Journal of Magnetic Resonance 138, 232243 (1999)

Article ID jmre.1999.1727, available online at http://www.idealibrary.com on

Pulsed Electron-Nuclear Double Resonance (ENDOR) at 140 GHz

M. Bennati,* C. T. Farrar,* J. A. Bryant,* S. J. Inati,* V. Weis,* G. J. Gerfen,
P. Riggs-Gelasco,* J. Stubbe,* and R. G. Griffin* ,1
*Francis Bitter Magnet Laboratory and Department of Chemistry, Massachusetts Institute of Technology, Cambridge, Massachusetts 02139; and
Department of Physiology and Biophysics, Albert Einstein College of Medicine, Bronx, New York 10461
Received September 25, 1998; revised February 3, 1999

We describe a spectrometer for pulsed ENDOR at 140 GHz,

which is based on microwave IMPATT diode amplifiers and a
probe consisting of a TE 011 cavity with a high-quality resonance
circuit for variable radiofrequency irradiation. For pulsed EPR we
obtain an absolute sensitivity of 3 3 10 9 spins/Gauss at 20 K. The
performance of the spectrometer is demonstrated with pulsed
ENDOR spectra of a standard bis-diphenylene-phenyl-allyl
(BDPA) doped into polystyrene and of the tyrosyl radical from E.
coli ribonucleotide reductase (RNR). The EPR spectrum of the
RNR tyrosyl radical displays substantial g-anisotropy at 5 T and is
used to demonstrate orientation-selective Davies-ENDOR. 1999
Academic Press

INTRODUCTION

Electron nuclear double resonance (ENDOR) spectroscopy,

first proposed by Feher (1), is a well established technique for
the determination of hyperfine interactions and spin density
distributions in paramagnetic systems. With ENDOR, the
NMR spectrum in the vicinity of a paramagnetic center can be
indirectly detected through the electron spin, leading to much
higher sensitivity than in conventional NMR experiments. ENDOR also offers substantially improved resolution compared to
a typical EPR experiment and therefore represents a very
powerful method for hyperfine spectroscopy. Developments in
pulsed ENDOR techniques allow further possibilities for manipulating and extracting information from very complicated
CW-ENDOR spectra. Reviews of Davies (2) and Mims (3)
pulsed ENDOR and more advanced pulsed schemes are available (4 6).
In the past few years, improvements in high frequency
microwave technology stimulated the development of EPR
spectrometers operating in the millimeter regime (714). Although there are several attractive features of high field EPR
spectroscopy, one of the most important is the improved orientation-selective excitation that can be achieved. Indeed, high
field EPR spectra of systems with even small g-tensor anisot1

To whom correspondence should be addressed. E-mail: griffin@ccnmr.

mit.edu.

1090-7807/99 $30.00
Copyright 1999 by Academic Press
All rights of reproduction in any form reserved.

ropy are typically dominated by the g-dispersion. Thus, microwave pulsed irradiation at arbitrary field positions in the EPR
line excites only molecules with particular orientations of the
g-tensor principal axis with respect to the magnetic field. This
leads to a simplification of the data, where spectral information
from only a few molecular orientations is observed. Further,
high magnetic fields also lead to much higher resolution in
ENDOR spectroscopy of low-gamma nuclei. For instance, in
contrast to the situation at X-band, hyperfine spectra of 13C,
15
N, 14N, and 2H are completely resolved from one another at
our field of 5 T.
Therefore, the combination of high frequency EPR and
ENDOR techniques is an attractive spectroscopic tool, as has
been demonstrated at 95 GHz (10, 15, 16) and with CW experiments at 250 GHz (17). One challenging aspect of pulsed
EPR/ENDOR at high frequencies is the design and implementation of efficient double resonance probes. The paucity of
stable millimeter-wave power sources and amplifiers requires
the use of high quality microwave cavities in order to obtain
short microwave pulses (t p/ 2 ! 200 ns). In contrast to standard
9 GHz EPR, high quality cavities (Q $ 3000) can be employed at high frequencies without degrading the temporal
resolution (i.e., dead time) of the pulsed spectrometer. The
cylindrical TE 011 cavity, introduced by Lebedev et al. (18) at
140 GHz, has been employed for pulsed EPR at millimeter
wavelengths because of its high filling factor and ease of
tuning. One disadvantage of this cavity is its miniature size,
which requires small samples and external mounting of the
radio frequency NMR coil for ENDOR, leading to compromises between RF penetration and the quality factor of the EPR
cavity.
In this contribution we describe the design and operation of
a new spectrometer for pulsed ENDOR at 140 GHz and demonstrate its capabilities with studies of the tyrosyl radical from
E. coli ribonucleotide reductase (RNR). The microwave bridge
described in detail by Prisner et al. (11) and further by Becerra
et al. (12) was modified in order to achieve high-power microwave pulses. A double resonance probe for efficient EPR
and NMR irradiation has been implemented together with a

232

PULSED ENDOR AT 140 GHz

radiofrequency transmission line. Pulsed ENDOR performance

at 140 GHz is demonstrated with a standard BDPA/polystyrene
sample, where the ENDOR spectrum is rather simple because
it is determined by only two sets of nonequivalent protons.
Finally, the advantages of high-frequency orientation-selective
ENDOR are demonstrated experimentally and with simulations
of ENDOR spectra of the tyrosyl radical from E. coli RNR.
THE 140 GHz PULSED EPR/ENDOR SPECTROMETER

