0:00:01.246,0:00:04.830
Computers used to be as big as a room.

0:00:04.854,0:00:06.446
But now they fit in your pocket,

0:00:06.470,0:00:07.641
on your wrist

0:00:07.665,0:00:10.984
and can even be implanted[br]inside of your body.

0:00:11.008,0:00:12.289
How cool is that?

0:00:12.809,0:00:17.146
And this has been enabled[br]by the miniaturization of transistors,

0:00:17.170,0:00:19.662
which are the tiny switches[br]in the circuits

0:00:19.686,0:00:21.462
at the heart of our computers.

0:00:22.051,0:00:25.223
And it's been achieved[br]through decades of development

0:00:25.247,0:00:28.045
and breakthroughs[br]in science and engineering

0:00:28.069,0:00:30.741
and of billions of dollars of investment.

0:00:31.352,0:00:34.100
But it's given us[br]vast amounts of computing,

0:00:34.124,0:00:35.929
huge amounts of memory

0:00:35.953,0:00:40.895
and the digital revolution[br]that we all experience and enjoy today.

0:00:41.665,0:00:44.433
But the bad news is,

0:00:44.457,0:00:47.589
we're about to hit a digital roadblock,

0:00:47.613,0:00:51.963
as the rate of miniaturization[br]of transistors is slowing down.

0:00:52.471,0:00:55.345
And this is happening[br]at exactly the same time

0:00:55.369,0:00:59.367
as our innovation in software[br]is continuing relentlessly

0:00:59.391,0:01:03.151
with artificial intelligence and big data.

0:01:03.175,0:01:08.215
And our devices regularly perform[br]facial recognition or augment our reality

0:01:08.239,0:01:12.464
or even drive cars down[br]our treacherous, chaotic roads.

0:01:12.959,0:01:14.166
It's amazing.

0:01:14.618,0:01:19.285
But if we don't keep up[br]with the appetite of our software,

0:01:19.309,0:01:23.096
we could reach a point[br]in the development of our technology

0:01:23.120,0:01:27.330
where the things that we could do[br]with software could, in fact, be limited

0:01:27.354,0:01:28.625
by our hardware.

0:01:29.075,0:01:33.583
We've all experienced the frustration[br]of an old smartphone or tablet

0:01:33.607,0:01:36.771
grinding slowly to a halt over time

0:01:36.795,0:01:40.770
under the ever-increasing weight[br]of software updates and new features.

0:01:40.794,0:01:44.177
And it worked just fine[br]when we bought it not so long ago.

0:01:44.201,0:01:48.711
But the hungry software engineers[br]have eaten up all the hardware capacity

0:01:48.735,0:01:50.041
over time.

0:01:51.883,0:01:55.495
The semiconductor industry[br]is very well aware of this

0:01:55.519,0:01:59.403
and is working on[br]all sorts of creative solutions,

0:01:59.427,0:02:03.738
such as going beyond transistors[br]to quantum computing

0:02:03.762,0:02:07.974
or even working with transistors[br]in alternative architectures

0:02:07.998,0:02:09.601
such as neural networks

0:02:09.625,0:02:12.638
to make more robust[br]and efficient circuits.

0:02:13.270,0:02:16.609
But these approaches[br]will take quite some time,

0:02:16.633,0:02:21.260
and we're really looking for a much more[br]immediate solution to this problem.

0:02:22.899,0:02:27.681
The reason why the rate of miniaturization[br]of transistors is slowing down

0:02:27.705,0:02:32.391
is due to the ever-increasing complexity[br]of the manufacturing process.

0:02:33.142,0:02:36.392
The transistor used to be[br]a big, bulky device,

0:02:36.416,0:02:39.725
until the invent of the integrated circuit

0:02:39.749,0:02:42.440
based on pure crystalline silicon wafers.

0:02:42.946,0:02:45.725
And after 50 years[br]of continuous development,

0:02:45.749,0:02:49.122
we can now achieve[br]transistor features dimensions

0:02:49.146,0:02:51.675
down to 10 nanometers.

0:02:52.361,0:02:54.798
You can fit more than[br]a billion transistors

0:02:54.822,0:02:57.785
in a single square millimeter of silicon.

