Professional Documents
Culture Documents
i
“NLMS_484” — 2019/9/11 — 16:05 — page 2 — #2
i
i
“NLMS_484” — 2019/9/11 — 16:05 — page 3 — #3
i
CONTENTS
OBITUARIES In memoriam 45
i
“NLMS_484” — 2019/9/11 — 16:05 — page 4 — #4
i
4 NEWS
LMS NEWS
i
“NLMS_484” — 2019/9/11 — 16:05 — page 5 — #5
i
NEWS 5
CONFERENCE FACILITIES
De Morgan House offers a 40% discount on room hire to all mathematical charities and
20% to all not-for-profit organisations. Call 0207 927 0800 or email roombookings@
demorganhouse.co.uk to check availability, receive a quote or arrange a visit to our venue.
i
“NLMS_484” — 2019/9/11 — 16:05 — page 6 — #6
i
6 NEWS
LMS Honorary Members 2019 tureship in Applied Mathematics for his leadership in
numerical linear algebra, numerical stability analysis,
and communication of mathematics.
At the Society meeting on 28 June 2019, the Dr Toby Cubitt of University College London is
LMS elected Professor Edward Witten (Institute for awarded a Whitehead Prize in recognition of his out-
Advanced Study, Princeton) and Professor Don Zagier standing contributions to mathematical physics, in
(Max Planck Institute for Mathematics) as honorary particular the interconnections between quantum
members of the Society. information, computational complexity, and many-
body physics.
Edward Witten occupies an unrivalled position
amongst contemporary mathematical and theoreti- Dr Anders Hansen of Cambridge University is
cal physicists. He has made profound contributions awarded a Whitehead Prize for his contributions to
to the development of contemporary physics, includ- computational mathematics, especially his develop-
ing topological quantum field theory, string theory, ment of the solvability complexity index and its cor-
M -theory and quantum gravity. responding classification hierarchy.
Don Zagier is an outstanding mathematician who Professor William Parnell of the University of
has made major contributions in number theory, par- Manchester is awarded a Whitehead Prize for highly
ticularly to the theory of modular forms, and in its novel and extensive research contributions in the
interactions with other areas of mathematics and fields of acoustic and elastodynamic metamaterials
mathematical physics. and theoretical solid mechanics, as well as excellence
in the promotion of mathematics in industry.
i
“NLMS_484” — 2019/9/11 — 16:05 — page 7 — #7
i
NEWS 7
i
“NLMS_484” — 2019/9/11 — 16:05 — page 8 — #8
i
8 NEWS
Ken Brown awarded the 2019 David Beyond this long-term involvement with the LMS, Ken
Crighton Medal has made numerous further contributions to the UK
Mathematical Sciences, including as a member of the
Research Assessment Exercise (RAE) Mathematics
The LMS and IMA have subpanel in 1996, as Vice-Chair of this panel in 2001,
awarded the 2019 David and then as Chair of the RAE Pure Mathematics Sub-
Crighton Medal to Ken panel in 2008. He was also a member of the Research
Brown, Professor of Excellence Framework (REF) expert advisory group
Mathematics at the Uni- in 2008-09. He has been a member of the EPSRC
versity of Glasgow, for College since 1996 and served on the EPSRC Mathe-
his seminal contribu- matical Sciences Strategic Advisory Team (SAT) first
tions to noncommuta- as a member from 2013 to 2015 and subsequently
tive algebra and for his as its Chair from 2015 to 2017.
remarkable record of service and dedication to the
UK mathematics community. Read the full citation at tinyurl.com/y5ubvpg3.
i
“NLMS_484” — 2019/9/11 — 16:05 — page 9 — #9
i
NEWS 9
OTHER NEWS
Alan Turing honoured on new £50 The first excerpt is a table from page 240 which
banknote provides a schema for succinctly representing Turing
machines. The table gives a complete description
of how to specify such machines and therefore can
be thought of as one of the first examples of a pro-
gramming language.
i
“NLMS_484” — 2019/9/11 — 16:05 — page 10 — #10
i
10 NEWS
i
“NLMS_484” — 2019/9/11 — 16:05 — page 11 — #11
i
NEWS 11
mathematics funding. How can we counteract this identified the top three science priorities in order for
unfortunate development? First of all, there is no Japan to lead the fourth industrial revolution and to
reason to apply for the full possible budget if this even go beyond its limits: mathematics, mathematics,
is not appropriate for a research project, smaller and mathematics!”
proposals are very welcome, and second we mathe-
maticians should be more self-confident in writing
proposals. It is not a wasted time, even if one is not
Bernoulli Center (CIB)
funded. In several European countries there is even The CIB in Lausanne has announced the following sched-
financial support for proposals that make it to the ule of research programmes: Dynamics with Structures
second round but do not get funded due to budget (1 July – 31 December 2019); Functional Data Analysis (1
restrictions. January – 30 June 2020); Locally Compact Groups Acting
on Discrete Structures (1 July – 31 December 2020); Dy-
It is very important that applications are encour- namics, Transfer Operators, and Spectra (1 January – 30
aged throughout the mathematical community and June 2021). For more details see http://cib.epfl.ch.
the EMS is planning to create an initiative to sup-
port applicants. So please distribute this information
within your community. European Congress of Mathematics
Volker Mehrmann The preparations for the 8th European Congress of
President of the EMS Mathematics in Portorož, Slovenia, from 5 to 11 July
2020 are proceeding energetically. A full list of plenary
and invited speakers at for the 8th European Congress
Academia–industry roundtable of Mathematics in Portorož, Slovenia, is available at
8ecm.si/program.
The EMS Applied Mathematics Committee notes with in-
terest the minutes of an academia-industry roundtable
organised by the ministry of economy in Japan (March EMS News prepared by David Chillingworth
2019) (tinyurl.com/y3o54aud) which claims “we have LMS/EMS Correspondent
OPPORTUNITIES
EMS Prizes Events within Europe and speakers with strong rela-
tions to European academic institutions will be given
Calls for nominations of candidates for ten European preference. Closing date for applications is 30 Septem-
Mathematical Society Prizes as well as The Felix Klein ber 2019. For further details and nomination forms go
Prize and The Otto Neugebauer Prize for the History to the EMS website and click on Scientific Activities.
of Mathematics are still open: details can be found
at the European Mathematical Society website.
Maryam Mirzakhani Prize in
EMS Call for Proposals Mathematics: call for nominations
The European Mathematical Society (EMS) Meetings
Committee is calling for nominations or proposals for In recognition of Dr Mirzakhani’s remarkable life and
speakers and scientific events in 2020. The EMS is achievements, the National Academy of Sciences
willing to provide support to cover the cost of EMS has established a newly named Maryam Mirzakhani
Lecturers and Distinguished Speakers, and to give par- Prize in Mathematics (formerly the NAS Award in
tial support to the organization of EMS Weekends and Mathematics, established in 1988 by the American
of EMS Summer Schools of high scientific quality and Mathematical Society in honour of its centennial).
relevance. The EMS is committed to increasing the par- The prize will be awarded biennially for exceptional
ticipation of women in research in mathematics and contributions to the mathematical sciences by a mid-
its applications. Efforts to give opportunities to math- career mathematician. Nominations for the inaugural
ematicians of both genders and to integrating early Mirzakhani Prize are due by 7 October 2019. For more
career mathematicians will be particularly appreciated. information see tinyurl.com/yykobras.
i
“NLMS_484” — 2019/9/11 — 16:05 — page 12 — #12
i
12 NEWS
Women in Mathematics and other working in all areas of mathematics and mathematics
Diversity Events education, including teachers, researchers and post-
graduate students who can demonstrate a benefit
A call for expressions of interest in running Women in from attending ICME-14.
Mathematics Days, Girls in Mathematics Days and a
new event for 2019/20, Diversity in Mathematics Day, Download the application form at tinyurl.com/yyagyo23.
is now open. The deadline for submitting expressions Email the completed application form to Professor
of interest is Friday 20 September 2019. Chris Budd, University of Bath (mascjb@bath.ac.uk).
The deadline is 30 November 2019.
More details about Women in Mathematics Days are
available at tinyurl.com/y4wy8px3, Girls in Mathe-
matics Days at tinyurl.com/y44t4mdt, and Diversity
Clay Research Fellowships:
in Mathematics Day at tinyurl.com/y2a3pmbg. call for nominations
The Clay Mathematics Institute calls for nomina-
ICME-14 Bursaries tions for its competition for the 2020 Clay Research
The Joint Mathematical Council of the UK has Fellowships. Fellows are selected for their research
launched the ICME bursaries scheme, which is partly achievements and their potential to become lead-
funded by the LMS. The scheme will provide nine ers in research mathematics. All are recent PhDs,
bursaries of £500 each to attend the 14th Interna- and most are selected as they complete their thesis
tional Congress on Mathematical Education to be work. Terms range from one to five years, with most
held from 12 to 19 July 2020 in Shanghai, China given in the upper range of this interval. Fellows are
(tinyurl.com/yyxbvxth). employed by the Clay Mathematics Institute, which
is a US charitable foundation, but may hold their
The bursaries can fund travel, subsistence, child- fellowships anywhere in the USA, Europe, or else-
care, registration fee or preparation of a presentation. where in the world. Nominations should be received
They cannot fund salary or related costs. Applica- by 16 November 2019. To nominate a candidate see
tions are encouraged from the full range of those tinyurl.com/y6stxs6o.
VISITS
i
“NLMS_484” — 2019/9/11 — 16:05 — page 13 — #13
i
LMS BUSINESS 13
i
“NLMS_484” — 2019/9/11 — 16:05 — page 14 — #14
i
14 LMS BUSINESS
i
“NLMS_484” — 2019/9/11 — 16:05 — page 15 — #15
i
LMS BUSINESS 15
LMS Grant Schemes holder. For those mathematicians going to their col-
laborator’s institution, grants of up to £2,000 are
For full details of these grant schemes, and for infor- available to support a visit for collaborative research
mation on how to submit an application form, visit by the grant holder to a country in which mathemat-
www.lms.ac.uk/content/research-grants. ics could be considered to be in a disadvantaged
position.
MATHEMATICS RESEARCH GRANTS
Research Workshop Grants: Grants of between
The deadline is 15 September 2019 for applications £3,000 - £5,000 are available to provide support for
for the following grants, to be considered by the Research Workshops held in the United Kingdom, the
Research Grants Committee at its October meeting. Isle of Man and the Channel Islands.
Conferences Grants (Scheme 1): Grants of up to AMMSI African Mathematics Millennium Science
£7,000 are available to provide partial support for Initiative: Grants of up to £2,000 are available to
conferences held in the United Kingdom. Awards support the attendance of postgraduate students
are made to support the travel, accommodation, at conferences in Africa organised or supported by
subsistence and caring costs for principal speakers, AMMSI. Application forms for LMS-AMMSI grants are
UK-based research students and participants from available at ammsi.africa.
Scheme 5 eligible countries.
MATHEMATICS/COMPUTER SCIENCE RESEARCH
Visits to the UK (Scheme 2): Grants of up to £1,500 GRANTS
are available to provide partial support for a visitor
to the UK, who will give lectures in at least three Computer Science Small Grants (Scheme 7):
separate institutions. Awards are made to the host Grants of up to £1,000 are available to support visits
towards the travel, accommodation and subsistence for collaborative research at the interface of Math-
costs of the visitor. It is expected the host institu- ematics and Computer Science, either by the grant
tions will contribute to the costs of the visitor. holder to another institution within the UK or abroad,
or by a named mathematician from within the UK
Joint Research Groups in the UK (Scheme 3): or abroad to the home base of the grant holder.
Grants of up to £4,000 are available to support joint Deadline: 15 October.
research meetings held by mathematicians who have
a common research interest and who wish to engage GRANTS FOR EARLY CAREER RESEARCHERS
in collaborative activities, working in at least three
The deadline is 15 October 2019 for applications for
different locations (of which at least two must be
the following grants, to be considered by the Early
in the UK). Potential applicants should note that the
grant award covers two years, and it is expected that Career Research Committee in November.
a maximum of four meetings (or an equivalent level Postgraduate Research Conferences (Scheme 8):
of activity) will be held per academic year. Grants of up to £4,000 are available to provide partial
support for conferences held in the United Kingdom,
Research in Pairs (Scheme 4): For those mathemati-
which are organised by and are for postgraduate
cians inviting a collaborator to the UK, grants of up
to £1,200 are available to support a visit for collabo- research students. The grant award will be used to
rative research either by the grant holder to another cover the costs of participants.
institution abroad, or by a named mathematician Celebrating New Appointments (Scheme 9):
from abroad to the home base of the grant holder. Grants of up to £600 are available to provide partial
For those mathematicians collaborating with another support for meetings held in the United Kingdom
UK-based mathematician, grants of up to £600 are to celebrate the new appointment of a lecturer at
available to support a visit for collaborative research. a UK university.
Collaborations with Developing Countries Travel Grants for Early Career Researchers:
(Scheme 5): For those mathematicians inviting a Grants of up to £500 are available to provide par-
collaborator to the UK, grants of up to £3,000 are tial travel and/or accommodation support for UK-
available to support a visit for collaborative research, based Early Career Researchers to attend confer-
by a named mathematician from a country in which ences or undertake research visits either in the
mathematics could be considered to be in a disad- UK or overseas.
vantaged position, to the home base of the grant
i
“NLMS_484” — 2019/9/11 — 16:05 — page 16 — #16
i
16 LMS BUSINESS
Report: LMS Invited Lectures 2019 Thursday 23 May, organised on behalf of the Applied
Probability Section of the Royal Statistical Society.
Professor Asmussen presented applications of phase-
type distributions in life insurance, building from the
graduate-level material he discussed earlier in the
week to the state of the art in the area. Additional,
complementary presentations were given by Burak
Buke (University of Edinburgh), Ayalvadi Ganesh (Uni-
versity of Bristol) and Ronnie Loeffen (University of
Manchester) on various aspects of the theory of
stochastic processes, with applications to financial,
biological and other systems.
A relaxed schedule and the hospitable environment
offered by the ICMS allowed for plenty of interac-
tion between participants at various career stages
Professor Søren Asmussen throughout the meeting, and also time to enjoy the
city of Edinburgh in one of the better weeks of
The LMS Invited Lectures 2019 were delivered by weather this summer has given us.
Professor Søren Asmussen (Aarhus University) on
20-24 May at the International Centre for Mathemat- Fraser Daly and Sergey Foss
ical Sciences in Edinburgh, on advanced topics in Heriot-Watt University
life insurance mathematics. This course was based
on parts of Professor Asmussen’s forthcoming book
Risk and Insurance: A Graduate Text, co-authored with Report: LMS Popular Lectures 2019
Mogens Steffensen (University of Copenhagen).
Professor Asmussen gave a fascinating account of
inhomogeneous Markov models, semi-Markov mod-
els, and their applications in life insurance. Topics
covered in detail included unit-linked insurance, inter-
est rate and mortality rate modelling, and dividends
and bonuses. The course consisted of nine lectures
and two tutorials, and was complemented by an
opening lecture by Takis Konstantopoulos (University
of Liverpool) and a closing lecture by Andrew Cairns
(Heriot-Watt University).
