Optimized synthesis and comprehensive characterization of palladium nanoparticles

A team of researchers has developed a streamlined approach to making palladium nanoparticles—tiny metal spheres used as catalysts in industrial chemistry—by running multiple synthesis experiments in parallel and using statistical design to find the best conditions. The work, published in Nature in September 2026, demonstrates how to produce these particles more efficiently while generating less waste, a shift toward what chemists call green chemistry.

Palladium nanoparticles are workhorses in industry, speeding up chemical reactions in everything from pharmaceuticals to petroleum refining. But making them consistently, with uniform size and shape, has been a persistent challenge. The researchers tackled this by using a commercial multi-synthesis device that can run up to 40 reactions simultaneously under identical conditions, paired with a statistical method called Doehlert design optimization. Rather than changing one variable at a time—the traditional approach—they evaluated five parameters together: the concentration of palladium, the concentration of two different thiol ligands (molecules that coat and stabilize the particles), the amount of reducing agent, reaction time, and temperature. This multivariate approach allowed them to identify which factors actually mattered and how they interacted, using far fewer experiments than conventional methods would require.

The optimization process revealed that temperature and palladium concentration were the key variables influencing particle size distribution. The team settled on compromise conditions: 10 micromoles of palladium, 0.15 millimoles of ascorbic acid as the reducing agent, 0.08 millimoles of ligand, a temperature of 60 degrees Celsius, and a reaction time of 90 minutes. Under these conditions, they produced two types of particles: palladium nanoparticles capped with 3-mercaptopropionic acid (PdNP-MPA) and those capped with L-cysteine (PdNP-Cys). The particles were then purified through six washing cycles to remove unbound ligands, a step the researchers confirmed by monitoring sulfur concentration in the wash solutions.

Characterizing these particles required a toolkit of complementary techniques, each revealing different information. Transmission electron microscopy showed the actual shape and size of individual particles—PdNP-MPA averaged 96 nanometers with a standard deviation of 33 nanometers, while PdNP-Cys averaged 71 nanometers with a standard deviation of 30 nanometers. Both samples contained particles of mixed shapes: some quasi-spherical, others cuboctahedral or truncated octahedral, with some multi-pod structures. Energy-dispersive X-ray spectroscopy confirmed that palladium was evenly distributed within each particle and that sulfur from the ligands coated the surface uniformly. Dynamic light scattering and nanoparticle tracking analysis measured the hydrodynamic diameter—how large the particles appear when suspended in liquid—and confirmed that PdNP-Cys had a narrower size distribution than PdNP-MPA. Single particle inductively coupled plasma mass spectrometry, a technique that vaporizes individual nanoparticles and measures their elemental composition, provided element-specific size determination and particle counting, yielding results consistent with electron microscopy but analyzing thousands of particles rather than a few hundred.

One critical finding was that different measurement techniques can give different answers, and understanding why matters. When the researchers analyzed a platinum nanoparticle reference material by nanoparticle tracking analysis, they obtained a particle count of 5.6 × 10¹⁴ particles per liter—substantially higher than the certified value of 4.0 × 10¹³ particles per liter. Single particle ICP-MS, by contrast, gave 3.5 × 10¹³ particles per liter, much closer to the certified value. This discrepancy likely arises because nanoparticle tracking analysis has a detection bias toward larger particles and agglomerates, while single particle ICP-MS measures only the element of interest. For the palladium samples themselves, the two techniques gave statistically equivalent results, suggesting that for these particular particles, the bias was minimal.

The researchers also determined how densely the thiol ligands packed onto the particle surface by measuring the ratio of sulfur to palladium using bulk ICP-MS. For PdNP-MPA, the total palladium concentration was 343.80 milligrams per liter and sulfur was 0.39 milligrams per liter; for PdNP-Cys, palladium was 310.33 milligrams per liter and sulfur was 0.35 milligrams per liter. Using particle counts from single particle ICP-MS and assuming the particles were spherical, they calculated ligand density—the number of ligand molecules per square nanometer of surface. The two calculation methods they employed gave comparable results, with no statistically significant difference, though the researchers note that because the actual particles were irregularly shaped rather than spherical, the true ligand density was likely slightly lower than their calculations suggested.

The broader significance of this work lies in its demonstration that multivariate optimization and parallel synthesis can reduce both the time and the material waste involved in developing nanoparticle syntheses. Compared to conventional single-batch, one-variable-at-a-time approaches, the team's method required fewer experiments, consumed less reagent, and generated less waste—all principles central to green chemistry. The researchers propose that future work will link automated synthesis directly to real-time analytical feedback, creating closed-loop systems that continuously optimize production. For now, their framework offers a practical roadmap for researchers developing new nanoparticle syntheses and for those characterizing polydisperse, irregularly shaped particles that don't fit the idealized assumptions of simpler measurement techniques.