International Journal of Mass Spectrometry 215 (2002) 59–75

Fourier transform ion cyclotron resonance detection:

principles and experimental configurations
Alan G. Marshall∗ , Christopher L. Hendrickson
Center for Interdisciplinary Magnetic Resonance, National High Magnetic Field Laboratory,
Florida State University, 1800 East Paul Dirac Drive, Tallahassee, FL 32310, USA

Received 8 August 2001; accepted 26 November 2001

Abstract
Fourier transform ion cyclotron resonance mass spectrometry is based on image current detection of coherently excited
ion cyclotron motion. The detected signal magnitude and peak shape may be understood from idealized behavior: single
ion, zero-pressure, spatially uniform magnetic field, three-dimensional axial quadrupolar electrostatic trapping potential, and
spatially uniform resonant alternating electric field. In practice, deviation from any of the above conditions will shift, distort,
split, and/or coalesce FT-ICR mass spectral peaks. Fortunately, such peak distortions may typically be avoided by appropriate
experimental design and/or greatly minimized by internal frequency-to-m/z calibration. Various aspects of modern FT-ICR
detection (hardware and software) are discussed. (Int J Mass Spectrom 215 (2002) 59–75) © 2002 Elsevier Science B.V. All
rights reserved.
Keywords: Fourier transform; Ion cyclotron resonance; ICR; FTMS

1. Introduction detected over a more limited m/z range. As explained

below, the basis for that performance is that ions of
Broadband (e.g., ions of 200 < m/z < 1000 a given m/z are detected according to their cyclotron
detected simultaneously) Fourier transform ion cy- frequency, and ions need to be confined for a suffi-
clotron resonance mass spectrometry (FT-ICR MS) ciently long period (∼1 s) to determine that frequency
offers ultrahigh mass resolving power (m/m50% > with high precision.
300, 000 at 9.4 T, in which m is ion mass and m50% An FT-ICR MS experiment consists of a series of
is the mass spectral peak full width at half-maximum temporally separated events, typically: ion formation
peak height), mass resolution (m2 − m1 > 0.003 Da), external to the detector, ion cooling/focusing/ accu-
and mass accuracy (ppm rms), for complex mixtures mulation, ion transmission to a Penning trap, mass-
containing compounds of up to several thousand dif- to-charge ratio (m/z)-selective ejection to select ions
ferent elemental compositions. That performance can spanning a desired (m/z) range, dipolar excitation,
improve by more than an order of magnitude for ions dipolar analog detection with simultaneous analog-
to-digital conversion and storage of the time-domain
∗ Corresponding author. Present address: Member, Department of
analog signal, apodization, fast Fourier transforma-
Chemistry, Florida State University, Tallahassee, FL 32306 USA.
E-mail: marshall@magnet.fsu.edu. tion, magnitude computation, and frequency-to-m/z

1387-3806/02/$20.00 © 2002 Elsevier Science B.V. All rights reserved.

PII S 1 3 8 7 - 3 8 0 6 ( 0 1 ) 0 0 5 8 8 - 7
60 A.G. Marshall, C.L. Hendrickson / International Journal of Mass Spectrometry 215 (2002) 59–75

conversion with respect to external or internal spectral

peaks. Recent reviews describe the general features
and variants of such experiments [1] as well as the
milestones in historical evolution of each stage [2].
Here, we focus on principles and methods for detec-
tion of ion cyclotron rotation. Because ion cyclotron
motion must be spatially coherent to be detectable,
it is necessary to consider ion excitation as well.
Because ions must be confined for extended detec- Fig. 1. Ion cyclotron motion. Ions rotate in a plane perpendicular
tion periods, it is necessary to discuss means for ion to the direction of a spatially uniform magnetic field, B. Note that
positive and negative ions orbit in opposite senses.
trapping. Because the geometric requirements for
excitation/detection and trapping are different, it is
important to understand the tradeoffs in attempts to the magnetic field direction as shown in Fig. 1. The
optimize both aspects in a single geometric config- “unperturbed” cyclotron (rotational) frequency, ωc
uration. Finally, we shall briefly discuss the effects (SI units), is expressed in Eq. (1) [1].
of trap configuration, excitation mode, magnetic field qB
strength, ion charge density, ion-neutral collisions, ωc = (1a)
m
and digital data reduction on the appearance (po-
sition, shape, multiplicity, coalescence) of FT-ICR ωc 1.535611 × 107 B
νc = = (1b)
mass spectral peaks. 2π m/z
This discussion will be limited to image-charge in which νc is in hertz, B in tesla; m in microgram; z
detection, which has supplanted earlier detection in multiples of elementary charge.
based on charge collection (“omegatron” [3–6]) and A notable feature of Eq. (1) is that all ions of a given
power-absorption (“marginal oscillator” [7,8]) de- mass-to-charge ratio, m/q, rotate at the same ICR fre-
signs. We shall not describe single-ion detection of quency, independent of velocity. That property makes
ion axial oscillation (see below) by a superconducting ICR especially amenable to mass spectrometry, be-
quantum interference device (SQUID). Such exper- cause ion frequency is relatively insensitive to kinetic
iments have been used for ultraprecise atomic mass energy, so translational energy “focusing” is not es-
measurements [9], and to determine the (lack of) mass sential for precise determination of m/z. Moreover, at a
difference between the proton and antiproton [10]. common static magnetic field value of 7.0 T (at which
However, because such experiments require liquid he- the corresponding proton NMR Larmor frequency is
lium temperature, and report only a single m/z value at 300 MHz), ICR frequencies for ions of interest range
a time, they are not applicable to analytical mass spec- from a few kHz to a few MHz, a particularly con-
trometry in which a wide range of m/z values must be venient range for commercially available broadband
covered quickly. Another non-FT detection method electronics.
is axial ejection of the ions with time-of-flight mass
analysis by single-ion counting [11]: that method is
low-resolution but high-sensitivity. 3. Ion cyclotron excitation and detection

Treatment of resonant ion cyclotron dipolar excita-

2. Ion cyclotron motion tion and detection begins from the idealized model of
Fig. 2. The electric potential, V(y), between two in-
An ion of mass, m, and charge, q, moving in a finitely extended parallel conductor plates varies lin-
spatially uniform magnetic field, B, rotates about early with y. Thus, the corresponding electric field,
 A.G. Marshall, C.L. Hendrickson / International Journal of Mass Spectrometry 215 (2002) 59–75 61

Fig. 2. Excitation voltage difference and detection induced charge difference in FT-ICR MS, demonstrated for two infinitely extended
parallel flat electrodes located at y = ±d/2 m away from the z-axis. If a potential, V0 , is applied to the upper electrode and −V0 to
the lower electrode, the potential, V(y), anywhere between the electrodes is 2V0 y/d. Conversely, if a point charge, q, is located between
the same electrodes (in the absence of any applied voltage), then the difference, Q(y), between the charge induced on the upper and
lower electrodes is –2 q y/d. It turns out that the relationship, Q/q = –V /V0 , is true (by the principle of “reciprocity”, also known as
Earnshaw’s theorem [112]) for opposed electrodes of arbitrary shape. (Reproduced, with permission, from [1].)