Participants (particularly early-career researchers)
were given the opportunity to present some of
their own work. During the week we heard inter- Peter Higgins
esting presentations from Jason Anquandah (Uni-
versity of Leeds) on Optimal Stopping in a Simple On 26 June 2019 the Institute of Education, London
Model of Unemployment Insurance, Abdul-Lateef Haji- hosted the 2019 LMS Popular Lectures. The audience
Ali (Heriot-Watt University) on Multilevel Monte Carlo enjoyed an evening of mathematics with a journey
for Efficient Risk Estimation, and Lewis Ramsden (Uni- through a life in mathematics and a look at the com-
versity of Hertfordshire) on The Time to Ruin for a plexity of the human brain and some of the basic
Dependent Delayed Capital Injection Risk Model. principles at work in shaping our brains.
In addition, a workshop on Stochastic Processes in Peter Higgins (University of Essex) presented the first
Finance and Beyond was held on the afternoon of lecture and took the audience through his personal
i
“NLMS_484” — 2019/9/11 — 16:05 — page 17 — #17
i
LMS BUSINESS 17
mathematical journey. Going back to his roots and Report: LMS Undergraduate
recounting personal episodes that led to him becom-
ing a professional mathematician. He went on to tell
Summer Schools
the story of how he invented circular Sudoku. Hig-
gins recognised from his work that the knowledge he
had accumulated could be applied to mathematical The LMS Undergraduate Summer Schools are annual
problems in this case relating to Burrows–Wheeler two-week courses, which are held every summer at a
transform, which is an algorithm that rearranges char- UK university. The aim is to introduce modern mathe-
acter strings into runs of similar characters and is matics to the best UK undergraduates who will enter
important in data compression. their final year the following autumn, and encourage
them to think about an academic career in mathe-
matics. The LMS Undergraduate Summer School con-
sists of a combination of short lecture courses with
problem-solving sessions and colloquium-style talks
from leading mathematicians. 50 places are available
to students per year. Gwyn Bellamy, organiser of the
2018 LMS Undergraduate Summer School at Glasgow,
reports on the LMS Undergraduate Summer School
2018. The 2019 LMS Undergraduate Summer School
took place at the University of Leeds from 14–26 July
and the next LMS Undergraduate Summer School
will take place at the University of Swansea from
12-24 July 2020. Nominations for places will open in
Alain Goriely
October 2019.
The second lecture was presented by Alain Goriely In July 2018, 53 extremely eager students from across
(University of Oxford). The outer surface of the brain the United Kingdom descended on Glasgow for a
(cerebral cortex) is very convoluted (wrinkled) so that jam-packed two weeks of mathematical activities
a maximum amount of gray matter (brain neurons) which is the LMS Undergraduate Summer School.
can fit inside the skull, but how do these convolutions Each mathematics department in the UK elects up to
emerge and how is the brain’s geometry related to three applicants to attend, and from these we select
function? Professor Goriely began by looking at the approximately 50 participants. Almost the entire cost
relative sizes of the brains of some famous scientists of the two weeks is covered by the LMS.
and mathematicians, with some interesting results.
Einstein in particular had a small brain compared with In the first week, lecture courses were given by Pro-
Babbage, Helmholtz, Gauss and Dirichlet. He went on fessor Mike Prest (University of Manchester) on The
to describe comparisons in the convolutions of Gauss Compactness Theorem, by Dr Robert Gray (University
and De Morgan and discussed how the brain develops of East Anglia) on The Word Problem in Combinato-
its shape by using mechanics. Gauss had complex rial Group and Semigroup Theory and by Professor
convolutions and De Morgan had convolutions that, Shaun Stevens (University of East Anglia) on Local–
although ’voluminous’, were less intricate. Professor Global Principles in Number Theory. In the second
Goriely explained how different parts of the brain are week, lecture courses were given by Dr Martina Lanini
connected to each other in brain networks and how (Università di Roma Tor Vergata) on Introduction to
neurodegenerative diseases develop. In essence he Schubert Calculus, by Dr Derek Harland (University of
showed how some of the basic principles at work in Leeds) on Fun with Solitons and by Dr David Bourne
shaping the brain can be explained using geometry, (Heriot-Watt University) on Optimal Transport Theory.
scaling laws and network topology. As one would expect, the lectures were uniformly
fascinating and engaging, and a real pleasure to lis-
If you weren’t able to attend the London lectures, ten to. It was clear that the lecturers had spent a
there is another chance to attend the Lecture in Birm- great deal of time and energy in preparing their lec-
ingham on Thursday 19 September. You can register tures. Moreover, each lecturer had prepared a series
online at lms.ac.uk/events/popular-lectures. of exercises to go with the lectures. Together with
PhD and MSc students from the University of Glas-
John Johnston gow, the lecturers guided the students through the
LMS Communications Officer exercises during the exercise sessions.
i
“NLMS_484” — 2019/9/11 — 16:05 — page 18 — #18
i
18 LMS BUSINESS
In addition to the lecture series, there were collo- Surprisingly, around 10 of the students managed to
quium talks later in the afternoon on most days. solve all the problems.
These were:
There is a real hunger amongst undergraduate mathe-
• Arithmetic and Geometry of Conway Rivers by Pro- matics students in the UK to be exposed to research
fessor Alexander Veselov level mathematics. The LMS Undergraduate Sum-
mer School is one of the few ways in which we
• Chasing the Dragon: Tidal Bores in the UK and Else-
can effectively meet this need. Though it is, as one
where; Quantum and Hawking Radiation Analogies
can imagine, a great deal of work to organize an
by Sir Michael Berry
LMS Undergraduate Summer School, it is also an
• Faster than Fourier (pre)revisited: Vorticulture, Noise, extremely rewarding experience and an effective way
Fractals, Escape. . . by Sir Michael Berry to make a significant contribution to undergraduate
mathematics in the UK. For those LMS members in
• Unexpected Connections. . . by Dr Tom Leinster
a position to apply to organize a summer school, I
• A Tour of the Mandelbrot Set by Dr Holly Krieger would strongly encourage you to do so.
• The Birch and Swinnerton-Dyer Conjecture: From the I sincerely hope that the summer school will have
Ancient Greeks to a 1 Million Dollar Problem of the had a positive influence on each student’s long-term
21st Century by Dr Alex Bartel relationship with the mathematical sciences, and that
• Mathematics, Magic and the Electric Guitar by Dr we’ll see many of them studying for a PhD in the
David Acheson near future. I would also like to give my thanks to the
lecturers and colloquium speakers for all their hard
• Random Games with Finite Groups by Professor work in preparing their talks, and their enthusiasm
Colva Roney-Dougal in delivery.
As one can guess from the above list, these wonder-
Gwyn Bellamy
ful talks took the students on a journey through a
University of Glasgow
huge swath of exciting mathematics, all of which was
beautifully illustrated (and in one case accompanied
by live music!)
On the Sunday between the two weeks of lectures, Report: LMS Northern Regional
there was an organised bus trip to Balmaha on the Meeting and Workshop
shores of Loch Lomond. All the students braved the
weather to climb the nearby Conic Hill, where there
were spectacular views of the loch and nearby Ben
Lomond; at least this was the case until about half
way up, the summit itself was unfortunately shrouded
in mist for the day. Afterwards we retired to the pub
at the base of the hill for a well-deserved drink.
As organiser, by far the most enjoyable aspect of
the summer school was interacting with the partic-
ipants. It was energizing being surrounded by so
many bright, enthusiastic and hardworking young
people! They really gave it their all, and I think this
ensured that they got the most out of the two weeks.
It was particularly gratifying to see them very quickly
bond, seeming to be old friends by the end of the
first week. It was also a great experience to be in Conference participants
exercises classes where the students were asking
questions throughout and racing to try and solve the An LMS Northern Regional Meeting took place in
problems. To provide motivation for the students to the afternoon of Tuesday 28 May 2019, followed by
work on mathematics questions (as it turned out, a wine reception and dinner, and a Workshop on
this was not required at all!), a series of mathematical Higher Dimensional Homological Algebra was held on
problems were set, with prizes for the best solutions. Wednesday 29 May 2019 at Newcastle University.
i
“NLMS_484” — 2019/9/11 — 16:05 — page 19 — #19
i
LMS BUSINESS 19
The LMS Northern Regional Meeting and the work- The workshop was organised by Ilke Canakci and
shop brought to the UK international experts in the Peter Jorgensen. The organisers thank the LMS
representation theory of finite dimensional algebras and the School of Mathematics, Statistics and
to present their work on the most recent develop- Physics at Newcastle University for financial sup-
ments and the speakers included many of the most port. The website of the workshop can be found at
renowned experts in the area. The workshop also tinyurl.com/y4qc2gwt.
gave early career mathematicians such as PhD stu-
dents and postdocs an opportunity to present their Ilke Canakci
work. In total there were nine speakers, of whom one Newcastle University
was a PhD student and three were postdocs.
There were two general talks at the Northern
Regional Meeting on the latest developments in the Report: LMS Graduate Student
field. These talks were given by Karin Baur (University
of Graz and University of Leeds) on CM Modules for
Meeting and the General Meeting
Grassmannians and by Sibylle Schroll (University of of the Society & Aitken Lecture
Leicester) on Geometric Models and Derived Invariants
for Gentle Algebras.
The workshop was focused on Higher Dimensional
Homological Algebra. This area concerns d -cluster
tilting subcategories, which are structures found at
the cusp of algebra and combinatorics. They have a
rich algebraic structure, investigated by homological
methods, and there are many examples relating to
higher dimensional structures in combinatorics, such
as cyclic polytopes.
There were seven talks at the workshop, two of
which were expository talks by Steffen Oppermann Aitken Lecturer Bakh Khoussainov
(Norwegian University of Science and Technology)
on (d + 2)-angulated cluster categories and by Hugh The LMS Graduate Student Meeting and the General
Thomas (Universite du Quebec a Montreal) on Tropical Meeting of the Society & Aitken Lecture were held
Coefficient Dynamics for Higher-Dimensional Cluster on Friday 28 June 2019, a warm and sunny summer’s
Categories. Talks focusing on recent developments day, at Mary Ward House in London. The morning
of higher homological algebra were by Francesca session was kicked off by Dr Robert D. Gray, of the
Fedele (Newcastle University) on A (d + 2)-angulated University of East Anglia, whose lecture served as
generalization of a theorem by Bruning, by Martin Her- an introduction to the material which would be pre-
schend (Uppsala University) on Wide subcategories sented in lectures later in the day. Ideas surrounding
of d -Cluster tilting subcategories for higher Auslander finite presentations of algebraic structures, recur-
algebras, by Karin Marie Jacobsen (NTNU) on d-abelian sive enumerability, and notions such as the word
Quotients of d + 2-angulated Categories, by Gustavo problem were all introduced. This lecture was fol-
Jasso (Universitat Bonn) on Generalised BGP Reflection lowed by a number of graduate student presenta-
Functors via Recollements and by Sondre Kvamme tions, where the broad selection of topics ranged
(Université Paris-Sud) on d -abelian Categories are d - from climate models, modular representation the-
cluster Tilting. ory, and polyhedral combinatorics, among others. By
The atmosphere during the meeting was very pleas- popular vote, two graduate speakers were then each
ant and convivial. The frequent coffee breaks, the awarded a prize: Aras Asaad, for a presentation on
wine reception and dinner followed by the Northern using homology to detect computer generated faces,
Regional Meeting and the lunch at the workshop pro- and Carl-Fredrik Nyberg Brodda, for a presentation
vided great opportunities for discussions, informal on decision problems for groups and semigroups.
exchanges of ideas and for establishing professional After the morning’s Graduate Student Meeting, the
contacts. There were 31 participants, 10 of whom afternoon’s General Meeting was opened by Profes-
were PhD students. sor Caroline Series. While the morning meeting had
i
“NLMS_484” — 2019/9/11 — 16:05 — page 20 — #20
i
20 LMS BUSINESS
The meeting was held at the Herschel Building, Newcastle University, as part of the workshop on Higher
Dimensional Homological Algebra. Over 30 members and guests were present for all or part of the
meeting. The meeting began at 3.10 pm, with the Vice-President, Professor Cathy Hobbs, in the Chair.
There were 31 members elected to Membership at this Society Meeting. Four members signed the
Members’ Book and were admitted to the Society by the Vice-President during the meeting. There were
no Records of Proceedings signed at this meeting of the Society.
Dr Ilke Canakci introduced the first lecture, given by Dr Karin Baur (ETH Zurich/Leeds) on CM-Modules
for Grassmannians.
Dr Ilke Canakci then introduced the second lecture, given by Dr Sibylle Schroll (Leicester) on Geometric
Models and Derived Invariants for Gentle Algebras.
The Vice-President thanked the speakers for their talks, and further extended her warm thanks to
the local organisers, Dr Ilke Canakci and Professor Peter Jorgensen, for hosting such a well-structured
and interesting meeting, and additionally thanked the attendees for coming to the Northern Regional
Meeting.
Dr Ilke Canakci thanked the speakers, and invited all those present to attend a wine reception held in
the Penthouse of the Herschel Building. A Society Dinner was held following the wine reception in the
Herschel Building.
i
“NLMS_484” — 2019/9/11 — 16:05 — page 21 — #21
i
LMS BUSINESS 21
i
“NLMS_484” — 2019/9/11 — 16:05 — page 22 — #22
i
22 FEATURES
Mathematics is useful because we can find things to do with it. With this utility comes ethical issues relating
to how mathematics impacts the world. Now, more than ever, we mathematicians need to be aware of these,
as our mathematics, and our students, are changing society. In the first of a two-part series on Ethics in
Mathematics, we address why, as mathematicians, we need to consider the ethics of what we do.