0:02:58.273,0:03:00.295
And to put this into perspective:

0:03:00.319,0:03:04.145
a human hair is 100 microns across.

0:03:04.169,0:03:06.688
A red blood cell,[br]which is essentially invisible,

0:03:06.712,0:03:08.311
is eight microns across,

0:03:08.335,0:03:11.735
and you can place 12 across[br]the width of a human hair.

0:03:12.467,0:03:15.567
But a transistor, in comparison,[br]is much smaller,

0:03:15.591,0:03:19.439
at a tiny fraction of a micron across.

0:03:19.463,0:03:23.009
You could place more than 260 transistors

0:03:23.033,0:03:25.011
across a single red blood cell

0:03:25.035,0:03:29.499
or more than 3,000 across[br]the width of a human hair.

0:03:29.523,0:03:33.847
It really is incredible nanotechnology[br]in your pocket right now.

0:03:35.204,0:03:37.392
And besides the obvious benefit

0:03:37.416,0:03:41.250
of being able to place more,[br]smaller transistors on a chip,

0:03:41.984,0:03:45.476
smaller transistors are faster switches,

0:03:46.166,0:03:50.567
and smaller transistors are also[br]more efficient switches.

0:03:50.591,0:03:53.068
So this combination has given us

0:03:53.092,0:03:57.391
lower cost, higher performance[br]and higher efficiency electronics

0:03:57.415,0:03:59.478
that we all enjoy today.

0:04:02.415,0:04:05.179
To manufacture these integrated circuits,

0:04:05.203,0:04:08.411
the transistors are built up[br]layer by layer,

0:04:08.435,0:04:10.788
on a pure crystalline silicon wafer.

0:04:11.332,0:04:13.560
And in an oversimplified sense,

0:04:13.584,0:04:17.865
every tiny feature[br]of the circuit is projected

0:04:17.889,0:04:20.221
onto the surface of the silicon wafer

0:04:20.245,0:04:23.924
and recorded in a light-sensitive material

0:04:23.948,0:04:26.887
and then etched through[br]the light-sensitive material

0:04:26.911,0:04:29.932
to leave the pattern[br]in the underlying layers.

0:04:30.612,0:04:34.696
And this process has been[br]dramatically improved over the years

0:04:34.720,0:04:37.493
to give the electronics[br]performance we have today.

0:04:38.279,0:04:41.721
But as the transistor features[br]get smaller and smaller,

0:04:41.745,0:04:44.782
we're really approaching[br]the physical limitations

0:04:44.806,0:04:46.689
of this manufacturing technique.

0:04:48.515,0:04:51.620
The latest systems[br]for doing this patterning

0:04:51.644,0:04:53.947
have become so complex

0:04:53.971,0:04:58.701
that they reportedly cost[br]more than 100 million dollars each.

0:04:58.725,0:05:03.012
And semiconductor factories[br]contain dozens of these machines.

0:05:03.036,0:05:07.462
So people are seriously questioning:[br]Is this approach long-term viable?

0:05:08.441,0:05:12.121
But we believe we can do[br]this chip manufacturing

0:05:12.145,0:05:16.168
in a totally different[br]and much more cost-effective way

0:05:16.966,0:05:20.939
using molecular engineering[br]and mimicking nature

0:05:20.963,0:05:24.576
down at the nanoscale dimensions[br]of our transistors.

0:05:25.267,0:05:29.928
As I said, the conventional manufacturing[br]takes every tiny feature of the circuit

0:05:29.952,0:05:32.076
and projects it onto the silicon.

0:05:32.818,0:05:35.562
But if you look at the structure[br]of an integrated circuit,

0:05:35.586,0:05:37.560
the transistor arrays,

0:05:37.584,0:05:41.213
many of the features are repeated[br]millions of times.

0:05:41.237,0:05:43.845
It's a highly periodic structure.

0:05:44.331,0:05:47.399
So we want to take advantage[br]of this periodicity

0:05:47.423,0:05:50.120
in our alternative[br]manufacturing technique.

