iiinova — where your digital self keeps evolving

Evan Spotte-Smith (they/them) @ewcss.info · 19/12/2025

Come be my colleague @ucdchemistry.bsky.social! We're currently looking for an organic chemist (broadly defined) to start as a permanent (i.e., tenured) assistant professor. Job reference (www.ucd.ie/workatucd/jo...) is 019237. Due date Feb. 27, 2026 Reach out if you have questions!

Lecturer / Assistant Professor of Organic Chemistry, UCD School of Chemistry, Permanent

Applications are invited for a permanent post of a Lecturer / Assistant Professor of Organic Chemistry within UCD School of Chemistry.

The successful candidate will have research interests in any area of Organic Chemistry, and these include, but are not limited to, the development of novel methods in synthetic chemistry, sustainability, supramolecular chemistry, natural products chemistry, green chemistry, catalysis, and industrial chemistry.

The candidate will contribute to our undergraduate and postgraduate taught programmes in particular, curriculum development of our BSc programmes in (a) Chemistry, (b) Medicinal Chemistry & Chemical Biology and (c) Chemistry with Environmental and Sustainable Chemistry. The candidate will supervise and perform cutting edge research and contribute to administration of the school.

95 Lecturer / Assistant Professor (above the bar) 2010 Salary Scale: €64,753 - €102,539 per annum.

Appointment will be made on scale and in accordance with the Department of Finance guidelines

Closing date: 12:00 noon (local Irish time) on 27/02/2026.

Applications must be submitted by the closing date and time specified. Any applications which are still in progress at the closing time of 12:00 noon (Local Irish Time) on the specified closing date will be cancelled automatically by the system. UCD are unable to accept late applications.
UCD do not require assistance from Recruitment Agencies. Any CV's submitted by Recruitment Agencies will be returned.

Note: Hours of work for academic staff are those as prescribed under Public Service Agreements. For further information please follow link below:
https://www.ucd.ie/hr/t4media/Academic%20Contract.pdf

186

Evan Spotte-Smith (they/them) @ewcss.info · 27/11/2025

"Much Ado About Nothing: Unmotivating "Gender Identity"" by Hernandez and Bell gives @floralashes.bsky.social a run for their money in the "best/funniest footnote in a paper on transness" category. Paper: journals.publishing.umich.edu/ergo/article...

A screenshot from a scholarly paper.

The text reads: "On this picture, questions about the legitimacy of various gender classifications are both descriptive and normative. There is no context-independent fact of the matter about who is a man, a woman, or a sinner. 16 These things depend on contingent and localized social realities. As a result, our theoretical approach to gender is always guided by normative commitments—whether we are aware of that or not. Questions about which classification practices we ought to accept are not just about how the social world is organized, but about how it ought to be."

Footnote 16, referenced above, reads: "16. Ah, yes, the three genders."

1238

Evan Spotte-Smith (they/them) @ewcss.info · 19/11/2025

I suspect that soon, because of "AI", I'll be unable to use Google Scholar. What a loss. OpenAlex is incredible, but its algorithm has, for me, never been as effective for finding relevant papers. (note: OpenAlex does have a new engine; I haven't tested it much, but it might help) #AcademicSky ⚗️ 🧪

2191

Evan Spotte-Smith (they/them) @ewcss.info · 24/10/2025

I'm not surprised to find this in my e-mail inbox, but I am enraged. Springer Nature is now offering "AI" to "help" you "quickly understand any paper" and edit your manuscripts. Publishers should be pushing back against "AI" to protect scientific integrity and author's rights. (1/2) ⚗️🧪 #AcademicSky

4316

Evan Spotte-Smith (they/them) @ewcss.info · 22/08/2025

Yesterday, the US Supreme Court ruled that funding can be withheld for schools that continue DEI practices or recognize gender diversity. Today, I learned that @cmu.edu removed signs in its restrooms supporting trans folks' needs and autonomy. Every day validates my decision to leave the US.

Example of a sign previously posted in CMU bathrooms (both single-stall and multi-stall). It reads: "This restrooms is for everyone. Carnegie Mellon supports individuals using the facility of their choice. Please help us make campus a more welcoming and safer place for everyone. For more information visit: cmu.edu/all-gender"

181

Evan Spotte-Smith (they/them) @ewcss.info · 18/08/2025

Alt text of the open letter that I sent to ACS is continued here.