E(y) = −∂[V(y)]/∂y, is constant anywhere between and the packet accelerates along a spiral trajectory, as
the two plates. Conversely, the difference in charge, illustrated in Fig. 3 (left).
Q(y), induced by an ion of charge, q, at the same Following resonant irradiation of duration, Texcite ,
y-position, also varies linearly with y. the post-excitation ion packet cyclotron radius, r (SI
As for nuclear magnetic precession [12], ion cy- units), increases to
clotron rotation is initially “incoherent,” and does not
E0 Texcite
by itself generate an observable electrical signal (for r= (2)
2B
ICR, a net difference in the charge induced on two
opposed parallel electrodes). At their instant of forma- As described previously [1], a coherently orbiting
tion in (or injection into) the ion trap, ion cyclotron or- ion packet induces a differential current between two
bital phases are random—i.e., an ion may start its cy- opposed detection plates and may be modeled as a
clotron motion at any point around either circle shown current source. The current amplitude is proportional
in Fig. 1. Thus, for an ensemble of ions, any charge to the number of spatially coherent orbiting ions. The
induced on one detector plate will be balanced by an receiver plates and wiring that connect the ion trap to
equal charge induced on the other detector plate (i.e., the detection preamplifier have an inherent resistance
by an ion with opposite phase), so that the net dif- and capacitance in parallel (Fig. 3) [14,15]. At typical
ference in detected charge between the two plates is ICR frequencies (>10 kHz) the signal-to-noise ratio
zero. Moreover, the cyclotron radius of thermal ions is essentially independent of cyclotron frequency.
is too small to induce a detectable signal (i.e., propor- However, at sufficiently low-frequency (<10 kHz),
tional to y/d in Fig. 2), even if all ions had the same the signal is expected to vary directly with frequency
cyclotron phase. [14,16]. Therefore, throughout most of the frequency
FT-ICR detection is thus typically preceded by exci- range excited by a broadband waveform in a standard
tation produced by applying a spatially uniform elec- FT-ICR MS experiment, the detected signal-to-noise
tric field of amplitude, E0 , directed perpendicular to ratio reflects the relative current differential induced
the magnetic field direction, and rotating at the cy- on the detection plates. Furthermore, the detection
clotron frequency of ions of a particular m/z value limit (namely, the minimum number, N, of ions that
(i.e., “resonant”). During such excitation, the dimen- may be detected from an undamped signal in a single
sions of the initial ion packet remain the same [13], 1 s acquisition period to yield a S/N ratio of 3:1) may
62 A.G. Marshall, C.L. Hendrickson / International Journal of Mass Spectrometry 215 (2002) 59–75

Fig. 3. Incoherent ion cyclotron orbital motion (left) is converted to coherent (and therefore detectable) motion (right) by application of
an electric field that rotates in the same sense and at the ICR frequency of the ions of a given m/z value.

be calculated from [15] amplitude × duration product). Second, a Fourier

transform of the time-domain ICR response gives
CVd(p–p)
N= (3) the same “absorption” spectrum that is otherwise ob-
qA1 (r)
tained by measuring power absorption while sweep-
in which C is the capacitance of the detection circuit, ing slowly across the m/z range [19]. In addition, the
Vd(p–p) is the peak-to-peak amplitude of the detected “superposition” principle implies that the signals from
voltage (calibrated for a given spectrometer), and A1 (r) any number of ions of arbitrary m/z values simply
is a trap configuration-dependent coefficient that is ap- add at the detector; thus, ions of a wide m/z range can
proximately proportional to r and may be determined be detected simultaneously. The two prior statements
graphically [17]. For example, for some typical operat- combine to constitute the “multichannel” advantage
ing parameters, namely, a detection circuit capacitance of pulsed excitation followed by Fourier transforma-
of 50 pF, Vd(p–p) of 3 × 10−7 V, and A1 (r) = 0.5 (i.e., tion to yield a spectrum of N data points in 1/N the
the ion is excited to approximately half of its maximal time it would take to scan the spectrum one channel
cyclotron radius), an observed signal-to-noise ratio of at a time [19]. Finally, the detected signal increases
3:1 corresponds to a detection limit of ∼187 ions. linearly with ion charge, so that ICR is more sensitive
The ICR signal is proportional to the induced cur- for multiply-charged ions. For example, individual
rent, dQ/dt = −2q(dy/dt)/d [14,15,18] (note that DNA ions of ∼108 Da (each with ∼30,000 charges)
ICR signal is independent of magnetic field strength). have been detected by FT-ICR MS [20].
Further, the induced current increases linearly with Although image current detection at room tempera-
ion cyclotron radius (because the ion y-velocity ture is typically less sensitive than ion counting tech-
component, dy/dt, increases linearly with radius), niques characteristic of ion beam instruments, FT-ICR
so the ICR signal increases linearly with ion cy- detection is non-destructive. Summing N repeated data
clotron post-excitation radius. Linearity is important acquisitions from the same sample of ions can thus
for several reasons. First, the ICR response at any potentially improve the signal-to-noise ratio of the ac-
frequency is proportional to the excitation spectral cumulated data by a factor of N1/2 [19]. Ion remeasure-
magnitude at that frequency, because the ICR signal ment was first achieved by allowing collisions to relax
varies linearly with ion cyclotron post-excitation ra- the ion cyclotron post-excitation orbital radius back
dius (which depends linearly on the excitation voltage to near-zero before the next excitation/detection cycle
 A.G. Marshall, C.L. Hendrickson / International Journal of Mass Spectrometry 215 (2002) 59–75 63

