Eman AhmedRutgers BME

Research area

Running ATRP on a robot

Oxygen-tolerant photoinduced ATRP in open 96-well plates: reaction conditions, ligand and initiator screening, and reported dispersity.

Atom transfer radical polymerisation gives you control over chain length and composition, which is exactly what you need to build a polymer library worth modelling. Historically it also needed sealed, degassed glassware, because propagating radicals react with molecular oxygen far faster than with monomer.

That requirement quietly sets the ceiling on throughput. Degassing is manual, slow and does not parallelise. A dozen carefully controlled reactions in a day is not enough to map a reaction space with four or five interacting variables.

The reaction, as it actually runs

Published conditions for the automated photo-ATRP platform.

Two of these numbers carry the whole idea. A 200 µL reaction sitting under 100 µL of air is a reaction that has given up on being oxygen-free. The catalytic system scavenges the oxygen instead. That is what lets the plate stay in open labware and be handled by a robot.

Show the numbers
ParameterValue
Plate format96-well polypropylene
Reaction volume200 µL
Oxygen headspace100 µL
SolventDMSO
Metal catalystCuBr₂
PhotocatalystsZnTPP, Eosin Y
Primary light source560 nm red LED, 5 mW cm⁻²
Validation light source515 nm green LED, 2.61 mW cm⁻²
Ligands screenedMe₆TREN, PMDETA
Initiators screenedMBiB, BPN

Reported Conditions as published in Ramirez, Ahmed et al., ACS Polymers Au 6(1), 181–193 (2026).

What the screen is actually for

Acrylates propagate quickly and are relatively forgiving. Methacrylates are not. Methyl methacrylate has a substantially smaller propagation rate constant, so the activation–deactivation balance ATRP depends on has to be retuned, and the right ligand and initiator pairing is not something you can reliably reason your way to.

So you screen it. That is the case for throughput in one sentence: not that more is better, but that for this class of question the empirical answer is cheaper and more trustworthy than the theoretical one.

Different monomers want different reagents

The reagent pairing that gave good control, by monomer class. The same combination does not work across the board, which is the finding.

Acrylates HEA, MA and HPA pair with Me6TREN and MBiB; methyl methacrylate pairs with PMDETA and BPN.Me₆TREN + MBiBPMDETA + BPNHEAHEA · Me₆TREN + MBiBMAMA · Me₆TREN + MBiBHPAHPA · Me₆TREN + MBiBMMAMMA · PMDETA + BPN

Methyl methacrylate propagates more slowly than the acrylates, so the activation–deactivation balance that ATRP depends on has to be retuned. There is no reliable way to reason to the answer from first principles, which is exactly why screening it is worth the instrument time.

Show the numbers
MonomerClassReported best pairing
2-Hydroxyethyl acrylate (HEA)acrylateMe₆TREN + MBiB
Methyl acrylate (MA)acrylateMe₆TREN + MBiB
2-Hydroxypropyl acrylate (HPA)acrylateMe₆TREN + MBiB
Methyl methacrylate (MMA)methacrylatePMDETA + BPN

Reported As reported in the paper: Me₆TREN consistently produced high dispersity with MMA, while PMDETA and BPN emerged as the most suitable reagents for high-throughput MMA synthesis.

Dispersity, reported

Đ = Mᵥ/Mₙ. A perfectly uniform chain population would be 1.00; below about 1.3 is normally taken as well-controlled.

Acrylates optimised 1.15, acrylates well-controlled upper bound 1.30, MMA with PMDETA and BPN 1.26.1.001.101.201.301.40Đ = 1.00, perfectly uniformAcrylates, optimised: 1.15Acrylates, optimisedbest reported condition1.15Acrylates, well-controlled: 1.30Acrylates, well-controlledreported upper bound, Đ < 1.31.30MMA with PMDETA + BPN: 1.26MMA with PMDETA + BPNat target specifications1.26Dispersity (Đ)
  • Acrylates
  • Methacrylate (MMA)
Show the numbers
ConditionĐNote
Acrylates, optimised1.15best reported condition
Acrylates, well-controlled1.30reported upper bound, Đ < 1.3
MMA with PMDETA + BPN1.26at target specifications

Reported Only values stated in the paper are plotted. This is not the full screen. For the complete dataset see the publication.

The part that makes it reproducible

A platform that only works in the lab that built it is a demonstration, not a method. The obstacle is rarely the robot. It is translating an intended design into deck layout, volumes and transfer order without arithmetic errors propagating silently across ninety-six wells.

The paper ships a Python package for that planning step. It is not the scientifically interesting part, and it is probably what decides whether anyone else can run this.

More on oxygen tolerance in the research notes.

Other research areas