0:05:50.144,0:05:53.579
We want to use self-assembling materials

0:05:53.603,0:05:56.580
to naturally form the periodic structures

0:05:56.604,0:05:58.987
that we need for our transistors.

0:06:00.052,0:06:02.194
We do this with the materials,

0:06:02.218,0:06:05.655
then the materials do the hard work[br]of the fine patterning,

0:06:05.679,0:06:10.538
rather than pushing the projection[br]technology to its limits and beyond.

0:06:11.909,0:06:15.808
Self-assembly is seen in nature[br]in many different places,

0:06:15.832,0:06:19.242
from lipid membranes to cell structures,

0:06:19.266,0:06:22.321
so we do know it can be a robust solution.

0:06:22.345,0:06:25.906
If it's good enough for nature,[br]it should be good enough for us.

0:06:26.549,0:06:31.349
So we want to take this naturally[br]occurring, robust self-assembly

0:06:31.373,0:06:35.338
and use it for the manufacturing[br]of our semiconductor technology.

0:06:36.929,0:06:39.544
One type of self-assemble material --

0:06:40.388,0:06:42.635
it's called a block co-polymer --

0:06:42.659,0:06:47.442
consists of two polymer chains[br]just a few tens of nanometers in length.

0:06:47.466,0:06:49.517
But these chains hate each other.

0:06:49.541,0:06:51.025
They repel each other,

0:06:51.049,0:06:54.946
very much like oil and water[br]or my teenage son and daughter.

0:06:54.970,0:06:56.327
(Laughter)

0:06:56.351,0:06:59.125
But we cruelly bond them together,

0:06:59.149,0:07:01.844
creating an inbuilt[br]frustration in the system,

0:07:01.868,0:07:04.074
as they try to separate from each other.

0:07:04.716,0:07:08.001
And in the bulk material,[br]there are billions of these,

0:07:08.025,0:07:11.326
and the similar components[br]try to stick together,

0:07:11.350,0:07:14.159
and the opposing components[br]try to separate from each other

0:07:14.183,0:07:15.338
at the same time.

0:07:15.362,0:07:19.116
And this has a built-in frustration,[br]a tension in the system.

0:07:19.140,0:07:23.449
So it moves around, it squirms[br]until a shape is formed.

0:07:24.209,0:07:28.257
And the natural self-assembled shape[br]that is formed is nanoscale,

0:07:28.281,0:07:32.008
it's regular, it's periodic,[br]and it's long range,

0:07:32.032,0:07:35.890
which is exactly what we need[br]for our transistor arrays.

0:07:37.347,0:07:39.878
So we can use molecular engineering

0:07:39.902,0:07:42.966
to design different shapes[br]of different sizes

0:07:42.990,0:07:45.053
and of different periodicities.

0:07:45.077,0:07:47.808
So for example, if we take[br]a symmetrical molecule,

0:07:47.832,0:07:50.907
where the two polymer chains[br]are similar length,

0:07:50.931,0:07:53.602
the natural self-assembled[br]structure that is formed

0:07:53.626,0:07:56.555
is a long, meandering line,

0:07:56.579,0:07:58.389
very much like a fingerprint.

0:07:58.951,0:08:01.273
And the width of the fingerprint lines

0:08:01.297,0:08:03.307
and the distance between them

0:08:03.331,0:08:07.242
is determined by the lengths[br]of our polymer chains

0:08:07.266,0:08:10.560
but also the level of built-in[br]frustration in the system.

0:08:11.320,0:08:13.878
And we can even create[br]more elaborate structures

0:08:15.487,0:08:17.926
if we use unsymmetrical molecules,

0:08:18.839,0:08:22.924
where one polymer chain[br]is significantly shorter than the other.

0:08:23.749,0:08:26.459
And the self-assembled structure[br]that forms in this case

0:08:26.483,0:08:30.283
is with the shorter chains[br]forming a tight ball in the middle,

0:08:30.307,0:08:34.148
and it's surrounded by the longer,[br]opposing polymer chains,

0:08:34.172,0:08:36.220
forming a natural cylinder.