[21]. Much more efficient (and faster) remeasurement dependent of ion position within the trap [25]. Second,
was later performed by use of quadrupolar axializa- ion axialization [26,27] is optimally generated from
tion to collapse the post-excitation ion cyclotron or- the two-dimensional quadrupolar potential of Fig. 4,
bital radius to near-zero between acquisitions [22]. middle. Third, azimuthal (i.e., in a direction perpen-
dicular to the magnetic field) dipolar excitation is op-
timally produced by applying an alternating voltage
4. Penning trap difference between two infinitely extended parallel flat
conductors (Fig. 4, right), as noted in Fig. 2.
It is easy to show that mass resolving power, Any of the three potentials of Fig. 4 can in principle
m/m50% , is equal to frequency resolving power, be generated from a set of conductive surfaces that
ω/ω50% , in ICR mass spectrometry [5,23]. Stated follow the isopotential surfaces for a given symmetry.
another way, m/m50% can be thought of as the However, (a) it is clearly possible to achieve only one
number of cyclotron orbits an ion makes during the symmetry from one set of conductive surfaces, and (b)
data acquisition period [24]. Thus, it is desirable to each idealized isopotential surface extends to infinity
confine ions for as long as possible after excitation in whereas the dimensions of any actual trap must be fi-
order to maximize mass resolving power. Although, nite. Fortunately, any of several simple trap geometries
cyclotron rotation inherently confines ions radially, it (e.g., cubic or cylindrical, Fig. 5a and b) can approach
is necessary to apply an axial electric field gradient in all three of the desired potentials near the center of the
order to keep ions from escaping axially (i.e., parallel trap. Deviations from linearity or quadrupolarity re-
to the magnetic field direction). duce the magnitude of the desired ICR signal (by fac-
Fig. 4 shows isopotential surfaces for the three ideal- tors that depend (non-linearly) on ion position within
ized electric potentials employed in FT-ICR MS [18]. the trap and on the trap aspect (length-to-width) ratio
First, the three-dimensional axial quadrupolar electro- [1,28]). Non-linearities also generate signals at fre-
static “Penning trap” potential (Fig. 4, left) is optimal quencies obtained from sums and differences of the
for axial confinement of ions. The advantages of such cyclotron, trapping, and magnetron frequencies [29].
a trapping potential are: (a) the ion cyclotron motion A general approach to optimizing ICR trapping,
remains independent of the other ion natural motions excitation, and/or detection is to segment any or all of
(namely, axial “trapping” oscillation along the mag- the electrodes as shown in Fig. 5(c–e). For example,
netic field and rotational “magnetron” motion about the end-cap (“trapping”) electrodes may be cylindrical
the trap axis); and (b) the ion cyclotron frequency is in- in the “open” trap configuration of Fig. 5d, thereby

Fig. 4. Plot of isopotential surfaces for trapping (left), 2D quadrupolar (middle), and dipolar (right) potentials. The z-axis of the trap
coincides with the direction of the applied magnetic field. (Reproduced, with permission, from [1].)
64 A.G. Marshall, C.L. Hendrickson / International Journal of Mass Spectrometry 215 (2002) 59–75

Fig. 5. ICR ion trap configurations. E = excitation; D = detection; T = end cap (“Trapping”). (a) cubic [16,113]; (b) cylindrical
[45,114–116]; (c) end-caps segmented to linearize excitation potential (“infinity” trap) [117]; (d) open-ended, with capacitive rf coupling
between the three sections [118–120] and (e) “matrix-shimmed” [18].

allowing for easier access of ions or photons or elec- in which the value of ωc is given in Eq. (1a)
trons into the trap. Capacitive coupling between the

2qVtrap α
central and end-cap electrodes of the “open” cylindri- ωz = (4c)
ma2
cal trap then effectively shorts the central and end-cap
segments with respect to rf excitation, without affect- in which ωz is the axial “trapping” oscillation fre-
ing the static potential difference between the central quency (SI units), Vtrap is the potential applied to each
and end-cap segments. The capacitively-coupled open end-cap electrode, a is a characteristic trap dimen-
trap thus provides for near-perfect dipolar rf electric sion (e.g., the length of one side of a cubic trap), and
field for ion cyclotron excitation. Similar linearization α is a parameter determined by the trap geometri-
of the excitation field can be achieved by segmenta- cal configuration (e.g., 2.77373 for a cubic trap) [1].
tion of the end-cap electrodes of a closed cylindrical Note that
trap (Fig. 5c). Ultimately, if all six sides of a cubic
ω+ = ωc − ω− (5)
trap are segmented in grid fashion, and appropriate po-
tentials applied to each individual grid element (Fig. Thus, because ω− “magnetron” frequency is (ex-
5e), then all three desired potentials of Fig. 4 can be cept at very high m/z) independent of m/z, the shift in
approximated in a single geometric configuration. cyclotron frequency is also independent of m/z.
The three-dimensional axial quadrupolar electro- From Eq. (4), it is possible to derive [30]
static trapping potential lowers the ICR frequency of m A B
Eq. (1) to the “reduced” cyclotron frequency, ω+ , = + 2 (6)
z ω+ ω+
given by

  One might expect Eqs. (4) and (5) to be valid only in
ωc ωc 2 ω z 2
ω+ = + − (4a) the single-ion limit. Fortunately, it can be shown that
2 2 2 a large ensemble of ions of a given m/z settles into an

 
ωc ωc 2 ω z 2 overall shape that generates a potential (acting on ions
ω− = − − (4b) of other m/z values) of the same form (but opposite in
2 2 2
 A.G. Marshall, C.L. Hendrickson / International Journal of Mass Spectrometry 215 (2002) 59–75 65

