the structure of matter at the nano level
in a very ambitious area and I spent a lot of time on,
I'm going to have probably time to talk about it, reactivity,
the way in which molecules react on solids
and within their pores.
And we want to apply these methods to lots
of different systems, to things like oxides and silica.
Silica is materials at the earth's core,
and certainly here is mantle and crust are made out of.
We want to apply into semiconductors.
We want to apply into molecular crystals, for example,
pharmaceutical-- crystals of pharmaceutical compounds.
If I get time, I'll give an example there later on
and of course a wide range of metals and alloys
that we apply these methods to.
Now, I don't wanna kind of give you a lecture on chemistry,
I just want to give you a glimpse of how we try
and do this modeling in chemistry,
and we use two very general strategies.
The first, I wanna call interatomic potentials.
And the idea here is, you know, atoms, nuclei surrounded
by electrons, well up to a point we try and partly forget
about that, we just say atoms and perhaps
for all the squashy spheres and we'll develop models for the way
in which these squashy spheres interact with each other.
And over the last kind of 60 years,
we become very good at that.
And once you got these models,
so the way in which these atoms interact with each other,
lots of kind of computing games you can play.
You can run down helion energy,
just say now we know you interact
and I just find the lowest energy way
of arranging the yourselves.
You can imagine your eyes at Newton
and you can use Newton's classical equations of motion
to work at how atoms will move around,
given a certain amount of energy.
And then you can-- as I say, you roll dice,
to generate ensembles.
You can generate lots of different configurations
of your atoms and then try and calculate properties.
But then the other thing you can do,
which I've already referred to,
you really can't say we'll use our very basic knowledge
of how electrons behave in molecules in solids,
[inaudible] equation summarize now
and we'll calculate these properties in detail.
Well, were back to challenges instead of ground challenges,
we're gonna have key [phonetic] challenges
and probably we won't get through all these
but the first will be, can we predict the structures
of crystals and nanoparticles?
Then we'll look and see what can we do some guidance
for synthesis, then a very important problem
in both chemistry and chemical engineering, can we understand
at the molecular level how crystals grow.
And then the one I won't have time
for probably is how we determine the mechanisms whereby molecules
react on solids.
So let's look at the first one, can we predict the structures
of crystals and nanoparticles?
And we're gonna start off here with a quote.
This came from an article by John Maddox
who for many years was the Editor of Nature.
He published a very provocative news and views in 1988
and he said, "One of the continuing scandals
in the physical sciences is that it remains impossible
to predict the structure of even the simplest crystalline solid
from a knowledge of his composition."
So I'm saying, if you know that atoms are present in a solid,
you couldn't predict their structure.
>> Now, when that statement was made it actually was pretty
wildly inaccurate.
So it wasn't fairly good science.
It was fantastic journalism though
because he stimulated a huge response.
And in fact, he was really very good
because he stimulated the feel.
One of the earlier responses was a review by David Price
and myself in Nature in 1990
in which we partly answered this provocative comment of Maddox
and then a couple of years ago, Scott Woodley and myself kind
of looked at where we were following the Maddox challenge
20 years on.
And the onset, the position is now.
We actually can do a pretty good job in many cases
of predicting the structures of crystals.
I'm just gonna give you 1 or 2 illustrations.
I'm gonna give you some of the ideas that we try and use
in predicting structures.
The problem is, crystals, complex systems,
they've got lots of atoms and there are vast number of ways
of arranging them, how are we gonna decide what the best
one is.
So, one approach we can use is what we call genetic algorithms.
It's a really clever, neat idea and it's based like lots
of scientific ideas are on analogies.
It's analogy with evolutionary theories.
What you say is you've got a population, you got lots
of different ways of arranging atoms
and that's a kind of population.
And then you allow these populations to pass
through successive generations.
But the possibility of a structure procreating,
passing on it features to the next generations depends
on some rough and ready measure of how good that structure is.
And that's one of the key features.
But I say it's a really neat idea
and in a minute I'll show you that it works
and I'm gonna apply this to a copy material
that I've been interested in for many years.
These are systems called zeolites.
They are silicate so the basic topology is simple,
the build of other tetrahedra,
you can see there are silicon surrounded
by [inaudible] in atom.