The pulsed EPR spectrometer consists of three main units: a

magnet utilizing a 60.4 T superconducting sweep coil and a 2H
NMR field lock system, a 139.5 GHz microwave bridge with
four phase capabilities operating in reflection mode, and a
probe for EPR/NMR double resonance. These components are
described in more detail in the following sections.
Magnet and Lock System
The polarizing magnetic field consists of a two-coil superconducting magnet (Magnex Corporation) with a 130-mm
room-temperature bore. The larger main coil, operating in
persistent mode, provides a static field of 5 T; the smaller
superconducting sweep coil, located in the main coil bore,
allows field sweeps of 60.4 T and access to systems with
electron g-values in the range 1.84 2.16. The low current
required for a 60.4-T field sweep (613 A) minimizes the
helium boil-off rate.
For precise magnetic field measurements and field locking,
an NMR teslameter that permits g-factor accuracies of #10 26
was implemented. The system, which is similar to that described by Un et al. (19), utilizes the dispersive component of
the 2H NMR signal from a 2H 2O sample doped with MnCl 2 as
an error signal in a feedback loop. The feedback signal controls
the current amplifier that drives the sweep coil. When performing a field step, the RF excitation frequency of the deuterium
NMR probe is swept under computer control, causing a change
in the sweep coil current. Thus, the magnetic field step is
determined by the width of the deuterium NMR line, which is
;100 Hz (0.015 mT for deuterium). The field resolution is not
limited by the current stability of the sweep coil power supply
(KEPCO), specified to 0.01% of the current value, since the
large inductance of the sweep coil (L 5 24.0 H) and small
resistance of the sweep coil lead (R 5 0.5 V) act to filter out
any high-frequency power supply voltage fluctuations due to
the very long time constant ( t 5 L/R ; 48 s) for the current
response. For instance, for g 5 2 resonances a sweep coil
current of about 1 A leads to a specified current stability of 0.1
mA, corresponding to a predicted field instability of 3.1 mT,
which is greater than that observed experimentally. Field stability and resolution were tested experimentally with the narrow EPR line of perylene-PF 6, an organic conductor radical
cation salt. At 140 GHz the single-crystal (0.1 3 0.15 3 0.1

233

mm) EPR spectrum reveals a two-component EPR line with a

total width of 0.05 mT. CW-EPR spectra were recorded with a
minimum field step of 1 mT, representing a field resolution of
0.2 ppm of the central field. Consistently, the pulsed EPR
signal of perylene in the time domain (FID) displays a frequency noise of about 30 kHz between consecutive shots,
yielding a field stability of ;0.2 ppm in agreement with the
CW experiment.
Microwave Bridge
The microwave bridge described previously (11, 12) has
been significantly modified in order to achieve higher microwave pulse power and four-phase switching capability. Since
we use a heterodyne detection scheme, the main microwave
sources consist of two InP Gunn diodes (Millimeter-Wave
Oscillator) operating at 139.5 (main source) and 139.95 GHz
(reference) locked to a 50 MHz source (FM noise of 2150
dBc/Hz at 1 kHz offset). The difference between the main and
reference frequencies represents our intermediate frequency of
450 MHz. Microwave power amplification is achieved with an
IMPATT diode network (Donetsk Physico Technical Institute).
A combination of fast amplitude and 090, 0180 phase
switches with three IMPATT diode oscillators leads to 37 mW
output power and microwave pulses with #10 ns rise and fall
times. However, the FM noise of IMPATT diodes exceeds that
of frequency locked Gunn diodes, because of the intrinsically
different methods of generating microwaves (avalanche emission versus oscillation of thermally excited electrons). In order
to minimize any increase in the noise figure of the spectrometer, the 139.5 GHz IMPATT oscillator network is injection
locked by our 139.5 GHz Gunn diode, which was measured to
be free (,50 dB) of adjacent harmonics of the primary 50 MHz
source. The reference arm, which drives the high-frequency
mixers, is unaffected by the IMPATT amplifier. Finally, the
microwaves are conducted to the resonator using a combination of oversized and fundamental waveguides (WR-42 and
WR-8), tapers, and bends. Insertion losses in the circulator and
propagation pathway (8 9 dB) attenuate the incident power at
the resonator to approximately 5 mW.
ENDOR Probe
Our current resonator is a TE 011 cylindrical cavity suitably
modified to permit RF irradiation during the ENDOR experiment. The cavity radius r and length d are chosen so that
2r/d 5 1, the condition for optimal Q in the TE 011 mode (20).
Tuning is achieved with a fixed and a moveable plunger, while
microwave coupling into the cavity is established through a
circular, centered iris and can be adjusted by rotating the wave
guide with respect to the axis of the resonator. The performance of the resonator depends critically on the iris thickness,
diameter, and length of the fixed plunger. Generally, the length
of the fixed plunger is optimized for each sample depending on

234

BENNATI ET AL.

its dielectric loss. However, the samples used most frequently,

such as frozen solutions and powders, display low dielectric
constants and are investigated in 0.5 mm OD capillaries combined with a single plunger of fixed length. In contrast, aqueous
solutions require shorter plungers and thinner capillary tubes
(0.17 mm OD). An empirically optimized iris diameter of 0.66
mm and thickness of 0.12 mm were found for WR-8 wave
guide of 1 3 2 mm cross section.
For the ENDOR probe we first followed the design described by other authors (10, 15), where the NMR coil is
mounted external to the cavity and RF field penetration into the
resonator must be optimized. Since the skin effect prevents RF
penetration into a solid metal cavity, it is necessary to cut slots
in the resonator walls finding a compromise between RF penetration and microwave losses. RF penetration can be further
increased if the formation of RF induced eddy currents in the
resonator walls and plungers are suppressed by cutting slots
through the entire length of the metal cylinder, a concept tested
and discussed previously for a 95 GHz resonator (21). For our
140 GHz resonator we employ 3 slots 0.5 mm in width and
0.38 mm apart along the active cavity volume and further slots
0.25 mm in width through the entire length of the metal
cylinder. The additional slots are narrower in order to maintain
mechanical stability of the soft silver cylinder. The resonator
wall thickness amounts to 1.13 mm. Plungers are fabricated
from insulating Kel-F with copper plated end faces. The resonator assembly is shown in Fig. 1. A new improved design of
the cylindrical TE 011 cavity, which permits optimal RF penetration without power losses, has been implemented recently
and will be presented in a separate paper.
In ENDOR experiments it is necessary to sweep the RF over
a large frequency range (typically several megahertz) while
concurrently maintaining a high quality factor in the RF circuit.
Initially, broadband RLC resonant circuits were tested in combination with our TE 011 resonator but were of low RF efficiency, particularly at high frequencies in the region of the 1H
resonance ( Hn 0 5 212 MHz at our g 5 2 resonance field). We
therefore adopted the circuit shown in Fig. 2a, which is typical
in solid-state NMR and consists of a high-quality LC parallel
circuit with variable matching and tuning capacitors. Highvoltage variable capacitors are mounted on top of the probe,
outside of the cryostat, to prevent arcing and temperaturedependent capacitance changes when cooling with helium gas.
A semirigid cable of approximately n l / 2, where n is an
integer and l is the RF wavelength, connects the capacitors to
the NMR coil, which is a split solenoid with 2 4 turns
mounted perpendicular to the resonator axis. To maintain tuning and matching during the RF sweep, we utilize a computercontrolled high-precision micrometer and a digital encoder
(Oriel Corporation) that drives the tuning and matching capacitors. For a high-Q circuit, the ratio of the capacitors at resonance is given by a constant (22),