sign) to that of the axially quadratic trapping potential image-charge is typically negligible (e.g., ∼10−5 Hz
[31]. Thus, the form of Eq. (6) is preserved even in the per ion, at an ICR orbital radius equal to half the trap
presence of “space charge”, and the constants, A and radius in a cubic trap [36]). The equations of ion mo-
B, may be obtained by fitting to Eq. (6) a set of ICR tion also become solvable in the other extreme limit
mass spectral peak frequencies corresponding to ions that the ion density is much higher (“plasma” con-
of at least two known m/z values. Eq. (6) and its vari- ditions [37,38]) than in typical FT-ICR MS experi-
ants [32] are the standard frequency-to-m/z conversion ments.
(“calibration”) expression for FT-ICR MS. Various authors have attempted to treat the prob-
“Internal” calibration (i.e., calibrant ions are present lem of two packets of ions of different m/z. Because
along with analyte ions in the same sample) typically the problem cannot be solved analytically, it is neces-
improves mass accuracy by a factor of 3–10 compared sary either to make various limiting assumptions about
to “external” calibration (i.e., calibration is performed the packet shape and dynamics (e.g., point charge,
on a separate sample different from the analyte). Ex- ellipsoid of revolution, cylinder, dish, etc.) or solve
ternal calibration works best when calibrant ions are Poisson’s equation numerically for a limited number
excited to the same cyclotron radius as analyte ions of ions [39]. To date, neither approach has been par-
and when the number of ions in the trap is the same ticularly helpful in quantitating FT-ICR mass spectral
for both experiments. peak shape or Coulomb-induced ICR frequency shifts.
However, some insight has been gained into the ba-
sis for experimentally observed peak coalescence [40]
5. Ion–ion interactions of ICR signals for ions of closely-spaced m/z values.
Qualitatively, the difference in cyclotron velocity for
The effects of ion–ion interactions in a Penning trap two ions of the same charge, q and same cyclotron
are understood qualitatively, but not quantitatively. Ex- radius, r, is
perimentally, as the number of trapped ions increases,  
FT-ICR mass spectral peaks are observed to shift in 1 1
v2 − v1 = (ω2 − ω1 )r = qBr − (7)
frequency, broaden, and split or coalesce. Moreover, m2 m1
for a high population of trapped ions, the instanta-
neous ICR frequency can drift with time after exci- Conversely, the velocity of a packet of ions mov-
tation [33,34]. If ions are in constant supply, the best ing in a plane perpendicular to an applied magnetic
approach is to reduce the number of ions until all of field, B, is E/B, in which E is the electric field pro-
the above effects become negligible. However, if the duced by a second packet of ions of different m/z.
amount of sample is limited in amount or duration Ion cyclotron resonances tend to coalesce when the
(as for on-line chromatography/mass spectrometry), it common velocity due to the “E × B” (i.e., ion–ion
may not be possible to perform several independent interaction) exceeds the difference in cyclotron veloc-
measurements for each chromatographically separated ities for the two ion packets. Of course, the electric
component, and one must deal with the above “space field, E, is proportional to ion charge, q. The ratio,
charge” effects. (E/B)/qBr(1/m2 − 1/m1 ) = Em1 m2 /qB2 (m1 − m2 ),
The equations of motion for two ions of the same may thus, be thought of as a “coalescence” index—the
m/z in a Penning trap can be solved analytically, bigger the ratio, the higher the coalescence tendency.
and neither ion shifts the cyclotron frequency of the This simplified argument explains the experimentally
other [35]. Another Coulomb-mediated effect is that observed increased tendency for peak coalescence at
a packet of ions of the same m/z induces an “image” higher ion mass, lower magnetic field strength, and
charge in the electrodes that surround the ions; closer ICR MS peak separation, as confirmed by more
however, the ICR frequency shift induced by that detailed theoretical analysis [41,42].
66 A.G. Marshall, C.L. Hendrickson / International Journal of Mass Spectrometry 215 (2002) 59–75

6. Heterodyne detection at sum and difference frequencies of the reference

and ICR signals (see Fig. 6). The mixing (multipli-
When the first FT-ICR MS experiments were per- cation) process in the time-domain corresponds to
formed [43,44], analog-to-digital converters (ADCs) a convolution in frequency-domain—a spectrum of
were not sufficiently fast, and buffer memory was not two ICR signals (S1 at a frequency below the ref-
sufficiently large to sample and store a time-domain erence frequency, and S2 at a frequency above the
ICR signal spanning a wide m/z range (and thus, a reference frequency) is thus convolved with the ref-
wide range of ICR frequencies). It was therefore, nec- erence spectrum to produce components at the sum
essary to employ “heterodyne” detection, in which the and difference frequencies [19]. The sum frequency
time-domain ICR signal is first multiplied (“mixed”) signals are then suppressed by a low-pass filter, and
with a single-frequency reference oscillator sinu- the low-frequency band of difference frequencies can
soidal time-domain signal, thereby, generating signals then be sampled and stored for subsequent discrete

Fig. 6. Step-by-step procedure for heterodyne single-phase detection, represented in the frequency-domain. (Reproduced, with permission,
from [49].)
 A.G. Marshall, C.L. Hendrickson / International Journal of Mass Spectrometry 215 (2002) 59–75 67

Fourier transformation [45]. Now that commercial (m/m50% >5,000,000) in a protein of ∼16 kDa in
ADCs are available at e.g., 20 Msample/s, with buffer mass [47].
memory to 32 Mword, most FT-ICR mass spectra are
now acquired in “direct” mode (i.e., direct sampling
of the time-domain ICR analog signal. 7. Quadrature heterodyne detection
Nevertheless, heterodyne-mode detection is still
preferable to direct-mode detection when the highest There are three problems with single-phase het-
mass resolution is needed over a restricted m/z range. erodyne detection. First, the inability to distinguish
For example, current world records for mass reso- between ICR signals whose frequencies lie above
lution are based on heterodyne detection: baseline or below the reference frequency results in spectral
resolution of two ∼904 Da peptides separated by less foldover (i.e., each signal on one side of the reference
than 0.0005 Da (i.e., less than the mass of a single frequency also appears as its mirror image on the other
electron) [46], and resolution of isotopic fine structure side of the reference frequency). Thus, only half of the

Fig. 7. Step-by-step procedure for quadrature heterodyne detection, represented in the frequency-domain. (Reproduced, with permission,
from [49].)
68 A.G. Marshall, C.L. Hendrickson / International Journal of Mass Spectrometry 215 (2002) 59–75