0:08:37.089,0:08:39.164
And the size of this cylinder

0:08:39.188,0:08:42.603
and the distance between[br]the cylinders, the periodicity,

0:08:42.627,0:08:46.221
is again determined by how long[br]we make the polymer chains

0:08:46.245,0:08:48.983
and the level of built-in frustration.

0:08:49.896,0:08:53.774
So in other words, we're using[br]molecular engineering

0:08:53.798,0:08:56.623
to self-assemble nanoscale structures

0:08:56.647,0:09:01.557
that can be lines or cylinders[br]the size and periodicity of our design.

0:09:02.369,0:09:05.666
We're using chemistry,[br]chemical engineering,

0:09:05.690,0:09:10.479
to manufacture the nanoscale features[br]that we need for our transistors.

0:09:13.611,0:09:17.660
But the ability[br]to self-assemble these structures

0:09:17.684,0:09:20.121
only takes us half of the way,

0:09:20.145,0:09:22.954
because we still need[br]to position these structures

0:09:22.978,0:09:26.528
where we want the transistors[br]in the integrated circuit.

0:09:27.246,0:09:29.984
But we can do this relatively easily

0:09:30.008,0:09:36.985
using wide guide structures that pin down[br]the self-assembled structures,

0:09:37.009,0:09:38.930
anchoring them in place

0:09:38.954,0:09:41.801
and forcing the rest[br]of the self-assembled structures

0:09:41.825,0:09:43.175
to lie parallel,

0:09:43.199,0:09:45.599
aligned with our guide structure.

0:09:46.510,0:09:51.149
For example, if we want to make[br]a fine, 40-nanometer line,

0:09:51.173,0:09:55.311
which is very difficult to manufacture[br]with conventional projection technology,

0:09:56.274,0:10:01.059
we can manufacture[br]a 120-nanometer guide structure

0:10:01.083,0:10:03.587
with normal projection technology,

0:10:03.611,0:10:10.202
and this structure will align three[br]of the 40-nanometer lines in between.

0:10:10.226,0:10:14.995
So the materials are doing[br]the most difficult fine patterning.

0:10:15.790,0:10:19.697
And we call this whole approach[br]"directed self-assembly."

0:10:21.586,0:10:24.340
The challenge with directed self-assembly

0:10:24.364,0:10:28.840
is that the whole system[br]needs to align almost perfectly,

0:10:28.864,0:10:34.145
because any tiny defect in the structure[br]could cause a transistor failure.

0:10:34.169,0:10:37.138
And because there are billions[br]of transistors in our circuit,

0:10:37.162,0:10:40.390
we need an almost[br]molecularly perfect system.

0:10:40.977,0:10:42.982
But we're going to extraordinary measures

0:10:43.006,0:10:44.173
to achieve this,

0:10:44.197,0:10:47.189
from the cleanliness of our chemistry

0:10:47.213,0:10:49.539
to the careful processing[br]of these materials

0:10:49.563,0:10:51.134
in the semiconductor factory

0:10:51.158,0:10:55.730
to remove even the smallest[br]nanoscopic defects.

0:10:57.311,0:11:02.501
So directed self-assembly[br]is an exciting new disruptive technology,

0:11:02.525,0:11:05.094
but it is still in the development stage.

0:11:05.680,0:11:09.541
But we're growing in confidence[br]that we could, in fact, introduce it

0:11:09.565,0:11:11.252
to the semiconductor industry

0:11:11.276,0:11:14.233
as a revolutionary new[br]manufacturing process

0:11:14.257,0:11:16.324
in just the next few years.

0:11:17.014,0:11:20.048
And if we can do this,[br]if we're successful,

0:11:20.072,0:11:21.603
we'll be able to continue

0:11:21.627,0:11:24.885
with the cost-effective[br]miniaturization of transistors,

0:11:24.909,0:11:28.662
continue with the spectacular[br]expansion of computing

0:11:28.686,0:11:30.568
and the digital revolution.

0:11:30.592,0:11:34.137
And what's more, this could even[br]be the dawn of a new era

0:11:34.161,0:11:36.392
of molecular manufacturing.

0:11:36.416,0:11:37.947
How cool is that?

0:11:38.519,0:11:39.677
Thank you.

0:11:39.701,0:11:43.910
(Applause)