Illuminating the Past, Present, and Future of Photonics

Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
Speaker 1 (00:02):
Illuminated by IEEE.
Photonics is a podcast seriesthat shines light on the hot
topics in photonics and thesubject matter experts advancing
technology forward.
Hi everyone and welcome totoday's episode of Illuminated.
I'm Akhil and, as the pastAssociate Vice President of the
Young Professionals, it's mypleasure to be your host today.

(00:25):
I'm a biomedical physicist andengineer working at the
University of Strathclyde as aChancellor's Fellow and a Levy
Hume Early Career Fellow.
In my role for the IEEEPhotonics Society, I support and
promote initiatives much likethis podcast to raise the
profile of valuable youngprofessionals within various
sectors, of valuable youngprofessionals within various
sectors.
The Young ProfessionalsInitiative is for graduate

(00:46):
students, postdoctoralcandidates and early career
professionals up to 15 yearsafter their first degree.
This affinity group within theExoPre Photonics Society is
committed to helping one pursuea career in photonics.
We're here to help evaluateyour career goals, better
understand technical pathwaysand subject matters, refine
skills and grow professionalnetworks through mentorship.

(01:09):
It is very interesting to havethis conversation today as the
Photonics Society turns 60 in2025.
Includes six decades ofinnovation, growth and
leadership in the field ofphotonics, making a journey from

(01:31):
its origins in quantumelectronics to becoming a global
leader in photonics technology.
On to our podcast.
In this podcast we're chattingwith Professor Sir David Payne.
We'll reflect, along with yourhost, dr Richard Pitwin, on the
last 50 years of opticalinterconnects, fiber
technologies, opticalcommunications.
We'll look at the evolution ofECOC as a conference that it is

(01:51):
today.
We'll talk about Sir David'sexperiences as a knight
experiences outside sciencethroughout his career.
Let me introduce Richard first.
Dr Richard Pitwin is a director, scientist and engineer who has
been instrumental over the past25 years, mostly at Seagate, in
the development of system-leveloptical and photonic

(02:13):
technologies, in particular, forhyperscale data center
environments, which are thebedrock of modern AI
infrastructures.
Today he is the CEO of ResolutePhotonics, which he founded in
2018, developing specializedphotonic integrated circuits as
subsystems for communicationsensors and quantum and space

(02:34):
applications.
He's also a chartered engineerfellow of the Institute of
Physics and the IET and has over55 patents on a broad range of
technologies and the IET and hasover 55 patents on a broad
range of technologies.
He's the current chair of theIEEE UK and Ireland Photonics
Chapter and also the chair formany international standard
committees in the BSI and theIEC on fiber optics and

(02:56):
connectors.
He recently established and nowchairs a new committee on
optical interconnects, onquantum optical interconnects in
the IEC, which will develop thestandards for the next
generation of very sensitivefiber optic connectors and
extremely low-loss opticalfibers.
Over to you, richard.

Speaker 3 (03:16):
Thank you very much, akhil.
So it's my great pleasure tointroduce Professor David Payne,
who I've had the privilege ofknowing for over 15 years.
Now.
This is my first podcast and tohelp me frame my introduction,
I read David's online bio, andit starts with the sentence Sir
David is an internationallydistinguished research pioneer
in photonics, having been in thefield for over 50 years.

(03:38):
Well, I'd say that this isprobably one of the great
understatements of the last 50years.
Now this year, we celebrated the60th anniversary of the
internet, and I mentioned thisbecause outside the photonics
community, not many people willhave heard of David Payne and
thus not many people will knowthat without his work and the
work of his many colleagues overthe last 50 years, there would

(04:01):
simply be no internet.
There would be notelecommunication as we know it,
no means to convey the vastamounts of information over long
distances that we takecompletely for granted today,
and this is not an exaggeration,and the reason for this is
optical fibers.
Optical fiber technology is oneof the greatest scientific
successes of the last sixdecades and, as we'll be hearing

(04:22):
today, david Payne's role inthe formation and evolution of
optical fiber technology hasbeen indispensable.
David's pioneering work inoptical fiber fabrication in the
70s resulted in almost all thespecial fibers we use today.
He led the team that, in 1985,first announced the silica fiber
laser and the erbium dopedfiber amplifier, edfa.

(04:42):
And the EDFA is widely regardedas one of the foremost and most
significant developments inmodern telecommunications, and
that work, that invention,singularly fueled the explosive
growth in the Internet byenabling the transmission and
amplification of vast amounts ofinformation.
Now I won't list all the awardsbecause there simply isn't

(05:03):
enough time, but I must mentionthat in 2013, he was knighted
for services to photonics in theQueen's New Year's Honours list
and became Sir David Payne.
And in 2018, with hiscolleagues from ORC, he won the
Queen's Anniversary Prize thatcelebrates excellence,
innovation and public benefit inwork carried out by UK
universities benefit in workcarried out by UK universities.

(05:24):
Now the theme of this podcast ispast, present and future, with
focus on the anniversary of theinternet, and I think we only
have 90 minutes and that'sbarely enough time to scratch
the surface, especially withsomeone like yourself, david,
but I'm going to try Now.
Along with your amazingcolleagues at the ORC in

(05:47):
Southampton, you're responsiblefor most of the key technical
achievements which underpin theold, current and future internet
.
And these achievements spandiverse areas of photonics which
we'll touch upon in thispodcast, from telecommunications
that's your seminal work inoptical fiber, communication
sensors and optical materials,to nanophotonics, in particular
silicon photonic integratedcircuits which now form the

(06:07):
basis of all modern opticaltransceivers.
Optical storage, for example,the famous 5D optical storage in
glass, which can storeinformation until, I think, the
end of the universe.
I hope we talk about that aswell.
All of these form part of avast ecosystem which underpins
the modern internet.
So let's start at the verybeginning, and sorry for the

(06:34):
boring question to start you off, but when did you first start
working on optical fibers?

Speaker 2 (06:36):
Well, thank you, richard, for that very generous
introduction, and I'm reallyexcited about this opportunity
to provide my thoughts on past,present and future.
Starting with the past, Ibelieve I was actually the very
first PhD student on opticaltelecommunications in the wake

(07:02):
of the famous Charles Cowellpublication.
So I started my PhD in 1967under the great Professor Alec
Gambling, and he came to me andhe said, well, why don't you do
a PhD?
And I said, well, I'm notreally planning on doing that.

(07:27):
I want to go into industry anddo some real stuff.
And he said well, you know,we've got this really
interesting project where wewant to create optical fibers
that will span the globe, andthis intrigued me.
So I said, well, sounds aninteresting project.
This intrigued me.
I said, well, sounds aninteresting project.

(07:49):
How far would you like to go?
He said maybe from Southamptonto London.
I said, okay, that's about 100kilometers.
How far can we go at the moment?
He said maybe about a meter.
So this was a challenge whichonly the arrogance of youth

(08:17):
would step up to, and indeed,within a few years at
Southampton, with our makeshiftfiber drawing machines and so on
, we had actually got theworld's record of the lowest
loss fiber ever since, and Ifound myself in the midst of
this maelstrom of developmentand excitement as we fiberized
up the entire world to createwhat today is called the

(08:39):
internet.
So that's how it all started.

Speaker 3 (08:43):
Well, you mentioned Sir Charles Cowell.
Charles Cowell, I believe, ishis seminal paper in 1964.
He was he's basically, I thinkit's safe to say he's the father
of optical fiber technologiesand I think you knew him very,
very well.

Speaker 2 (08:57):
Yes, that's right.
I was extremely privileged toknow Charlie extremely well,
partly because, of course, hiswork was done here in the UK, at
Harlow, essex, which wasstandard telecommunications
laboratories there, and he didhis seminal work there

(09:21):
describing how we could usesilica as a means of
communication.
And frankly, everybody thoughthe was crazy at the time,
because everybody else in theworld was working on microwave
guided waves, the H01 wave guide, for example, in copper and

(09:43):
Bell Labs.
The great Bell Labs, whichusually led the world, was
actually working on a lenssystem, periodic lenses every
few kilometers, refocusing alight beam.
And Charles said no, no, no, wecan do this with silica.
And you've got to remember thatin those days one had no idea of

(10:06):
the loss of glasses Becausenobody had ever measured them at
the ultimate limit of how lowloss they could be.
And take an example the averagewindow would only allow
something like 90% of the lightthrough it, so you're losing 10%
just to the thickness of awindow.