available bandwidth is used, because the other half for optimal removal of noise without distorting the
yields no new information. Second, it is not possi- ion relative abundances.
ble to distinguish between positive and negative ions,
because linearly polarized dipolar irradiation may be
analyzed as the sum of two counter-rotating com- 8. Distinguishing positive and negative ions
ponents of the same frequency. One component will
excite (or detect) positive ions and the other will ex- Ordinarily, ions of only one charge sign are held
cite (or detect) negative ions. Third, noise at frequen- in a Penning trap, because application of a positive
cies (say) above the reference frequency is folded over (negative) electrostatic potential to the end-caps will
into the other half of the spectrum, thereby increas- trap only positive (negative) ions. However, it is pos-
ing the rms noise by a factor of 21/2 , and reducing sible to trap ions of both charge signs by applying an
S/N ratio by the same factor, relative to direct-mode alternating potential to each of the end-caps [50–52],
detection. or by application of “nested” positive and negative
Foldover of signal and noise is eliminated in static voltages [53–55]. In the former mode, ion ax-
“quadrature” heterodyne detection, by distinguish- ial motion is governed by a Mathieu equation, whose
ing between signals whose frequencies lie above and solution leads to a z-stability diagram for which opti-
below the reference frequency, as shown in Fig. 7. mal results are obtained at z-stability parameter, qz =
The ICR signal is divided into two equal components (11.095)qVac /(md2 Ω 2 ) ≈ 0.5, in which m/q is the ion
referred to as real and imaginary data. A cosine wave- mass-to-charge ratio, Vac and Ω are the zero-to-peak
form is convolved with the real channel and a sine amplitude and frequency of the applied alternating
waveform is convolved with the imaginary channel. electric potential, and the numerical factor is for a cu-
The reference waveforms (cosine and sine) have the bic trap of edge length, d. The latter technique creates
same frequency but differ in phase by 90◦ . The data separate positive and negative potential wells inside
are treated with a low-pass filter, and the convolved the ICR cell volume.
real and imaginary data are summed to produce the Ion polarity may be distinguished based on opposite
final spectrum. The residual foldover peaks present in rotational phase by use of a different form of quadra-
single-phase heterodyne detection are eliminated by ture detection. In this case, the signal from one pair
that addition, because the foldover peaks have equal of opposed detection electrodes is stored as mathe-
amplitude but opposite phase in the two channels matically “real” and the signal from an orthogonal
(shown in Fig. 7). Adding the spectra from the two pair of detection electrodes is stored as mathematically
channels doubles the signal (thereby exactly compen- “imaginary”. Fourier transformation then produces a
sating for halving the original signal when it was di- spectrum in which positive (negative) ions appear at
vided into two channels). The S/N ratio for quadrature positive (negative) frequency. Thus, it becomes easy
heterodyne detection is thus, equal to that for direct to distinguish positive and negative ions of the same
detection, and exceeds that for single-phase hetero- nominal mass in a single spectrum. For example, C60 +
dyne detection by a factor of 21/2 . Finally, although and C60 − differ by only the mass of two electrons
quadrature detection can be performed by analog [48] (∼0.001 Da) in single-phase detection, but differ by
or digital [49] means, digital quadrature detection is (+720) − (−720) = 1440 Da in quadrature detection
preferable because (a) it eliminates noise from added [56]. We denote this experiment as “physical” quadra-
analog oscillator, mixer, and filter components; (b) ture (because the signals are detected on physically
“image” peaks [19] resulting from unequal scaling separate pairs of electrodes), whereas the “digital”
of the two quadrature signals are eliminated; and (c) quadrature of the preceding section is performed by
on-line digital filtering ensures that the filter band- software manipulation of the signal from a single pair
width automatically matches the Nyquist bandwidth of detection electrodes. (Conversely, by reciprocity
 A.G. Marshall, C.L. Hendrickson / International Journal of Mass Spectrometry 215 (2002) 59–75 69

[57], ions of a single charge sign may be isolated by collision model is much more appropriate for the
quadrature excitation [58] rather than quadrature de- much higher velocities of ions during FT-ICR data
tection.) acquisition.) At intermediate collision frequency (e.g.,
collision lifetime of the order of the time-domain
data acquisition period), the mass spectral line shape
9. Ion-neutral collisions represents a convolution between the zero-pressure
and high-pressure limiting line shapes [19,59] (Fig. 8,
The magnitude-mode FT-ICR mass spectral peak middle). In fact, ICR time-domain data are typically
shape for an undamped time-domain signal (i.e., zero “windowed” (i.e., multiplied by any of several weight
neutral pressure) is a “sinc-squared” function, with a functions) to yield an “apodized” peak shape that is
series of decreasing-amplitude lobes on either side of usually designed to narrow the otherwise broad skirts
the peak maximum [23] (see Fig. 8, top). The effect of the peak at the expense of slight increase in the
of ion-neutral collisions (reactive or non-reactive) width at half-maximum peak height [19].
is to damp the time-domain ICR signal. In the Ion-neutral collisions broaden and distort FT-ICR
“high-pressure” limit that the ICR time-domain signal mass spectral peaks (but produce only a negligi-
is collisionally damped essentially to zero during the ble frequency-shift). Thus, such collisions do not
acquisition period, the FT mass spectral peak shape is change the m/z values determined by frequency-to-m/z
either Lorentzian (for the Langevin collision model) calibration [30,32], as long as the peaks are still
[59] (see Fig. 8, bottom), or a non-analytic function well-resolved. However, even for well-resolved peaks,
that is narrower at half-height but broader at the bot- collisional broadening does reduce the precision of
tom than a Lorentzian (hard-sphere collision model) mass measurement, because that precision is propor-
[60]. (Although, the Langevin model offers a good tional to the product of peak height and the square
description for room-temperature ions, a hard-sphere root of the number of data points per peak width [61].
Since, FT-ICR peak height varies inversely with peak
width, m50% , precision varies as (1/m50% )1/2 .

10. Fourier transform aspects of ICR excitation

and detection

The most unique advantage of FT-ICR as a mass

analyzer is that ion mass-to-charge ratio is experimen-
tally manifested as a frequency. Because frequency can
be measured more accurately than any other exper-
imental parameter, ICR MS offers inherently higher
resolution (and thus, higher mass accuracy) than any
other type of mass measurement.
Various aspects of Fourier transform data reduction
Fig. 8. Simulated time-domain ICR signals (left) and
related to ICR peak shape and position have been
frequency-domain magnitude spectra (right) for low-pressure, discussed elsewhere: Nyquist sampling and foldover
τ
T acq’n (top), intermediate-pressure, τ ≈ T acq’n (middle), [62], fast Fourier transformation [63], zero-filling
and high-pressure, τ  T acq’n (bottom) limits. τ is the lifetime
[64], “windowing” or apodization [19], deconvolu-
for ion-neutral collisions, Tacq’n is the data acquisition period,
and the collision mechanism is assumed to be ion-induced dipole tion [65–67], oversampling [68], two-dimensional
(Langevin [121]). (Reproduced, with permission, from [1].) Hadamard [69] or Fourier [70] MS/MS, etc. Non-FT
70 A.G. Marshall, C.L. Hendrickson / International Journal of Mass Spectrometry 215 (2002) 59–75

methods for obtaining a frequency-domain (and thus are significantly enhanced at higher magnetic field
mass-domain) spectrum from a digitized time-domain strength. (c) The upper mass limit at which ions can
ICR signal include: the Hartley transform (a way of be detected increases as B2 [30]. (d) The maximum
performing a Fourier transform on real-only data) number of ions that can be confined without coales-
[71,72], the Bayesian maximum entropy method cence of their FT-ICR signals increases as B2 [41],
(MEM) [73,74], linear prediction [75,76], and filter and the highest mass at which two packets of equally
diagonalization [77]. However, FT data reduction charged and equally abundant ions (independent of
is still overwhelmingly preferred, in part because charge state) can be resolved to 1 Da varies as B [42].
non-FT methods typically require an order of mag- (e) The maximum period that ions can be confined
nitude more data storage; their computation time for in a Penning trap varies as B2 . (f) The maximum
an N-point time-domain data set typically scales as post-excitation ion translational energy varies as B2 ,
N2 or N3 (vs. N loge N for the fast Fourier transform for more efficient and/or extensive collisional frag-
[63]); and they typically perform best when the num- mentation at higher magnetic field strength. (g) Ion
ber of spectral peaks is small and the peak shapes are cyclotron thermal (i.e., initial) radius varies inversely
uniform and known. with B, so that an ion packet can be made more spa-
Finally, it is worth noting that Fourier transforms tially coherent at high magnetic field strength. Finally,
work backward (from frequency- to time-domain) as these various advantages may be exploited in combi-
well as forward (from time- to frequency-domain). nation, so as to produce even higher enhancement in
Thus, it is possible to specify a desired magnitude a particular parameter: e.g., signal-to-noise ratio can
vs. frequency excitation spectrum, and apply an improve by more than a factor of B2 if mass resolving
“inverse” Fourier transform to generate the corre- power is fixed at the same value as at lower magnetic
sponding time-domain waveform to be applied to field [83].
the excitation electrodes of an ICR trapped-ion cell. For a given spatial homogeneity, a larger magnet
Such “stored waveform inverse Fourier transform bore diameter, d, is also advantageous. (a) Dynamic
(“SWIFT”) excitation [78] is now widely used in both range (i.e., ratio of most abundant to least abundant
FT-ICR [79] and Paul (quadrupole) [80,81] ion traps, ions in the mass spectrum) increases as d2 . (b) Effi-
for mass-selective excitation and ejection. ciency of removal of gases from the vacuum chamber
increases as d3 . (c) A larger magnet bore facilitates in-
troduction of multiple electrical feed-throughs as well
11. Effect of magnet strength and size as photon optics, and heating or cooling of the ion
on ICR detection trap.