Mathematics and the world beauty and intrinsic interest, rather than its applica-
tions in science and industry. It’s as though we are
We study one of the most abstract areas of human studying a form of abstract art; far from real world
knowledge: mathematics, the pursuit of absolute impact or considerations, and only fully appreciated
truth. It has unquestionable authority. But, in some by a small number. Despite all this, government and
sense, absolute truths have absolutely no meaning. industry pay for our work; one suspects they don’t
The statement “2 + 3 = 5” is an absolute truth, just do this for the sake of our intellectual stimula-
but what does it mean? Its meaning and utility tion. If our work is completely abstract and detached
are added later when people who understand the — an art form, so to speak — then shouldn’t we seek
statement reconcile it with the physical world. It is funding from those who fund abstract art? So what
the mathematically-trained who interpret and apply might be the value that science councils and industry
mathematics to the real world and thereby assign it see in what we do? It is not only the mathematical
meaning; through this it becomes useful. results that we produce, but also the mathemati-
cians we train. Our mathematics makes a difference,
Indeed, it is clear that mathematics is one of the and our students go out and do real things with
most useful and refined tools ever developed. When their training. If it is the case that our work is being
something is useful, however, it can often also be funded because it has perceived impact, then surely
harmful; this can be either through deliberate misuse we should query and understand why we are being
or ignorance. The humble knife provides an illus- paid to do it.
tration of the principle; in order to use such a tool
responsibly, one must be made aware — often by There is already much discussion of ethics in the
those who first introduce you to it — of the poten- mathematical community. However, these discus-
tial dangers. If a primitive tool like a knife can be sions usually focus exclusively on issues within the
so useful and harmful at the same time, then what community. These are important and many of us are
is mathematics capable of? Mathematics has many already familiar with them: from improving diversity
more applications, and by the same reasoning must and inclusivity, to widening participation in mathe-
also have a greater potential to do ill. As mathemati- matics, to addressing instances of plagiarism and
cians we are seldom warned of this. Other disciplines publishing irregularities. These are pressing concerns.
such as law, medicine, and engineering have, for a Every discipline engages with such intrinsic ethical
long time, addressed the potential for harm within issues. These, however, are not our focus in this
their field. We, as the practitioners and wielders of article. Mathematics is one of the few disciplines that
mathematics, need to be similarly aware, and adjust fails to address extrinsic ethical issues; those con-
our actions accordingly. Otherwise we can, and some- cerning how the community impacts wider society.
times do, cause harm with our work. But how could This particularly includes ethical implications relat-
mathematics possibly be harmful, and what exactly ing to the applications of mathematics and work of
might this harm be? mathematicians. It is these extrinsic ethical issues
that we are trying to raise awareness of. Our concern
In this article our emphasis is on the experiences of is not so much that mathematicians are deliberately
pure mathematicians, although our arguments apply malign, but instead that they fail to recognise these
equally to applied mathematicians, statisticians and extrinsic ethical issues. Indeed, most of the mathe-
computer scientists. Many of us (although certainly maticians we have come across would baulk at the
not all) are motivated to study mathematics by its thought of acting unethically; the problem, instead, is
i
“NLMS_484” — 2019/9/11 — 16:05 — page 23 — #23
i
FEATURES 23
that many do not recognise that mathematical work as such it’s now possible to specify who doesn’t see
can have such an effect. an advert. This allows advertising campaigns that
are selective, that contain adverts that contradict
each other, and that are impossible to externally
Some case studies scrutinise. In short, adverts can now be used to ma-
nipulate individual people. This becomes particularly
Having recognised that mathematics is useful dangerous when applied to political advertising. Us-
because it can be applied, and that with these appli- ing large data sets obtained through social media,
cations come extrinsic ethical issues, we now con- it is possible to profile the political persuasions and
sider two concrete examples: the global financial preferences of an individual. Machine learning has
crisis (GFC) of 2007–8 and targeted advertising. become the main tool of the trade here, and it is
The GFC was one of the defining events that shaped the mathematically trained doing it [3]. These ad-
the modern global economy. Its repercussions have verts can even deceive by appearing non-partisan.
been felt around the world, with many suffering For instance, one can send an advert saying “Voting
a decline in living standards. The causes of the is important; make sure you vote” only to those who
GFC are complex, however, there is consensus that might be inclined to vote for your party. Whatever
mathematical work played a vital role. An impor- the strategy, these types of adverts are increasingly
tant factor is thought to have been the misuse of prevalent, and it is thought that such tactics influ-
Collateralised Debt Obligations (CDOs). These saw enced the 2016 US election and UK referendum on
mathematicians pool large collections of interest- EU membership. It is we mathematicians who make
bearing assets (mostly mortgages), then ‘cut the pool all this possible. Cambridge Analytica, one of the
into pieces’ to form a collection of interest-bearing organisations alleged to have been involved in such
products. Mathematically these products had less advertising, had a small team of no more than 100
overall risk and thus higher value than the original data scientists [4], some of whom were trained math-
assets. They were traded wildly. The mathematics ematicians. Regardless of one’s political persuasion,
behind their construction is highly non-trivial, requir- it is clear that this sort of work is deceptive and
ing stochastic calculus, differential equations, etc. dangerous, and that mathematicians are enabling it.
Research mathematicians, beginning with the work Ultimately, it is mathematicians who make up part
of Black and Scholes, and later Li, derived a model of the teams specifying how such targeting works
and pricing formula for CDOs. Though it took a deep and carrying it out.
understanding of mathematics to derive these mod-
els, only a more superficial understanding (at the
undergraduate level) was required to apply and to The impact of mathematicians
trade them. As a result, their users may not have
fully appreciated their limitations or inner workings. As a result of the pace and scale at which mod-
Mathematics — which by itself is sure and certain ern technology operates, through use of internet
— seemed to explain their value, and so most were connectivity and readily-available fast computation,
happy. Unfortunately, some of the assumptions did the consequences of the actions of mathematicians
not hold. For example, the model assumed there are more quickly realised and far-reaching than ever
wasn’t tail dependence in the default risk of under- before. A mathematician in a big tech company can
lying assets, but there was; for instance when two modify an algorithm, and then have it deployed
mortgaged houses were on the same street. In the almost immediately over a user base of possibly bil-
end, the risk was not properly accounted for, and lions of people. Even on a smaller scale, we have
when house prices declined it led to the write-down seen that a small number of mathematicians, despite
of $700billion of CDO value from 2007 to 2008. The limited resources, can have a vast impact on the
rest is history. world; targeted advertising exemplifies this.
Our next example is targeted advertising. Adverts If you model a physical system, such as gravity, then
have always been placed so as to catch the eye of your model is falsifiable. If the model does not accu-
their desired audience. However, now that people rately reflect the physical system, then on application
possess portable internet-connected devices and it clearly fails — your rocket doesn’t launch prop-
social media accounts, it has become possible to tar- erly. You know when a model was good because the
get adverts at the individual level. Nowadays, these rocket makes it to the moon and back. Modelling a
can be tuned to fit very specific demographics, and financial system is more difficult, as the system is
i
“NLMS_484” — 2019/9/11 — 16:05 — page 24 — #24
i
24 FEATURES
affected by the application of the model. A pricing point that most people will not be interested, it must
algorithm, if widely-used to buy or sell a product, be remembered that some people will be sufficiently
influences the market for the product in question. desperate for credit that there will always be some
How does a model model its own impact? takers. These companies scrape an applicant’s social
media looking for actions they perceive to reflect
So now what happens if you are modelling the future creditworthiness. These could include places the per-
behaviour of people by predicting something like son visits, the hours they sleep, the ‘quality’ of the
‘How likely is a particular individual charged with friends they have, and so on. This approach is unfalsi-
a crime to reoffend with a serious offence, a non- fiable, lacks proper regulation, and has the potential
serious offence, or not reoffend, in the next 24 to harm society since the extension of credit is a
months?’ Furthermore, what if that is being used to mechanism of social mobility. If such a process, one
determine what prosecution and sentencing mecha- that is enabled by mathematically-trained people,
nisms are applied to that person?1 If you predict that starts having negative impact, who is accountable?
a person will reoffend seriously in 24 months, and Ultimately, we must live in the world that we and our
they don’t (after being released or acquitted), then students create, and we must ponder whether there
you might observe that. But what if they are found is a sense in which we are partially responsible.
guilty and sentenced to 25 months, with the choice
of judicial process based on your prediction? How do
you test whether your prediction was correct? Now
we have a serious ethical issue: we are using math-
ematical reasoning to make decisions about people
Do these ethical issues arise in academia?
that impact their lives, and in many of these cases
we can never know whether the decisions made were But what about mathematicians working in academia;
desirable or appropriate. Is it right to use mathemat- are any of these ethical issues relevant to them? Con-
ics in such a way without careful reflection? sider a pure mathematician, a number theorist, say.
We now face an ethical dilemma. Do we limit our- Suppose they develop an algorithm for fast factorisa-
selves to falsifiable claims, or do we allow ourselves tion. Should they publish it? If so, when, where, and
to make claims, make decisions, and initiate actions how? If not, what should they do? Should they have
that are unfalsifiable? We are of course entitled to do thought about it beforehand? We have asked many
the latter, however we should then bear in mind that mathematicians this exact question, and a typical
we have lost mathematical certainty. Furthermore, if response is “I would publish it on arXiv immediately.
we do this, we should broaden our perspective and It’s my right to publish whatever mathematical work
training so that we can incorporate as many aspects I do.” (Not all mathematicians give such a response,
of society as possible. but many do.) When pressed on the consequences
of publishing such an algorithm in that way — for
instance the breaking of RSA encryption in a chaotic
Concerns for the future manner and the ensuing collapse of internet com-
merce and the global economy that would follow —
So what is on the horizon for mathematicians? Is it one explained “Well, it’s their fault for using RSA.
sufficient to simply look at the above list of cases and It’s not my problem.” Of course, responsible disclo-
avoid those specific actions or industries entirely? sure is a complicated topic, and one that is heavily
Unfortunately not; new mathematics produces new debated by security researchers. But with an exam-
ethical issues every day. Such a future example may ple like this, ethics has crept into the world of the
lie in alternative credit scoring. This is starting to be pure mathematics researcher in academia. If an area
done by new companies who lack access to standard as abstract as number theory is not ‘safe’ from ethi-
datasets that established credit-scoring agencies cal considerations, is there any mathematical work
have (such as financial records, bill payment history, that is? Can a pure mathematician hide from ethical
etc.). They instead use different datasets such as issues in academia? What about a statistician, or an
social media profiles, in some cases requesting full applied mathematician? Or do ethical questions arise
access to social media accounts by asking for login for all mathematicians regardless of where we do our
credentials [2]. While this sounds undesirable to the mathematical work?
1 The
Harm Assessment Risk Tool (HART) developed by the Durham Constabulary, which uses random forest machine learning, is used to
make such predictions, and then determine if an accused criminal is to be offered the opportunity of going through the Checkpoint
program (tinyurl.com/y4vxrd77) which is an alternative to criminal prosecution aimed at reducing re-offending.
i
“NLMS_484” — 2019/9/11 — 16:05 — page 25 — #25
i
FEATURES 25
Why management can’t guide us might”. True, someone else might, but they may not
be easy to find, or even exist at all. Now the power
Some mathematicians (academic or industrial) may of meaningful objection has returned to the mathe-
think that, since they are not directly involved in the matician. Whether you choose to take the pragmatic
application of their work, they need not consider its perspective that there are not many mathematicians,
extrinsic ethical implications. After all, we just do the or the logical perspective of the contrapositive, your
maths, and so it’s ‘not our problem’. This oft-held objection means something. Some mathematicians
belief is generally associated with the perception that take this idea even further, and make a conscious
there exist people and structures above us (man- decision to take a seat at the table of power, effect-
agers, supervisors, advisory boards, etc.) who will ing positive change from the managerial level. This
intervene to prevent us from doing anything that we happens in various areas: in academia, in industry,
ought not to. We work on the abstract problems, they and even in politics. This is discussed in more detail
worry about why. But can we rely on management to in [1] as ‘the third level of ethical engagement’.
do this effectively? Will they vet our work, to ensure
that its use is aligned with the values of society? At
each stage of separation from mathematical work,
some understanding of it is lost. It is difficult for Why the law can’t guide us
a manager to understand all of the mathematical
work we do, and its limitations when applied and The problem extends beyond management. We may
used. It is the nature of management that managers think that the law provides a clear description of
will only have partial knowledge of the work being what is and is not acceptable to society, and thereby
done. There would be no point in a manager repro- presumably what is and is not ethical. However, this
ducing the work of all of the people under them, misses the point for several reasons. Firstly, the law
and mathematics is such that if you don’t ‘do it for is not an axiomatised system; it is interpreted by
yourself’ then there is a chance you may not fully courts rather than by machines. This is a type of sys-
understand it. Given this fact, there is always an onus tem with whose details mathematicians are generally
on the individual mathematician to consider the eth- not familiar. Furthermore, there is the problem that
ical implications of what is being done. Of course, it the law will always lag behind technological develop-
must also be considered that managers might have ment; we cannot expect lawmakers to have done our
other values, perhaps more aligned with the objec- mathematics before we do it ourselves. Additionally,
tives of the organisation than of wider society. We the processes by which laws are made are (deliber-
should understand and anticipate this. ately) slow, requiring public consultation, votes, and
implementation periods. Consider the case of the
Some managers may go so far as to try to manipu- General Data Protection Regulation (GDPR). It started
late us. For instance, if we voice objection at what to be written in 2011, only came in to effect in 2018,
we have been asked to do, they may try to quash it and is thought by many to be already out of date.
with the classic argument: “if you don’t do it, then Finally, it can be the case that lawmakers lack a full
someone else will”. At a first glance, this seems understanding of the fine details of the subject at
convincing, however, it fails on two counts when hand. For example, a member of the UK Science
referring to mathematical work. First, there are not and Technology Select Committee, Stephen Metcalfe,
that many mathematicians in the world. We possess declared at a public outreach event that “one solu-
a unique set of skills and abilities, and it requires tion to algorithmic bias is the use of algorithms to
years of training to produce a good mathematician, check algorithms, and the use of algorithms to check
even when starting with someone who has the right training data”. Ultimately, the law is not there to serve
interests and reflexes. Given the scarcity of mathe- as moral advice; there are plenty of immoral things
maticians, this argument fails in practice. Moreover, one can do that do not break any laws. As such, it is
as mathematicians we understand its contrapositive; not well-suited as a source of ethical advice.
the original statement is equivalent to “If no-one else
does it, then you will”. This is, of course, absurd. The Thus, if we can’t rely on management and we can’t
argument has as the implicit underlying assumption rely on lawmakers and regulators, then who can we
that the task being requested will definitely be com- rely on? The answer is as obvious as it is difficult
pleted. If no-one else builds me a nuclear bomb, then to admit: ourselves. The only way mathematicians
will you? What we should really be considering here is can try to prevent their work from being used to do
the argument “If you don’t do it, then someone else harm is if they think about it themselves. No one
else can, so we must.
i
“NLMS_484” — 2019/9/11 — 16:05 — page 26 — #26
i
26 FEATURES
A growing awareness of Ethics in Mathematics of the Newsletter, we’ll further explore why such
teaching has not yet occurred in mathematics, and
Awareness that mathematicians need to consider ex- outline how one might go about giving such directed
trinsic ethical issues is building in the community. In teaching of Ethics in Mathematics.
2018 the head of mathematics at Oxford, Professor
Mike Giles, commented at a panel discussion event:
“Cambridge Analytica is interesting from one point of
view in that, if you’d asked me 20 years ago whether Acknowledgements
mathematicians at the PhD level needed to be ex-
posed to ideas of ethics, I would have said ‘Clearly,
that is irrelevant to mathematicians’. Now I really We wish to thank Piers Bursill-Hall and Dennis Müller
think that this is something we have to think about. for their valuable discussions over the past few years,
In the same way that engineers have courses looking and Dennis in particular for his comments and sug-
at ‘What it means to be a professional engineer’, and gestions on previous versions of this article.
‘Ethics, and your responsibilities as an engineer’, I
think that is something that we have to think about
as mathematicians now.” Moreover, arxiv.org is cur- FURTHER READING
rently revising the description of their mathematics
tag History and Overview to include “Ethics in Mathe- [1] M. Chiodo, P. Bursill-Hall, Four levels of ethi-
matics” as a sub-category. cal engagement, Ethics in Mathematics Discussion
Papers, 2018/1 (2018).
