Simulations of Chemical Exchange Lineshapes in CP/MAS Spectra
Using Floquet Theory and Sparse Matrix Methods
P. Hazendonk, Alex D. Bain,
1
H. Grondey,* P. H. M. Harrison, and R. S. Dumont
Department of Chemistry, McMaster University, 1280 Main St. W, Hamilton, Ontario L8S 4M1, Canada; and
*Department of Chemistry, University of Toronto, Toronto, Ontario M5S 3H6, Canada
Received October 11, 1999; revised April 28, 2000
This paper presents a general method for simulating the effect
of chemical exchange on MAS NMR spectra of solid samples. The
complication in MAS spectra is that the Hamiltonian itself is
time-dependent, due to the spinning of the sample. The approach
taken in this work is to use Floquet theory to convert the problem
into a time-independent form, and then use established methods
(used in liquid NMR simulations) to calculate the lineshape. Flo-
quet theory has been admired for its elegance, but criticized for its
computational inefficiencies. This is because it removes the time
dependence of the system by expanding the problem in a Fourier-
like series. This makes a relatively small, time-dependent calcu-
lation into a much larger time-independent one. Typically, we use
twice as many Floquet blocks as there are spinning sidebands, so
the increase in size is substantial. The problem that this creates
stems from the fact that the usual Householder methods for
diagonalizing a matrix scale as the cube of the size of the matrix.
This would make a Floquet calculation prohibitively long. How-
ever, the Floquet matrix is inherently sparse, so sparse matrix
methods can produce substantial computational savings. Also,
fully diagonalizing a matrix is expensive, but converting the ma-
trix to a tridiagonal form (using iterative Lanczos methods) is
much cheaper. The use of the Lanczos methods makes the Floquet
calculations feasible as a general method forsystems of more than
one spin. We show how to set up the full matrix describing
chemical exchange in a spinning sample, but the details of how the
Lanczos methods work are not included—they are described else-
where. We then validate the theory by simulating the MAS spectra
of dimethyl sulfone both with natural abundance
13
C and with
methyl groups labeled with
13
C. The latter system has both dipolar
and chemical shielding anisotropy terms contributing to the
spectrum. © 2000 Academic Press
Key Words: chemical exchange; CP/MAS; Floquet theory; spin-
ning sidebands.
INTRODUCTION
The use of cross-polarization (CP) and magic-angle spinning
(MAS) in obtaining the NMR spectra of solid samples is well
known (1). The lineshape changes in the NMR spectra of liquid
samples, due to chemical exchange, are also very familiar
(2, 3). However, there are relatively few rigorous studies of the
effects of chemical exchange on the patterns of spinning side-
bands observed in CP/MAS spectra of complex spin systems.
This is because the Hamiltonian is time-dependent, and the
spectrum is relatively complicated to calculate (compared to
liquids). For simple solid systems (4), exchange effects are
often treated in the same way as liquid spectra, using the
standard Kubo and Tomita (5) or Gutowsky and Holm formal-
ism (6). To provide a full and general description, it is neces-
sary to combine the rigorous treatment of liquids with the
complexities of the MAS experiment.
In this work we use Floquet theory (7–17) to convert the
time-dependent MAS problem into a time-independent (but
much larger) description. This is not a new approach (18), but
it has been hampered by the size of the matrices generated.
However, the Floquet matrices are inherently sparse. As such,
modern sparse matrix methods reduce the numerical difficul-
ties dramatically. We use the dual Lanczos method (19 –24) in
Wassam’s formulation (25, 26) to tridiagonalize the resulting
matrices. For large matrices, this is much more efficient than
the usual Householder method (27). The general formalism is
developed and applied to the simulation of the MAS
13
C NMR
spectra of doubly
13
C-labeled dimethyl sulfone.
Floquet theory was introduced to spectroscopic problems by
Shirley (28), who applied this theory to compute the propaga-
tor corresponding to a time-dependent non-self-commuting
Hamiltonian. This is achieved by changing to a Fourier-spin
space, where the Hamiltonian is time-independent, but infinite
in dimension. Using similar methods, Vega described multiple-
quantum effects in double-frequency pulsed NMR experiments
on spin-1/2 and 1 systems (29 –31). Later Zax and Vega used
Floquet methods to design broadband pulses (32, 33). Vega
also applied Floquet theory to compute sideband patterns of
rotating solids. This led to expressions for sideband intensities
that were similar to those by Herzfeld and Berger (34) and
Maricq and Waugh (35). For multispin systems, numerical
methods are required. Based on a perturbation method pro-
posed by Maricq (36), spectra of coupled spin pairs were
computed to model rotational resonance (13–15, 17, 37) and
REDOR dephasing curves (12, 38). Schmidt and Vega applied
Floquet theory to uncoupled exchange in rotating solids using
the Bloch–McConnell approach (18, 39, 40). In this case, only
1
To whom correspondence should be addressed.
Journal of Magnetic Resonance 146, 33– 42 (2000)
doi:10.1006/jmre.2000.2111, available online at http://www.idealibrary.com on
33
1090-7807/00 $35.00
Copyright © 2000 by Academic Press
All rights of reproduction in any form reserved.