As noted in Section 2, the detected ICR signal

magnitude is formally independent of applied mag- 12. Harmonics and multiply segmented
netic field strength. Nevertheless, the quality of the detector electrodes
detected signal improves dramatically with increas-
ing magnetic field, B, for several other reasons [82]. Non-linearity in ICR time-domain signal magni-
(a) Mass resolving power (and mass measurement tude as a function of ion cyclotron orbital radius
accuracy) in FT-ICR MS increase linearly with in- leads to signals at (typically odd-numbered) multi-
creasing B [23,59]. (b) Because ICR frequency in- ples of the “fundamental” reduced ICR frequency of
creases linearly with B, the length of time needed to Eq. (4a) [84]. The non-linear component can pro-
acquire data for a mass spectrum of a given resolving vide an independent measure of the post-excitation
power varies inversely with B. Thus, data acquisition ICR orbital radius [85]. However, harmonic signals
period-limited experiments, such as on-line LC/MS, are generally viewed as undesirable in FT-ICR MS,
 A.G. Marshall, C.L. Hendrickson / International Journal of Mass Spectrometry 215 (2002) 59–75 71

because the fundamental ICR signal magnitude is re- One way is to measure the m/z separation between the
duced by the combined magnitudes of the harmonic quasimolecular ion whose carbon atoms are all 12 C
signals; and harmonics increase the number of peaks and the same species in which one of the carbons is
in the mass spectrum and can interfere with identi- 13 C—the m/z separation will be 1/z [95]. A second way

fication of other fundamental (but low-magnitude) is to change the trapping potential: multiply-charged
signals. and singly-charged ions will exhibit the same fre-
Nevertheless, more than a decade ago, it was pro- quency shift, whereas the signal at a harmonic fre-
posed that FT-ICR mass spectral resolution could be quency will shift by n times as much as the signal
improved by segmenting a Penning trap into multi- at the fundamental frequency. The latter method was
ple pairs of detection electrodes, and combining the used in the first demonstration of a fullerene gas-phase
signals from alternate pairs [86–88]. For n pairs of dianion [96].
opposed electrodes, the detected ICR signal for ions
of a given m/z would execute n cycles of oscillat-
ing signal, compared to just 1 cycle for the usual 13. New developments and directions
single pair of opposed detector electrodes. Thus, the
ICR signal frequency would increase by a factor of With current (year 2002) FT-ICR MS detector tech-
n (“harmonic”). If the ICR signal duration were the nology, it is already possible to resolve and identify
same, then the FT-ICR spectral peak width would up to thousands of components of a complex mixture,
be the same, and mass resolving power would in- often without prior wet chemical separation, thereby
crease by a factor of n. Although, this general idea has potentially changing the whole approach to dealing
been examined repeatedly since its inception [89–92], with chemical and biological complexity. More than
it has yet to be adopted for any chemical or bio- 360 FT-ICR mass spectrometers have been installed
chemical applications. The main problems are: (a) it worldwide. We now consider some projected future
is difficult to configure a detector that will produce improvements.
just a single multiple-harmonic signal—usually sev-
eral harmonics are present, further complicating anal- 13.1. Magnets
ysis; (b) the extent of improvement depends on the
mechanism for damping of the time-domain ICR sig- Given the multiple independent advantages offered
nal [90]; and (c) harmonic signal magnitude is small by higher magnetic field, an obvious direction for fu-
relative to that at the “fundamental” frequency unless ture ICR technology is toward higher-field magnets.
the ICR orbital radius is very near to the cell max- Superconducting magnets up to 11.5 T [97] are already
imum (at which non-linearities in electric and mag- in use for ICR, with at least one 15 T system sched-
netic field lead to other problems) [93]. (Signals at uled for installation in 2002 at NHMFL. FT-NMR su-
other “combination” frequencies (e.g., sums and dif- perconducting magnets up to 21.1 T (900 MHz for 1 H
ferences of integer multiples of ω+ , ωc , and ω− , can NMR) systems are already commercially available.
result from non-linearities in the electric fields applied FT-ICR MS has been performed at 20 T [98] and 25 T
to the ion trap [29]. These signals may be enhanced [99] with resistive magnets, and even better perfor-
by appropriate detection schemes [94].) mance can be anticipated from a hybrid magnet whose
At this point, the reader might wonder how one can superconducting and resistive coils are connected in
distinguish signals at harmonic frequencies (i.e., n ω+ , series (so that the high inductance of the supercon-
in which n = 2, 3, 4, . . . ) for a singly-charged ion ducting magnet will limit temporal fluctuations in the
from the fundamental cyclotron frequency (ω+ ) for low-inductance resistive coil). There are no technical
multiply-charged dimers of the same ion (e.g., [2M + barriers to such a hybrid magnet up to at least 27 T
2H]2+ vs. [M + H]+ or [2M + 2e]2- vs. [M + e]− ). with 110 mm bore diameter.
72 A.G. Marshall, C.L. Hendrickson / International Journal of Mass Spectrometry 215 (2002) 59–75