As part of their formal training, few mathematicians [2] Systems and methods for using online social
have ever been told about extrinsic ethics before. footprint for affecting lending performance and
Previous generations of mathematicians have evaded credit scoring, Patent No. US 8,694.401 B2 , 8 April
this crucial point, and in the process have possibly let 2014.
society down. It rests on the current and upcoming [3] Cambridge Analytica: how did it turn clicks into
generations to pick up this idea, before it’s too late. votes?, theguardian.com, tinyurl.com/yy89e5vl,
Mathematicians always take a generation or more to May 6th 2018.
accept a new and fundamental idea about the nature [4] Secrets of Silicon Valley, Series 1 Episode 2:
of their subject; debates about the admissibility of The Persuasion Machine, see recording at 30:17.
zero as a number provide such an example. We’re BBC documentary, 2017
at a similar juncture again. Now some say “Surely
there’s no use in considering ethical issues in mathe- Maurice Chiodo
matics”, but by the time our students are professors
and industry leaders, they may well be saying “Of Maurice is a postdoc-
course we should be considering ethical issues in toral research fellow
mathematics!” But why hasn’t the mathematical com- at the University of
munity taken this on board already? Why wasn’t this Cambridge, and lead
done 100 years ago, by the likes of Gödel and Russell? investigator of the Cam-
Two reasons come to mind. Firstly, the dangers were bridge University Ethics
less proximate, since much of today’s technology in Mathematics Project
simply did not exist. Secondly, every mathematics (ethics.maths.cam.ac.uk). He’s developing a teaching
undergraduate was already exposed to philosophy, programme and curriculum to teach mathematicians
as it formed part of every university education. Thus, about the ethical implications of their work.
exposure to Ethics in Mathematics, in its own right,
was less urgently needed. Toby Clifton
So if Ethics in Mathematics has become so important Toby is a recent grad-
to mathematicians, then how might we teach it to uate of astrophysics,
them in a relevant and useful way, without foisting and is the current pres-
an entire philosophy degree upon them? Disciplines ident of the Cambridge
such as law, medicine, and engineering have long University Ethics in
taught their undergraduate students about extrinsic Mathematics Society
ethics in their respective fields. In the next issue (cueims.soc.srcf.net).
i
“NLMS_484” — 2019/9/11 — 16:05 — page 27 — #27
i
FEATURES 27
The arithmetic of Drinfeld modules has recently yielded novel algorithms for factoring polynomials over finite
fields; a computational problem with applications to digital communication, cryptography and complexity
theory. We offer a gentle invitation to these developments, assuming no prior knowledge of Drinfeld modules.
Factoring polynomials modulo a prime uct of irreducible polynomials of the same degree.
By Fermat’s little theorem, x q − x = a ∈Fq x − a. To
Q
Let Fq denote the finite field of integers modulo an extract the product of linear factors of f(x), take the
odd prime number q . Polynomials Fq [x] over Fq share greatest common divisor gcd(x q − x, f(x)). To extract
striking analogies with integers, yet we begin with products of degree two factors, degree three factors
an algorithmic distinction. While factoring integers and so on iteratively, look to the succinct expression
remains a notoriously difficult problem, factoring poly- d
Y
nomials in Fq [x] is long known to be easy, at least xq − x = p(x)
with access to randomness. p:deg(p)|d
Polynomial factorisation over finite fields is not a for the product of monic irreducible polynomials p(x)
mere curiosity, but has many applications. In num- of degree dividing d . At the d th iteration, with smaller
d
ber theory, finite fields arise as residue fields of degree factors already removed, gcd(x q − x, f(x))
global fields such as number fields. While deter- yields the product of degree d irreducible factors.
mining the splitting of a prime in a number field, d
one factors a polynomial defining the number field Care in handling x q is required for its degree is expo-
d
modulo the prime. Several instances of polynomial nential in n log q . All we need is x q mod f(x), easily
factorisation appear while factoring integers using accomplished by a sequence of q th powers modulo
quadratic/number field sieve algorithms or while per- f(x), each performed by repeated squaring. Better
d
forming index calculus to compute discrete loga- still, x q mod f(x) can be rapidly computed with a
rithms, both foundational problems in analysing the fast algorithm to compose two polynomials modulo
security of cryptographic systems. In digital com- f(x) in concert with q th powers. Kaltofen and Shoup
munication, polynomial factorisation aids the con- devised an ingenious improvement over this naive
struction of certain error correcting codes (BCH and iteration resulting, in a significant speed up. The
cyclic redundancy codes), structures vital to reliable Kaltofen–Shoup algorithm implemented using the
transmission of information in the presence of noise. modular composition algorithm of Kedlaya–Umans
performs distinct degree factorisation with run time
Let us recount a simple polynomial factorisation exponent 3/2 in the degree n and 2 in log q .
algorithm. We are given a monic f(x) ∈ Fq [x] of
degree n whose factorisation into irreducible poly-
Equal degree factorisation
nomials f(x) = im=1 pi (x) is sought. Assume f(x)
Q
is square free, that is the pi (x) are distinct. This is Distinct degree factorization leaves us with the prob-
without loss of generality for there are algorithms to lem of factoring polynomials all of whose irreducible
rapidly reduce to this special case. It takes n log2 q factors are of the same known degree d . All known
bits to write down f(x) and we seek algorithms that algorithms for this task with polynomial runtime in
run in time polynomial in n and log q . Berlekamp was log q are randomized. Even for the simplest case
the first to show there is a randomized polynomial of factoring a quadratic polynomial into two linear
time algorithm, but we follow a different two step factors, no unconditional deterministic polynomial
process. time algorithms are known. It is closely related to the
problem of finding a quadratic nonresidue modulo a
Distinct degree factorisation given large prime.
The first step, known as distinct degree factorisation, The following randomized algorithm can be traced
decomposes f(x) into factors each of which is a prod- backed to ideas of Gauss and Legendre. For a uni-
i
“NLMS_484” — 2019/9/11 — 16:05 — page 28 — #28
i
28 FEATURES
i
“NLMS_484” — 2019/9/11 — 16:05 — page 29 — #29
i
FEATURES 29
i
Y
• If A Fq [x]/(p(x)) for a monic irreducible p(x), = pi (x) + tφ,pi (x) = f(x) + tφ,f (x),
• If 0 → A1 → A → A 2 → 0 is exact, then χ(A) = for some tφ,f (x) of degree at most s f /2, where s f
χ(A1 ) χ(A2 ). denotes the degree of the smallest degree factor of
f(x). Thus, we have an extension of Gekeler’s bound
to reducible polynomials
For the Fq [x] module φ(Fq [x]/(f(x))) featuring in
our algorithms, the Euler–Poincaré characteristic χ φ Fq [x] (f(x)) = f(x) + tφ,f (x),
.
χ(φ(Fq [x]/(f(x)))) has a simple linear algebraic inter- |{z}
≤s f /2
pretation: the characteristic polynomial of the linear
map φx on Fq [x]/(f(x)). In particular, it is a degree n implying f(x) and χ(φ(Fq [x]/(f(x)))) agree at the high
polynomial that can be computed efficiently. degree coefficients! The number of agreements tells
us information about the smallest factor degree.
Frobenius distribution of Drinfeld modules For a randomly chosen φ, tφ,f (x) likely has degree
Let us put our newly defined Fq [x] modules and car- exactly bs f /2c (with probability at least 1/4). The
dinality measure χ to use. Take an elliptic curve E number of agreements does not merely bound but
over the rational numbers and reduce it at a prime p. determines the degree of the smallest factor. To
The Fp -rational points E(Fp ) famously form a finite claim this probability, one needs to prove for a ran-
abelian group with cardinality p + 1 up to an error domly chosen φ, the Frobenius traces corresponding
determined by the Frobenius trace tE,p . The Hasse to the irreducibles of smallest degree do not conspire
bound, considered the Riemann hypothesis for ellip- yielding cancellations. To this end, we seek equidistri-
√
tic curves over finite fields, asserts that |tE,p | ≤ 2 p. bution formulae for the Frobenius traces. Analogous
Thereby, to elliptic curves, there is a correspondence between
the number of isomorphism classes of Drinfeld mod-
#(E(Z/(p))) = p + 1 − tE,p . ules with a given trace and Gauss class numbers in
√ √ certain imaginary quadratic orders. The latter can be
|{z}
−2 p ≤ ≤2 p
computed using analytic class number formulae.
Gekeler established the following Drinfeld module
An algorithm to estimate the degree of the smallest
analogue. Take a Drinfeld module φ, a monic irre-
degree factor of a given f(x) is now apparent. Pick
ducible polynomial p(x), and consider the resulting
a Drinfeld module φ (by choosing gφ, ∆φ at random
Fq [x] module φ(Fq [x]/(p(x))). Its Euler–Poincaré char-
of degree less than n). Compute the Euler–Poincaré
acteristic equals
characteristic χ(φ(Fq [x]/(f(x)))) and count the num-
ber of high degree coefficients it agrees in with f(x).
χ φ Fq [x] (p(x)) = p(x) + tφ,p (x) ,
.
|{z}
≤deg(p)/2
Drinfeld module analogue of Lenstra’s algorithm
which is p(x) plus an error determined by the Frobe-
nius trace t φ,p (x) of degree at most half that of p(x). It is instructive to begin with Lenstra’s elliptic curve
The analogy with the Hasse bound is striking. The integer factorisation algorithm before seeing its Drin-
error in each case takes roughly half the number of feld module incarnation. Pollard designed his p-1
bits as the estimate. algorithm to factor an integer that has a prime factor
modulo which the multiplicative group has smooth
order. But this smoothness condition is rarely met.
Factor degree by Euler–Poincaré characteristic Lenstra recast the role of the multiplicative group
with the additive group associated with a random
Gekeler’s bound concerns Drinfeld modules φ at an elliptic curve. If the integer has a prime factor mod-
irreducible p(x). What happens at f(x) = i pi (x), ulo which the randomly chosen elliptic curve has
Q
i
“NLMS_484” — 2019/9/11 — 16:05 — page 30 — #30
i
30 FEATURES
smooth order, the algorithm succeeds in extracting and apply the Drinfeld action φr (α) on α. It is likely
that factor. For a random elliptic curve, this smooth- that φr (a) is a zero at some but not all factors pi (x)
ness condition is met with a probability depending of f(x) and gcd(φr (a), f) gives a random factorisation.
sub-exponentially on the size of the smallest prime
factor. Consequently, it is among the most popular A brief outline of why this is the case follows. As with
algorithms for integer factorisation, particularly as groups, the order of an element divides the cardinal-
an initial step to extract small factors. ity of an Fq [x] module. That is, Ord(α) divides the
Euler–Poincaré characteristic of φ(Fq [x]/(f(x))). In
fact, with high probability, Ord(α) equals the Euler–
Pollard’s p − 1 algorithm
Poincaré characteristic
Fix a positive integer B as the smoothness bound
and denote by m, the product of all prime powers Y
Ord(α) = pi (x) + tφ,pi (x) .
at most B. Given an N to factor, choose a positive
i
integer a < N at random. Assume a is prime to N
for otherwise gcd(a, N ) is a nontrivial factor of N . If the factors on the right have/don’t have a linear fac-
If N has a prime factor p with every prime power tor independently and roughly uniformly at random,
factor of p − 1 at most B, then the algorithm yields a random factorisation. This
a m −1 = (a p−1 )m/(p−1) −1 0 mod p =⇒ p | a m −1 is indeed the case!
and gcd(a m − 1, N ) is likely a nontrivial factor of N . The factors on the right lie in the short intervals
The running time is exponential in the size of B. Ii centred at pi (x) with interval degree bounded
For typical N , B needs to be as big as the smallest by deg(pi )/2. The Frobenius trace distribution as-
factor of N and thus the running time is typically sures a certain semi-circular equidistribution of
exponential in the size of the smallest factor of N . pi (x) + tφ,pi (x) in the short interval Ii . Remarkably,
unlike over integers, factorisation patterns in the
short intervals Ii are unconditionally proven to be
Lenstra’s algorithm
random enough. In summary, for random φ, the fac-
Lenstra’s elliptic curve factorization algorithm factors torisation patterns of each pi (x) + tφ,pi (x) is like that
every N in (heuristic) expected time sub-exponential of a random degree deg(pi ) polynomial. Further, they
in the size of the smallest factor p of N . A key insight are independent.
of Lenstra was to substitute the multiplicative group
(Z/pZ)× in Pollard’s p − 1 algorithm with the group The computation of Ord(α) dominates the runtime
E(Fp ) of Fp rational points of a random elliptic curve and can be performed efficiently through linear
E over Fp . The running time depends on the smooth- algebra. This is in stark contrast to integers, where
ness of the group order |E(Fp )| for a randomly chosen finding the order of an element in the multiplicative
E. The Hasse bound guarantees that ||E(Fp )| − (p + group modulo a composite appears hard.
√
1)| ≤ 2 p and Lenstra proved that his algorithm runs
in expected time sub-exponential in the size of p as-
suming a heuristic on the probability that a random Drinfeld modules with complex multiplication:
√ √
integer in the interval [p + 1 − 2 p, p + 1 + 2 p] is
smooth.
Our last algorithm is distinguished in that it samples
Drinfeld module analogue from Drinfeld modules with complex multiplication.
A Drinfeld module φ has complex multiplication if its
In the ensuing Drinfeld module version, random ellip- endomorphism ring
tic curves will be recast with random Drinfeld mod-
ules to factor polynomials. As before, let f(x) ∈ Fq [x] EndFq (x) (φ) ⊗Fq [x] (Fq (x))
denote the polynomial to factor. Pick a Drinfeld mod-
ule φ at random and a random element α in the is isomorphic to a quadratic extension of Fq (x). A typ-
Fq [x] module φ(Fq [x]/(f(x))). The order, Ord(α), of ical Drinfeld module has an endomorphism ring only
α is the smallest degree polynomial b(x) ∈ Fq (x) that isomorphic to Fq (x). Complex multiplication is the
annihilates α, that is φb (α) = 0. Extract and divide rare case where the Drinfeld module is more symmet-
away the linear factors of Ord(α), ric than typical. As with reducing elliptic curves over
rational numbers at primes, Drinfeld modules can
r(x) := Ord(α)/ gcd(Ord(α), x q − x) be reduced at irreducible polynomials. The reduction
i
“NLMS_484” — 2019/9/11 — 16:05 — page 31 — #31
i
FEATURES 31
i
“NLMS_484” — 2019/9/11 — 16:05 — page 32 — #32
i
32 FEATURES
1
Mathematicians currently use computers to do tedious calculations which would be unfeasible to do by hand.
Abstract. Mathematicians
In the future, currently
could they be helpinguse
us computers to do tedious
to prove theorems, or to calculations which
teach students would
how be unfeasible
to write proofs? to do
by hand. In the future, could they be helping us to prove theorems, or to teach students how to write proofs?