(10:27):
And what Charles proceeded todo was to select silica which
was an amazing precedent thoughtand then proceed to show that
it had losses below 20 dBs perkilometer, which, when you think
about even today, is anincredibly challenging thing to

(10:50):
measure in the lengths available, which were maybe 10
centimeters or something likethat.
So this is why Charles got theNobel Prize.
But also he was a greatadvocate for optical
telecommunications and he touredthe world telling everybody

(11:12):
that this was the way to go.
And everybody said no, no, no,charles, that's never going to
happen.
But look where we are today.
We have 1 billion kilometers ofsilica fiber installed.

Speaker 3 (11:25):
Actually, I looked this up this morning and it
turns out it's more than 4billion kilometers of silica
fiber installed.
Actually, I looked this up thismorning and it turns out it's
more than 4 billion kilometers.
So it really scales up reallyquickly.
And that is the distance to thesun and back 22 times.
And that's the distance toNeptune, our furthest planet.
So that shows you theincredible impact of optical

(11:45):
fiber technology.
You mentioned that.
Oh, go on, sorry.

Speaker 2 (11:51):
RAOUL PAL.
Yeah, interesting that thosenumbers Corning Glass and YOC in
China both celebrated this yeartheir one billionth kilometer
sold.
So that kind of adds up becausethose two companies probably
have about half the world'smarket and the current

(12:17):
installation rate is at abouthalf a billion kilometers per
year.
So this is just an amazing,amazing achievement.

Speaker 3 (12:23):
You mentioned losses, Of course.
One of your most notableachievements is the Erbium dope
fiber amplifier.
Can you tell us what led tothat breakthrough?

Speaker 2 (12:39):
Yes, one of the things that has characterized
our work at Southampton is thatwe are very aware of what it is
that the market needs.
A number of people haveremarked that that's quite rare
for a university.
Universities tend to do themore fundamental research
without paying too muchattention to the marketplace.
It became very apparent in theearly 80s that we could get

(13:06):
through about 100 kilometers offiber before the light dimmed
and there was insufficient lightto overcome the noise of the
detector.
What you needed to be able todo was to boost the light back
up again, because thealternative was to detect the
light and use an electronicamplifier and then retransmit

(13:30):
the light.
It's not rocket science tofigure out.
Well, the problem with thatapproach is, first of all, the
electronics is nowhere near asfast as the optics and secondly,
electronics detectors cannotdistinguish between different
colors of light, and so itscrambles them all up, and it

(13:53):
therefore was apparent worldwidethat we needed an amplifier.
So a lot of people started towork on things like
semiconductor amplifiers, onRaman amplification, and we
started to think about thesematters as well, and it's
interesting to note that weactually came out of a laser

(14:17):
group at Southampton, so we werelate to telecommunications, but
our roots was in lasers.
So we were late totelecommunications, but our
roots was in lasers, and so itdidn't take us very long to
think about.
Well, you know, a laser is anamplifier with feedback around
it, so why don't we think aboutdoing that?
And we started to put rareearths.

(14:38):
I have to confess that at thetime we were driven not just by
amplification but by sensors,and a lot of our early work was
on making a distributed fibersensor, which was also another
Southampton invention, usingrare earth doped optical fibers.

(14:59):
And then we hit on the ideathat we could make erbium lasers
, and it's a little anecdotethat I don't often tell we were
actually making lasers becausethat was our roots.
Right, we wanted to make lasers.

(15:19):
And one day the PhD student,robert Me, took the mirrors off
the laser, the erbium-dopedlaser and realized that the gain
was in excess of 20 dBs.
And we all got very excited andjumped up and down about this
and said well, you know whatthis is going to be an amplifier

(15:41):
then?
And it took a while for us tofully sink in because all of the
other attributes of a fiberamplifier, compared to the
competing uh diode uh amplifiersand and the raman, which are
the very slow dynamic gain,meaning that you pump the thing

(16:05):
up and then the signal that iscoming through it doesn't change
the gain very much, and inthese other devices you get a
modulation of the gain dependingupon the magnitude of the
signal.
That's obviously undesirable.
It tends to mix up the channels, and so it was also realized
that its polarization inagnostic.
It tends to mix up the channels.
It was also realized that it'spolarization in agnostic and it

(16:31):
was extremely efficient and itrequired a very small pump, and
so on and so on.
The rest is history.
But the anecdote I was about totell us that we published 27
publications on erbium dopedfibers before we realized it was
going to be an amazingamplifier and we announced that

(16:54):
in 1987, january at OFC, etchedin my mind and of course, course
, the world largely said nah,it's never going to work.
Because at the time it waspumped by 100 kilowatts of argon
laser, because that's the onlysource we had and Bell Labs.

(17:17):
My old friend Emmanuel deSevier followed immediately
afterwards and because he hadthe fibers and he had the laser
as well in his labs and equally,his took 100 kilowatts.
But the next big step was tomake it diode pumped and

(17:40):
actually it was BT that did thefirst diode pumping and that's
not widely recognized, largelybecause it was not very
efficient and it was done at 800nanometers before the Japanese
came in with 1480 pumping.

(18:11):
With 1480 pumping, but it was980 nanometers and it was
Southampton, in collaborationwith Bell Labs, that did that
first and pointed out that thatwas the way to go, and the world
today uses 980 nanometerpumping and Obeam dope fibers in
, and even today, after morethan 30 years, there's no
serious competitor to theObeam-Tope fiber amplifier, the

(18:33):
basis of the internet, as peoplehave often said.
You know what?
What it actually did was itmade us globalize networks.
Networks up to then, because wecould only get through 100
kilometers or so, were small.
I often joke, that's whythey're called local area
networks.

(18:54):
And then it became WANs widearea networks, because we now
have amplification and we couldget across an ocean.
So it became one network of aglobal size and maybe the next
big thing is to make itintergalactic, but I don't think
we're going to use fibers forthat.

Speaker 3 (19:17):
Well, funnily enough, there is a call now for low
Earth orbit.
Satellites are so ubiquitousnow for communication and they
need such high bandwidth withinthem.
There is the emerging need foroptical transceivers for
communication within satellitesthemselves.
So transceivers and fiberswithin the enclosures of
satellites simply to accommodateall these high bandwidths of

(19:40):
information.
But these transceivers need tobe completely space resilient.
They have to survive cosmicradiation, high energy particles
, and fibers, of course, arefantastic for that.
Fibers are very radiation hard.
The weak points of thetransceivers and the lasers
there which, if you have quantumwell lasers, they can be
destroyed by a well-placed highenergy particle and so on, and

(20:02):
that's why we're seeinginnovations such as quantum dot
lasers.
I think where you get theseredundant micrometers Space is
now.
It is certainly the nextfrontier for optical
communication.
Peter DIAMANDIS, absolutely.

Speaker 2 (20:15):
With the networks of communication satellites that
are the next big thing.
Inter-satellite communicationsusing optics is, of course,
well-known because it gives youthe bandwidth.
But you still can't do theup-and-down link with optics

(20:36):
easily because of clouds and soon.
But maybe one day we'll figureout how to do that.

Speaker 3 (20:41):
Yes, now Erbium.
So EDFAs they amplify, but theyamplify a certain range of
wavelengths, I think in the Cband I think around 1550, if I'm
not mistaken, but I thinkyou've been looking.
Is there work to amplify otherwavelength ranges as well, so we
can tap into more of thespectrum?

Speaker 2 (21:04):
Absolutely, despite, as we have observed earlier, the
enormous amount of fiber that'sbeen installed and its
incredible bandwidth per fiber.
We're forever chasing more andmore bandwidth, because, as fast
as we put in fiber to satisfythe demand, somebody comes up

(21:25):
with a new bandwidth hungryapplication, Of course, the
latest one being artificialintelligence AI, which is
scraping all the servers allover the world to to gather its
information, and this is reallydriving a bandwidth demand again
, really driving a bandwidthdemand again.

(21:46):
One of the ways that you canincrease the available bandwidth
is to use different wavelengthwindows.
Now, in conventional fibers,this is not that attractive an
approach, because the loss is byfar the lowest at 50 nanometers
, the wavelength of operationthat you mentioned and so you go

(22:07):
away from that.
Further into the infrared orshorter wavelengths towards the
visible, the loss of the fibergoes screaming up.
And, yes, it's OK for shortlinks, maybe in data centers,
but not for telecommunications.
However and I'm sure we're goingto talk about this in a little

(22:32):
while, the new holocore fiberswhich have just appeared have
become wavelength agnostic, andso now you can make fibers which
are pitched to much shorter orlonger wavelengths, and this has
opened up and as I often calledthe tyranny of 1550, we no

(22:57):
longer have it.
You know, we can think atdifferent wavelengths, we can
think about red wavelengths and,you know, maybe even eventually
down into the blue.
But, as you rightly point out,that requires wavelengths for
amplifiers, different wavelengthamplifiers, and it turns out

(23:18):
that the erbium-doped fiberamplifier, although its primary
attribute is that it actuallyworks at the wavelength window
of lowest loss in silica, ismatched by a number of other
wavelength amplifiers, such asthe thulium-doped amplifier and

(23:40):
there are several others, theholmium-doped fiber amplifier,
and they work extremely well andsome would say even better than
the erbium-doped fiberamplifier.
So we do have fiber amplifiersto match the wavelengths which
are emerging with hollow corefibers, raoul, pal, we'll
definitely talk about hollowcore fibers later, because

(24:01):
that's the creme de la creme 's.

