I'm a physicist with interests in out-of-equilibrium phenomena and statistical physics. My recent projects have largely been in astrophysics, but I also work on tensor networks, biological self-assembly, and quantum transport processes.

Ph.D. candidate in Astronomy • *2015-18*

I am advised by Professors Christopher Tout and Gordon Ogilvie. My focus is on out-of-equilibrium and transport phenomena
in stars, discs, and planets. This includes scaling relations in turbulence, the effects of turbulent anisotropy on stellar evolution,
thermal feedback in tides, and quasi-steady-state evolution of discs.

B.S. with Honors in Physics • *2011-15*

In my degree I focused on statistical physics and condensed matter theory. My academic advisors
were Tom Tombrello and Jason Alicea, and I completed my senior thesis with Sterl Phinney on the atmospheric
dynamics of pulsar companions. I also worked on a number of other projects, including
plasmonic carrier generation in Harry Atwater's group, various topics in biophysics with Milo Lin, and analysis of parafermion zero modes with Jason Alicea.
Some of these projects are still ongoing.

Postdoctoral Scholar • *2018-Present*

I am a postdoctoral scholar at the Kavli Institute for Theoretical Physics at the University of California at Santa Barbara. My work here focuses on stellar physics but also extends to accretion disks, turbulence and methods development in statistical physics.

Convection in the cores of massive stars becomes anisotropic when they rotate. This anisotropy results in heat flowing in a different direction from the thermal gradient, which in turn results in baroclinicity and circulation currents in the upper radiative zone. Along with Christopher Tout and Shashikumar Chitre, I showed that this induces a much stronger meridional flow in the radiative zone than previously thought
[MNRAS][arXiv]. This drives significantly enhanced mixing, though this mixing does not necessarily reach the surface. The extra mixing takes on a similar form to convective overshooting, and is relatively insensitive to the rotation rate above a threshold, and may help explain the large overshoot distances inferred from observations.

Many problems in astrophysics involve turbulence.
To deal with this several tools may be used including simulations and analytic closure models.
Along with Pierre Lesaffre, Christopher Tout and Shashikumar Chitre I developed a new semi-analytic closure model.
The model associated with each linear mode an amplitude proportional to its growth rate and scaled to match either a convective or MHD spectrum.
Linear growth rates are computed in a perturbative formalism that permits the inclusion of shear.
We provide results for a wide variety of scenarios including rotating, magnetised convection as well as sheared convection and baroclinic fluids [MNRAS][arXiv].
We also provide a computational implementation of our scheme which can be readily employed in novel circumstances [Github].

The spectra of stars which are actively accreting material may not reflect the native/bulk chemistry of the star.
Mihkel Kama and I developed a formalism relating the accreting chemistry, the accretion rate and the surface chemistry of main-sequence stars, though the formalism is also readily applicable to giants and even accreting planets [MNRAS][arXiv].
We predict the fraction of accreted material in the photospheres of a wide range of disk-hosting stars and, in cases where the accretion rate and surface chemistry are known, characterize the composition of the accreting material.

One of the most common techniques in astronomical data analysis is fitting a line or curve to data. Unfortunately this is often done in ways that lead to systematic biases and uncontrolled errors. Charles Steinhardt and I wrote a tutorial explaining how to avoid these pitfalls and introducing two new curve-fitting methods [PASP][arXiv]. The first, ZeBRA, provides a new means for fitting piecewise linear functions through data with one-dimensional errors. The second, GIRAfFE, generalizes this to the case of errors in both variables. We benchmark these and other methods and provide comparisons as well as Python implementations of the various methods appearing in the paper [Github].

Tensors are a natural way to express correlations among many physical variables, but storing and manipulating tensor networks in computers naively requires exponential cost in the size of the network. To remedy this I've developed an algorithm for automatically contracting unstructured tensor networks
[arXiv][GitHub],
as well as two algorithms for identifying near-optimal tensor decompositions
[arXiv][GitHub].
Taken together, these should enable investigations of a wide range of phenomena through the lens of tensor networks.

A number of numerical methods including numerical differentiation and identifying eigenvalues depend on the analytic structure of complex-valued functions.
Of particular importance to these methods are the locations of any poles.
I developed an algorithm for bounding the distance to the nearest pole (i.e. the radius of convergence)
[arXiv]
. This method combines binary search with a robust protocol for identifying whether or not there are poles in a given circular contour, and provides accurate answers even with a small (100-1000) number of function evaluations.

Many physical phenomena are characterised by non-equilibrium transport and scattering of particles. For instance radiation in stellar atmospheres, or charge transport in metals. We've developed a new hybrid method, NESSE
[arXiv]
, for calculating the outcomes of such transport processes in complex geometries. The method is hybrid in the sense that we track distribution functions (a la classical Boltzmann) but we do so using Monte Carlo sampling to make it computationally tractable. This lets us answer physical questions like "How many scattering events separate this carrier from the initial perturbation?" and "Which kind of carrier is deposited where and at what energy?". Using this method we've made several comparisons with experiment, probing the importance of ballistic transport
[Nat. Comm.][arXiv]
and field localization
[Nat. Comm.], [arXiv].