FIG. 1. Top: Side view of the 140 GHz cylindrical resonator used for
ENDOR displaying the different slots for improved RF penetration, the fixed
and movable plungers, and the sample tube. Bottom: Tuning curve of the
ENDOR resonator loaded with a 0.4 mm (ID) capillary and solid BDPA/
polystyrene sample.

C2
5
C1

r
,
RT

[1]

where r is the resistance of the coil and R T is the impedance of

the transmission line. Therefore, a single micrometer can control both capacitors simultaneously through a gear system with
the appropriate gear ratio. With this arrangement we typically
cover a sweep range of about 20 MHz around the 1H free
Larmor frequency. A calibration curve allows the setting of the
micrometer position for tuning and matching during a frequency scan and typical RF steps of 50 kHz are achieved in ;3
s. For low gamma nuclei, such as 13C, 15N, 14N, and 2H,
broadband circuits with characteristics of low-pass filters also
permit highly efficient RF irradiation, as described by Hofer
(23) and Disselhorst et al. (15) and are preferentially used
because of the ease of construction.

235

PULSED ENDOR AT 140 GHz

GHz) and steps of 62 MHz. The reflected power was monitored

with a crystal detector. A typical measurement is displayed in
Fig. 1 (bottom) for the loaded ENDOR cavity, containing a
BDPA/polystyrene powder sample loaded in a 0.5 mm (OD)
quartz tube. We find a Q . 2200 and 1400 for unloaded and
loaded cavities, respectively. Quality factors larger than ;2200
could not be measured because of the limited step size of the
sweeper.
In addition, the 90 microwave pulse length (t p/ 2 ) was
estimated in two pulse spin-echo experiments. We used dilute
samples of a symmetric trityl radical (NYCOMED Innovation
AB), which displays very long relaxation times at low temperature (T 1e ; 1 s at T 5 15 K; the spin-echo intensity is
constant for pulse spacings # 800 ns). The 90 pulse length
was obtained by varying the incident microwave power for
fixed pulse lengths as well as by varying the pulse lengths for
a fixed maximum power. Accordingly, the cylindrical resonator without slots yielded 90 pulses of ;60 and 70 ns at low
and room temperature, respectively. With the ENDOR resonator we obtained values in the range 90 110 ns.
A knowledge of the loaded Q and incident power into the
resonator (5 mW, see above), and assuming a typical conversion factor c for cylindrical cavities of 30 mT/=W (experimental values are between 20 and 50 mT/=W (24)), permits an
estimate of the theoretical field strength B 1 for the ENDOR
cavity using
B 1 5 c Q z P.

FIG. 2. (a) Schematic representation of the NMR circuit for the ENDOR
probe. C 1 and C 2 are variable capacitors controlled by a high-precision micrometer (M). The micrometer position is set by the computer through a digital
controller. (b) Diagram of the NMR transmission line for variable frequency
ranges and four-phase capability.

The frequency variation in the RF is generated with a PTS

synthesizer. Pulses and four orthogonal phases are formed at an
intermediate frequency of 50 MHz and are controlled by a
pulse programmer (Interface Technology). RF amplification is
performed over 10 300 MHz with an American Microwave
Technology amplifier (P max ; 500 W in pulsed operation). The
transmission line is represented schematically in Fig. 2b.

[2]

We arrive at B 1 5 0.08 mT, in agreement with the experimental 90 pulse length of 100 ns.
The efficiency of the RF irradiation was first tested by
measuring the Rabi oscillation of the nuclear sublevels in an
ENDOR experiment. As a standard sample we use the solid
organic radical bis-diphenylene-phenyl-allyl (BDPA) doped in
polystyrene (1%) at room temperature. The Rabi oscillation is
usually monitored on the maximum of the ENDOR line at 209
MHz using a Mims-ENDOR type of EPR preparation detection scheme and recording the electron spin echo intensity as a
function of the RF pulse length. A typical curve is displayed in
Fig. 3 for an ENDOR probe with a four-turn RF coil and an RF
power of 450 W. The RF field strength B 2 can be estimated
considering the expression for the nutation of a spin-21 nucleus
under the effect of a magnetic field B 2 5 v / g perpendicular to
the external field B 0 (23):

SPECTROMETER PERFORMANCE

ENDOR Probe
The performance of the ENDOR probe was assessed in
different ways. Initially, quality factors of the cylindrical cavity
were measured using a microwave frequency sweeper (Insight
Products Corp.) with a calibrated frequency output (20 180

^I z & 5

D v 2 1 v 2 z cos~ v efft RF! z exp(2t RF /T r)

,
v 2eff

[3]

where v eff 5 =Dv 2 1 v 2. Here Dv is the resonance offset

with respect to the irradiation frequency and t RF is the RF pulse
length. RF-field inhomogeneities and spectral diffusion lead to

