Saptarsi Mondal Ultrafast & multidimensional IR spectroscopy
Menu

RESEARCH / ■ THEME IV · ONGOING · COLLABORATIVE

Vibrational probes, methods and analysis

What is a vibrational probe actually reporting, and how do we extract a trustworthy dynamical number from a noisy, congested, overlapping spectrum?

The two surface-specific methods that underpin the interfacial work: SEIRAS for potential-resolved structure, surface-enhanced 2D IR for picosecond dynamics.

Most of my method work started as an obstacle in someone’s experiment. An azide probe whose lineshape would not fit a single band. A free-volume measurement that gave a physically implausible cone angle. Ultrafast data whose signal-to-noise put the interesting waiting times out of reach. In each case the fix turned out to generalise beyond the immediate problem, which is the useful thing about method work: a corrected model or a cleaner probe improves every measurement that uses it afterwards.

The machine-learning benchmark belongs here for the same reason. Computing frequency-fluctuation correlation functions from first principles is what limits how large and how heterogeneous a system can be simulated alongside a 2D IR measurement. Showing that a general-purpose potential reproduces those functions without retuning — across water, methanol and tetrahydrofuran — removes a real constraint on the interfacial and confined systems I care about, where the reference calculations are hardest.

Methods

  • Isotope substitution and labelling strategies
  • Two-dimensional infrared spectroscopy
  • Discrete wavelet transform denoising
  • Frequency-fluctuation correlation function analysis
  • Machine-learned interatomic potentials

Publications in this theme

6 records → publications

Projects in this theme

  • Water structure and free volume in biocompatible polymers Two connected questions about amorphous polymers: what water does at their surface, and how large the empty spaces inside them actually are — both answered by ultrafast infrared measurement rather than by inference from bulk properties. Completed
Why it matters
Every result in the other three themes rests on a probe molecule and a fitting model. If the probe's lineshape is contaminated by a Fermi resonance, or if the model attributes to whole-molecule motion something that is really internal bond rotation, the extracted timescale is wrong in a way no amount of signal averaging will reveal. Method work is not separate from the science here; it is what makes the science reliable.
My contribution
I used isotope labelling and 2D IR to identify hidden Fermi resonances in azide-derivatised amino acids, giving design guidance for cleaner biological probes. I recast the wobbling-in-a-cone model by separating a nitrile probe's internal bond rotation from whole-molecule diffusion — the correction that makes free-volume measurement accurate. I developed wavelet-transform denoising for ultrafast spectra, and benchmarked a general-purpose machine-learned potential against measured 2D IR frequency-fluctuation correlation functions.
What we found
Azide probes carry Fermi resonances that distort their lineshapes and can be removed by isotope substitution. The standard wobbling-in-a-cone analysis systematically misassigns internal rotational fluctuation as whole-molecule motion, and correcting it changes the inferred free-volume element size. And the Universal Model for Atoms reproduces empirical solvent-dependent frequency fluctuation correlation functions without system-specific tuning, at a fraction of the cost of conventional electronic-structure methods.
Where it goes next
Extend machine-learned potentials from bulk solvation to interfacial and confined environments, where reference electronic-structure calculations are most expensive and least tractable — the regime where the computational cost is currently the binding constraint on interpreting surface-enhanced 2D IR data.

Where this work was done

Other themes

Loading larger view…