Speaker 3 (24:05):
It's, uh, uh, hugely interesting.
It's one of the rare situationswhere I where I actually got
goosebumps earlier this year atoc when that, that, that recent
breakthrough, was announced byfrancesca paletti.
Um, but one, one smalldigression, since we're talking
about amplification, uh, one ofyour, uh, one thing you invented
was the single mode silicafiber laser.

(24:25):
This is, I think, not forcommunication, this is for
conveying very, very highoptical powers.
I think you broke the kilowattbarrier many years ago.
I was wondering how much powercan we send over fiber today?

Speaker 2 (24:39):
I think the world record, which I think was IPG
company, is in the region of 15kilowatts per fiber and that's
in a solid core fiber, of course, because fiber lasers tend to
be solid core fiber, because youneed somewhere to put the

(25:03):
dopant, of course, and that's inthe core.
So I think the answer is inround terms about 10 kilowatts
per fiber, which is a staggeringnumber Because actually, when I
first started working on this,I looked up the books, of course

(25:23):
, and textbooks all said thatthat should be closer to a few
hundred watts.
But the textbooks were wrongand you know they were an order
of magnitude above what thetextbooks would tend to tell you
about the breakdown of silicaglass itself with power.

Speaker 3 (25:46):
It's extraordinary.
What is the limit?
How much power would?
At which point would a singlefiber melt?
Peter BAKER.

Speaker 2 (25:53):
Well, it tends to.
It's not so much melting,although it can be.
Melting, comes from heat, ofcourse, and you don't get much
more heat resistance than silica.
It tends to be breakdown.
Also, it tends to be at thesplices, because I've often

(26:14):
remarked that a fiber lasercompany is actually a fiber
splicing company, because it'sthe fiber splices which are the
hard part of the whole thing.
Of course, if you get a littlebit of leakage in a 10 kilowatt
fiber laser of just a fractionof a dB, that's an awful lot of
watts at that point.
So, yeah, this brings us back,of course, to the holocore fiber

(26:40):
, because we've been able todemonstrate multi-kilowatts of
transmission of power in ahollow core fiber as well, and
that's got a lot of people'sinterest because it means that
potentially, you could deliveran awful lot of power over, say,
100 kilometers and powersomething I mean it's not an

(27:04):
alternative to power lines, ofcourse not, but there are
applications where you mightwant to remote power, for
example, an amplifier or asensor or something like that on
an island or under the sea.
It's opening up thatpossibility as well.

Speaker 3 (27:24):
Now we've mentioned holocore fiber a few times.
It is an incredible moderninnovation for a reflection of
the future of the internet.
I know there's been thetechnology the photonic bandgap
fibers were invented a very,very long time ago, but only

(27:45):
very recently.
These new designs by ORC, theseNAMF fibers, these DNAMF fibers
, have allowed incredibly lowloss, lower loss than the lowest
solid core fiber, which is anincredible milestone.
I think that was reportedearlier this year.
Please tell us about how ORCcame to develop these fibers.

Speaker 2 (28:10):
Well, the history of that is interesting because it
stems back to the time whenPhilip Russell who's widely
regarded as a pioneer of theholocore fiber, in particular
the photonic bandgap fiber wasat Southampton, and so the first

(28:31):
fibers were made at Southampton, and he then moved to Bath
where he did amazing work,developing different designs,
different designs, but the realbreakthrough came with the
so-called you mentioned italready NANF, the anti-resonant
design, which are nested tubesinside the core of a fiber, and

(29:00):
these, apart from being muchsimpler to make, have the
amazing characteristic that thelight is expelled out of the
glass by the anti-resonancethat's why they're called
anti-resonant and so very smallamounts of light are present in
the glass structure.

(29:20):
The glass structure which isperforming the, the guidance,
like, uh.
My colleague francescofrancesco um poletti sometimes
describes this as the bars of acage which are holding the
gorilla in and, uh, you know,they.
That means that all power iscontained in the quite large

(29:44):
core, and this means that youbecome independent of the
structure of the fiber and youeffectively potentially got the
losses of vacuum, which, ofcourse, are almost negligible,
if not negligible.
So this is what led to thebreakthrough, and the design of

(30:04):
those fibers is what Sal Hansenhas pioneered in, with David
Richardson, my colleague,leading a big group which was
funded by EPSRC, and they formeda company called Luminicity as

(30:42):
well, which was you what theimportance that they put on the
invention of these ultra lowloss fibers?
The figure the current figureis the lowest loss that has been
measured is below 0.1 dB perkilometer and people are

(31:03):
projecting that that should getdown to about 0.05 dBs per
kilometer.
If you theoretically add up allthe known loss mechanisms, when
you think about it, that wouldallow 1,000 kilometers without
an amplifier.
I think my ambition is toeliminate the EDFA.

Speaker 3 (31:28):
One of the great other characteristics of hollow
core fiber is, of course, thespeed of light.
Everyone says light is so fast,but in a solid core fiber light
is traveling at two thirds thespeed of light in the vacuum.
But in a hollow core fiberyou're traveling at the speed of
light in a vacuum.
But in a hollow core fiberyou're traveling at the speed of
light in a vacuum.

(31:48):
So that means light takes 50%down.
Your latency is reduced byabout 50%.
And I remember working in theHypescale industry.
We were developing system leveloptical interconnect devices and
we were toying with the side.
We had hundreds of meters offiber and there are various
protocols which have what arecalled time of flight
restrictions.
So you need a certain tiny,minuscule amount of time for

(32:12):
something to be transferred toanother node, another CPU,
before other problems come intoplay.
So even sending light overhundreds of meters of
solar-cooled fiber can belimiting.
But the ability to send it muchfaster over solar-cooled fiber
means you can expand thedistance between these different

(32:34):
nodes, and that might be one ofthe reasons that Microsoft has
such an interest in this lowlatency.

Speaker 2 (32:41):
No, you're absolutely correct, raoul PAL.
No, you're absolutely correct.
Putting it crudely, it's just,the refractive index of air or
vacuum is lower than that ofglass and latency has come to.

(33:06):
We are at a point, amazinglyenough, where the size of the
world has become a problem.
It's just, it takes time.
Even the speed of light it'stoo slow, and there are people
that care about these, and youmentioned the data centers and
so on, where it's criticallyimportant, but even things like
gamers, it really is important.

(33:27):
You shrink the world withhollow core fiber.
Everything seems a littlecloser than it does otherwise.
That's important and it'simportant to people like traders
, because automated computertrading it depends upon how

(33:47):
close you are to the stockexchange.
The initial sales of the fiberhave been very much into
interconnecting the traders tothe stock exchanges because that
brings them closer, which is alot cheaper than moving your
entire office and exchange.

Speaker 3 (34:05):
Raoul PAL, yes, I remember we were having a
discussion I think I justremembered, as you're saying,
that we were having.
I think we were having lunchtogether at one of the photon
X's and we talked about how canwe refer to different
nomenclatures rather thanphotonics.
I think we talked about optics,of banking, to account for the

(34:27):
nanoseconds of difference thatthis Holocore 5 can make and how
critical that is to traders andbanks.

Speaker 2 (34:38):
PETER BARRON Absolutely, the Holocore brings
the confluence of a large numberof attributes because, as well
as the fact that the delay isless, it also, of course, has
incredibly low nonlinearity andtherefore you can put more power
in which we've alreadydiscussed the amount of power

(34:59):
that you can put down the fiber.
But in the telecommunicationsfield that's incredibly
important because the normallimit to a solid core fiber that
limits the amount of bandwidthyou can get out of the fiber is,
in fact, the nonlinear mixingbetween all the channels, and

(35:20):
that is virtually non-existentin a hollow core fiber.
So we're expecting to be ableto get much closer to the
theoretical limits with thehollow core fiber.

Speaker 3 (35:36):
That is very exciting .
I mean really pushing thoselimits, and I'm greatly looking
forward to seeing how much youcan approach absolute perfection
over the next years for that.
So it's going to be excitingwhen these holography fibers
start to enter generalcirculation.
I know it's now probably likehen's teeth, but when these

(36:00):
fibers start to longer lengths,start to be fabricated and
deployed, it's going to reallyreally change the world.