Recently, one of the invited speakers for my symposium reached out to me. This speaker, who is an immigrant and a person of color, did not feel safe traveling to D.C. and attending ACS in person. They (here I am using gender-neutral pronouns to respect and protect anonymity) requested the option to present virtually so that they could participate in the conference without risking their safety. This is beyond reasonable, and it is certainly possible to accommodate. ACS meetings generally have strong internet connection, allowing for video-conferencing. While I understand that virtual presentations might introduce some additional technical logistics, the extraordinary circumstances justify the relatively minor headache that those logistics would bring. Even so, I was told that ACS would not be allowing any virtual talks in the Fall Meeting.

ACS claims to prioritize the "safety and well-being" of its members and participants, but your actions tell a different story. At this very moment, you are asking researchers to either put their bodies at risk by entering a hostile territory in the midst of a fascist power-grab, or else to be sidelined from valuable scientific dialogue. This is an unacceptable choice that no chemist, and especially no marginalized chemist, should be forced to make. Our community deserves better, and I demand better, from ACS.

Your choices have already damaged the present ACS meeting, and there is perhaps nothing that can be done, now that these choices have been made. Going forward, ACS must make the changes necessary to truly ensure safety and well-being, including making your conferences fully accessible.

I would be open to meeting with your staff to discuss this matter further.

Sincerely,

Evan Walter Clark Spotte-Smith

131

Evan Spotte-Smith (they/them) @ewcss.info · 18/08/2025

The 2025 @acs.org Fall Meeting in Washington, D.C. while D.C. is in the midst of a take-over by the Trump administration. In spite of the circumstances, ACS is not allowing any speakers to switch to virtual presentations. This threatens the safety of vulnerable community members. ⚗️ 🧪 #AcademicSky

To whom it may concern,

Hello. My name is Evan Spotte-Smith (they/them). I am a Carnegie Bosch Institute Postdoctoral Fellow at Carnegie Mellon University (CMU), an incoming Adjunct Professor of Chemcial Engineering (CMU), and an incoming Ad Astra Fellow and Assistant Professor of Digital Chemistry at University College Dublin (UCD) in Ireland. I am an ACS member and an organizer for the 2025 ACS Fall Meeting, for which I co-organized the symposium "Chemical reaction networks, retrosynthesis, and reaction prediction" alongside Dr. Samuel Blau and Prof. Brett Savoie.

I am writing to you with concerns regarding the 2025 ACS Fall Meeting in Washington, D.C.

The District of Columbia is currently in a state of _de facto_ martial law. Armed federal agents roam the streets. In some cases, these agents are masked and refuse to self-identify, making them indistinguishable from vigilante militias or secret police. Coupled with the Trump administration's extralegal, authoritarian, and violent detentions and "deportations" of immigrants, visitors, and in some cases even United States citizens, this makes D.C. an unsafe place for many ACS members and meeting attendees...

Alt text is continued in next post

1227

Evan Spotte-Smith (they/them) @ewcss.info · 26/12/2024

These optimized particles included some with visible/UV emissions 6.5x higher than any UCNP in our training set! To ensure that these far-OOD predictions were reasonable, we validated by performing additional kMC simulations, finding generally good agreement. (9/10)

a) A plot of "# of regions, core + shell" vs. "maximum nanoparticle radius (in nm)". Each box in the plot has two sections, with the lower section representing the ML predicted intensity of the optimal particle in that region of the search space and the upper section representing the kinetic Monte Carlo-simulated ultraviolet emission of the same particle. The color of each section of each box represents the ultraviolet/blue intensity (in counts per second or cps). The maximum intensity within the training set is around 20,000 cps, while the brightest particle observed has a kMC-predicted intensity > 100,000 cps; b) four selected optimized particles represented by layered pie charts indicating the doping in each region of the particles: i) 4 regions, with a maximum radius of 10nm. This particle has 2x UV intensity compared to the best particle in the training set; ii) 5 regions, with a maximum radius of 12nm; iii) 7 regions, with a maximum radius of 14nm; iv) 10 regions, with a maximum radius of 15nm. This particle had 6.5x UV intensity compared to the best particle in the training set.

130

Evan Spotte-Smith (they/them) @ewcss.info · 26/12/2024

While testing out these different models, we realized that UCNPs and similar core-shell particles display a unique "subdivision invariance". You can define layers arbitrarily, and the particle behavior doesn't change, but the representations (including for heterographs) do change. (6/10)

a) A visual depiction of subdivision invariance. Two versions of the same particle are shown. One has only a single layer; the other is split in two layers. These two structurally equivalent particles have non-equivalent hetero-graph representations; this is shown through an explicit depiction of their associated graphs. b) A comparison of un-augmented and augmented training. Unaugmented training leads to an out-of-distribution (OOD) MSE of 2.3%, while on-the-fly data augmentation, with new, randomly-subdivided particles added in each training epoch, leads to an OOD MSE of 1.7%.

