Twelfth Statistical Mechanics of Soft Matter Meeting

The Twelfth SM2 Meeting will be held at the University of Auckland from 2-3 December, 2026, with a summer school immediately beforehand on 1 December.

logo

SM2 is a discussion meeting on all aspects of the equilibrium and nonequilibrium statistical mechanics of soft condensed matter, from fundamentals to applications, relating to liquids, colloids, polymers, gels, biological molecules, liquid crystals and other forms of soft matter.


If you wish to be added to our mailing list please send an email to: smsq.meeting@gmail.com.



Registration

Registration is now open! Note that registration is free but essential.

Please register early to help us make plans for catering and the conference dinner.

There will be regular talks (of approximately 20 minutes, including questions) and lightning talks (5 minutes, including questions) in the program.

Abstract submission is not required for now, and it is likely that all people who want to give a regular talk will be able to do so. But, abstracts will be assessed in the order of submission, so to be absolutely sure of getting a regular talk slot you are encouraged to submit early.

Registration deadline: 31 October, for abstract submission to be considered for a talk.

If we get more abstract submissions than can be accommodated, then some abstracts may be accepted as lightning talks rather than regular talks.

Alternatively, if there’s space in the program we may offer speakers the opportunity to give a lightning talk in addition to a regular talk.

Registration form: To register for the conference and/or the summer school please complete this Google form.

Make sure to keep the link that you are sent so that you can edit your title and abstract if necessary.


Summer School

We will run a Summer School on Tuesday 1 December, 2026 on Atomic Descriptors, Machine Learning, and Unsupervised Discovery: From SOAP to UMAP. This involves will be training machine learning models with local atomic descriptors and visualising them with low-dimensional representations.

The summer school is targeted at honours students, PhD students, and postdocs.

Please register for the summer school via the registration form.


Summer School Details

Atomic Descriptors, Machine Learning, and Unsupervised Discovery: From SOAP to UMAP.

This one-day session introduces participants to atomic descriptors as the bridge between raw atomic coordinates and machine-learned models of matter. We begin with the Smooth Overlap of Atomic Positions (SOAP) framework, building an intuition for how local atomic environments can be encoded as rotationally and permutationally invariant fingerprints suitable for machine learning. Participants will see how these descriptors are constructed from expansions in radial basis functions and spherical harmonics, and why this representation has become a workhorse for machine-learned interatomic potentials.

From there, we turn to supervised learning: using SOAP (or related descriptors) as input features to train a model that predicts structural or energetic properties, illustrating the general pipeline from descriptor generation to model training and validation.

The second half of the day shifts to unsupervised learning. We take the same descriptor vectors — now computed for a range of crystal structures and/or colloidal (Janus) particle assemblies — and project them into two dimensions using UMAP. Participants will see, hands-on, how structurally distinct environments naturally separate into clusters in this reduced space, without any labels or supervision. This demonstrates that the same descriptor doing the heavy lifting in supervised force-field training can, on its own, distinguish and classify structural motifs.

Together, the two halves of the day tell a coherent story: a single, well-designed atomic representation can support both predictive (supervised) and exploratory (unsupervised) machine learning, opening the door to automated structure classification, phase identification, and discovery of order in complex assemblies.

No prior machine learning experience is required; basic familiarity with atomic/molecular structure is helpful.



Organisers

Kannan Ridings University of Auckland
Elke Pahl University of Auckland
Enquiries: smsq.meeting@gmail.com


Venue

The meeting will be held at the University of Auckland. More details to be added closer to the date of the conference.



Transport

From the airport, one can catch the Airport Link Bus or the Skydrive Auckland Express. Alternatively, taxi (not sure of price, perhaps NZ$80 or NZ$100) and Uber rides (NZ$65 to NZ$80 according to Uber app) are available from the airport.

Accommodation

The following hotels are all within 15 mins walk of the University of Auckland campus and provide easy access to the restaurants of Queen Street, Britomart, or the Viaduct Harbour.
The Quadrant
The Pullman
Quest Auckland
Quest on Queen
Quest on Eden
Waldorf Celestion Apartment
Other accommodation options can be found by searching for example on expedia, airbnb, stayz, wotif. Suburbs which are local (~ 30 mins walk) to the University of Auckland include Parnell, Newton, and Ponsonby.

Past SM2 meetings

2013 Inaugural meeting at RMIT, Melbourne
2015 Second meeting at Swinburne University of Technology
2016 Third meeting at the University of Melbourne
2017 Fourth meeting at the University of Sydney
2018 Fifth meeting at the University of Auckland
2019 Sixth meeting at the University of Adelaide
2020 Seventh meeting hosted virtually by Griffith University and the University of Queensland
2022 Eighth meeting at Monash University
2023 Ninth meeting at Murdoch University
2024 Tenth meeting at the University of Sydney
2025 Eleventh meeting at RMIT