Speaker 2 (36:07):
Yes, I would agree.
As I pointed out, thespecialist applications, such as
installations in London whichthey are carrying live traffic
at the moment.
There are a number ofinstallations around the world.
I believe there's one coming upinterconnecting the exchanges
in Shenzhen and Hong Kong, again, where latency is super

(36:30):
important.
So, yeah, as my colleagueFrancesco Belletti points out,
there's a remarkable mirroringof the development of solid core
, and this is wonderful becauseyou and I, richard, both know,
then, what's going to happen,because we were around at the

(36:51):
time when the solid core wasrolling out.
So we know that the next stepsare going to be the scale up and
the cost reduction of thesedevices, the final tweaking of
the designs and then a range ofspecialist applications for
hollow core fiber.

(37:11):
We're beginning to see thesedeveloping right now.
So, for example, the gyroscopewhich currently uses a solid
core polarization, maintainingfiber in a Sagnac configuration,
and we've published some workwith our sponsors in Honeywell

(37:34):
on resonant gyros, fiber gyros,on resonant gyros, fiber gyros,
which is very difficult to do ina solid core fiber because the
fiber in general statement islimited by non-linearity in the
core and, tiny though that is, alot of development effort for

(37:57):
solid core fiber has gone intoovercoming that non-linearity
which produces an asymmetry andan offset in the gyroscope and
of course these virtually don'texist the non-linearities in the
hollow core fiber.
So a lot of people are workingon that for much improved

(38:23):
gyroscopes for the futurenavigational grade gyros.

Speaker 3 (38:27):
So one other area is one other.
We've spoken about theinterconnects, the fibers being
the interconnects oftransmission but the endpoints,
the fibers being theinterconnects of transmission
but the endpoints.
Southampton OSCS has been acritical player in developing
photonic integrated circuits,silicon photonics, and these
silicon photonics started off asan academic curiosity many

(38:48):
years ago.
Now the silicon photonics formsthe basis of all modern
transceivers that you'd find ina high-skill data center modern
transceivers that you'd find ina high-skill data center.
I mean, what can you tell usabout photonic integrated
circuits and what we'll expectto see in the future?

Speaker 2 (39:09):
Yeah, so this is of course been pioneered by my
colleague, graham Reid,initially at Surrey and now at
Southampton, where he movedabout 15 years ago, if I
remember correctly, along withhis team.
And yeah, it's a very excitingintegration technology because

(39:31):
the degree of integration thatoccurs in photonics is sadly not
anywhere near as much as wewould like it to be, and most
people would agree that it'sintegration which gives you the
cost reductions.
Integration in optics is hardbecause there isn't the

(39:53):
equivalent of silicon inelectronics, the equivalent of
silicon in electronics, whereevery microprocessor on the
planet is made from siliconmicrocircuits.
As a consequence, it tends tobe a hybrid integration of
different material systems andthat is very limiting.

(40:15):
You mentioned earlier, actually, when I think of, one of the
most exciting developments isthe use of quantum dots on
silicon photonics to overcomethe fact that nobody's yet
successfully made an emitter insilicon, an optical emitter in
silicon.
That has been limiting thefield, because you had to do

(40:38):
this, as I mentioned, hybrid, soyou had to pick and place three
, five chips for lasers andlight sources and amplifiers and
so on, and that's not a cheaptechnology because of the
alignment requirements and so on.
But we are making some seriousprogress there, and I think
people are also coming up withmulti-material platforms where

(41:03):
you are able.
Well, we already mentioned thequantum dots, but there are
other technologies that peopleare working on to be able to mix
, for example, silicon nitridewith silicon.
Silicon itself is is and Ialways kid my colleague graham,
read about this silicon itselfis a lousy optical material.

(41:25):
Okay, compared to silica, theoxide.
We love the oxide.
We've already mentioned that100 kilometers is possible, but
for silicon itself, you'retalking about millimeters,
centimeters.
We'd like to overcome thatproblem as well.
It's a very exciting area.

(41:49):
As you said, we have a largeteam at Southampton and there
are other large teams around theworld Europe, us, china, japan
all working on this integrationin the hope that one day we will
have little chips that costtens of dollars maximum.
We're a long way from that atthe moment.

Speaker 3 (42:14):
So one other area connected as part of this whole
ecosystem of data processing,data communication there's also
data storage, and one excellentinnovation from ORC is what's

(42:34):
referred to as 5D opticalstorage.
This is optical storage inglass is what's referred to as
5D optical storage.
This is optical storage inglass, very dense storage, but
the key characteristic of thisis that the length of time over
which you can store information,which I believe, is close to
the age of the universe, if Iremember correctly.
What can you tell us about thistechnology?

Speaker 2 (42:54):
Yeah, I love this technology.
This was developed by mycolleague, peter Kazansky, and
he's been working on this for 20odd years, so it's not kind of
an overnight sensation.
Little tiny, what we callvoxels, with a focus laser in

(43:21):
the substrate, which is onceagain silica.
So you take a little chip ofsilica and you focus a high
power pulse laser onto it andthis creates a tiny void inside
the glass.
They're a little like you mayhave seen these consumer

(43:46):
commercial plastic blocks whereyou can write your kid's picture
in it or something like that.
It's not a million miles fromthat, but there are some
differences.
So what Peter Kzansky found wasthat these little tiny voids
had a characteristic which wecall birefringence, whose

(44:12):
direction depended upon thepolarization of the light that
wrote it for reasons which arestill not perfectly clear, and
as a result, you got anadditional dimension.
So you mentioned that thistechnology is called 5D,
standing for five dimensions.
And what are those dimensions?

(44:33):
The three are x, y and z, asusual, but an additional two
dimensions are obtained becauseof the directionality of the
polarization of the light thatwrites these little voxels,
these voids.
And the fifth dimension is thestrength of the birefringence of

(44:55):
these little tiny voids canalso be measured.
So this gives you fivedifferent parameters and this
immediately, of course, givesyou a much larger data storage
capability.
And then there's one additionalthing, because this is a very

(45:15):
highly nonlinear approach, youcan write multiple layers, each
layer of these voxels, a littlebit like the tiny little pits in
a CD, but you can write and Ithink Peter has written up to
200 layers of these, because youcan write through the previous

(45:39):
layers and you can read throughthe previous layers.
And you mentioned the uh, theheadline which attracts so many
people, which is that, uh,because it's silica, uh, it will
last for billions and billionsof years.
In fact, the world is aboutfour billion years old, so it'll

(46:06):
last much longer than the ageof the Earth.
Up to now, unless, of course,you accidentally Kazansky loves
demonstrating is that you cantake a Bunsen burner and you can
heat these chips up untilthey're red hot, and they still

(46:29):
work perfectly well.
And so this is what is calledarchival storage, which is, as
you point out, incrediblyimportant, because we've talked
up to now about the ability totransmit these huge quantities
of data around the world and thedata centers which store them.

(46:53):
Currently, the storage is donein archival storage in tape
units, and this is 50-year-oldtechnology, and the problem with
tape units is that they need tobe refreshed because they don't
last forever and that's a veryexpensive business to go back

(47:15):
every 10 years or so and refreshall the data on the tape units.
And this is archival storage,meaning that there are many
things that you need to storeforever, or at least for 50
years, and these are, forexample, bank transactions and

(47:37):
so on, historical events whichhave to be stored, or even for a
human lifetime.
You won't like it if yourpictures of your grandchildren
are lost after 10 years, whichyou've stored up there in the
cloud.
So archival storage isincredibly important and this,
we believe, is the technologywhich solves this problem.

(48:02):
Once again, microsoft has pickedup this work, is running with
it in what they call ProjectSilica.
We understand that they are ata prototype stage and we need to
remember that ultimately, youalso need to be able to go and
fetch these little tiny silicachips.

(48:23):
They've developed machines tobe able to do that and go, find
the chips and bring them back inand read out the data if you
demand it.
It takes a few seconds to dothat.

Speaker 3 (48:37):
RAOUL PAL, if you call.
I used to work down the road astone's throw away from you and
haven't in Seagate.
Previously it was Zyrotex.
That's all about data storage.
Archival storage was veryimportant.
We were in a project a longtime ago with the BBC.
They had, as you mentioned,problems that they had all these

(49:00):
mountains of tape which wassimply rotting away and so much
has been lost.
So much very valuableinformation has been lost simply
because the media has rottedaway.
I think one of the importantthings to rescale this
technology is the speed withwhich you can write data onto
these silica blocks.
How quickly can a machine writethis information and will this

(49:26):
speed up in future?