236

BENNATI ET AL.

achieved in the EPR experiment, which is reflected in the

ENDOR spectrum. This is particularly pertinent when small
amounts of sample are available, as is often the case with
biological compounds. Therefore, we provide a brief discussion of the sensitivity of our spectrometer. A detailed sensitivity analysis of the previous 140 GHz spectrometer was reported
by Prisner et al. (11). We measured the sensitivity of the CW
experiment using standard water/glycerol solutions of the trityl
radical at room temperature contained in 0.1 mm ID quartz
tubes in order to minimize dielectric losses. The experimental
sensitivity, E [spins/Gauss], can be obtained from the formula
(20)
FIG. 3. Rabi oscillation of the protons in BDPA for RF excitation at n 5
209 MHz, T 5 293 K. Mims ENDOR, MW p/2 pulse: 110 ns, pulse interval
t1 5 200 ns, RF power 450 W.

a damped oscillation and are included in the time constant T r.

In our case the pulse excitation width is much smaller than the
ENDOR line; therefore, all ^I z & i contributions at a frequency
offset D v i covered by the excitation width have to be integrated to obtain a total ^I z &. Different nutation frequencies of
single ^I z & i lead to a higher observed nutation frequency compared to the line at Dv 5 0. The experimental curve displayed
in Fig. 3 shows a more complicated behavior than represented
by integrating Eq. [3] over the frequency offset D v i , likely due
to fast electron and nuclear relaxation effects at room temperature. However, the position of the first maximum is well
resolved at 8.6 ms. Taking into account the effect of offresonance contributions, this value should lead to a 180 pulse
length on the order of 10 15 ms. In order to determine the 180
pulse length more precisely, we used the ENDOR probe together with a 200 MHz solid-state NMR magnet and spectrometer (12). A liquid H 2O sample was inserted into the resonator
and the 180 pulse length of the protons was observed directly.
The on-resonance, narrow proton line could be inverted using
a 24 ms 180 pulse with 140 W RF power delivered to the
probe. This corresponds to about 13 ms 180 pulses with the
450 W power available in the ENDOR experiment, in good
agreement with the estimate from the ENDOR Rabi oscillation.
For standard Davies- and Mims-ENDOR experiments,
which suffer from power broadening, pulse lengths of this
magnitude are a good compromise between sensitivity requirements and linewidth resolution. For coherence transfer ENDOR experiments, much shorter pulse lengths and correspondingly larger excitation bandwidths are desirable, which will
require improvement of the resonator design.
EPR Sensitivity
One advantage of performing double resonance experiments
at high microwave frequencies is the high absolute sensitivity

E5

N
,
~S/N! z n z DH 2pp

[4]

where N is the total number of spins, S/N is the peak-to-peak

signal-to-noise, n is the number of lines, and DH pp is the peak-topeak width of the derivative EPR spectrum. We arrive at E 5
1.4 3 10 8 spins/Gauss, a result that compares favorably with
previous experimental values of 10 9 spins/Gauss (11, 12). The
improvement in sensitivity is primarily due to improved cavity
construction and to the incorporation of a new low-noise 50 MHz
source to which the microwave sources are locked.
The sensitivity of the pulsed EPR spectrometer was determined in a two-pulse spin-echo experiment, using a trityl
radical solution with an inhomogeneous linewidth of ;1.5 mT.
We follow the definition for E proposed by Prisner (25),

E5

N
e 22 t /T 2,
~S/N!~2m I 1 1!DH 1/ 2

[5]

where S/N is the signal-to-noise (in a single shot) of the

peak-echo amplitude compared to the rms noise with a detection bandwidth matched to the echo shape. We arrive at E 5
3 3 10 9 spins/Gauss at T 5 20 K. This sensitivity limit
permits pulse EPR/ENDOR studies of samples in the 10 23
10 24 M concentration range, which is typical for biological
proteins and opens a wide range of applications for pulsed EPR
techniques in the 2 mm regime.
As expected, the absolute sensitivity of the pulsed mode is
lower than CW mode operation because of the narrower detection
bandwidth of the latter experiment. Nevertheless, the measured
sensitivities cannot be compared directly because of the different
time constants associated with the two experimental modes.
Moreover, we found that the CW experiment is sensitive to
baseline distortions due to microphonics arising from the B 0
modulation coils. The modulation coils are located in the cryostat
inserted in the narrow bore of the superconducting magnet and are
difficult to mechanically decouple from the probe. This can lead to
a large baseline offset at the lock-in detection frequency.

PULSED ENDOR AT 140 GHz

ENDOR Performance
Two standard pulsed ENDOR experiments, Davies- and
Mims-ENDOR, have been employed in order to demonstrate
the performance of the spectrometer. As typical experimental
parameters we use p/2 microwave pulses of 100 ns and RF p
pulses of 10 ms. Baseline correction is performed by subtracting two consecutive sequences and toggling the RF pulse.
Spectra from 1% BDPA in polystyrene at room temperature
were obtained from a single scan of the RF frequency with an
integration averaging of 100 shots/point.

237

BDPA is a stable radical with an alternating spin density

localized on the diphenylene rings (Fig. 4, top). Two hyperfine
couplings result from two sets of eight magnetically equivalent
ring protons, yielding a pair of doublets in the ENDOR spectrum. The EPR line is nearly Gaussian with a width of ;1 mT,
which is determined by the hyperfine splitting of the 16 protons
and a small g-anisotropy visible upon deuteration. Although
the ENDOR lines are broad because of hyperfine anisotropy
and power broadening, two splittings are easily distinguished,
and therefore the spectra are useful for comparing lineshape

FIG. 4. 1H ENDOR spectra of BDPA in polystyrene (1%) at room temperature, n 0 5 211.7 MHz. Mims-ENDOR (top), Davies-ENDOR (middle),
Davies-ENDOR simulation (bottom). Experimental parameters: MW p pulse: 200 ns, t (Mims) 5 200 ns, t (Davies) 5 400 ns, RF p pulse: 10 ms, RF power:
400 W, 100 averages/point, single scan spectra. Simulation parameters used are: g x 5 2.00263, g y 5 2.00260, g z 5 2.00257. The following hyperfine
constants are in megahertz and represent the absolute values: 1 A x 5 1.0, 1 A y 5 1.0, 1 A z 5 1.26, 2 A x 5 7.7, 2 A y 5 5.3, 2 A z 5 2.0. Inset: Structure of BDPA
radical.