Mathematics from the future analysis, topology and so on. Were software like this
Mathematics from the future to
to be
be adopted
adopted by by aa broader
broaderclass
classof
ofmathematicians,
mathematicians,
we
we might
might see
see aa future
future where
where these
these systems
systems start
start
Take a look at the following piece of computer code. to
to become
become useful
useful for
for aa broader
broaderclass
classof
ofresearchers
researchers
Take a look at the following piece of computer code. too.
too.
lemma continuous_iff_is_closed
{f : α → β} : In
In this
this article
article we
we will
will see
see an an overview
overviewof ofwhywhythese
these
continuous f ↔ (∀s, is_closed s → systems
systems existexist and
and what
what theythey arearecurrently
currentlycapablecapable
is_closed (f −1 ’ s)) := of.
of. They are getting
They are getting better,
better,faster,
faster,andandsmarter
smarterevery ev-
hassume hf s hs, hf (−s) hs, ery
year,year,
and and I believe
I believe thatthat
it is itonly
is only
a mattera matterof timeof
assume hf s, by rw [←is_closed_compl_iff, time
←is_closed_compl_iff]; exact hf _i until until mathematicans
mathematicans willforced
will be be forced
to sittoup sitand
up and
take
take
notice.notice. Note however
Note however that computers
that computers will not will not
be prov-
be
ingproving
theorems theorems by themselves
by themselves any timeany soon. time
All soon.
of the
This
This isis aa proof
proof that
that aa function
function is is continuous
continuous ifif and and All of the mathematics I have
only
mathematics I have done in done
Lean in Lean myself
myself is mathe- is
only ifif the
the preimage
preimage of of every
every closed
closed setset isis closed.
closed. mathematics which I could have done
matics which I could have done (in many cases more (in many cases
It
It is
is written
written inin aa computer
computer language
language called
called Lean,
Lean, aa more
programming easily)easily)
with with
pen pen
and andpaper.paper.TheThekeykey difference
difference of
programming languagelanguage developed
developed by by Leonardo
Leonardo de de of course is that with the pen
course is that with the pen and paper method,and paper method,
I could
Moura
Moura at at Microsoft
Microsoft Research.
Research. There
There isis aa “:=”
“:=” some-
some- Ihave
couldmade
have slips,
madeorslips, or forgotten to everything
check ev-
where forgotten to check
where in in the
the middle
middle of of the
the code;
code; the
the statement
statement of of erything which needs checking; Lean
that needs checking; Lean will make sure everything will make sure
the
the theorem is before this token, and itit is
theorem is before this token, and is basically
basically everything is correct before it will compile your proof.
human-readable. is correct before it will compile your proof. There
human-readable. The The proof
proof ofof the
the theorem
theorem is is after
after There are typecheckers
are typecheckers writtenwritten
in otherinlanguages
other languages
that can
the
the token, and is incomprehensible if you do not
token, and is incomprehensible if you do that
know
not check Lean proofs, so the chances of errorsofslipping
can check Lean proofs, so the chances errors
know how how totospeak
speakLean.
Lean.A A computer
computer withwithLean in-
Lean slipping through the system nowadays
through the system nowadays are remote — cer- are remote –
stalled
installed would have no trouble checking that this ais
would have no trouble checking that this certainly an order of magnitude less than the chance
valid tainly an order of magnitude less than the chance
a validproof,
proof,and
anda ahuman
humanmathematician
mathematician who who cancan that
thataahuman
humanwill willmake
makeaamistake.
mistake.As AsI Iget
getolder,
older,I Ifind
find
speak
speak Lean would vouch for the fact that the proof
Lean would vouch for the fact that the II value
is
proof value this certainty more. Perhaps this is somekind
this certainty more. Perhaps this is some kind
is just
just the
the obvious
obvious proof.proof. of
of mid-life
mid-life crisis—but
crisis — butLean Leanisisfreefreeand
andopen opensource
source
We software,
software, and hence much cheaper than buying
and hence much cheaper than buyingaa
We see
see from
from this
this example
example that
that computers
computers can can now
now Ferrari.
check proofs of simple theorems in pure mathemat-
check proofs of simple theorems in pure mathemat- Ferrari.
ics,
ics, as
as long
long as
as you
you are
are prepared
prepared toto learn
learn how
how toto talk
talk
to
to them in their own specialised language. In fact this
them in their own specialised language. In fact this
has
has been
been the
the case
case for
for decades,
decades, and
and evidence
evidence has Proofs
appeared more recently that computers can check
has Proofs and
and computer
computer programs
programs
appeared more recently that computers can check
quite complicated proofs. One observation which I
quite complicated proofs. One observation which I
find particularly striking is that even though these Most of us now rely on computer code to do all
find particularly striking is that even though these Most of us now rely on computer code to do all sorts
computer proof verification systems give us a very sorts of things. Bugs in code can cause chaos, al-
computer proof verification systems give us a very of things. Bugs in code can cause chaos, allowing
cool new way of doing mathematics, which also has lowing hackers into systems or just making stuff
cool new way of doing mathematics, which also has hackers into systems or just making stuff stop work-
applications to teaching and outreach (indeed I now stop working. It is hence not surprising that software
applications to teaching and outreach (indeed I now ing. It is hence not surprising that software devel-
use it for both), the number of people in mathemat- developers have made tools which can check com-
use it for both), the number of people in mathemat- opers have made tools which can check computer
ics departments across the UK who can use this sort puter programs for bugs. But what is a computer
ics departments across the UK who can use this sort programs for bugs. But what is a computer program?
of software is extremely small. Moreover, most of program? It is just is a sequence of statements fol-
of software is extremely small. Moreover, most of It is just is a sequence of statements following a
them are doing research in foundational areas such lowing a precise syntax, which performs a certain
them are doing research in foundational areas such precise syntax, which performs a certain task when
as category theory or type theory rather than more task when the statements are interpreted within the
as category theory or type theory rather than more the statements are interpreted within the framework
“mainstream” areas in pure mathematics such as framework of a particular programming language
“mainstream” areas in pure mathematics such as of a particular programming language (say, Python
analysis, topology and so on. Were software like this (say, Python or C++). A mathematical proof is also a
i
“NLMS_484” — 2019/9/11 — 16:05 — page 33 — #33
i
FEATURES 33
or C++). A mathematical proof is also a sequence expressed concern that this part of the argument
of statements following a precise syntax, which per- was not human-checkable. They have now all been
forms a certain task when the statements are inter- formally verified by computer.
preted within the logic and language of pure math-
ematics. This observation is more than an analogy The system Coq was the system used in George
and goes back to Kolmogorov. Gonthier’s formal proof of the four colour theorem,
and Gonthier also led the team which in 2013 formal-
Initially it was observed that certain kinds of mathe- ized (again in Coq) the proof of the Feit–Thompson
matical proofs (namely, proofs using something called theorem that every group of odd order is solvable.
“constructive mathematics”) really are exactly the These two theorems are of different natures. The
same thing as certain kinds of computer programs. four colour theorem is another example of a theo-
As a consequence, the tools developed by computer rem where the standard argument ends up having
scientists to verify that code has no bugs can also to go through thousands of tedious case checks, but
be used to show that a proof in constructive mathe- this is not what the Feit–Thompson proof looks like;
matics has no gaps or errors. the latter proof is a long group-theoretic argument
for which Thompson recieved a Fields Medal in 1970.
However, often mathematicians do not work purely The modern mathematical proof of the odd order
in constructive mathematics. One key difference theorem is written up in two books by Peterfalvi and
between a proof and a program is that proofs are Bender–Glauberman, both published in the LMS lec-
allowed to use non-constructive things such as the ture note series [1, 3]. The Coq proof was created by
axiom of choice, which can be viewed in this optic a team of 15 people over a period of 6 years, during
as a function for which there is no algorithm, i.e., which time they formalized the 328 pages of research
no program. But some of the newer computer proof level mathematics in the two books as well as all the
checking tools (such as Lean) have been designed prerequisites. Perhaps surprisingly, over two thirds
with mathematics in mind, and have simply added of the code base is the prerequisites for the books
things like the axiom of choice as an axiom of the rather than the books themselves. The prerequisites
system. This means that tools like Lean can now covered basic results in the representation theory
check “normal” mathematical proofs. In short, if you of finite groups, Galois theory, some commutative
can prove something using classical logic and ZFC set algebra, and a little number theory. Moreover, most
theory (which is what most pure mathematicians just of the people involved did not work anywhere near
call “mathematics”) then you can, in theory, prove it the full six years that the project lasted. Ultimately
in Lean. there were under five lines of computer code corre-
sponding to each line of the book on average, a figure
which perhaps better reflects the effort, although it
should be said that writing a correct line of code in
What mathematics is formalized? one of these languages can take a long time.
The work of Hales et al., and Gonthier et al., are hence
What however can we prove in practice in Lean or in evidence that these computer systems are now capa-
these other systems? Which results are already avail- ble of handling very long and complex proofs about
able to us? Let us start at the top, and go through relatively simple objects such as finite groups or
some of the traditional breakthrough results. spheres.
The systems HOL Light and Isabelle/HOL were used In 2016, Ellenberg and Gijswijt established a new
in the formally verified proof of the Kepler conjec- upper bound for the cap set problem, resolving the
ture (“HOL” stands for Higher Order Logic). The for- cap set conjecture. The cap set conjecture is a com-
mal verification was done by a team led by Tom binatorial question about the growth of the size of
Hales. Hales’ 2017 talk at the Big Proof workshop a certain sequence related to an n-dimensional ver-
at the Newton Institute is available on the INI web- sion of the card game “Set”, as n increases. The
site (see tinyurl.com/y4avuzj9) and tells the interest- Ellenberg–Gijswijt proof was published in the Annals
ing story of how he ended up having to formalize of Mathematics. In 2019 Sander Dahmen, Johannes
the paper proof which he and Ferguson had con- Hölzl and Rob Lewis formalized the proof in Lean.
structed in the late 1990s. The argument involved, at Dahmen is a number theorist; Hölzl’s background
some point, checking thousands and thousands of is in computer science and Lewis’s is in logic. The
nonlinear inequalities, and the human referees had mathematical proof was remarkable in the sense
i
“NLMS_484” — 2019/9/11 — 16:05 — page 34 — #34
i
34 FEATURES
that many of the methods used were elementary, degree. Formalising undergraduate level mathematics
making the result a very good candidate for formali- is a very different ball game. It takes far less time,
sation. The Lean proof is modern mathematics being and can be worked on by undergraduates who are
formalized in a realistic time frame. learning about the subject for the first time. All of
the systems I have mentioned so far contain a lot
The next example I will speak about here is of a of undergraduate-level mathematics, but one cannot
different nature; it is the work of Sébastien Gouëzel easily port theorems from one system to another
and Vladimir Shchur from [2]. The Morse Lemma (the problem is analogous to porting computer pro-
is a result about the Hausdorff distance between a grams from one language to another), so when new
geodesic and a quasi-geodesic with the same end- systems come along these libraries need to be built
points in a Gromov-hyperbolic space. In 2013 Shchur from scratch. Each system contains some, but not all,
gave a quantitative (i.e., with explict constants) proof of the pure mathematics component of a typical UK
of the lemma, which was published in the Journal undergraduate degree; the union of all the systems
of Functional Analysis. Gouëzel decided to formalize might well contain most of it nowadays, however
the proof in Isabelle/HOL, and during the process he because of the portability issue no one system has
found a serious error missed by both author and access to all of it.
referee. A completely new argument was found by
Gouëzel and Shchur to fix the problem, and Gouëzel I started learning Lean by formalising the questions
formalized the new proof in Isabelle/HOL. The result and solutions in Imperial’s first year “introduction to
was written up as a joint paper which was published proof” course, a course which I was lecturing at the
in the Journal of Functional Analysis earlier this year. time. I typically got stuck several times per question,
Again this is modern mathematics being formalized I asked hundreds of questions in the Lean chat-
in a realistic time frame, but this time the objects room on Zulip and they were answered promptly and
are more complex. politely, often by computer science PhD students and
post-docs. When I had got the hang of things a little
Last year I got interested in exactly how complex the more, I started encouraging Imperial’s undergradu-
objects handled by this sort of software could be. ates to join in, and we have made good progress.
Perfectoid spaces were discovered by Peter Scholze Over the last two years, mathematics undergradu-
in 2011 and his applications of the theory to fun- ates at Imperial have been formalising results which
damental questions in the Langlands program and they have been taught during their degree, or learnt
arithmetic geometry won him a Fields Medal in 2018. independently. For example the theorem of quadratic
Earlier this year Patrick Massot, Johan Commelin and reciprocity, Sylow’s theorems, the fundamental theo-
myself formalized the definition of a perfectoid space rem of algebra, matrices, bilinear maps, the theory
in Lean. The formalisation process took around ten of localisation of rings, the sine, cosine and exponen-
months in total, and we wrote well over 10,000 lines tial functions, tensor products of modules and many
of code in order to formalize this definition, involving other things have been implemented by Imperial’s
(for example) formally proving many of the theo- mathematics undergraduates in Lean. Chris Hughes
rems in Bourbaki’s “Topologie Générale”. We have and Kenny Lau in particular are two students who
as yet proved nothing at all about perfectoid spaces have formalized many results from our undergrad-
and furthermore cannot yet produce any non-trivial uate curriculum. There is more to come; we are
examples, because we still need to formalize the gen- working on the Galois theory course and after that
eral theory of discrete valuations on finite extensions we will start on the representation theory course.
of the p-adic numbers. Formalising this theory would One possible future project would be to turn some
be an ideal problem for a PhD student who wanted of these formalisation projects into some sort of
to learn the material in an innovative way, however formally verified book or website; imagine a book on
it would be far less efficient in terms of time. Galois theory which was guaranteed to contain no
errors, and which every proof which the reader did
not follow could be “unfolded” more and more until
the penny dropped.
Undergraduate level formalized mathematics
In 2018–19 I also supervised BSc student Anca
Ciobanu, who formalized the basics of group coho-
As well as these “complex” formalisations, there are mology in Lean, and MSc student Ramon Mir who
formalisions of much of the pure material which is formalized Grothendieck’s notion of a scheme in
typically taught in a UK undergraduate mathematics
i
“NLMS_484” — 2019/9/11 — 16:05 — page 35 — #35
i
FEATURES 35
Lean. I was very surprised to discover that schemes just as knowledge of LaTeX now is. Another purpose
had not been done in any other formal system. As is an ongoing project to completely digitise the pure
a result it seems that Mir’s work is of a publishable mathematics part of Imperial’s undergraduate cur-
standard. Nowadays we see students typing up their riculum.
lecture notes in LaTeX; it would be wonderful to see,
in the future, students typing up their course notes I have also used Lean for outreach, showing
in Lean or one of these other systems. schoolkids how to prove things like 2 + 2 = 4 from
the axioms of Peano arithmetic and then raising the
question of how much more mathematics can be
done like this.
Using Lean for teaching.
In the 2018–19 academic year, I taught Imperial’s Using Lean to help with research?
introduction to proof course again, this time using
Lean live in the lectures for some of the proofs. For
example, I proved that a composite of surjective func- Currently, the only thing that Lean can offer the
tions was surjective in Lean, I proved various results researcher is a “certificate” that any theorem they
about equivalence relations in Lean, and I proved have formalized in Lean is indeed correct. Many
Cantor’s diagonal argument in Lean. I wrote the code humans feel that they do not need this certificate. I
live during the lectures, projecting my laptop onto find it interesting that more and more mathematics
the screen, but I also circulated “traditional” proofs relies on unpublished work; talk to any pure mathe-
in the course notes afterwards. All of this was done matician in any big area and they will know of some
under the watchful eye of Athina Thoma, a specialist theorems which rely on results in the “grey literature”
in mathematical education, who interviewed and sur- (a letter or an email from X to Y) or even results
veyed the students and is writing up her conclusions which are not in the literature at all (“a forthcoming
for a forthcoming publication with Paola Iannone, paper”, sometimes announced years ago and which
another such specialist based in Loughborough. This never came forth). Humans are good at forgetting
approach was completely new to both me and the to go back and fill in these details. Lean will not
students, and I note with interest that their perfor- let you forget. The computer-generated red squiggly
mance on the (conceptually complicated) question line underneath your theorem reminds you that your
about binary relations on the end of year exam was proof is not complete, even if your work relies on
done very well. Of course one has to be careful here; results announced by others and hence “it is not
not all the students will want to see mathematics your job to complete it”. During our work formalizing
being done with computers, and of course these perfectoid spaces, it got to the point where I knew
students should not be disadvantaged by such an “we had done it”, but until the definition compiled we
approach. However I did think it was important that knew we had not convinced Lean. Another issue is
the students learnt, as early as possible, what a logi- the fact that formalising the proof of a big theorem
cal argument looked like, and how a logical argument like Fermat’s Last Theorem, given the state of the
is something which can be explicitly checked by an art, would surely take more than 100 person-years.
algorithm. This is not how mathematics is taught in Getting money for this would involve a grant proposal
schools but it seems to me to be a crucial message which no grant agency would fund, and which would
which an introduction to proof course in a mathe- yield a result which no mathematician would care
matics department should convey. Patrick Massot is about anyway — we all know Fermat’s Last Theorem
also using Lean in the mathematics department at is true, because we know the experts have accepted
Orsay, in his introduction to proof course. the result.