120

Evan Spotte-Smith (they/them) @ewcss.info · 26/12/2024

We also designed a new heterogeneous graph representation, with one set of nodes representing UCNP dopants and another set representing interactions within a layer or between layers. We found that our hetero-GNN outperformed other models in-distribution and especially out-of-distribution. (5/10)

A depiction of heterogeneous graph representation for upconverting nanoparticles (UCNPs) (a-b) and a hetero-graph neural network architecture (c-e). a) A directed graph is shown, with green circles representing dopant nodes, red triangles representing intra-layer interaction nodes, and blue triangles representing inter-layer interaction nodes between a core layer and a shell layer. Directed edges link dopant nodes to interaction nodes. Node feature vectors, represented by gray boxes, are shown. Node features, shown in a box on the right, include dopant type, dopant concentration, inner and outer layer radius, and interaction type. b) Integrated interaction is a way to account for the variation in interaction strength between dopants as a function of distance. We integrate over pairwise distances; this is shown graphically using a circle with a line segment representing the pairwise distance between two points, as well as with an equation for the integrated interaction phi. c) The overall hetero-GNN architecture. Dopants, intra-layer interactions, and inter-layer interactions all go through separate embedding layers. Then, there are three "hetero-convolution" layers. These pass through a mean aggregation, followed by a fully connected neural network layer that produces the log(intensity). d) A more detailed depiction of the dopant node embedding, including a species embedding, feature-wise linear modulation (FiLM) layers and batch normalization; e) a more detailed depiction of the interaction node embedding, which includes a type-based embedding, the integrated interaction layer, a batch normalization layer, and a FiLM layer.

140

Evan Spotte-Smith (they/them) @ewcss.info · 26/12/2024

With SUNSET, we can learn structure-property relationships for these complex nanostructures. We considered a number of different representations for nanoparticles - simple feature matrices, voxels (for convolutional neural networks), and graphs (for graph neural networks or GNNs). (4/10)

a) A 3D depiction of a core-shell upconverting nanoparticle (UCNP), with a shell of undoped NaYF4 and core layers doped with Yb3+, Nd3+, and Er3+; b) the heterostructure - where the dopants are, and what their concentrations are - determines the photophysics. This is depicted in two ways: i) a circle representing a UCNP contains three smaller circles representing dopants, with arrows between them indicating energy transfer. Two of the dopants (Nd and Er) engage in "cross-relaxation" or quenching, which prevents near infrared (NIR) light from being converted to ultraviolet light. ii) Two NIR photons hit two Nd dopants, which then pass energy to Yb dopants until eventually reaching Er, which emits UV light; c) The core and each shell layer can be doped differently. This is shown by a layered pie chart, with white representing Y, orange representing Er3+, green representing Nd3+, and blue representing Yb3+; d) to construct a dataset, multiple nanoparticles (represented by the layered pie charts) are used as inputs to kinetic Monte Carlo simulations, which produce a database of upconversion spectra; e) a plot with "model error" as a y-axis and "representation development" as an x-axis is shown. Three points are plotted. From right to left and top to bottom: tabular data, images (n-D tensors), and graphs. The graph point links to a box labeled "hetero-GNN"; the hetero-GNN passes gradients and predicted intensities to subfigure f; f) We preform gradient based optimization. A simple optimization surface is shown in blue, with an orange star at the maximum point. An arrow links this orange star to a pie chart representing the composition of an optimized UCNP.

120

Evan Spotte-Smith (they/them) @ewcss.info · 26/12/2024

Towards this end, we performed thousands of high-fidelity kinetic Monte Carlo simulations (using RNMC, recently published in JOSS) and compiled them into a dataset called SUNSET - Simulated Upconverting Nanoparticle Spectra for Emission Tuning. (3/10)

A depiction of the SUNSET dataset; a) a table shows the breakdown of four sub-datasets (SUNSET-1, SUNSET-2, etc.) in terms of incident wavelength (800nm for SUNSET-1, 980nm for the others), presence of dopants (only SUNSET-1 has Nd, but all four have Yb and Er and only SUNSET-4 has Tm) and surface effects (only present in SUNSET-2 and 4). A plot showing the number of datapoints in each sub-dataset is also shown, with SUNSET-1 having about 6,000, SUNSET-2 and SUNSET-3 having somewhat less, and SUNSET-4 having the most at about 20,000. b) A further depiction of SUNSET-1. In-distribution train and test data contains core-only particles and particles with one, two, and three shells. The core size varies from a radius of 1nm to 4nm, and the shell thickness varies from 1nm to 2.5nm. The out-of-distribution test set contains 206 4-shell nanoparticles. At the bottom of the sub-figure are plots showing the number of particles versus nanoparticle radius (a distribution centered at around 8nm), shell thickness (with a somewhat even distribution between 1.0nm and 2.5nm), dopant concentration (a distribution which decays from 0.0 to 1.0), and log(UV intensity) (a sharp peak around 2, then a relatively flat distribution until about 3.5, at which point the distribution decays).

