Date:

Monday, May 05, 2008,
11:00 am - 12:00 pm.

Place:

Mathematics Conference Room 626, UCD Building, 1250 14th St., Denver.

Speaker:

Brad Bendiak

Affiliation:

Cell and Developmental Biology, University of Colorado Denver School of Medicine.

Title:

Nuclear magnetic resonance spectroscopy: conversion of multidimensional time-domain data to a correlated frequency.

Abstract:

Nuclear magnetic resonance spectroscopy is a type of spectroscopy where radiofrequency is first absorbed by a molecule from a transmitter pulse, and is later emitted from individual nuclei within the molecule as a function of time. The emitted radiofrequency is detected as the sum of a series of overlapping sinusoids, each nucleus typically emitting at a unique frequency. However, if time delays are also incorporated between multiple transmitter pulses, the data is related as a function of more than one time period. The emitted signal then reflects energy transfers occurring between nuclei during time periods between transmitter pulses. These energy transfers are related to whether nuclei are connected to each other (coupled) through chemical bonds, or through physical space. Transforming the data into a frequency-domain "picture" is important so that the relationship between nuclei in some unknown molecule can be established, thus the overall primary and three-dimensional structures of molecules can be deduced. For many years, transforming the data through multiple orthogonal Fourier transforms was the best approach available, but more recently, a conversion of the time-domain signal using multiple time domains directly to a multidimensional frequency-domain spectrum can be accomplished as a mathematical eigensystem using a technique called the filter diagonalization method. The advantages can be dramatic, but there are still issues, both computational and mathematical, that complicate convergence to a frequency-domain solution, especially when signal/noise is limiting.