On Thursday evenings during term time I run a Lean Because Lean is offering something that pure math-
club at Imperial called the Xena Project (with asso- ematics researchers do not generally want (a fully
ciated blog at xenaproject.wordpress.com), where formalized verification of their theorems) at a price
undergraduates learn to formalize the mathematics which researchers generally will not pay (the time
which they are learning in lectures, and any other taken to learn how to use the software, and then the
mathematics of interest to them. One purpose of the time taken to formalize their own proofs), in practice
project is to make knowledge of formalisation more one has to rule this out as a viable practical applica-
prevalent amongst the undergraduate community, tion of these tools at this point in time. Things might
i
“NLMS_484” — 2019/9/11 — 16:05 — page 36 — #36
i
36 FEATURES
change in the future, but rather than speculate on computer scientists what we are actually doing, in a
how good computers can or will become at under- format which is more useful to them than tens of
standing natural language (i.e., arXiv preprints), let thousands of arXiv preprints where we use words
me simply observe that currently they are nowhere like “normal” and “complete” to mean five differ-
near good enough, and instead turn to other things. ent things and where we write in the idiomatic and
bizarre English language, thus adding another layer
Tom Hales has suggested a completely different use of obfuscation to the material. Formalisation solves
of these tools, and has won a multi-million dollar this problem — it makes us say what we mean in a
Sloan grant to implement his idea, which is called clear and coherent manner. This is what these people
FABSTRACTS (Formal Abstracts). Hales suggests that need to make tools for us — and we are not giving
we should create a mathematical database containing it to them. Mathematicians do not seem to have any
not proofs but formalized statements of theorems, desire to formalise a proof of Fermat’s Last Theorem,
cross-referenced to the traditional literature (journal however computer scientists are desperate to train
articles etc). He proposes that Lean be used for this their AI’s on a formalised proof, to see how much
database, and indeed that is why I chose Lean. The further they can take things. If any mathematician
reason this idea is a game-changer is that if all the reading this decides that they would be interested in
definitions are there, then formalising a statement taking up the challenge of formalising the statement
of a new theorem can be easy. The fact that Lean of one of their theorems in Lean, then they can find
can handle the definition of a perfectoid space is me at the Zulip lean chat.
surely evidence that it can handle any pure mathe-
matical definition which we can throw at it. There Computer scientists do not understand a lot of what
will also be a natural language component to the we as mathematical researchers do, or how we do it.
project; the current plan is that theorems will be Let’s tell them in their own language, and see what
presented in a controlled natural language, i.e., in happens next. Computer scientists have developed
human-readable form, understandable to a mathe- tools which can eat mathematics, but until these
matician with no Lean training. One can imagine in tools get more self-sufficient they will need to be
the near future such a database being constructed, hand-fed. It will be very interesting to see what these
initially concentrating on the areas of speciality of tools become as they grow.
the mathematicians already involved. What would be
the potential uses of such a database?
The first obvious use would be search. Say a FURTHER READING
researcher is wondering what is known about a math-
ematical statement. They formalize the statement, [1] H. Bender and G. Glauberman, Local Analysis
look it up in the database, and get either a perfect hit for the Odd Order Theorem, Cambridge University
(a reference to the literature) or perhaps some partial Press, 1994.
results (several references, perhaps to variants of [2] S. Gouezel and V. Shchur, A corrected quanti-
the statement proved under certain hypotheses). If tative version of the Morse lemma, J. Funct. Anal.
we can train PhD students and post-docs to learn 277 (2019) 1248–1258.
enough about Lean to state their results, then this [3] T. Peterfalvi, Character Theory for the Odd Or-
database could grow quickly. der Theorem, Cambridge University Press, 2000.
The second obvious use would be AI. Already com- Kevin Buzzard
puter scientists are trying to develop tools which
can actually do mathematics better than humans. Kevin Buzzard is an
They want to train their AI on databases — but algebraic number theo-
these databases are few and far between, and sev- rist and a professor of
eral of the databases which are used in practice are pure mathematics at Im-
databases of solutions to gigantic logic puzzles which perial College London.
bear no relation to modern mathematics. I believe He tweets occasionally
that it is up to us as a community to explain to at @XenaProject.
i
“NLMS_484” — 2019/9/11 — 16:05 — page 37 — #37
i
FEATURES 37
Reciprocal Societies:
Belgian Mathematical Society
i
“NLMS_484” — 2019/9/11 — 16:05 — page 38 — #38
i
Microtheses and Nanotheses provide space in the Newsletter for current and recent research students to
communicate their research findings with the community. We welcome submissions of micro- and nano-theses
from current and recent research students. See newsletter.lms.ac.uk for preparation and submission guidance.
A graph is F -saturated if it doesn’t contain a copy of a graph F but adding any edge creates a copy of F . The
maximum number of edges an F -saturated graph can have is well-studied and called a Turán number. Our
topic here is the minimum number of edges an F -saturated graph can have, which behaves quite differently.
A graph is F -free if it does not contain a copy of Turán’s Theorem and the Erdó´s–Stone Theorem tell
the graph F . Given a fixed graph F and a positive us that for a (non-empty) graph F ,
integer n, Turán’s extremal number, ex(F, n), is the
maximum number of edges an F -free graph G on n 1
! !
n
vertices can have (see figure 1). What if we were to ex(F, n) = 1 − + o(1) ,
χ(F ) − 1 2
ask about the minimum number of edges?
If we just replace the word ‘maximum’ with ‘mini- where χ(F ) is the chromatic number of F . In partic-
mum’ in the definition above we get a trivial function, ular, ex(F, n)/ n2 converges to a limit as n tends to
since a graph with no edges will be F -free. But notice infinity.
that in any maximal F -free graph, adding any edge Can we say anything similar about the satura-
creates a copy of F . This inspires the following defi- tion number? Kászonyi and Tuza [2] proved that
nition: we say a graph G is F -saturated if it does not sat(F, n) = O (n). Tuza [5] went on to conjecture that
contain a copy of F as a subgraph but adding any for every graph F the limit limn→∞ sat(F,n) exists.
edge creates a copy of F . n
For example, if F is the triangle then the maximal For a family of graphs F we have sat (F, n) = O (n)
2
triangle-saturated graph has b n4 c edges. In contrast, [2], just as we did for single graphs. However, the
the minimal triangle-saturated graph is a star, which generalisation of Tuza’s conjecture to finite families
has only n − 1 edges. of graphs is not true, as shown by Pikhurko in [4]:
i
“NLMS_484” — 2019/9/11 — 16:05 — page 39 — #39
i
Theorem 1. There exists a finite family F of graphs contains two types of r -graph: a pair of intersect-
such that sat (F, n)/n does not tend to a limit as n ing complete r -graphs; or some disjoint complete
tends to infinity. r -graphs together with a ‘bridge’ edge between them.
i
“NLMS_484” — 2019/9/11 — 16:05 — page 40 — #40
i
40 REVIEWS
i
“NLMS_484” — 2019/9/11 — 16:05 — page 41 — #41
i
REVIEWS 41
ming — are related to fluid dynamics, meaning that My final thoughts are that this book is an excellent
there is some repetition between articles. The article resource for any mathematics departmental library.
on “A Symmetric Framework with Many Applications” The articles cover a vast array of different applica-
outlines a nice example of a unifying viewpoint for tion of mathematics and are generally well-written
minimisation problems of various kinds. As a com- with useful reference lists. In particular, the book rep-
putational applied mathematician, I welcomed the resents an excellent launch point for individuals such
contributions in Part VII related to historical develop- as project students looking for an area of Applied
ments in programming languages (and the confusion Mathematics to investigate in greater depth.
to which different conventions can lead) and the
future in terms of high-performance computing. David Graham
A particularly interesting feature of the book is the David is a Senior Lec-
last Part (VIII), which offers some “Final Perspectives”. turer in Applied Mathe-
I found many of these contributions to be extremely matics at the University
thoughtful and useful, including advice on how to of Plymouth. His main
read Mathematics articles or to write articles or even research interests are
general interest books. As an author of reasonably in developing and using
large fortran codes myself, I found the sections on numerical methods for
“Reproducible Research” and “Experimental Mathe- fluid dynamics. David is a keen footballer and, in
matics” to be especially thought-provoking. decreasing order of competence he also plays guitar,
banjo, ukulele and bouzouki.
For several decades and work, The Turing Guide is an excellent contribu-
after his tragic and tion to this development and the growing literature
untimely death in 1954 on Turing.
at the age of just 41, very
little was known about The book consists of forty-two chapters divided
Alan Turing’s important into eight sections, with the first section providing
work at Bletchley Park biographical material. The first chapter provides a
during the second world brief sketch of Turing’s life, with a helpful timeline
war. However, as the of key events, while in the second chapter, entitled
full scale of his achieve- ‘The man with the terrible trousers’, Turing’s nephew,
ments in codebreaking Sir John Dermot Turing, provides a unique family
and the relevance of the wartime effort to the history perspective. The third chapter is a compilation of
of computing became clearer, Turing’s reputation extracts from papers and reminiscences of the late
has increased dramatically. By providing a very wide- Peter Hilton, who worked with Turing for the last 12
ranging and yet accessible account of Turing’s life years of Turing’s life. According to Hilton, Turing was
i
“NLMS_484” — 2019/9/11 — 16:05 — page 42 — #42
i
42 REVIEWS
a ‘warm, friendly human being’, ‘an approachable paranormal (Chapter 32), which provides context to
and friendly genius’ and he gives a sense of Turing’s Turing’s equally surprising comments about the pow-
‘quirky and infectious sense of humour’. In addition ers of extra-sensory perception in his famous article
to providing insights into Turing’s early life, school- in Mind in 1950.
days, and working life, the first section also discusses
his conviction for homosexuality, the terrible chem- Sections six and seven look at Turing’s contribu-
ical treatment he had to undergo as a result, the tions to biology and mathematics, with the former
subsequent apology from Gordon Brown in 2009 and providing an introduction to his fascinating work on
royal pardon in 2013, as well as the circumstances of morphogenesis in which he used reaction-diffusion
his death. As for whether Turing committed suicide, equations to model the development of biological
‘We shall most probably never know.’ structures. In Chapter 33, Margaret Boden explores
the influence of Turing’s work on research in self-
Those interested in the history of computing and organisation, artificial life, and structuralism in biol-
the codebreaking work at Bletchley Park will find ogy. Interestingly, she argues that Turing’s innovative
a lot of useful material in sections two to four. In ideas were ‘out of line with the biological orthodoxy’
Chapter 6, Jack Copeland provides an introduction of neo-Darwinism and associates his approach with
to Turing’s greatest contribution: the universal Turing that of the structuralist D’Arcy Thompson. In Chap-
machine. He also addresses the origin of the concept ter 35, Bernard Richards describes how he carried
of the stored program computer, an issue which is out work to validate Turing’s theory by showing how
further explored in Chapter 20 in the context of the solutions to his equations in three dimensions could
development of Baby, the first electronic universal explain the shapes of marine creatures known as
stored program computer, in 1948. Chapters 7 and 37 Radiolaria. He had arranged to meet Turing on 8 June
provide very readable introductions to Turing’s impor- 1954 to show him his intriguing findings, but sadly
tant work on Hilbert’s famous Entscheidungsproblem, Turing died the previous day.
while Chapter 24 explores congruences between Tur-
ing’s work and the earlier ideas of Charles Babbage Chapter 36 provides a very helpful overview of Tur-
and Ada Lovelace. ing’s work on the central limit theorem, group theory
and the Riemann hypothesis, while other chapters in
The Enigma machine is the subject of Chapter 10 with this section explore further topics such as Turing’s
Turing’s bombes (special purpose electromechanical work on randomness, which ‘anticipated by nearly
computers used to decipher Enigma messages) the thirty years the basic ideas behind the theory of
subject of Chapter 12. Both of these chapters pro- algorithmic randomness’. In Chapter 38 (and also 13),
vide helpful explanations about how the machines Edward Simpson, who worked as a cryptanalyst at
worked while Chapter 14 explores Tunny, the encryp- Bletchley Park (and after whom Simpson’s paradox
tion machine used by Hitler later in the war. Turing’s is named), provides a very interesting and insightful
contribution, as well as that of Bill Tutte, Max New- explanation of Banburismus, a process for reducing
man and Tommy Flowers, including the essential role the work the bombes had to do to decipher Enigma
of Flowers’ Colossus computer, are all discussed. A messages. He also gives a nice account of Turing’s
number of people who worked as codebreakers or use of Bayes factors and logarithmic scoring to weigh
operated Colossus or the bombes provide fascinating evidence.
accounts of their time at Bletchley Park.