238

BENNATI ET AL.

effects arising from preparation and detection pulses. Roomtemperature spectra are displayed in Fig. 4.
Mims-ENDOR, which is a stimulated echo sequence with a RF
pulse at a nuclear transition inserted between the second and third
mw-pulses (p/2tp/2RFp/2), is known to preferentially detect small hyperfine couplings. This is demonstrated by the experimental comparison of Mims- with Davies-ENDOR spectra in
Figs. 4a and 4b. Additionally, a strong 1H matrix peak is observed
at the Larmor frequency. However, the Mims-ENDOR intensity
suffers from a periodic dependence on the pulse spacing t, given
by I (ENDOR) ; 1 2 cos(2pAt), where A is the hyperfine constant
and t is the time between the two p/2 excitation pulses. The
spectrum represented in Fig. 4a was recorded with t 5 200 ns,
which leads to clearly visible intensity minima at Dn > 62.5
MHz. Davies-ENDOR (pRFp/2tp) (2) is the more attractive technique because of limited lineshape distortion, essential for
accurate simulations. Spectra are affected by an intensity hole
centered at Dn 5 0, which reflects the excitation band width of the
preparation pulse. Note that the doublet at 60.7 MHz is much
weaker than the second one at 62.5 MHz, although they are
determined by the same number of protons. We find that the
intensity of the center increases for longer and weaker preparation
pulses, but the sensitivity also degrades. The overall signal-tonoise is comparable to that of Mims-ENDOR, and therefore
Davies-ENDOR was typically used. A more detailed discussion
of the lineshape is given in the next section.

For calculating Davies ENDOR spectra, we utilize analytic

expressions for the ENDOR lineshapes as a function of the preparation and detection pulse as outlined by Grupp and Mehring (4).
The computation involves simulation of the EPR spectrum in
order to obtain the g-values and an initial set of hyperfine couplings. During computation of the EPR spectrum the orientations
that match the pulse excitation profile at a given external field are
extracted and the corresponding ENDOR frequencies are computed. The resulting ENDOR spectrum is a sum of the orientationselected hyperfine spectra convoluted with preparation and detection functions. EPR and ENDOR frequencies are obtained by
diagonalizing the Hamiltonian for interacting electron and nuclear
spins and are given by the expressions

n iI~6! 5 A i ~6!

O m ~ A ~1! 2 A ~2!!
n

i
I

for EPR

[6]

i51

for ENDOR

A~6! 5 60.5A zz 2 n oI,

I ENDOR~D v I! 5 I N(DvI)

[7]
[8]

where g and A are the angular-dependent g-factor and hyperfine constant, respectively, the summation runs over the i

1`

dD v Sg~D v S!

2`

3 f p(DvS, DvI)f d(DvS),

[9]

where the integral runs over all EPR offset frequencies Dv S, I N

is the normalized hyperfine spectrum given by the foregoing
transition frequencies, f p and f d are preparation and detection
functions as reported in (4), and g(D v S) is the EPR lineshape
function. In the continuum approximation used here, the excitation bandwidth is assumed to be much smaller than the EPR
linewidth, and g(D v S) can therefore be neglected because it is
nearly constant over the excitation profile. For a 180 preparation pulse and a spin echo detection sequence, the previous
equation can be approximated by (23)
I ENDOR 5

SIMULATION OF DAVIES ENDOR SPECTRA

g b eB 0
nS 5
2
h

nuclear spins, and the symbol 6 refers to the sign of electron

spin manifold m s. The angular dependence of A is numerically
computed with two consecutive Euler rotations, first into the
g-tensor principal axis frame (R a,b,g ), then into the laboratory
frame (R u,w ).
Effects produced by the preparation and detection pulses in
Davies-ENDOR can be derived from Eq. [3]. The complete
ENDOR response is written as

v *21s
,
Dv 2I 1 v *21s

[10]

with v*1s ' 0.7 v 1s, where the phenomenological scaling factor
of 0.7 includes relaxation by spin diffusion effects. Finally, RF
power broadening leads to ENDOR linewidths on the order of
1/t RF, where t RF is the RF pulse length. This can be derived
directly from Eq. [3]. For a 10 ms pulse the ENDOR linewidth
is therefore ;0.1 MHz. For faster computation, the ENDOR
lineshape is approximated by a Lorentzian line and the lineshape convolution is performed in the time domain.
An example of the calculated Davies-ENDOR spectrum of
BDPA for a 200 ns 180 preparation pulse is displayed in Fig.
4 together with the experimental spectrum. The calculation is
performed for 16 nuclear spins coupled to the electron spin.
Although there are only two inequivalent sets of protons, the
full spin space must be considered for computation of the EPR
line and for proper orientation selection. The weak intensity of
the doublet at 60.7 MHz compared to the doublet at 62.5
MHz is a direct consequence of the preparation pulse. The
matrix peak is simulated by a simple Gaussian line and is
weighted in order to obtain the observed intensity. The experimental spectrum was recorded in the center of the EPR line,
which results in excitation of a wide range of molecular orientations. This is reflected in the broad, almost featureless
powder pattern for the larger hyperfine coupling. The calcu-

PULSED ENDOR AT 140 GHz

FIG. 5. Top: Chemical structure of the tyrosyl radical. The directions of

the g-tensor principal axes and the labeling of the protons are shown as
discussed in the text. Bottom: 140 GHz stimulated echo-detected EPR spectrum of 1 mM tyrosyl radical from E. coli RNR, T 5 15 K. Experimental
parameters: MW p/2-pulse: 110 ns, pulse interval t 5 300 ns, 300 averages/
point, single scan spectrum.