Speaker 2 (49:28):
Well, it is the one area where we need to improve
here.
If you want to write anarchival copy of a movie at what
I think is called thedirector's level of definition,

(49:50):
which is very high indeed.
It's higher than Blu-raydefinition it's going to take
you a day or so and although youdo this, of course, with
multiple machines, writing itmultiple times, it could be
faster.
And there are many things thatyou can do and we have done and

(50:11):
we are continuing to work onthis to improve it.
Once again, you've mentionedwhat we do at Southampton kindly
on a number of occasions.
A lot of what we do atSouthampton is materials-based,
and I often say that inphotonics you never have the

(50:34):
material you want, with thepossible exception of silica.
But there's an awful lot youcan do on materials research.
Is there a better material thansilica for this application,
the 5D archival storage?
Is there a magic material or amagic dopant that you will put

(50:56):
into the silica?
That will change everything andmake it much faster to write,
and we are working on that as wespeak Now one thing with from
the photonic integrated circuitindustry.

Speaker 3 (51:10):
One of the big challenges is how to couple
fibers to photonic integratedcircuit chips PIC chips and this
is done in a variety ofdifferent ways vertically,
through grating couplers oralong the edge.
But one limitation I've heardabout optical fibers is that a
standard optical fiber is thatthere's a limitation on how

(51:34):
closely you can put the fiberstogether to get the cores,
because each core is in singlefibers.
Each core is surrounded by acertain cladding.
But there are innovations toaddress this.
There's multi-core fibertechnology coming out.
Is this something that, uh,that you're working on in our
orc?

Speaker 2 (51:51):
yes, um, we've talked , I think we've we've hinted
throughout this discussion thatan ecosystem is required for
each of the technologies thatare emerging or are in the
marketplace already.
And I think there are those whobelieve that multicore fibers

(52:14):
are an important part of thefuture marketplace.
Fibers are an important part ofthe future marketplace, but the
question you asked, which is,what about the splicing of them?
And of course it is moredifficult than when you have a
single core, because now youhave to orientate during the
splicing procedure.
I take an open-minded approachto whether or not the multi-core

(52:39):
fiber is going to make it inthe marketplace.
One part of me says well,actually the fiber is always
dirt cheap compared to the totalinstallation costs.
Current costs for a kilometerof standard single-core fiber is
about $3 a kilometer, which Ijust find an incredible number.

(53:03):
It's $3.
Of course, it'll be much higherthan that for multi-core,
because most of the cost in afiber is actually in the core
anyway, is actually in the coreanyway, and so scaling.
You could argue that a 10 corefiber is going to cost 10 times

(53:25):
as much.
And then, of course, you needto be able to have amplifiers
for each of the cores and youreally don't want to split them
all out and have singleamplifiers and then put them all
back again.
So we have published work onmulti-core amplifier fibers as

(53:45):
well, but all this is getting alittle bit complicated and a
little bit costly.
And to what advantage?
Well, there's a clear advantageif you have very full ducts,
which some cities do have, andyou don't want to put multiple
fiber cables in because the ductis already full, and so the

(54:10):
spatial density, if you like, ofinformation in a multicore
fiber is of interest.
But, as I say, I've got an openmind on it and I suspect that
maybe some links will appear inmulticore fiber, but I suspect

(54:30):
not many.

Speaker 3 (54:33):
I tend to completely agree with you.
One of the reasons I broughtthis up is that I chair a
standards group in the IEC, asubcommittee on optical
connectors, which is the largeststandards group in the IEC,
which reflects the fact thatthere are many, many experts
part of this group, and for itto make it a standard, it has to
be very well established.

(54:54):
And only this year I remarkedon how a huge number of
multi-core fiber proposals seemto have emerged, most of them
from Japan, actually includingone for EDFA.
Multi-core fiber from NTT, Ithink, and that's but I share
with you unless you absolutelyneed it, why would you the

(55:17):
complexity of splicing two ofthese together, the rotational
tolerances and so on.
I would have thought they'd beprohibitive, but if they can
manage to achieve it, it wouldbe quite a feat of engineering.

Speaker 2 (55:32):
Indeed, and if anybody can achieve that, it
would be the Japanese.
So good luck to them, say I.
We have worked with a number ofJapanese groups in this area,
so we have an insight into it.
But it's interesting to see howthe hollow core fiber impacts

(55:53):
on that particular market,because it brings a number of
other advantages which we'vealready discussed.
You can, of course, makemulti-core hollow core fibers as
well if you wish, but that's acomplexity that we're not even
yet quite at the scale upposition where we'd like to be

(56:13):
with a single core fiber.
So that's for the future.
But think of it this way,Richard this is what pays our
salaries, right, and, as Ipointed out earlier, we know
what's going to happen becausewe just are following what
happened for solid core fibers,so you can almost predict what
you're going to be doing in fiveyears time.

Speaker 3 (56:34):
Indeed, indeed, it's exciting, it's a very exciting
time to be alive, well, I mean,you can say that for 60 years.
But Indeed, indeed, it'sexciting, it's a very exciting
time to be alive, well, I mean,you can say that for 60 years,
but it really is mind-blowing.
I wanted to just mentionhyperscale data centers.
All of this has related, Ithink, in one respect or another

(56:56):
, to hyperscale data centers,and this is the term we give for
very, very large data centerswhich typically comprise at
least 100,000 servers and theyare used to administer cloud
services, and now, increasingly,they're used to incubate AI
clusters.
So all the AI we hear about thechat, gbt and so on, it's all
incubated in these hyperscaledata centers GBT and so on, it's

(57:19):
all incubated in thesehyperscale data centers.
And, as of three years ago,hyperscale data centers are the
dominant form of data centersand they really are critical to
everything we do.
But with the emergence ofartificial intelligence as I
mentioned, chat, gbt we've seenthese kind of frivolous
applications come up which arevery amusing, seen these kind of
frivolous applications come upwhich are very amusing.

(57:40):
But the fact of the matter isthis wide-scale pattern
recognition can be used forindispensable applications such
as healthcare and preventativecare and so on, and they can be
a tool for incredible,incredible good.
So it is a lot of people sayit's a bubble.
I think it'll become tooindispensable to be a bubble.
So the relevance of thisartificial intelligence in these

(58:02):
high-scale data centers is veryimportant.
But the hyperscalers Microsoft,google, meta are all saying the
big problem there to make thisoptical is power consumption.
And I think Microsoft, they areworking.
They hired a reactor on ThreeMile Island and Google want to
hire a nuclear reactor simply topower their data centers,

(58:24):
because the power consumption ofthis is astronomical.
And we're starting to seeinnovations such as immersion
cooling, where you immerse theentire server into a mineral oil
to really reduce your powerconsumption.
So there's a strong pressurefrom the hyperscale to strongly
reduce power consumption on allaspects.

(58:45):
And one of the things we'reseeing is for these networks and
data centers is WDMcommunications using different
wavelengths of light in certainarchitectures One architecture,
for example you have a tunablelaser at a source and you simply
change the wavelength veryquickly and then it passes
through passive filters likearrayed waveguide gratings to

(59:06):
instantly move from onedestination to another.
And I think in UCL there was aspin out to Aureola Networks,
which is leveraging this kind ofarchitecture and I think
Southampton also plays a largepart in this kind of ecosystem,
for example, arrayed waveguidegratings.
What are your views on usingthese new WDM architectures for

(59:29):
AI in the future?

Speaker 2 (59:31):
I think this is an incredibly important topic and
most of us have been careeringalong ignoring it for the last
decade or so.
And it was brought home to merecently when somebody mentioned
, in a talk I was at, that atypical single cabinet in a data

(59:55):
center takes a goodly fractionof a megawatt A megawatt.
I just was stunned by thatnumber.
And not only that, but I'veheard talks of predictions that

(01:00:16):
the IT of the world will consumethe entirety of all energy
generation by about the year2040.
And this is obviously,obviously unsustainable.
And what are we going to doabout it?
And very few people have hadthe insight that you just

(01:00:40):
mentioned up to now to say whatabout the energy consumption of
that wonderful new device?
We just talked about someamazing things like the archival
storage and so on.
What does it mean for energyconsumption?
What does it mean for energyconsumption ought to now be a

(01:01:04):
question that everybody askswhenever they see a new
technology emerging.
So there are some simple things, apparently, that we can do,
which is just improve ourcooling systems, which you've

(01:01:25):
mentioned, and make them muchmore efficient, just as your
electric vehicle does, you know,and uses a heat pump instead of
the old ways of doing things.
But going back to thefundamentals which is the point
you were making, I think, whichis to go and say no, that's an
unacceptable technology, we needto do better.
And when I started to look intothis, I realized that the way
that we've structured theoverall global communication

(01:01:46):
system is not optimal.
For example, we use an awfullot of wireless, and wireless is
very energy inefficient becauseit sprays energy in all
directions, of which you onlyreceive a small amount, whereas
optics is very focused on thattiny little fiber and you use
pretty well all of the lightthat you transmit.