Then you link the tetrahedra together by sharing corners
and then you start to buildup these lovely networks,
but the important thing is though, they're open networks,
they can change sheet and pages and channels.
So there are lots of different structures here.
We're interested in them because of their beautiful
crystal architectures.
But there's a great deal we can do with them.
They are fantastic catalysts.
That means they promote chemical reactions.
I can't go into all the catalytic properties
but they're absolutely key in the petrochemicals industry,
for instance, they break down tars into molecules
of about the right size for petrol.
So they're absolutely vital for in many ways
for energy security, lots of other catalytic properties.
They're also used in very important area in industry,
gas separation, these channels are about the same size
as many molecules so you can use them to sieve out,
separate one molecule from another
and they are an older example is ion exchange.
In fact, a lot of them are still used in detergents.
I'm gonna give you one example now.
A prediction of the structure using this genetic algorithm
method is, [inaudible] didn't really a prediction.
It's just showing that the method works
to a known structure, cellulite.
So I'll let it run again, and what's happening
in the early stages, this is the genetic algorithm phase.
It's playing around, finding better and better ways as we go
through these successive generations
of organizing the atoms then it ends up with a very good way
and then there's a kind of final push downhill in energy
and generates that structure.
So, lots other examples I can give you but I wanna move
on to another approach and this is using ideas from topology.
Probably lots of people in this audience know more
about topology than I do but topology is all about the way
in which different shapes relate to each other.
Now, crystal structures contain complex shapes and so
for a long, long time, topology has been used
to generate possible models for structures.
But this process was taken one step further in very, I think,
important work by [inaudible] and coworkers [inaudible]
in the chemistry department at UCL.
I just want to illustrate the relationship between topology
and the crystal structure.
We've got topology on the right-hand side.
You can think of it as a way of connecting vertices or a way
of connecting polyhedra.
It's also kind of mathematical construct.
You can turn that as it were into a model
of a crystal structure, you put a--
in this case a silicon atom at each of those vertices.
If [inaudible]oxygen atom along the edges,
then that is a crystal structure,
in fact it's a crystal structure of a well-known material.
Now, I can't go to any details and I say [inaudible]
and many other colleagues used this
to predict entirely new structures
and moreover structures that were calculated to be stable
and this is a beautiful example of one
of these new predicted structures.
It's a lovely topology by the 12 and 8 rings
that is predicted to be stable.
This has worked about a couple of years ago.
It remains a challenge to see if we can synthesize it.
Just a few words later on about how we might respond
to that charge.
Now, the final example, again,
work from the Chemistry Department,
this is like a Sally Price, Derek Tucker and his student,
Ashley Hugh, and they're, again, trying to predict--
they're trying to predict the structures of crystals made
up out of quite complex molecules
and they use actually a rather simple idea.
They systematically tried and packed these molecules together
in lots of different ways and they found out the best way
of doing it and they had a really fantastic success story.
About five years ago, they are very much built on that.
It concerned this material, 5-fluorouracil
and if I got time, I'll say a few more words
about that material, molecule in a few minutes.
It's an important pharmaceutical compound.
This is the crystal structure that was known
for it for many years.
You see here it's a molecule and very quite interesting ways
in which the molecules fit together.
That was the only one that was known.
They then predicted using their methods
that there was another way
in which you could pack these molecules together
in this crystal, here it is.
And they-- that haven't been seen before but then
in some excellent experimental work,
they are able to grow form too.
So that was the real prediction.
They've tried lots of different solvents for this
and they found nitromethane which is a very dry solvent,
succeeded in growing that crystal structure.
Now, if I've got a few minutes,
I'll explain how we've solved the problem
of why they need this particular solvent to grow that crystal.
Before I do that, a word about nanochemistry.
I promised you to say something about nanochemistry
so here is some of the work in my group
on zinc sulfide's important semiconductor compound
and it's got a lovely nanochemistry.
These are the structures predicted by kind of some
of the methods I've been very briefly describing for the way
in which zinc and sulfur atoms are arranged at the nanoscale
from the beautiful open structures.
They look absolutely nothing like the crystal structure
of zinc sulfide and they are open bubble like structures.
This one on the right here is absolutely amazing.