239

GHz orientation-selective ENDOR with spectra of the tyrosyl

radical from E. coli RNR.
Figure 5 shows a schematic representation of the tyrosyl
radical together with the 140 GHz echo-detected EPR spectrum
obtained at 10 K. The sample consists of 0.25 ml of an aqueous
protein solution containing 1 mM tyrosyl radical. Note the
excellent signal-to-noise obtained in a single-scan spin-echo
experiment (experimental parameters are given in the figure
caption), which allows the performance of pulsed ENDOR
even on the low-intensity edges (B 0 5 4.96 and 4.979 T) of
the EPR line shape. The shape of the EPR spectrum is mainly
determined by g-anisotropy, with values of g x 5 2.00912, g y
5 2.00457, and g z 5 2.00225 (26). The orientation of the
g-tensor principal axis system was proposed by Gerfen et al.
(26), with the principal axis belonging to g x lying parallel to
the C 4O bond, and g z being normal to the ring plane (see Fig.
5). Six protons are coupled to the unpaired electron spin, two
magnetically inequivalent b-methylene protons and two sets of
equivalent ring protons (2,6 and 3,5), resulting in four different
hyperfine couplings. Figure 5 displays the labeling of the
protons as described in the text. ENDOR studies at 9 GHz
(27, 28) revealed that the b 1-methylene proton and the 3,5-ring
protons are strongly coupled to the electron spin, resulting in
very large isotropic hyperfine splittings ( A iso . 18 MHz),
which are directly observable in the EPR spectra. In contrast,
analysis of the weakly coupled protons is much more complex,
since these are detectable only in ENDOR spectra and the
resonances strongly overlap the matrix peak. However, at 140
GHz the congested small coupling region can be resolved with

lated spectrum shows slightly sharper features, a difference

that could arise from power broadening effects and finite
angular resolution in the powder pattern. Neither the experimental or simulated ENDOR spectra change considerably at
other EPR excitation fields, consistent with a very small ganisotropy.
ORIENTATION-SELECTIVE ENDOR OF THE TYROSYL
RADICAL FROM E. coli

A major application of high-field ENDOR consists of performing hyperfine measurements at different field positions in
the EPR line corresponding to excitation of highly selected
molecular orientations. The g-anisotropy dispersion scales
with the external magnetic field, and therefore the 140 GHz
ENDOR spectra reveal field-dependent powder patterns that
provide a large number of constraints for the determination of
the hyperfine couplings. We demonstrate the advantages of 140

FIG. 6. Orientation-selective Davies 1H-ENDOR of the tyrosyl radical of

RNR from E. coli. Spectra are displayed starting from the low-field edge ( g x )
to the high-field edge ( g z ). Experimental parameters: T 5 15 K, MW p pulse:
240 ns, RF p pulse: 10 ms, RF power: 400 W, 100 averages/point, single scan
spectra.

240

BENNATI ET AL.

the center, which is due to pure dipolar interactions with the

protein surroundings. Experiments with a D 2O exchanged matrix lead to analogous spectra with slightly narrower lines,
indicating that all relevant features are related to the 2,6-ring
and b 2-methylene tyrosyl protons and no additional hydrogen
bonded protons are involved. This is consistent with recently
reported 2H-ENDOR Q-band data of van Dam et al. (29).
When hyperfine spectra of different protons overlap, rough
assignment of hyperfine couplings can be done if the hyperfine
tensor principal axes are approximately collinear with the
g-tensor axes. At magnetic fields close to g z and g x this leads
to single-crystal-like spectra; however, at intermediate fields
the powder pattern can still be complicated. We expect this to
be the case for both 2,6- and b 2-methylene protons of the
tyrosyl radical, where the y-axis of the hyperfine tensor points
along the CH bond and the z-axis is approximately normal to
the ring plane. Sharp features are recorded at magnetic fields
close to g z . The 2,6-proton peaks should appear twice as
intense as those from the b 2-methylene, and therefore the
FIG. 7. Simulated ENDOR spectra of the tyrosyl radical from the lowfield edge ( g x ) to the high-field edge ( g z ). Simulation parameters for the
hyperfine couplings are given in Table 1.

orientation-selective measurements and unambiguous assignment of the hyperfine couplings is performed.

Davies ENDOR spectra were recorded at ;15 K for different excitation positions in the EPR line, with approximately 2
mT steps from g x to g y (low-field edge to center of the
spectrum) and 1 mT steps from g y to g z (center to high-field
edge). In Fig. 6 we show the magnetic field dependence of the
ENDOR spectra in the small coupling region, i.e., for A , u5u
MHz. The shape of the powder pattern changes dramatically
with the magnetic field, suggesting strongly anisotropic hyperfine couplings. A very intense matrix peak is also observed in
TABLE 1
Principal Hyperfine Values and Orientations a
Used in Simulations of the ENDOR Spectra

Protons

Ax
(MHz)

Ay
(MHz)

Az
(MHz)

f
(deg)

2 or 6
2 or 6
b2
b2
3 or 5
b1

15.0
14.9
12.1
22.1
226.4
58.0

17.6
17.7
25.0
6.2
28.1
55.4

12.1
11.9
24.0
21.0
219.3
52.4

44
25
226
230

c
(deg)

u10u
28

Ref.
28
This work
28
This work
This work b and 26
This work b and 26

The angle f describes a rotation in the xy-plane of the proton bond

direction toward the direction of the g y tensor principal axis. The angle c
describes a rotation in the yz-plane of the proton bond direction toward the ring
plane normal, i.e., the direction of the g z -tensor principal axis.
b
Hyperfine values obtained from simulation of the 140 GHz echo detected
spectrum, in agreement with ref. (26).
a

FIG. 8. Comparison of experimental and simulated ENDOR spectra of the

tyrosyl radical at the low-field edge ( g x ) and at the high field edge ( g z ). The
assignment of peaks to the b 2-methylene and to the 2,6 ring protons is
indicated.