(01:02:08):
This is why the whole globalinternet is structured with all
heavy lifting being done.
The figure, I think, is 99% ofall data is carried in optical
fibers, and then wireless hasits role, of course, as the last
drop to your mobile phone or inyour Wi-Fi in your home or

(01:02:30):
whatever, but it is not theinfrastructure and it's very
efficient, so inefficient.
One of the things we could bethinking about is more optics,
less wireless.
What does that mean?
It means shrinking the size ofthe wireless cells and feeding

(01:02:51):
them with optics is an obviousthing that we could be doing,
using less wireless masks thanwe do, which are, you know.
That seems the opposite of whatyou want to do with microcells,
but the wireless masksthemselves currently can take

(01:03:12):
tens of kilowatts, which is anawful number.
So this is not my area ofexpertise by any means, but I
think we need to rethink the waythat we structure the global
internet and telecommunicationsgenerally, with a much greater
focus on the energy consumption.
And ways of doing that, forexample you've already raised,

(01:03:34):
would be integration.
There are those that aretalking about having a
desegregation of microprocessors, electronic microprocessors
with optical interconnectsbetween them, which is not a new
idea.
But one of the things that Ilearned was that every idea has

(01:03:55):
its time window, and if you'retoo early or too late, well,
it's obvious, if you're too late, you've missed the boat.
But if you're too early, theother technologies which are
required to achieve it might notbe ready yet, and so it fails.
So optical interconnects goesback to the 80s, actually Early
80s.

(01:04:16):
I was working on some opticalinterconnect problems, but we
couldn't do it.
We didn't have integration.
So that's another way that wecan reduce the power consumption
of the internet and many othersthat will emerge, I'm sure,
because, as you point out, thisis a very fast-moving technology

(01:04:36):
, but one thing we can becertain of is it will be based
on photonics, absolutely.

Speaker 3 (01:04:44):
One thing we've addressed is we need fibers to
convey large amounts ofinformation, but the amount of
information that needs to beconveyed is getting larger and
larger all the time.
Today we send videos and wesend movies, and this takes a
lot of information.
But what's the future of theinternet?
What if we do have an internetwhere virtual reality becomes

(01:05:05):
more widespread, where we'd haveto convey not just
two-dimensional videos but we'dhave to convey the information
capturing the entirethree-dimensional, constantly
changing environment over this?
So the question is the amountof information we can send over
a fiber, and I know ORC andothers like UCL have done a huge

(01:05:27):
amount of work in increasingthe information carrying
capacity of a fiber throughdifferent multiplexing schemes
like wavelength divisionmultiplexing and QAM and so on.
What are your views on how farwe can push this, like
wavelength division multiplexingand QAM and so on?
What are your views on how farwe can push this?

Speaker 2 (01:05:41):
I think we are pretty well at the limit and have been
for close to a decade now onhow far we can push a single
conventional solid core fiber.
We see every single one of themajor conferences tweaking it a

(01:06:03):
little bit and getting just atiny little bit extra.
And there are fundamentallimits, of course, to the amount
of information you can pushdown a fiber, determined by the
wavelength window that you have.
We've already talked aboutopening up wavelength windows

(01:06:26):
and that would give us possiblyanother order of magnitude of
information capacity per fiber.
But that could only be donewith hollow core fiber because,
as we've already mentioned, thewavelength window in a silica
core fiber is fixed and has beensince the early 80s.
We are running out of bandwidthin conventional fibers.

(01:06:50):
The hollow core fiber willsatisfy that to some extent.
Then, when that is all over, inanother 20 years' time or so
and we've tweaked that as far aswe can we can always fall back
to more fiber, more and moreparallel fibers, more and more

(01:07:11):
cores inside a single fiber,which we've already mentioned.
We've got a lot of legs and alot of runway yet before we
completely saturate thecapability of photonics.
Happily, because what is goingto happen, and there's a theme I

(01:07:32):
think emerging here, richard,which you've touched on, namely
that it's the provision of thehardware and the capacity I
think this applies in a largenumber of technologies which
generates the applications.
It's not the other way around,as most people think it's.

(01:07:53):
Actually, somebody wakes up oneday and said you know what, look
at all this bandwidth.
We could use that and we couldbe profligate and we could do
silly things like we coulddevelop emails where we just
send back a one word answer yes,and we copy the entirety of
everything in the previous emailback again to the sender.

(01:08:16):
That's what this profligacy isabout.
If you don't have to think itthrough too much, then suddenly
an application can appear.
Using bandwidth in a wastefulmanner is just the way things
are, and long may it continue.
Frankly, I don't want to writean email thinking about can I

(01:08:38):
remove a word here, because it'sgoing to use too much bandwidth
.
There are many examples of that.
We want to send high-definitiontelevision signals.
We want to send photos of ourkids at very high definition.
We want big screens to viewthem on.
So profligacy in bandwidth isgoing to happen.

(01:09:05):
But a key question which acolleague of mine, will Stewart,
raised to me the other daywhich our listeners might like
to think about is what is thenext big application that's
going to eat up bandwidth?
And we talked about AI.
What follows next?

(01:09:26):
Is somebody going to come upwith a killer application which
suddenly requires vast amount ofextra data?
Because up to now in thehistory of the internet,
somebody has done that, cloudcomputing being a perfect
example of that, and the sortsof things which you could never

(01:09:47):
have done in the old days.
You and I are possibly oldenough to remember the good old
dial-up internet, where you usedyour phone nine and it went yes
.
And you wouldn't have thoughtabout.
You really did think about thenumber of words in your email in

(01:10:08):
those days.

Speaker 3 (01:10:10):
Yes, so, oh, one thing that was one thing I was
mentioning was ECOC, theEuropean Conference on Optical
Communication, and that'scelebrating its 50th anniversary
, and I was actually sat next toWill Stewart at dinner with you
there at the conference.
I was actually sat next to WillStewart at dinner with you
there at the conference.
I was sat next to Will Stewartwhen they played this song,

(01:10:30):
which was, you know, veryinteresting.
So completely AI generated thissong about.
They simply fed him theinformation about 50 years of
light and they played this andthey manufactured this complete
song from scratch.
That that that seemed like areal song.
And also you yourself a coupleof days ago, you sent me a

(01:10:51):
someone fed in the informationof your bio, and a podcast
between two, two fictionalAmerican podcasters was
manufactured discussing you, andit was so unbelievably
authentic that the power of AIin this is terrifying.
It's terrifying what can beachieved.
But if you combine that withthe possibility of virtual

(01:11:20):
environments where you havethese and I'm not sure this is a
utopian vision but where youhave these vast virtual
realities where you have toconvey information depicting a
moving, constantly changingenvironment shared by everyone,
that would astronomically eat upbandwidth, I don't know if

(01:11:41):
that's a good thing, but thatmight be where your next
bandwidth hog is going to comefrom.

Speaker 2 (01:11:47):
And of course, you're absolutely right and I too am
quite stunned by the capabilityof AI.
But what we haven't touched onquite yet is machine learning in
photonics.
On quite yet is machinelearning in photonics Because
another application of well,going back to my theme, it's

(01:12:08):
always about the hardware, andthe hardware in this case was
the GPU, the graphics processors, which have come on in leaps
and bounds, because the originsof AI?
I remember when I was doing myPhD back in the early 70s, we

(01:12:33):
were using optimizationalgorithms for intractable
problems, just to find asolution for simple things is to
find a solution for simplethings, but of course, at the
time it was limited by theability to have these clusters
of very high-powered GPUs.
So, once again, it's thehardware that changes everything

(01:12:53):
, and then along comes thealgorithms that you put onto
them, and I don't think we'vereally scratched the surface yet
.
So, to get back to my point,what can machine learning do for
the way that we use photonics,the design of, for example, a
low energy network, and thereare a number of people beginning

(01:13:16):
to work on this and decide whatthe optimum structures are
going to be A piece of workwhich I know is going on.
And returning to the idea of adata center, what is the best
way to communicate between allthe elements of the data center,
all the servers, with energy inmind and starting with a blank

(01:13:39):
piece of paper and with notechnology on it whatsoever?
So what's the best technology?
Almost certainly, as we'veagreed, it's uh, it's going to
be photonics, but whatwavelength?
And nowadays we've opened thatspace up.
So, and what are we going touse for the transmitters, which,
in optics, is often the energyconsuming part?

(01:14:01):
And what are we going to use?
How do we get rid of the ASICson the incoming fiber
communication links to the datacenter?
Because those ASICs which arecorrecting the impairments in
the communication medium arevery power hungry.