The final section has a very interesting chapter
Section five explores Turing’s contributions to artifi-
on whether the universe is computable. This wide-
cial intelligence and the mind. These chapters include
ranging chapter covers topics such as computability,
discussions of topics such as the Turing Test, his
Turing’s thesis (or the Church–Turing thesis) and mis-
use of heuristic search techniques in the bombe, his
conceptions about it, and the physical computability
early work on connectionist models, child machines,
thesis. It ends by arguing that Turing was not com-
his contribution to cognitive science, and his chess
mitted to the idea that the universe or even the brain
program. Particularly interesting is Diane Proudfoot’s
is computable. The final chapter highlights Turing’s
discussion of Turing’s concept of intelligence (Chap-
impressive legacy both within science and beyond.
ter 28). She argues that Turing thought of intelligence
as an emotional concept that is partly in the eye of The Turing Guide is extremely informative, highly read-
the beholder and she also argues against the com- able, and well produced with many photographs and
mon view that Turing was a behaviourist. Perhaps useful figures to aid exposition (though some colour
the most surprising contribution is a chapter on the
i
“NLMS_484” — 2019/9/11 — 16:05 — page 43 — #43
i
REVIEWS 43
would have been nice). The preface states the book David Glass
was ‘written for general readers, and Turing’s sci-
entific and mathematical concepts are explained in David Glass is a senior
an accessible way’. This has been achieved with lecturer in the School
great success. However, those working in a range of Computing at Ulster
of fields will also benefit a lot from articles written University. He carries
by experts and pointers to the extended literature. out research in artificial
In other words, it does exactly what a good guide intelligence, philosophy
should do. Alan Turing’s legacy has grown enormously of science, and the mod-
in recent years and The Turing Guide will serve to elling of complex systems. He enjoys football and
further enhance understanding of his remarkable rugby (watching not playing) and has an interest in
achievements. the relationship between science and religion.
i
“NLMS_484” — 2019/9/11 — 16:05 — page 44 — #44
i
44 REVIEWS
indeed ‘Boltzmann’s constant’ itself were in fact writ- over the two halves of a lecture room, atoms over
ten down by Max Planck! the two halves of a volume. He divides 7 units of
energy over 4 atoms and calculates the resulting dis-
The enormous importance of Boltzmann’s work, so tribution. He studies entropy by analysing different
obvious today, was unfortunately not at all clear numbers of bedbugs moving between the two halves
to a number of extremely influential scientists and of an auditorium. For just 4 bedbugs the entropy
philosophers, and disagreements were expressed initially moves towards what might be called an ‘equi-
increasingly aggressively from around 1890, when librium value’ but clearly that terminology is hardly
Boltzmann was in his mid-forties. The three chief appropriate because it continues to fluctuate wildly
antagonists were Ernst Mach, Planck and Ernst Zer- about this value. For 1000 bedbugs we see similar
melo, who was Planck’s assistant. behaviour but the fluctuations are far smaller. In
As an arch-positivist, Mach was diametrically both cases, though, there are clearly periods over
opposed to any idea of atoms, and clashes with Boltz- which the entropy decreases. All Johnson’s analysis is
mann were inevitable. Boltzmann hated them. In 1895 extremely helpful to anybody wishing to understand
Mach became Professor of the History and Philos- Boltzmann’s achievements.
ophy of Sciences at Vienna, where Boltzmann had He also presents a series of vignettes of Boltzmann’s
become Professor of Theoretical Physics the year life — some amusing, such as his walking down the
before. Boltzmann’s unease was such that he moved main street accompanied by a cow, purchased to
to Leipzig in 1900, returning to Vienna only in 1902, provide his children with fresh milk, but more often
after Mach’s retirement due to illness. sad, including his disastrous manoeuvers attempting
Planck was (until his dramatic conversion) the last to take up a chair in Berlin without informing his
of the important exponents of classical thermody- current employer, the Austrian education minister,
namics, following in particular Rudolf Clausius and and a study of his self-destructive concern about
Hermann von Helmholtz. As such he detested Boltz- a student lecture to be given the following day, a
mann’s statistical approach to matters of heat flow lecture he should have been able to give practically in
and particularly the use of probability in physics. his sleep. He even provides a ‘happy ending’ in which
Boltzmann dies at peace, soothed by his daughter
Zermelo concentrated on the nature of entropy. A playing the Moonlight Sonata. Sadly, of course, this
crucial point in classical thermodynamics was that was not to be.
entropy never decreases. In certain very special cir-
Ironically, as Johnson points out, it was at precisely
cumstances it might remain the same, but otherwise
this period that Boltzmann’s main ideas were being
it increases thus providing an arrow of time. On the
justified. It was argued, long before Einstein’s detailed
statistical definition, on the other hand, the entropy
analysis in 1905, that Brownian motion, the irregular
of very small systems would often decrease, but
motion of particles in a fluid, was due to bombard-
as the size of the system increased the probability
ment by atoms; and more specifically, Planck’s epoch-
of entropy decreasing became smaller and smaller,
making study of black-body radiation in 1900, which
becoming infinitesimally small for systems of macro-
introduced the quantum theory, was based soundly
scopic size but never exactly zero. Zermelo plagued
on Boltzmann’s probabilistic arguments, which Planck
Boltzmann on this point
had long excoriated. In repentance, as Johnson men-
tions, Planck twice proposed Boltzmann unsuccess-
It is well known that Boltzmann tragically hanged
fully for the Nobel Prize for Physics. For Boltzmann
himself in September 1906. The community of physi-
this was all too late. This book is an interesting and
cists has traditionally absolved itself of blame for
thoughtful account of Boltzmann’s life and work.
this event, providing as evidence his various ill-
nesses, both physical, including enormously dimin- Andrew Whitaker
ished vision, asthma, chest pains and headaches, and
mental, depression and in particular what Johnson Andrew Whitaker is
characterises as anxiety. Readers of Johnson’s slim Emeritus Professor of
book may judge for themselves to what extent physi- Physics at Queen’s Uni-
cists contributed to these illnesses. Johnson provides versity, Belfast, and
detailed analysis of relatively small systems demon- Chair of the Institute
strating Boltzmann’s main achievements. He studies of Physics History of
the probabilities of students distributing themselves Physics group.
i
“NLMS_484” — 2019/9/11 — 16:05 — page 45 — #45
i
OBITUARIES 45
Obituaries of Members this did not stop him continuing to receive visitors,
attend concerts and read widely. He never forgot his
gratitude to University College School and Trinity Col-
lege Cambridge, for their support given to a refugee
Ernst Sondheimer: 1923 – 2019 from Nazi Germany in difficult times.
Ernst Sondheimer, who
was elected a member of
the London Mathemati- Gerald Gould: 1925 – 2019
cal Society on 28 June
1974, died on 9 June 2019, Gerald Gould, who was
aged 95. elected a member of
the London Mathemati-
Julian Sondheimer writes: cal Society on 18 March
Ernst was born in 1923 1954, died on 23 Febru-
in Germany. In 1937 the ary 2019, aged 93.
family emigrated to London, and Ernst started his
new life in England at University College School in David Edmunds and Des
Hampstead. In 1942 he went up to Trinity College Evans write: Gerald was
Cambridge and remained there as research student born in Bermondsey,
and fellow until 1952. In 1946 he became a British London, on 9 September 1925. From Wilson’s Gram-
citizen and in 1950 he married Janet Matthews, a his- mar School he won an open scholarship to study
tory fellow at Girton College, Cambridge. In 1951 he physics at Christ’s College, Cambridge, an event
was appointed as a lecturer at Imperial College and celebrated by the school with a day off. After grad-
in 1955 he became a reader at Queen Mary College. uating, he served his military service working for
Tube Alloys, the code name for the top security
At the age of 37, in 1960, Ernst was appointed Pro- research and development programme to develop
fessor of Mathematics at Westfield College in the nuclear weapons. That done, Gerald enrolled for a
University of London, where he remained for 22 years mathematics degree in Birkbeck College, University
until his retirement in 1982. For most of this period of London. There he came under the influence of
he was Head of Department and he was instrumental Lionel Cooper, and after graduating started a PhD
in creating a very successful mathematics research under Cooper’s supervision. In 1950 Cooper was ap-
department. Ernst cared deeply about his students pointed to the chair in what was then the University
and staff and is fondly remembered by them. College of South Wales and Monmouthshire (now
Earlier, as a young researcher in the field of mathe- Cardiff University), and Gerald went with him; he was
matical physics, his main achievement was a seminal to spend the rest of his career in Cardiff. Apart from
and groundbreaking paper in 1948 with G.E.H. (Harry) his mathematics, Gerald had many interests and tal-
Reuter (also a German refugee) on the anomalous ents. He was an accomplished musician, being highly
skin effect. This was in the area of the optical prop- regarded and sought after by orchestras as a tim-
erties of metals and involved the difficult solution panist, bassoonist and flautist. Bridge was another
of an integro-differential equation. Later Ernst co- passion at which he excelled, winning a worldwide
authored the books Green’s Functions for Solid State competition in 2006 involving 1,000 players on four
Physicists and Numbers and Infinity. continents. In 1954 Gerald developed TB and spent
time in an isolation hospital. He claimed that to
Towards the end of his time at Westfield, in 1981, contend with the boredom, he learned to smoke,
Ernst and his family were hard hit by the death of became a master of Meccano, and learned Russian.
his younger brother Franz, the eminent chemist (a He would continue to smoke with obvious pleasure
Fellow of the Royal Society, Franz was distinguished until 1984, and his knowledge of Russian had a big
for his work on the synthesis of natural products). part to play in the contribution he made to the
After retiring from Westfield College, Ernst revelled mathematical community; he went on to act for the
in his two passions of mountains and flowers, taking LMS as English Edition Editor of Sbornik, and trans-
up the editorship of the Alpine Journal and devoting lated into English various Russian books, notably two
himself to his wonderful garden at his home in North that appeared in the Encyclopaedia of Mathematical
London. At age 88, an unexpected difficulty with a Sciences series: Dynamical Systems by D.V. Anosov
medical procedure left him in a frail condition, but and General Topology III by A.V. Arhangel’skii.
i
“NLMS_484” — 2019/9/11 — 16:05 — page 46 — #46
i
46 OBITUARIES
Gerald’s early work evolved from his PhD on where he was awarded a personal chair and spent
integration spaces, his principal contribution to the rest of his career. He was challenged by dif-
research being perhaps his paper Integration Over ficult questions and, although he wrote only 17
Vector-Valued Measures which appeared in the Pro- papers, virtually all of them contain deep and
ceedings of the LMS. This introduced a type of inte- clever ideas. He continued to work in geometric
gral (the Gould integral) of a bounded real-valued measure theory, relating rectifiability of sets (i.e.
function with respect to a finitely additive set func- whether they are ‘curve-like’ or ‘surface-like’) to
tion taking values in a Banach space. Gerald had a local measure densities. But he had a broad inter-
great appetite for new areas of mathematics and est across mathematical analysis. He constructed
was able to absorb challenging new material with a clever counter-example to Khinchin’s conjecture
enviable ease. Much of his research was gener- on the uniform distribution of certain sequences
ated by problems and flaws discovered through his modulo 1. He was fascinated by variants of the
extensive reading. His later work on test-function Kakeya set and curve packing problems and he
spaces and various aspects of spectral theory are solved the problem of ‘the minimal comfortable liv-
examples of this. ing quarters for a worm’ by showing that given any
plane set of zero area (Lebesgue measure) there
Outside his mathematics and his many interests, are arbitrarily short smooth curves that cannot
Gerald had a rich family life. He is survived by his be mapped into the set by a similarity (or, much
beloved wife Enid, his children Nina and Ben and more generally, an analytic) transformation. Out-
three grandchildren. They will miss him sorely, as standing was his geometrical proof of the ‘circle
will his friends and colleagues. packing conjecture’, of fundamental importance
in harmonic analysis: given a plane set E of zero
area, every circle with centre x intersects E in
John Marstrand: 1926 – 2019 zero perimeter length, apart from a set of cen-
tres x of zero area. In a probabilistic direction,
John Martin Marstrand, John, together with Geoffrey Grimmett, answered
who was elected a mem- in 1990 the major outstanding question on site
ber of the London Math- percolation on lattices in 3 or more dimensions.
ematical Society on 18
June 1953, died on 29 John was a highly regarded and engaging lecturer.
May 2019, aged 93. Once, lecturing on ‘random walks’, he jumped onto
Kenneth Falconer writes: the front bench, blindfolded himself and stepped
Although he took his forwards and backwards at random until he even-
DPhil at Oxford, John tually fell off the end of the bench! He was very
was essentially supervised by Cambridge Professor careful in all his duties, often spending hours
Abram Besicovitch. His thesis contained a number on tasks others would take a few minutes over.
of remarkable results, notably ‘Marstrand’s projec- Once when examining a PhD thesis his attention
tion theorems’ on the Hausdorff dimensions of or- to detail resulted in a five and a half hour viva!
thogonal projections of subsets of the plane onto
lines. Apart from generalisations by Pertti Mattila and Outside mathematics, John went through a range
Robert Kaufman, this attracted little attention un- of passionate interests, from fast sports cars to
til the late 1970s when Mandelbrot popularised and extreme skiing. In his 40s he took up fell-racing
unified the notion of fractals. It was then realised and in 1982 became British over-50s fell-running
that John’s work was just what was needed for the champion. He regularly participated in the Ben
study of the geometry and dimensions of fractals, Nevis Race and on several occasions won the over-
and his projection theorems became the prototype 40s or over-50s sections, with his fastest time in
for the now flourishing area of fractal geometry. In- 1983 an incredible 1 hour 48 minutes 54 seconds.
deed, the paper based on John’s thesis that appeared
in the LMS Proceedings in 1954 was highlighted in John was a kind and considerate person and is
2015 as one of six ‘landmark’ papers published in remembered by his former colleagues as a ‘love-
the Proceedings during its first 150 years. able eccentric’. He was modest about his achieve-
ments but his determination and insight when he
After short periods in Cambridge and Sheffield, became absorbed by a project led to some quite
John moved to the University of Bristol in 1958 remarkable work of enduring quality.
i
“NLMS_484” — 2019/9/11 — 16:05 — page 47 — #47
i
OBITUARIES 47
Christopher Hooley: 1928 – 2018 to these ideas. In particular the use of exponential
sums, as pioneered by Hooley, is now ubiquitous.
Christopher Hooley, who
was elected a member of Hooley’s writing was well known for his love of such
the London Mathemati- arcane phrases as ‘the dexter side of the ante-
cal Society on 28 June penultimate equation’. Equally however his papers
1974, died on 13 Decem- were a model of clarity and accuracy. They were often
ber 2018, aged 90. couched in seemingly old-fashioned terms, both lin-
guistically and mathematically, but they never failed
Roger Heath-Brown and to enlighten.
Thomas and Adam
Hooley write: Professor Christopher Hooley had many interests, most of
Christopher Hooley, FRS, FLSW was one of the lead- which were pursued with vigour. From early on as
ing analytic number theorists of his day. He won a pupil at Abbotsholme School during the Second
Cambridge University’s Adams Prize in 1973, and the World War he developed a love not only of mathemat-
LMS Senior Berwick Prize in 1980. He was elected ics, but also of classical, military, and naval history. In
a Fellow of the Royal Society in 1983, and was a years to come he would delight in discussing the his-
Founding Fellow of the Learned Society of Wales. He tory of western and eastern Roman empires with his
gave a plenary lecture at the 1983 ICM in Warsaw. colleagues from the history department at Cardiff,
who would ruefully admit they were out of their com-
Christopher Hooley did his undergraduate degree at fort zone, using that standard excuse ’not my period’!