PULSED ENDOR AT 140 GHz

strong doublet at 61.0 MHz arises from the 2,6 protons as

suggested previously (28). In addition, our spectra show wellresolved shoulders and a narrow doublet at 60.5 MHz. A
correct analysis of these features requires simulations of the
ENDOR lineshape. At magnetic fields close to the g x -edge we
observe strong peaks at around 63 MHz arising from the
2,6-protons. The line shape of the peaks deviates from a
single-crystal-like pattern. In fact, the y-axis of the 2,6proton hyperfine tensor is rotated by about 30 in the xy-plane
and mixing of the A x and A y tensor components is expected.
We anticipate that the ENDOR spectra in the field range
between g x and g y are primarily determined by the 2,6 protons
because the couplings of the b 2-methylene protons are much
smaller and overlap with the matrix peak. Therefore, the high
spectral resolution in this field region allows the angular dependence of only one hyperfine tensor to be followed with
accuracy.
Figure 7 shows the field dependence of the calculated ENDOR spectra of the RNR tyrosyl radical at different field
positions in the EPR line. The calculation is performed with all
six protons, but only the small coupling region is displayed.
For a better representation, the matrix peak was simulated by a
simple Gaussian line, and its width and intensity were varied at
different field positions. As a starting parameter set we used the

241

hyperfine tensor values reported by Hoganson et al. at 9 GHz

(28), which are listed for reference in Table 1. The hyperfine
parameters were iteratively varied in order to fit our experimental data. The values proposed by Hoganson et al. provide
an excellent fit of our data in the magnetic field region between
g x and g y . However, discrepancies were found for the region
between g y and g z , which necessitated reassignment of the
hyperfine couplings of the b 2-methylene proton. Our optimized
parameter set is assembled in Table 1.
First, we note that a salient feature of the experimental
spectra is the presence of two doublets and shoulders at magnetic fields close to g z . Previous studies with 9 GHz ENDOR
(27, 28) were not able to satisfactorily resolve the doublet at
60.5 MHz, and therefore the shoulders at 62.1 MHz were
assigned to the A z couplings of the b 2-methylene proton.
However, our simulations indicate that this assignment is not
compatible with our experimental spectra and their magnetic
field dependence. Instead, we assign the narrow doublet at
60.5 MHz to the b 2-methylene proton, and we show that the
shoulders at 62.1 MHz arise from the 2,6 protons, if mixing of
A z and A y tensor components is assumed due to small torsional
angles of the CH bond out of the ring plane (;10). An
experimental spectrum and simulation plotted together in Fig.
8 (top) are indicative of the sensitivity of the 140 GHz ENDOR

FIG. 9. Simulated ENDOR spectra of the tyrosyl radical at the low magnetic field edge ( g x ) for different angles f of the 2,6-proton bond (which defines
the direction of the A y principal axis) to the direction of the g y principal axis. Simulation curves (dotted lines) are displayed for f 5 10, 20, 25, 30, 35,
40. The angle f is related to the Euler angle a by a 5 90 2 f.

242

BENNATI ET AL.

spectra to such small angle variations. Simulations also reveal

that the A y hyperfine coupling of the b 2-methylene proton is
larger than previously reported. Specifically, the A y coupling
gives rise to edges at 63.1 MHz, which are weakly visible only
at a few magnetic field orientations (B 0 5 4971049730 G)
and cannot be easily recognized because of the complicated
structure of the spectra. We note that our new hyperfine tensor
values for the b 2-methylene proton, as reported in Table 1, lead
to a pure dipolar tensor of 23.13 (X), 5.17 (Y), 22.03 (Z)
MHz by subtracting an isotropic hyperfine coupling of 1.03
MHz. The pure dipolar tensor is in very good agreement with
the theoretical values proposed by Bender et al. (22.9 (X),
4.9 (Y), 21.9 (Z) MHz, (27)) and by Hoganson et al. (22.73
(X), 4.78 (Y), 22.05 (Z) MHz, (28)), which supports our
new assignments.
The ENDOR spectra at intermediate fields are rich in structure. For n $ u1.5u MHz the simulations reproduce all details
with high accuracy. However, for n # u1.5u MHz we are not
able to simulate details around the matrix peak. Interestingly,
both A y couplings of the 2,6 and b 2-methylene protons are
large (7.7 and 26.2 MHz, respectively) and do not influence
the n # u1.5u MHz region. The origin of the matrix peak shape
is still unclear; further investigation is needed to clarify this
point.
Finally, the low magnetic field region of the spectrum is
accurately reproduced by our simulations. We find that the
ENDOR lineshape is very sensitive to the rotation angle f of
the 2,6-protons in the xy-plane. An experimental spectrum and
a simulation for f 5 25 are displayed in Fig. 8 (bottom).
Further simulations for different angles f varying from 10 to
40 are plotted in Fig. 9. Since the chemical structure of this
radical is well known, it is useful to compare experiment and
simulations. The noise of the spectra limits the angular resolution to only 5 and the experimental spectrum is best fit with
f 5 2530, which is expected for an aromatic ring proton in
an sp 2 orbital.

ments. The magnetic field dependence of the lineshape is

complex, but the simulations are highly constrained and sensitive to the structural information reflected in the relative
orientation of the hyperfine tensor to the electron-Zeeman
g-tensor. From the analysis of experimental and simulated
spectra, we obtain a new assignment for the hyperfine couplings of the weakly coupled protons of the tyrosyl radical.

SUMMARY

11. T. F. Prisner, S. Un, and R. G. Griffin, Isr. J. Chem. 32, 357363

(1992).