13.2. FT data reduction capacitor to null the excitation signal at the detec-
tor electrodes [106] It should be possible to apply
FT-ICR mass spectra are currently reported as that experiment to phase a broadband FT-ICR mass
“magnitude”-mode, i.e., the square root of the sum spectrum, to yield significant improvement in mass
of the squares of the “real” and “imaginary” spectra resolution.
that result from FT of a time-domain signal. From The combination of digitizers of higher preci-
the outset, it was recognized that significantly higher
√ sion (>20 bit/word), speed (>10 MHz), and storage
mass resolving power (factor ranging from ∼ 3 (>16 Mwords) with high-speed (1 Mword FFT in
(Lorentzian peak shape) to 2 (sinc peak shape) [59] <0.4 s) desktop computers means that it is increas-
may be realized by appropriate “phasing” (i.e., linear ingly possible to acquire most FT-ICR MS data in
combination of the raw “real” and “imaginary” spec- direct-mode, without the need for heterodyning. For
tra to yield an “absorption”-mode spectrum [100]. example, we have recently demonstrated an aver-
Resolving power improves even more near the base age mass resolving power of >300,000 over a range,
of each peak. Moreover, access to absorption-mode 200 < m/z < 1000, at 9.4 T, based on direct-mode
display makes it possible to discriminate between dif- detection (4 Mword). That approach becomes increas-
ferent collisional line-broadening mechanisms [101]. ingly desirable for LC/MS applications, in which all
There have been a few reports of absorption-mode of the ICR data must be acquired during the pas-
FT-ICR mass spectra [101,102], and the improvement sage (a few seconds) of the LC peak into the mass
in mass resolving power was clearly demonstrated. spectrometer.
However, the two big problems have been to cor-
rect for the (non-linear) phase variation introduced 13.3. FT-ICR operating modes
by the broadband (e.g., frequency-sweep) excitation
itself and the time delay between the end of the The evolution of FT-ICR MS shares many fea-
excitation event and the beginning of the detection tures in common with the (generally prior) develop-
event (to provide for “ringdown” of the excitation ment of FT nuclear magnetic resonance spectroscopy
signal at the detector electrodes). A time delay longer [12,107]. In fact, there are deep homologies that
than one-half of one “dwell” period (i.e., the interval connect virtually every kind of FT-NMR experi-
between acquisition of successive time-domain data ment with an analogous one in FT-ICR MS, ranging
points) introduces uncorrectable baseline modulation from “spin-locking” [108] (“ion-locking” [109]) to
into the final FT spectrum [19]. Deconvolution of two-dimensional NOESY (NMR) or SWIM (ICR)
the non-linear magnitude vs. frequency spectrum of for identifying coupled spins (NMR [110]) or cou-
the excitation signal has been demonstrated theoreti- pled ion-molecule reactions (ICR [70]). However, one
cally for frequency-sweep excitation [65]. However, FT-NMR experiment that has not yet been applied
such deconvolution requires that excitation and de- to FT-ICR is time-shared excitation/detection, intro-
tection are simultaneous. Simultaneous excitation and duced in NMR as “tailored” excitation [111]. Like
(non-FT) single-frequency detection was introduced SWIFT excitation, tailored excitation begins from a
in 1971, but never followed up [103]. Simultaneous time-domain waveform generated by inverse FT of
excitation/narrow-band detection was explored briefly the desired frequency-domain excitation spectrum.
in the 1970’s [104] (by analogy to “correlation” NMR However, that waveform is then turned on and off as it
[105]), but was soon abandoned (in both NMR and proceeds, so that detection may take place during the
ICR) because the experiment was much slower (and intervals that the excitation is turned off. The excita-
more difficult to “tune”) than the FT experiment. Re- tion envelope may consist of pulses of equal duration
cently, simultaneous broadband excitation/detection and varying amplitude or (electronically simpler) of
has been achieved, by careful tuning of a small air-gap constant amplitude and varying duration. In this way,
 A.G. Marshall, C.L. Hendrickson / International Journal of Mass Spectrometry 215 (2002) 59–75 73