(01:14:24):
And so then we use machinelearning to look at all of this
and learn how to make theoptimum energy consuming, low
energy consuming structure.
And I have a colleague of minecalled Ben Mills at Southampton,
who's a young and upcominggenius in AI and machine

(01:14:50):
learning.
He just is staggering all of usbecause he just tends to wander
into a room and say is that theway you do it at the moment?
Is that the way you make things?
Just give that to me, I'll puta machine learning algorithm on
it and it'll come up with abetter way of doing it.
And we'd go oh God, so yeah,it's breaking all the rules and

(01:15:16):
designing integrated circuits.
You know, infotonics the nextgeneration of those is a very
exciting area.
We're currently using it, forexample, in an algorithm for a
project that we have onextremely high power lasers.

(01:15:36):
By extremely high power, I meana megawatt, continuous.
The applications of these arefairly self-evident in the
defense area, but also inincredibly fast manufacturing.
To do that, you have toparallel up vast numbers of

(01:15:58):
lasers, because we've alreadytalked about the maximum power
you can get out of one fiberlaser.
But now let's talk about athousand fiber lasers all being
combined together, and that's achallenge, because we know how

(01:16:19):
to do that.
From a physics point of view,we know that we have to phase
them together and that producesa single beam, and it's a
steerable beam, which is, ofcourse, exactly what you want.
But doing that phasing togetherin a way which is efficient and

(01:16:39):
fast because you've got amillisecond or something that
you want to be able to phase allthese lasers together as a
complexity problem that's a veryhard problem, but we've come up
with ways in which you can dothat using machine learning and
an algorithm which will justlock them all together suddenly
and very quickly.

Speaker 3 (01:17:00):
So that's just a beautiful example, I think of
the importance of machinelearning in photonics Agreed.
One other area is inversedesign.
Again in the photonicintegrated circuit field, people
put in their requirements and adesign is created which

(01:17:24):
fulfills those requirements.
And the design can becompletely counterintuitive.
It can look like a chaotic messbut actually it gives you the
perfect optical transmissionprofile and that's a lovely,
incredible example of what'spossible with it Absolutely
Machine-led design.

Speaker 2 (01:17:45):
Wow, it's going to put us all out of work actually.
Yes, exactly.
Well, the way I see it is,somebody's got to be able to
drive the machine learningdevice itself so maybe there's
room for us.

Speaker 3 (01:18:01):
yet that's right.
One hope I have actually iswith AI.
There is this adage garbage in,garbage out.
So you always need a source ofreliable expertise, otherwise
you won't get anything sensibleout there.
So hopefully there'll be theneed for human beings with
proper expertise to continuallyrefresh these large language
models.

Speaker 2 (01:18:20):
Definitely there's hope for us yet language models
DAVID BASZUCKI.
Definitely there's hope for us.
Yet.

Speaker 3 (01:18:25):
RAOUL PAL.
Okay, thank you, david.
I think I'll just finish upwith one last question.
What is the personal question?
What's the future for yourselfnow, david?

Speaker 2 (01:18:37):
BASZUCKI, who can predict the future.
I've stepped down as the headof the ORC in favor of my
colleague, graham Reid, thesilicon photonics expert that we
talked about earlier.
I'm having a lot of fun becauseit means I don't have to do all

(01:19:00):
the bureaucracy and all thatkind of stuff.
That it means I don't have todo all the bureaucracy and all
that kind of stuff that comesfrom being the director of the
operation, which is, by the way,is about a 300-person operation
at Southampton in photonics.
So I tend to focus my effortsat the moment on enterprise.
I've often observed thatthere's no point in doing

(01:19:25):
research unless you could have aroute to getting it into the
marketplace and test it, becausethere's all too many of these
crazy ideas that come up and apublication is made and that's
the end of it.
I've always felt it's important.
It I've always felt it'simportant.
We owe it, if you like, to oursponsors, to our governments, to

(01:19:48):
use the funding that has beentaxpayers' money, after all,
that has been paid to us, andshow that we can create jobs,
wealth and to the betterment ofmankind, which the internet, of
course, is a perfect example ofthat.

(01:20:08):
As a consequence, atSouthampton we have a dozen or
so companies which owe theirexistence to the work at the ORC
.
We have another three inprospect at the moment and
because I've been involved formany years in the transfer of

(01:20:32):
technology from research intocompanies and into products and
I enjoy that, it's having a footin both camps.
And right at the beginning ofthis podcast, I pointed out that
I had intended to go intoindustry but got seduced by my

(01:20:54):
professor to developing thefibers for the internet, and as
a result, I said, well, ok, I'llstay in the university, but I
think I have an idea, which isthat I'll have a foot in
industry as well, and that'sbeen very exciting and I've
thoroughly enjoyed that, and soI can intend to continue to do

(01:21:16):
that.
And to be able to do that, thefirst thing you need is to work
with incredible people, and I'veoften observed that my first
rule in life is always work withpeople smarter than you are,

(01:21:36):
and that includes the studentsthat I work with, and you know
these are young minds that areyet to be polluted with older
people's thinking, so I findthat hugely refreshing, and my
style of working is actually todiscuss in groups and just ping

(01:22:02):
off everybody, because that way.
We mentioned earlier the GPUcluster.
That's the human equivalent ofa GPU cluster.
It's a group of really smartpeople bouncing ideas off each
other, sharing them in an openfashion and not being afraid to
be wrong with stupid comments.
That's my environment.

(01:22:25):
To cut myself off from that andgo grow roses is not something
that I feel that I'm going to doanytime soon.

Speaker 3 (01:22:36):
Raoul PAL, md, phd.
Excellent, I'm looking forwardto meeting you in a couple of
weeks.
Actually, david, you've kindlyagreed to give the plenary talk
at the conference.
I chair British and IrishConference on Optics and
Photonics in the IIT and Ireally appreciate the
conversation we've had today.

Speaker 2 (01:22:51):
Thank you so much because this has been an
absolute pleasure, and it's arare privilege to be able to
chatter away with colleaguessuch as yourself and just bounce
ideas around, so I hope ourlisteners have enjoyed it.

Speaker 3 (01:23:08):
Thank you, david.

Speaker 1 (01:23:10):
Just before I'll let the both of you go, I've got a
couple more questions aroundcareer development and career
progression and the industry andacademic circles.
The conversation that you andRichard have had, Sir David, is
absolutely incredible.
I honestly do not want to askanything on top of these
questions, but I've just got acouple of questions to close

(01:23:31):
things out.
So and this is a question tothe both of you and hopefully
opens up a discussion what arethe characteristics and skills
in research, industry andacademia that have remained
timeless and what are the thingsthat have changed recently?
What are today's challenges?

Speaker 2 (01:23:50):
On the basis of the discussion which has been about
photonics, it leads meimmediately to a favorite topic
of mine, which isinterdisciplinarity.
Favorite topic of mine, whichis interdisciplinarity.
Photonics is amultidisciplinary field.

(01:24:10):
We need mechanical engineers,we need chemists, we need
physicists, to name just a few,and we still tend to silo the
way that we educate people.
I'm not sure I have a betterway of doing it, frankly, but

(01:24:33):
when we hire people in the ORC,we don't care too much about
what their background is.
Surprisingly enough, and some ofthe brightest people that we've
ever had have been with nonobvious backgrounds, like
mechanical engineering.
For example, one of the maincontributors to the hollow core

(01:24:55):
fiber that we talked about was amechanical engineer that we
recruited, greg Jassian, and itturned out he's the only person
that knew how to take a look atthe fiber drawing process and
say I can analyze that and comeup with models which did, and up
to then we'd just been doing it.
We'll try this, try that, trythe other.

(01:25:16):
So it's very, very important tohave those range of disciplines
.
It's very, very important tohave those range of disciplines
and I think there is anincreasing recognition in the
education system that we need todo better and even I say even,

(01:25:38):
that's probably not the rightword but to involve our
colleagues in humanities,because the characteristic of
most scientists and engineersand inventors is that we kind of
invent something and then chuckit over the fence and say, well
, it's your problem now, notmine.
Actually, we've seen how theinternet can be abused and maybe

(01:26:02):
we should have predicted thatand it might have changed the
way that we designed it, forexample.
So those are my initialthoughts in response to your
question.

Speaker 3 (01:26:15):
I tend to agree.
The humanities comment is veryimportant.
I think, in many respects,scientists and engineers, we
should be held to the same moralobligations that doctors are,
is very important.
I think in many respects,scientists and engineers, we
should be held to the same moralobligations that doctors are,
and doctors have to swear theHippocratic Oath do no harm.
The work we do has equally thatlevel of impact on humanity.

(01:26:38):
So ethics, ethicalresponsibility for how we design
, I think that's important.
You asked about what weconsider timeless.
One important thing isengineering.
The UK is extremely strong inengineering.