There's a big open bubble structure
and it doesn't remotely resemble the way
in which the atoms are arranged and crystalizing sulfide
that for the way in which they would be arranged
in crystalizing sulfide for a cluster
about as big the one I showed you,
that our worked showed has higher energy.
So the arrangement of atoms
at the nano level is quite different from that
which you get when you've got very large number
of atoms and crystals.
Here's an even more amazing structure.
Now you got 60 zinc, 60 sulfur atoms
and they arrange themselves in an onion-like class structure.
You've got a little cluster inside a big cluster.
So you're getting inside this nested onion structures.
Again, nothing like the arrangement
in the [inaudible] getting this one.
Now, let me go on in the remaining few minutes
to my second challenge, that is can we guide synthesis
and understand the factors that control synthesis?
One just brief illustration, and we're going back
to the zeolites, these wonderful complex crystal structures
that have all these applications in industry
when which we synthesize them is
to cut a large extent of blackout.
They contain silicon and aluminium so we chuck sources
of silicon aluminium into a kind
of synthesis brew we have in plastic soda.
And then, very interestingly,
we add organic molecules, big molecules.
And what we think happens is the zeolite, all these silicons
and aluminiums begin to link together
and they kind of crystalize.
They grow around these organic molecules that we chuck
into the synthesis [inaudible].
So, some time ago, one of my colleagues,
Darwin Lewis [phonetic], working in collaboration
with Dave Willet [phonetic], decided he would try [inaudible]
but he tried to get some guidance to, he tried
and predict a kind of organic molecule that you needed
to grow a particular zeolite structure
and he had fantastic success.
He grew molecules on the computer.
So he started off with something simple
like the template you see there.
It's just a methane molecule.
All kinds of chemical groups [inaudible] the computer starts
to fit them on.
It manipulates them and it ends up with a prediction
of a molecule that will that will sit nice
and snugly inside the target material.
>> And then you say, well,
that is probably gonna be a good template and it worked.
The very first example,
that molecule there successfully predict--
synthesized that particular cage that I'm showing you
and he was a real fantastic success.
This lovely cage here known
as Levine [phonetic] structure cage, we predicted
by this computational method that the molecule there
on the right will be really good for synthesizing and it worked.
It synthesized.
It absolutely like a dream.
So you can use computer modelling to guide synthesis.
Now, in the remaining 2 or 3 minutes, I'm just gonna look
at this third challenge, can we gain a molecular level
of understanding of crystal [inaudible]
and should say crystal [inaudible] one
of the understanding it's one of the really big challenges
in physical chemistry and chemical engineering.
Now, I'm gonna go back to the theme I had earlier,
the work of Sally Price.
But here's some contributions from [inaudible] worked with me
and they're concern with this phenomenon of polymorphism.
Now, the molecular crystals,
crystals when you pack molecules together, often can grow in lots
of different structures.
There are lots of different ways of packing them together.
Now this, in the pharmaceutical industry can be a disaster.
A pharmaceutical is a solid compound regulatory approval
only applies to one particular solid form.
And what sometimes happen during production
of a pharmaceutical is that it starts to crystalize
in a different form, not in the one
for which it's has been given regulatory approval.
So, certainly, it's been one of the really big problems
in the pharmaceutical industry.
So there's a great deal of incentive
for understanding this phenomenon.
So I'm gonna go back to this system, 5-fluorouracil,
important pharmaceutical compound, chemotherapy,
50 years, remind you
that Sally's group predicted this new form
and there are just very briefly to draw your attention to--
sorry, I better get it back--
to important structural principles here.
If you look at that thing that thing that I've ringed there,
those are fluorine atoms close together
and I just want you to remember that.
That's a feature of this form.
Look at this one, you see those dotted lines,
those are what we call hydrogen bonds
which are holding two molecules together.
So those are the kind of two structural signatures.
Now, form1, you can grow that from water.
Form 2, you only grow from nitromethane.
Now, let's look at what you do when you do a simulation.
You look at how the molecule behaves first in water.
And you can see here, this kind of greeny-blue thing,
that's the fluorine atom
and it's not interacting with the water.
And what you find is if you have species in water
but don't interact with the water, then they start
to interact with each other.
It's what we call a hydrophobic effect.
So, when you have this molecule in water,
as you allow the system to evolve,
the molecules come together but they come together
with their fluorine atoms pointing to each other.