Alex Rasmussen and I were intrigued by the work in high energy physics on soft modes, and
thought that there might be connections with low-energy condensed matter physics. What we found
is that soft modes are precisely the power-law low-energy modes which characterize gapless topological order
[PRB][arXiv]
. We showed that in non-trivial topologies these modes encode information not only about the amount of charge or mass wound through a system but
also about the winding direction. We also argued that this suggests a relatively simple resolution to the Firewall
paradox by introducing quantum mechanical violations of the equivalence principle which are not observable classically.

Along with my advisors Christopher Tout and Gordon Ogilvie, I uncovered a novel mechanism for
tidal heating in hot Jupiters
[MNRAS],
[arXiv]
. This mechanism takes advantage of the fact that regions where
heat is transported radiatively can act as resonant cavities, and enhance the dissipation of tidal
g-modes. This may explain the hot Jupiters which are observed to be severely bloated.

For my senior thesis[Caltech Library] I investigated the influence that pulsars orbiting with a low-mass companion star have on the companion. I focused on the way in which radiation from the pulsar heats their atmospheres as well as on the transport processes (radiation, convection, winds, etc.) that move this heat around the companion. In the process I found that feedback between accretion and deep companion insolation explains the decade-timescale switching behaviour seen in low-mass X-ray binaries. The same models quantitatively explain the temperature difference between the day (pulsar-facing) and night (opposite) sides of the companion, and readily extend to exoplanets and brown dwarfs. I also developed a new stellar structure code Acorn[Github] and proposed a novel mechanism of accretion-induced collapse.

I studied the stability of parafermionic zero modes in one-dimensional chains with Jason Alicea and Roger Mong.
Using a combination of analytic arguments, exact diagonalisation, and density matrix renormalization group (DMRG) methods we categorically ruled out exact zero modes in a large swath of parameter space, even if there is an exponentially protected ground state degeneracy
[PRB],
[arXiv],
[Github]
. In the perturbative regime this arises because parafermion models allow for several kinds of domain walls which mix when scattering off the system boundaries, resulting in power-law splitting.
In the non-perturbative regime the gap closes, the ground state manifold becomes enormous, and similar splitting results.

PyTNR [GitHub] is a library that allows efficient contraction and decomposition of tensor networks and tensors. I developed this to help with a variety of my projects, ranging from analysis of turbulent flows to computing Bayesian evidence and understanding biophysical lattice models.

ArrFunc [GitHub] is a library that allows vector-valued functions which vectorize over vector-valued inputs to be treated as NumPy arrays. I found myself wanting to do this frequently and across several projects, so it made sense to package separately. ArrFunc makes use of lazy evaluation, so that array operations are delayed until the shape of the input (and hence of the output) is known. This means that you can compose, combine, and modify complicated functions without having to worry about the shape of the input.

As part of my senior thesis I developed Acorn[Github], a Python library for stellar structure and evolution. Acorn uses a basic equation of state (based on a code by Bohdan PaczyĆski) with support for modern opacity tables and supports modelling the envelopes of stars, both in steady state and out-of-equilibrium. The library supports a day-night circulation model I developed for my thesis[Caltech Library] as well as irradiation with varying deposition depth. The library does not include any details of nuclear reaction rates and so cannot be used to model the cores of stars.

During an internship in Professor Markovic's lab at Johns Hopkins University, Justin Silverman and I developed several algorithms for analysing atomic force microscopy (AFM) images of carbon nanotubes. We were later awarded a patent for these algorithms [USPTO 13534428], and packaged them into the NanoImage library. The library produces robust statistics on length and aspect ratio distributions and automatically identifies and excludes sample defects. Requests for licensing should be directed to the Johns Hopkins Technology Transfer office [JHTV].

Churchill College

Cambridge, Cambridgeshire CB3 0DS

+44-7490-397012

adamjermyn@gmail.com

Github

I've written a few things, starting with science fiction short stories [Github]. I'm also active in chainwriting with the Cambridge University Science Fiction Society (CUSFS), and some of the pieces I've written appear in TTBA (the CUSFS magazine).

Ditch Day is a tradition at Caltech in which seniors build puzzle-scavenger-activity hunts known as Stacks. On the day-of, underclassmen spend a full day role-playing, puzzle-solving, and generally following their Stacks to completion. In 2015 Rachel Gates, Kalya McCue and I built the Stack 'Legend of Zelda: Friendship is Magic', as a mashup between the videogame 'The Legend of Zelda: The Wind Waker' and the TV series 'My Little Pony: Friendship is Magic'. Stacks are generally quite involved, so alumni often help construct them, and this one benefited greatly from Jesse Salomon, Jen Greco and Eduardo Gonzalez. The following year I flew back to help with Nick Schiefer's Stack 'The Martian', where I played the role of crew photographer. In 2017 one of my main contributions was a paper puzzle which appeared on four stacks including Tim Maxwell's. In this puzzle a variety of mathematical statements written in an "alien language" are presented without context, and the underclassmen have to use them to identify a four-digit code to undo a padlock.