Corpus Christi College, Cambridge, and continued — Hooley counted all European history of the two
there with a PhD under Albert Ingham. His career millennia A.D. as his period. This interest went along
then took him to Bristol, to Durham, and finally to with a love of antiques and especially of collecting
Cardiff, where he served as Head of the School of West Country friendly society brasses. Many days
Mathematics from 1988 to 1995. were spent scouring antique shops and following up
He was the author of nearly 100 papers. The best leads to remote places in Somerset during the 1970s
known of these concerns Artin’s conjecture on primi- and 80s. His wife, Birgitta, denied pets as a child in
tive roots. The conjecture proposes that primes with Sweden, made sure that the family always had plenty
a given primitive root have a definite density, and of dogs — at one time six, sometimes cats as well:
gives a formula for what the density should be. In this interest led to another hobby, that of travelling
particular it would follow that 10, for example, is a to terrier shows in the summer months, from which
primitive root for infinitely many primes p, so that the Birgitta, Christopher, and their two sons, Thomas
decimal expansion for 1/p would recur with period and Adam, often returned proudly bearing coloured
p − 1. Hooley’s work, in 1967, proved the conjecture rosettes for their dogs’ successes. Supportive as a
in full, assuming the Riemann Hypothesis for a class father he was nonetheless quick to spot and point
of Dedekind zeta-functions. These would have been out inaccuracies of memory in the reminiscences of
highly abstruse objects to analytic number theorists his sons at his 90th birthday party, held in Bristol
in those days, and even now it seems remarkable just four months before he died in December 2018.
that they could have any bearing on such a concrete Christopher’s grandfather had taught carpentry at
problem about decimal fractions. Macclesfield and on inheriting his tools Christo-
Many of Hooley’s papers made use of estimates for pher set about using them to restore or completely
exponential sums originating in the work of Weil and remake a large number of period sash windows in
Deligne. Thus for example, he was able to establish the family home in Somerset, often at some con-
the Hasse Principle for non singular integral cubic siderable height. Once done, he turned his hand to
forms in 9 or more variables, by using information panelling, lath and plaster, and similar skills with great
about the associated exponential sums. Another success.
recurrent theme was the introduction of novel sieve The practical side was also present when it came to
methods, and his Cambridge Tract “Applications of motoring. The first family car, a 1950s’ Wolseley 6/80,
sieve methods to the theory of numbers” showcases was kept and later given a thorough restoration in
the impressive variety of sieve tools he developed. the 1980s. Christopher showed no fear in creating
Modern analytic number theory owes a great deal an entire wiring loom from scratch!
i
“NLMS_484” — 2019/9/11 — 16:05 — page 48 — #48
i
48 OBITUARIES
These interests and his family life provided the coun- publications at that time was his book on Degree
terpoint to Hooley’s academic career. At times uncon- Theory in the Cambridge Tracts in Mathematics se-
ventional, for instance researching deep into the night ries; this book was particularly well received and
with a beloved whisky and cigarette alongside an became an established text and was widely con-
eclectic mixture of steam railway magazines, Dante, sulted. During the 1980s his interests moved towards
Milton, and paperback thrillers, he had a ready wit the study of the second part of Hilbert’s Sixteenth
and excellent sense of humour. He will be remem- Problem on the number and relative configuration
bered with fondness by colleagues and family alike. of limit cycles of polynomial systems in the plane.
This is acknowledged to be a very difficult area of
research but it is internationally recognised that he
Noel G. Lloyd: 1946 – 2019 and his research group have been responsible for
substantial developments, and his loss is strongly
Professor Noel Lloyd, felt amongst researchers in this area. Their progress
who was elected a mem- involved innovative use of computer algebra, as well
ber of the London Math- as analytic approaches.
ematical Society on 15
February 1973, died on 7 In the 1980s Noel was a member of the Mathematics
June 2019, aged 72. Committee of the SERC where his detailed knowl-
edge and his meticulous fairness came to the fore.
Alun Morris writes: Noel Also, in 1986–90 he was Joint Editor-in-Chief with
sadly died after a pro- me of the Journal of the LMS — a dramatic period
longed fight against for the journal with an overnight need to move the
prostate cancer which he met with his usual quiet printing to the Cambridge University Press. He was
composure. This followed a full life distinguishing also a long-serving Editor of the Mathematical Pro-
himself both as a mathematician and in his later ceedings of the Cambridge Philosophical Society. He
years as a senior university administrator and in was promoted to Senior Lecturer, Reader and then
public life. in 1986 to a Personal Chair. Although at that point
Noel was born in Llanelli, Carmarthenshire. After not heavily involved in administration, he was highly
attending Llanelli Grammar School, where his father respected and was for a short time Head of the
was a mathematics teacher, he entered Queens’ Col- Department of Mathematics, Dean of the Faculty of
lege, Cambridge in 1965 having obtained an Entrance Science and then Pro Vice-Chancellor, then Registrar
Scholarship. He graduated with a B.A. in 1968 with and Secretary of the University and finally in 2004
a first class in each year and an MA in 1972. On its Vice-Chancellor. He held that position with great
graduating, he worked for his PhD in a research area distinction until 2011. In 2010 he was appointed CBE
which was not too fashionable at the time, ordinary for Services to Higher Education in Wales and in 2011
differential equations, under the supervision of Sir a Fellow of the Learned Society of Wales.
Peter Swinnerton-Dyer. His PhD thesis dealt with a The many tributes that have appeared after his
case of van der Pol’s equation not considered by untimely death are an indication of the esteem he
J.E. Littlewood and Mary Cartwright. In addition to was held as Vice-Chancellor. He was meticulous in
numerous college prizes he also received the pres- his preparation, quietly firm and fair. He was univer-
tigious Rayleigh Prize. He proceeded to a Research sally regarded as a generous and compassionate man
Fellowship at St John’s College, Cambridge. After two and a person of great integrity, respected equally by
years in that position he was enticed to a Lectureship his academic colleagues and fellow administrators.
in Pure Mathematics at what was then the University During his tenure as Vice-Chancellor he was Chair
College of Wales, Aberystwyth (now Aberystwyth Uni- of Higher Education Wales and Vice-President of
versity), which he took up on 1 January 1975. At that Universities UK. He also served on many other UK
time, it was not anticipated that the university was university bodies. On his retirement, his talents were
appointing its future Vice-Chancellor — certainly not in great demand. He became an independent mem-
by him. He did not need to be persuaded; he had ber of the Silk Commission established by the UK
long expressed his wish to come to his father’s alma government to look at the future of devolution in
mater. His research work flourished in his chosen Wales. He also served as a lay member of the Judicial
area of applied analysis or non-linear analysis. He Appointments Commission. For six years he was
established a strong school in that area supervising Chair of Fair Trade Wales.
a number of good PhD students. Among numerous
i
“NLMS_484” — 2019/9/11 — 16:05 — page 49 — #49
i
OBITUARIES 49
Throughout his life, his church was central to his In 1998, Stuart again changed direction and built
life. He served not only his church with distinction on his research experience to found and edit a
but also his denomination and other religious causes BMJ publication Clinical Evidence, which assessed
both locally and nationally. Above all, he was a bril- the clinical significance of medical treatments and
liant organist; as one tribute said, students would drugs. In 2002 Stuart returned to GP practice in
see their VC relaxing on a Sunday morning at the London. With Dr Elizabeth Robinson, he took over a
organ playing his favourite Bach. Noel married Dilys failed GP practice in the Kings Cross area and they
in 1970; next year would have been their golden wed- worked together there for 8 years providing excel-
ding. They had a son Hywel and daughter Carys and lent health care to some of the most disadvantaged
two grand-daughters Sioned and Catrin. communities in London.
Stuart retired from GP practice in 2010, in order to
devote much of his time to mathematics, his first
academic love. He began to study with the Open
Stuart Brian Barton: 1952 – 2019 University, completing a first class degree in mathe-
Dr Stuart Barton MD, matics in 2017. His tutors speak highly of his assign-
DPhil, who was elected a ment work, which was always of excellent quality
member of the London and often contained material that went well beyond
Mathematical Society what the assignments officially required. After mov-
on 10 November 2017, ing to live in Cambridge, he regularly attended the
passed away in hospital Open University’s maths tutorials there and also
in Cambridge on 12 April weekend schools, where he greatly enjoyed meeting
2019, aged 66. other people with similar interests, and is remem-
bered by tutors as one of the brightest and liveliest
Phil Rippon writes: As a students. His partner Elizabeth Robinson reports
young person at Oldham Hulme Grammar School, that he read widely about the subject, and was
Stuart excelled at mathematics and physics, but especially influenced by the life of Paul Erdó´s, as
was advised that medicine was a more appropriate described in the book The Man Who Loved Only
career. Accordingly, he read medicine at Balliol Col- Numbers by Paul Hoffman. Like many mathemati-
lege, Oxford, and took a DPhil there in physiology cians Stuart would enjoy long walks (with their dog
in 1977, working much of the time in Stony Brook, Mabel), while thinking about mathematics problems.
New York with Professor Ira Cohen. They had been
friends at Oxford and in fact published together a Stuart then embarked on an MSc in Mathematics
rather mathematical paper on transmitter release at the Open University, his fascination with number
from nerve terminals (appearing in Nature in 1977). theory (and especially Goldbach’s conjecture) lead-
Stuart then began a career as a GP in the UK and ing him to study modules based on Tom Apostol’s
also qualified as a consultant physician though he book on analytic number theory. It was at this time
chose to remain a GP. that he joined the LMS, confessing to me (as his LMS
proposer) that his dream was to add publications in
In 1993, Stuart changed direction and became a pure mathematics to his publications in medicine
lecturer at the University of Liverpool, specialis- and biology. His tutors say that he would surely have
ing in evidence based medicine, where he could achieved this dream in time, but sadly it could not
bring into play his love of mathematics (espe- happen. In April 2019, he suffered a brain aneurysm
cially statistics). He became a leader in that field, and died in hospital some days later. He is survived
publishing many articles, with co-authors such as by his partner Dr Elizabeth Robinson, and by his
Professor Tom Walley CBE. two adult children Jonathan and Andrew.
i
“NLMS_484” — 2019/9/11 — 16:05 — page 50 — #50
i
Benefits of LMS membership include access to the Verblunsky Members’ Room, free online subscription
to the Society’s three main journals and complimentary use of the Society’s Library at UCL, among
other LMS member benefits (lms.ac.uk/membership/member-benefits).
If current members know of friends or colleagues who would like to join the Society, please do
encourage them to complete the online application form (lms.ac.uk/membership/online-application).
Details of all the courses MAGIC provide can be found at: https://maths-magic.ac.uk
50
i
“NLMS_484” — 2019/9/11 — 16:05 — page 51 — #51
i
EVENTS 51
LMS Meeting
i
“NLMS_484” — 2019/9/11 — 16:05 — page 52 — #52
i
52 EVENTS
Edge Days 2019: Fano Varieties, Cone Young Researchers in Algebraic Number
Singularities and their Links Theory
Location: University of Edinburgh Location: University of Warwick
Date: 4–8 November 2019 Date: 6–8 November
Website: tinyurl.com/y3xj764z Website: tinyurl.com/YRANT2019
The workshop will focus on three closely related The conference is aimed at PhD and post-docs work-
classes of objects: Fano varieties, log-terminal singu- ing in or interested in algebraic number theory and
larities and links with a positive Sasakian structure. arithmetic geometry. All participants are encouraged
It aims to bring together experts from singularity the- to give talks. There will be three keynote talks, given
ory, birational geometry and mathematical physics. by Jack Thorne (University of Cambridge), Lynne
Supported by an LMS Conference grant. Walling (University of Bristol) and Sarah Zerbes (UCL).
LMS Meeting
Website: lms.ac.uk/events/meeting/agm tend both the Graduate Student Meeting and the
LMS AGM. The AGM will be followed by a wine re-
The meeting will include student presentations of ception. The Society’s Annual Dinner will be held
their current work, with a prize awarded for the at 7.30 pm at Goodenough College. The cost of
best talk. This meeting is intended as an intro- the dinner will be £58.00, including drinks; email
duction to the AGM later in the day. Travel grants AnnualDinner_RSVP@lms.ac.uk to reserve a place.
of up to £100 are available for students who at-
i
“NLMS_484” — 2019/9/11 — 16:05 — page 53 — #53
i
EVENTS 53
LMS Meeting
British Postgraduate Model Theory LMS South West & South Wales
Conference 2020 Meeting & Workshop
Location: University of Leeds Location: University of Bristol
Date: 8–10 January 2020 Date: 15 January 2020
Website: tinyurl.com/bpgmtc20 Website: tinyurl.com/y37pked7
This meeting aims to bring together young re- This meeting forms part of the South West & South
searchers interested in model theory. It will feature a Wales Regional Workshop on Interactions between
mini-course, invited talks by established academics, Geometry, Dynamics and Group Theory on 16–17
and contributed talks by postgraduate researchers. January. Speakers are Martin Bridson FRS (Oxford),
Supported by the LMS, the University of Leeds and Corinna Ulcigrai (Bristol) and Yves Benoist (Paris). Par-
the British Logic Colloquium. tial funding is available.
i
“NLMS_484” — 2019/9/11 — 16:05 — page 54 — #54
i
January 2020
October 2019
15 South West & South Wales Regional
19 DeMorgan@21, London Meeting, Bristol
Calendar of Events
This calendar lists Society meetings and other mathematical events. Further information may be obtained
from the appropriate LMS Newsletter whose number is given in brackets. A fuller list is given on the Society’s
website (www.lms.ac.uk/content/calendar). Please send updates and corrections to calendar@lms.ac.uk.
i
“NLMS_484” — 2019/9/11 — 16:05 — page 55 — #55
i
17 Branching Processes and their Applica- 6-8 Young Researchers in Algebraic Number
tions, University of Sussex (483) Theory, University of Warwick (484)
18 Randomness, Symmetry and Free Proba- 13 Computer Science Colloquium 2019, De
bility, University of Sussex (483) Morgan House, London (484)
18 A Day of Lie Theory in Kent, University 18-19 Category Theory and its Applications
of Kent (484) Postgraduate Conference, University of
18-19 Philip Maini’s 60th Birthday Workshop, Leicester (484)
Mathematical Institute, Oxford (483) 21 Joint LMS Meeting with the IMA, Reading
18-20 Southampton–Bielefeld Geometric Group (484)
Theory Meeting, University of Southamp- 21 LMS/BCS-FACS Evening Seminar, De Mor-
ton (483) gan House, London (484)
18-21 Lattice Polytopes, with a View towards 29 Graduate Student Meeting, London (484)
Geometry and Applications, ICMS, Uni- 29 Society Meeting and AGM, London (484)
versity of Edinburgh (483)
19 LMS Popular Lectures, University of Birm-
ingham (484)
December 2019
25-26 Novel Approaches to the Multi-Armed
Bandit Problem, Imperial College London 6 PDE Models for Cancer Invasion, Queen’s
(484) University Belfast (484)
29-4 Oct Clay Research Conference and Work- 9-10 Probabilistic Coupling and Geometry
shops, Mathematical Institute, Oxford Workshop, University of Warwick (484)
(482) 9-13 Computational Complex Analysis, INI,
30-4 Oct Structure Preservation and General Rela- Cambridge (484)
tivity, INI, Cambridge (482) 13 Integrability, Algebra and Geometry, Uni-
versity of Glasgow (484)
October 2019
10 Dependence Modeling and its Applica-
January 2020
tions in Insurance and Finance, Heriot-
Watt University (483) 8-10 British Postgraduate Model Theory Con-
7-11 Dirac Operators in Differential Geome- ference 2020, University of Leeds (484)
try and Global Analysis, Bedlewo, Poland 15 South West & South Wales Regional Meet-
(482) ing, Bristol (484)
19 DeMorgan@21, De Morgan House, Lon-
don (483)
25 Sir Michael Atiyah: Forays into Physics,
July 2020
INI Cambridge (484)
28-1 Nov Complex Analysis in Mathematical 5-11 8th European Congress of Mathematics,
Physics and Applications, INI, Cambridge Portorož, Slovenia
(483) 12-19 14th International Congress on Mathemat-
ical Education Shanghai, China
November 2019
4-8 Edge Days 2019: Fano Varieties, Cone Sin-
August 2020
gularities and their Links, University of
Edinburgh (484) 17-21 IWOTA 2020, Lancaster University (481)
i
“NLMS_484” — 2019/9/11 — 16:05 — page 56 — #56
i