120

Evan Spotte-Smith (they/them) @ewcss.info · 27/11/2024

I'm Evan (they/them). I live with asthma, an autoimmune disease, and a host of mental illnesses (depression, anxiety, and more!). Also neurodivergent. I'm a postdoc and incoming ChemE faculty at Carnegie Mellon University. I study reactivity, sustainable chemistry, chemical data science.

A picture of Evan (white person with long, wavy brown hair) sitting on a staircase in front of a shrub and smiling at the camera. They're wearing an orange spaghetti strap dress, a black beanie reading "Slim Dilly Dogs", and sunglasses. In their hand is a biodegradable single-use cup, and on their lap is an N95 mask. A cell phone and hand sanitizer are visible in their dress pocket.

0211

Evan Spotte-Smith (they/them) @ewcss.info · 25/09/2024

I just got back today from a kickoff event at TRI HQ in CA. Pictured here are several of the other SARC PIs - Aditi Krishnapriyan (Berkeley) and Daniel Schwalbe-Koda (UCLA) - and the TRI crew. Not pictured: Bingqing Cheng (Berkeley) and Tonio Buonassisi (MIT). Very excited to work with these folks!

000

Evan Spotte-Smith (they/them) @ewcss.info · 05/06/2024

Name: Evan Spotte-Smith Pronouns: they/them Field: #CompChem, electrochemistry, chemical data science Picture: I have so few pics of myself! It's a problem Hobbies: hiking, reading, tasting wine and tea Random fact: my first educational gig was teaching a high school Model UN group in Brooklyn

A shot of a pale-skinned genderqueer chemist chest-up, shot from below. They're wearing a name tag from the 2023 AIChE Annual Meeting, a black blouse, an orange blazer, and an N95 mask. A light fixture above them is causing some strange effects with the camera.

051

Evan Spotte-Smith (they/them) @ewcss.info · 08/01/2024

While our findings are preliminary, we suggest that oxygen anions superoxide (O2^-) and peroxide (O2^-2) could be to blame. Both anions can react readily with EC, especially peroxide. We hope this stimulates more studies into cathode-side electrolyte decomposition, especially with oxygen anions! 5/5

Energy diagram for reactions between ethylene carbonate (EC) and peroxide anion. Three reactions are shown: a proton transfer (shown in light blue); a nucleophilic substitution (dark blue); and an addition to form a tetrahedral complex (purple). All pathways are thermodynamically favorable and have low barriers, with proton transfer being the most favorable, with a barrier of 0.08 eV.

000

Evan Spotte-Smith (they/them) @ewcss.info · 08/01/2024

Similarly, when we looked at chemical oxidation of EC by evolved singlet oxygen, we found that the mechanisms reported in the prior literature were kinetically limited and basically inacessible at room temperature. There must be some other, unexplored mechanism driving electrolyte degradation. 4/5

An energy diagram for two reactions between ethylene carbonate (EC) and singlet oxygen (1O2): a multi-step reaction forming a dicarbonyl and water, shown in red; and a concerted, single-step reaction forming hydrogen peroxide (H2O2) and vinylene carbonate (VC). Both pathways suffer from extremely high barriers and are kinetically inaccessible at moderate temperatures.

100

Evan Spotte-Smith (they/them) @ewcss.info · 08/01/2024

We find that direct electrochemical oxidation of the common solvent ethylene carbonate (EC) is thermodynamically infeasible at essentially any potential reached during normal Li-ion battery operation, even when accounting for concentration effects. 3/5

(a) Depiction of the stepwise oxidation reaction ethylene carbonate (EC) → EC+ + e– , with 3D structures for EC and EC+. The oxidation potential is Eox = 6.98 V. (b) Depictions of the concerted dissociative oxidation reaction 2EC → (EC+H)+ + (EC–H) + e–, with 3D structures of the reactants and products as clusters (top, Eox = 5.80 V) and isolated molecules (bottom, Eox = 5.84 V). (c) ΔG for the stepwise electrochemical oxidation of EC, as a function of potential and relative concentration. Oxidation is unfavorable until at least ~6.5 V (d) ΔG for the concerted dissociative oxidation of EC as a function of potential and relative concentration, where we assume that (EC+H)+ and EC–H are at the same concentration, denoted as [product]. This reaction is still unfavorable but could occur at ~4.8 V given extremely low product concentrations.

100