We describe the design and characteristics of a new pulsed

ENDOR spectrometer operating at 140 GHz. The essential
features consist of a superconducting magnet system with a
60.4 T sweep range and 0.2 ppm field resolution, a microwave
bridge for pulsed EPR/ENDOR, and a double resonance probe.
The efficiency of the probe is reflected in a absolute pulsedEPR sensitivity of 3 3 10 9 spins/gauss, which should permit
routine application of pulsed ENDOR, especially to samples
available in limited quantities such as biologically relevant
proteins. We demonstrated the capability of the spectrometer
with orientation-selective ENDOR spectra of the tyrosyl radical in E. coli RNR. Magnetic field dependent ENDOR spectra
at 140 GHz were obtained with high signal-to-noise and revealed essential new features not discerned in X-band experi-

12. L. R. Becerra, G. J. Gerfen, B. F. Bellew, J. A. Bryant, D. A. Hall,

S. J. Inati, R. T. Weber, S. Un, T. F. Prisner, A. E. McDermott, K. W.
Fishbein, K. E. Kreischer, R. J. Temkin, D. J. Singel, and R. G.
Griffin, J. Magn. Reson. A 117, 8 40 (1995).

ACKNOWLEDGMENTS
Thanks are accorded to Professor D.J. Singel for conversations pointing out
the advantages of high frequency ENDOR. We are grateful to K. Kreischer for
the loan of the frequency sweeper used in the cavity quality measurements and
to Jan Henrik Ardenkjaer-Larsen (NYCOMED Innovation AB) for the gift of
the trityl radical. We would like to acknowledge P. Hofer for helpful discussions and comments on this manuscript. M. Bennati thanks the Deutsche
Forschungsgemeinschaft for a postdoctoral fellowship. C.T. Farrar was supported by a National Institutes of Health postdoctoral fellowship (GM-18790).
This research was supported by the National Institutes of Health (grants
GM-38352 and RR-00995).

REFERENCES
1. G. Feher, Phys. Rev. 103, 834 (1956).
2. E. R. Davies, Phys. Lett. 47A, 12 (1974).
3. W. B. Mims, Proc. Roy. Soc. A283, 452 457 (1965).
4. A. Grupp and M. Mehring, in Modern Pulsed and CW Electron
Spin Resonance (L. Kevan and M. Bowman, Eds.), p. 195, Wiley,
New York (1990).
5. C. Gemperle and A. Schweiger, Chem. Rev. 91, 1481 (1991).
6. H. Thomann and M. Bernardo, Isr. J. Chem. 32, 323337 (1992).
7. W. Bryan Lynch, K. A. Earle, and J. H. Freed, Rev. Sci. Instrum. 59,
13451351 (1988).
8. R. T. Weber, J. A. J. M. Disselhorst, L. J. Prevo, J. Schmidt, and
W. T. Wenckebach, J. Magn. Reson. 81, 129 144 (1989).
9. A. Y. Bresgunov, A. A. Dubinskii, V. N. Krymov, Y. G. Petrov, O. G.
Poluektov, and Y. S. Lebedev, Appl. Magn. Reson. 2, 715728
(1991).
10. O. Burghaus, M. Rohrer, T. Gotzinger, M. Plato, and K. Mobius,
Meas. Sci. Technol. 3, 765774 (1992).

13. L. C. Brunel, A. L. Barra, and G. Martinez, Physica B 204, 298 302

(1995).
14. C. Kutter, H. P. Moll, J. van Toll, H. Zuckermann, J. C. Maan, and
P. Wyder, Phys. Rev. Lett. 74, 29152928 (1995).
15. J. A. J. M. Disselhorst, H. van der Meer, O. G. Poluektov, and J.
Schmidt, J. Magn. Reson. 116, 183188 (1995).
16. M. Rohrer, M. Plato, F. MacMillan, Y. Grishin, W. Lubitz, and K.
Mobius, J. Magn. Reson. A 116, 59 66 (1995).
17. L. Paschedag, J. van Tol, and P. Wyder, Rev. Sci. Instrum. 66, 5098
(1995).
18. Y. S. Lebedev, in Modern Pulsed and Continuous-Wave Electron

PULSED ENDOR AT 140 GHz

19.
20.
21.
22.
23.
24.

Spin Resonance (L. Kevan and M. K. Bowman, Eds.), pp. 365

436, Wiley, New York (1990).
S. Un, J. Bryant, and R. G. Griffin, J. Magn. Reson. A. 101, 9294
(1993).
C. P. Poole, Electron Spin Resonance, Wiley, New York (1983).
M. Rohrer, Ph.D. dissertation, Department of Physics, FU Berlin
(1991).
D. D. Traficante and J. A. Simms, J. Magn. Reson. 15, 484 (1974).
P. Hofer, Ph.D. dissertation, Department of Physics, University of
Stuttgart (1988).
T. F. Prisner, M. Rohrer, and K. Mobius, App. Magn. Reson. 7, 167
(1995).

243

25. T. F. Prisner, Adv. Magn. Opt. Reson. 20, 245 (1997).

26. G. J. Gerfen, B. F. Bellew, S. Un, J. M. B. Jr., J. Stubbe, R. G.
Griffin, and D. J. Singel, J. Am. Chem. Soc. 115, 6420 (1993).
27. C. J. Bender, M. Sahlin, G. T. Babcock, B. A. Barry, T. K. Chandrashekar, S. P. Salowe, J. Stubbe, B. Linstrom, L. Petersson, A.
Ehrenberg, and B.-M. Sjoberg, J. Am. Chem. Soc. 111, 8076 8083
(1989).
28. C. W. Hoganson, M. Sahlin, B. M. Sjoeberg, and G. T. Babcock,
J. Am. Chem. Soc. 118, 4672 (1996).
29. P. J. van Dam, J. P. Willems, P. P. Schmidt, S. Poetsch, A. L. Barra,
W. R. Hagen, B. M. Hoffman, K. K. Andersson, and A. Graeslund,
J. Am. Chem. Soc. 120, 5080 5085 (1998).