And that's just what you see in form 1.
So what's driving the formation
of that first form is what's happening very early
on as the molecules begin to come together
and if that's driven by the way
in which the molecules interact with the solvent.
Let's look at nitromethane.
Here it is.
It actually love to interact
with the surrounding molecules a total much.
Nitromethane doesn't particularly interested
interacting with this molecule and so the molecules interact
with themselves and you can see at the top.
You see those two dotted lines bringing these molecules
together and that's just the feature that we saw in form 2
and form 2 grows out of nitromethane.
So, again, what's happening really, really early
on is driving a polymorphic outcome.
So what we're showing here is we understand
when we get these different polymorphs and they depend
on what's happening in the [inaudible].
Fortunately, I'll have to skip the last three slides.
I just wanna say that the chemistry part of this talk,
lots of different people have contributed, many thanks to them
and we've been well funded by a number of agencies.
I'll stop at that point but I hope I've given you some glimpse
into the way in which the power among the computers is--
[ Noise ]
>> -- do anything serious.
So where else in chemistry--
actually there is not enough power to--
>> Sorry, I've [inaudible].
>> You're referring to the use of a computer modeling
in biomolecular modeling.
I mean, there is a huge, I mean that's not my area
and I didn't have time to talk about it
but it is hugely important in very active area of the field.
Essentially, using the same kind of techniques that people
like me use in modeling materials,
people use those techniques in modeling complex biomolecules,
proteins, DNA, and interactions between molecules.
I mean, there are huge challenges intrinsic
in the complexity of the structures
that they're looking at.
The other-- one of the other major--
I say it's not my area of expertise but in know that one
of the other major areas of challenge
for computational modeling
of those systems is understanding the interaction
between the molecules and the solvent.
I mean, I just alluded to that topic in the end
when I was talking fluorouracil.
But it's important in certain areas in materials modeling,
absolutely crucial in biomolecular modeling
and it's not at all an easy problem.
But methods are very, very widely used
in biomolecular modeling, as widely used as they are
in the materials field that I have discussed.
So thanks for that.
>> And actually, would you mind if I continued a little
with the same question 'cause I would like you
to answer just a little bit further that one
of the main questions about biomolecular systems is
that they're often very much larger than molecules.
Where do we stand these days in terms
of actual computer power that's available
for addressing these very large scale-- multiscale problems?
>> Oh, we stand actually in pretty good shape.
I mean, one of the aspects that has made these fields
so exciting to work with in my career has been the exponential
growth in computer power which still continues.
So, you really now can track all very, very complex systems,
systems with many thousands of atoms, even millions
of atoms using the simpler [inaudible] potential approach,
but even with quantum-mechanical methods, which really get
to the heart of how the electrons behave in the system,
you can tackle large systems.
So, there's a great deal it can be done and the horizons
of the field continue to expand
because of the growth in computer power.
>> We have a question here.
Excuse me.
>> Bill [inaudible], you talked briefly about your work
with genetic algorithms and you talked--
kipped over the cost function
but I got the impression there was more you wanted
to say about that.
>> Yeah, I skip over [inaudible].
The cost, there are a number different ones you can use.
You can use very simple ones, though it's just based
in crystal structure prediction on expected coordination numbers
and bond lengths or you can use more sophisticated ones,
a very approximate estimates of the lattice cohesive energy
and we've used both in the type of work I was describing.
But I mean, you're absolutely right,
the choice of cost function
in the GA approach is absolutely key to its efficacy.
>> Okay, one last question then.
[ Noise ]
>> Would the development
of the computational models be detrimental to the development
of analytical solutions to some of these problems?
>> Well, one should use analytical solutions
when you can.
There's no conflict between computer modeling
and analytical solutions.
We use computer models when the systems are of that complexity,
which the ones I was talking about,
that one cannot use analytical models.
You may be able to use analytical models for some
of the very broad features but if you want to get
into the details, for example, if you want to do the kind
of predictive the Alpha [phonetic] did calculating those
temperatures in the earth's core, you need this kind
of sophisticated simulation method.
>> But I analytical models [inaudible] replacement
for analytical models.
Analytical models continue to be of great importance.
>> And I think we should close now
because there's a class coming in.
Let's thank Professor Catlow once more.
[ Applause ]