(01:26:59):
Of course, modern engineeringthe UK has been the leading
player there.
But in the 80s we went throughthis phase, starting in the 80s,
where engineering became therewas almost a stigma attached to
engineering.
In the 90s in the UK, whileother countries like Germany,
countries around the world,engineers were treated as high

(01:27:21):
professionals, in this country,in the UK, there was almost a
stigma attached to it, which wasreally tragic because it is one
of our proudest traditions.
I'm glad to say that now,certainly in the last 15 years,
this stigma is now lifted, theimportance of hardware
engineering especially not justsoftware engineering but

(01:27:41):
hardware engineering and thatknow-how is becoming appreciated
again.
Engineering in the UK is beingseen as the asset, as it is
Personally.
In my company I prize.
We have nine engineers andexperience is the most valuable.

(01:28:01):
We have mechanical engineers,electronics engineers and a lot
of people ask me well you know,is this something that you
recommend after your PhD?
In my case I've said PhD is nota prerequisite, experience is a
prerequisite.
So the German apprenticeshipmodel that works so well for
them in the 70s and 80s isbearing fruit.

(01:28:23):
That works so well for them inthe 70s and 80s is bearing fruit
.
I think training up in thesedifferent areas and developing
actual expertise is crucial.

Speaker 1 (01:28:30):
I think that's very interesting because it shows,
for example, where the focus is,the need is and where the
direction is going towards aswell.
A second question I had andthis can be from both
perspectives given all of yourexperience between the both of
you and individually how do youperceive mentor mentee

(01:28:52):
relationships to be?
How do you perceive your roleas a mentor?
Have you had incredibleexperiences both as a mentor and
being mentored by somebody else?
Do you have any stories, anyanecdotes of these experiences,
um, from your careers?

Speaker 2 (01:29:09):
uh, it's a fascinating question, um,
because I think, just as humanbeings are incredibly diverse
diverse, their need or otherwisefor mentorship is very
different.
I've known people who reallydon't need any, or even resent

(01:29:34):
any, form of mentorship, becausethey know where they're going
and they want you just gettingin the way.
These are fairly rare type ofpeople, but I have known a
number of people like that andI've just enjoyed letting them
get on with it.
On the other hand, there arethose who are perhaps what's the

(01:29:58):
word I'm looking for lessdriven or have a personality
which is less robust, perhaps,who benefit enormously from
mentorship.
And then there are also ethnicdifferences, gender differences,

(01:30:20):
um, I have mentored a number ofwomen, for example, who appear
to really appreciate it, um, forreasons which it's too big a
topic for us to get into now,but I think um to to have

(01:30:40):
somebody that's been there, donethat and show them the way,
because many women are jugglingwith families at the same time
and their priorities are notalways obvious to them and a
mentor that helps them throughthis and helps them to make
career decisions.
And I have many examples of thatwhere I've said to a female in

(01:31:05):
particular well, where do youwant to get to.
And if they say I'd love to besomebody that's really famous in
the field, I said, well, okay,then I can tell you how to do it
.
But if they say, no, no, no, Iwant to spend more time with my
family and I want a middle-rangejob which is not quite so
demanding, I say, well, you'retalking to the wrong mentor then

(01:31:26):
it's equally valid, I think.
By the way, that choice andtalking about gender politics, I
think women have an advantagethat they have that choice and
we should respect that.
But equally, if they choose togo the route of I want to be

(01:31:48):
number one, then there shouldn'tbe any barriers in their way.

Speaker 3 (01:31:52):
Richard, in my case, mentor.
Well, about 20 years ago and Istarted 20, 20, 24 years ago and
I started in Zardex, down theroad from David in Heaven.
24 years ago, when I started inXardix, down the road from
David in Heaven, even though Icame from a photonics background
, I was mentored by a verystereotypical rough glass Ouija
you live in Glasgow, south Akiland he taught me electronics and

(01:32:17):
he really put me through mypaces and today that guy is my
chief engineer at ResolutePhotonics.
He's a good friend and he's thebest engineer I know and I'm
delighted that he is leading myengineering team in an office in
Glasgow.
To really, you know, you needsome of the experience to show

(01:32:45):
the ropes.
That will always be.
That always be the case.

Speaker 1 (01:32:47):
And, exactly as david says uh, you know, you have to
have the right mentor for theright for the right uh ambition
I think it's really interestingthose perspectives, because
personally, I have just startedsupervising my first two PhD
students, so I'm at the otherend of the spectrum, where the
students are looking forward tofeedback, they're looking

(01:33:08):
forward to advice, they'retrying to build a multifaceted
career and I have very much thesame conversations, as you said,
sir David, because it's alwayssit down at the start or do this
multiple times during a PhD.
Where you go?
Which way do you want to head?
If you want to go into academia?
These are do you want to head?
If you want to go into academia?
These are the things you need.
If you want to go into research, innovation and industry, this
is where you have to go.

(01:33:29):
So I think those sort ofmentorship steps and advice is
always useful.
I've received the same as wellat every stage of my career, so
I think this input is really,really useful.
To end, I'd like you might havealready mentioned this in your
conversation, but I think I'dlike to end on this note.

(01:33:49):
Listening to this conversationwould be industry personnel,
academics would be students,early career researchers,
postdocs of differentbackgrounds.
If there's one single piece ofadvice you'd like to leave for
them, what would that be?

Speaker 2 (01:34:06):
Well, I'm going to use a quote for that, because my
interest, as I pointed out andI'm a fairly unusual academic
which is the commercial as wellas the academic and the research
side very advanced research,and the quote is never confuse

(01:34:36):
the art of the possible with theart of the profitable, because
a lot of academics tend to learnthat it's not the really fancy
piece of physics you just did,it's whether the world actually
wants it and what price is itgoing to be.
Those are the important thingsand those are the difference

(01:34:58):
between a successfulentrepreneurial activity and a
successful company and one thatfails early because nobody wants
your products.
That's very good advice.

Speaker 3 (01:35:11):
I will have to heed that advice now, as I like doing
the cool stuff in our company.
We're in these different EU andUK projects at quantum, looking
at communication sensors and soon, and it's very, very cool,
but I have to make sure that theworld needs it and that it's
profitable, otherwise it won'tlast long.

(01:35:32):
So that's very good advice.
In my case, all I would say isI'll touch upon something that
David remarked upon before theimportance of
interdisciplinarity.
If you're in a field, get toknow people in different fields,
and some of the greatestinnovations I know have come
from talking to people fromcompletely different disciplines

(01:35:53):
and getting together and havinga conversation about what we
all do.
That's incredibly valuable.
So I say that's useful.

Speaker 1 (01:36:02):
Fantastic, and I'll echo something that sir david
mentioned earlier as well, andthis is something I received
from a mentor of mine who youmight know, based in glasgow
professor miles bandit um.
His advice was.
His advice was always thepeople that you work with are
almost more important than thework you actually do.
If you enjoy going to workevery single day, you'll always

(01:36:22):
do incredible work.
This is something that youtouched upon earlier as well,
and I think that's incredibleadvice that I'll probably hold
on to, and I should hold on tofor the rest of my career anyway
.
So I'd like to thank the bothof you for your time.
This has been an incredibleconversation.
We've talked optical fibers,interconnects.
We've talked about the internet, hollow core fibers, fibers,

(01:36:45):
photonics and nobel laureates.
If you haven't heard, we'vealso talked about intergalactic
optical fibers, so that might beinteresting to sort of look
back on and listen to as well.
And I'd like to also say I'vehad a dial-up connection that
has ruined phone conversationsbecause it's going.
It's using the internet to gothrough the phone lines back in
India.
Thank you very much for yourtime.
This has been an incredibleconversation and I'm very, very

(01:37:07):
glad that every one of you inthe audience has joined in to
listen to this conversation.
Thank you so much.
Have a wonderful day ahead andsee you next time.

All Episodes

Episode Transcript

Popular Podcasts

United States of Kennedy

Stuff You Should Know

Dateline NBC

.css-15opob5{left:0;position:absolute;top:0.8rem;} All Episodes

.css-14f5ked{margin:0;word-break:break-word;display:-webkit-box;-webkit-box-orient:vertical;box-orient:vertical;-webkit-line-clamp:2;overflow:hidden;}Illuminating the Past, Present, and Future of Photonics

Episode Transcript

Popular Podcasts

.css-r6mb8g{margin:0;word-break:break-word;display:-webkit-box;-webkit-box-orient:vertical;box-orient:vertical;-webkit-line-clamp:1;overflow:hidden;}United States of Kennedy

Stuff You Should Know

Dateline NBC

All Episodes

Illuminating the Past, Present, and Future of Photonics

United States of Kennedy