one achieves essentially simultaneous excitation and [20] R. Chen, X. Cheng, D.W. Mitchell, S.A. Hofstadler, Q. Wu,
detection, but without having to “null” the excitation A.L. Rockwood, M.G. Sherman, R.D. Smith, Anal. Chem.
67 (1995) 1159–1163.
by a large factor at the detector. Tailored excitation [21] E.R. Williams, K.D. Henry, F.W. McLafferty, J. Am. Chem.
thus, offers an alternate route to generation of the Soc. 112 (1990) 6157–6162.
absorption-mode spectrum discussed in Section 13.2. [22] J.P. Speir, G.S. Gorman, C.C. Pitsenberger, C.A. Turner, P.P.
Wang, I.J. Amster, Anal. Chem. 65 (1993) 1746–1752.
[23] M.B. Comisarow, A.G. Marshall, J. Chem. Phys. 64 (1976)
110–119.
Acknowledgements [24] A.G. Marshall, C.L. Hendrickson, Rapid Commun. Mass
Spectrom. 15 (2001) 232–235.
This work was supported by NSF (CHE-99-09502), [25] L.S. Brown, G. Gabrielse, Rev. Mod. Phys. 58 (1986) 233–
311.
NIH (GM/RR-31683), Florida State University, and [26] G. Savard, S. Becker, G. Bollen, H.-J. Kluge, R.B. Moore,
the National High Magnetic Field Laboratory in Tal- L. Schweikhard, H. Stolzenberg, U. Wiess, Phys. Lett. A
lahassee, FL. 158 (1991) 247–252.
[27] L. Schweikhard, S. Guan, A.G. Marshall, Int. J. Mass
Spectrom. Ion Proc. 120 (1992) 71–83.
References [28] G. Jackson, J.D. Canterbury, S. Guan, A.G. Marshall, J. Am.
Soc. Mass Spectrom. 8 (1997) 283–293.
[29] A.G. Marshall, P.B. Grosshans, Anal. Chem. 63 (1991)
[1] A.G. Marshall, C.L. Hendrickson, G.S. Jackson, Mass 215A–229A.
Spectrom. Rev. 17 (1998) 1–35.
[30] E.B. Ledford Jr., D.L. Rempel, M.L. Gross, Anal. Chem.
[2] A.G. Marshall, Int. J. Mass Spectrom. 200 (2000) 331–336.
56 (1984) 2744–2748.
[3] H. Sommer, H.A. Thomas, J.A. Hipple, Phys. Rev. 76 (1949)
[31] J.B. Jeffries, S.E. Barlow, G.H. Dunn, Int. J. Mass Spectrom.
1877.
Ion Proc. 54 (1983) 169–187.
[4] H. Sommer, H.A. Thomas, Phys. Rev. 78 (1950) 806.
[32] S.D.-H. Shi, J.J. Drader, M.A. Freitas, C.L. Hendrickson,
[5] H. Sommer, H.A. Thomas, J.A. Hipple, Phys. Rev. 82 (1951)
A.G. Marshall, Int. J. Mass Spectrom. 195/196 (2000) 591–
697–702.
598.
[6] P.V. Ameri, H.R. Krouse, H. Fichtner, Rev. Sci. Instrum. 72
[33] J.E. Bruce, G.A. Anderson, S.A. Hofstadler, B.E. Winger,
(2001) 2036–2042.
R.D. Smith, Rapid Commun. Mass Spectrom. 7 (1993) 700–
[7] N.I. Leont’ev, Instrum. Techn. (in English) 2 (1960) 275–
703.
281.
[34] S. Guan, M.C. Wahl, A.G. Marshall, Anal. Chem. 65 (1993)
[8] D. Wobschall, Rev. Sci. Instrum. 36 (1965) 466.
3647–3653.
[9] F. DiFillip, V. Natarajan, M. Bradley, F. Palmer, D.E.
Pritchard, Phys. Scripta T59 (1995) 144–154. [35] C. Gerz, D. Wilsdorf, G. Werth, Nuc. Instrum. Meth. Phy.
[10] G. Gabrielse, X. Fei, L.A. Orozco, R.L. Tjoelker, J. Haas, Res. B47 (1990) 453–461.
H. Kalinowsky, T.A. Trainor, W. Kells, Phys. Rev. Lett. 65 [36] X. Xiang, P.B. Grosshans, A.G. Marshall, Int. J. Mass
(1990) 1317–1320. Spectrom. Ion Proc. 125 (1993) 33–43.
[11] G. Bollen, H.-J. Kluge, M. Konig, T. Otto, H. Stolzenberg, [37] T.M. O’Neil, Phys. Scripta T59 (1995) 341–351.
R.B. Moore, G. Rouleau, G. Audi, Phys. Rev. C 46 (1992) [38] A.J. Peurrung, R.T. Kouzes, S.E. Barlow, Int. J. Mass
R2140–R2143. Spectrom. Ion Proc. 157/158 (1996) 39–83.
[12] A.G. Marshall, Acc. Chem. Res. 9 (1996) 307–316. [39] D.W. Mitchell, J. Am. Soc. Mass Spectrom. 10 (1999) 136–
[13] M. Wang, A.G. Marshall, Int. J. Mass Spectrom. Ion Proc. 152.
100 (1990) 323–346. [40] Y. Naito, M. Inoue, J. Mass Spectrom. Soc. Jpn. 42 (1994)
[14] M.B. Comisarow, J. Chem. Phys. 69 (1978) 4097–4104. 1–9.
[15] P.A. Limbach, P.B. Grosshans, A.G. Marshall, Anal. Chem. [41] D.W. Mitchell, R.D. Smith, Phys. Rev. E 52 (1995) 4366–
65 (1993) 135–140. 4386.
[16] M.B. Comisarow, Int. J. Mass Spectrom. Ion Phys. 37 (1981) [42] D.W. Mitchell, R.D. Smith, J. Mass Spectrom. 31 (1996)
251–257. 771–790.
[17] P.B. Grosshans, A.G. Marshall, Int. J. Mass Spectrom. Ion [43] M.B. Comisarow, A.G. Marshall, Chem. Phys. Lett. 25
Proc. 100 (1990) 347–379. (1974) 282–283.
[18] S. Guan, A.G. Marshall, Int. J. Mass Spectrom. Ion Proc. [44] M.B. Comisarow, A.G. Marshall, Chem. Phys. Lett. 26
146/147 (1995) 261–296. (1974) 489–490.
[19] A.G. Marshall, F.R. Verdun, Fourier Transforms in NMR, [45] M.B. Comisarow, A.G. Marshall, US Patent (1976).
Optical, and Mass Spectrometry: A User’s Handbook, [46] F. He, C.L. Hendrickson, A.G. Marshall, Anal. Chem. 73
Elsevier, Amsterdam, 1990. (2001) 647–650.
74 A.G. Marshall, C.L. Hendrickson / International Journal of Mass Spectrometry 215 (2002) 59–75

[47] S.D.-H. Shi, C.L. Hendrickson, A.G. Marshall, Proc. Natl. [77] P.B. O’Connor, V. Mandelshtam, M.R.T., in: Proceedings of
Acad. Sci. U.S.A. 95 (1998) 11532–11537. the 47th American Society Mass Spectrometry Conference
[48] F.R. Verdun, S.L. Mullen, T.L. Ricca, A.G. Marshall, FACSS on Mass Spectrometry and Allied Topics, Am. Soc. Mass
XIV, Detroit, MI, 1987 (Abstract #41). Spectrom., Dallas, TX, 1999, ThPB017.
[49] J.J. Drader, S.D.-H. Shi, G.T. Blakney, C.L. Hendrickson, [78] A.G. Marshall, T.-C.L. Wang, T.L. Ricca, J. Am. Chem.
D.A. Laude, A.G. Marshall, Anal. Chem. 71 (1999) 4758– Soc. 107 (1985) 7893–7897.
4763. [79] S. Guan, A.G. Marshall, Int. J. Mass Spectrom. Ion Proc.
[50] M.V. Gorshkov, S. Guan, A.G. Marshall, Rapid Commun. 157/158 (1996) 5–37.
Mass Spectrom. 6 (1992) 166–172. [80] S. Guan, A.G. Marshall, Anal. Chem. 65 (1993) 1288–1294.
[51] C.B. Jacoby, C.L. Holliman, D.L. Rempel, M.L. Gross, J. [81] R.K. Julian Jr., R.G. Cooks, Anal. Chem. 65 (1993) 1827–
Am. Soc. Mass Spectrom. 4 (1993) 186–189. 1833.
[52] V. Frankevich, R. Zenobi, Int. J. Mass Spectrom. Ion Proc. [82] A.G. Marshall, S. Guan, Rapid Commun. Mass Spectrom.
207 (2001) 57–67. 10 (1996) 1819–1823.
[53] G. Gabrielse, S.L. Rolston, L. Haarsma, W. Kells, Phys. [83] M.W. Senko, C.L. Hendrickson, L. Pasa-Tolic, J.A. Marto,
Lett. A (1988) 38–42. F.M. White, S. Guan, A.G. Marshall, Rapid Commun. Mass
[54] Y. Wang, K.-P. Wanczek, Rev. Sci. Instrum. 64 (1993) 883– Spectrom. 10 (1996) 1824–1828.
889. [84] P.B. Grosshans, P.J. Shields, A.G. Marshall, J. Am. Chem.
[55] V.H. Vartanian, D.A. Laude, Org. Mass Spectrom. 29 (1994) Soc. 112 (1990) 1275–1277.
692–694. [85] P.B. Grosshans, A.G. Marshall, Int. J. Mass Spectrom. Ion
[56] J.J. Drader, L. Schweikhard, S.D.-H. Shi, C.L. Hendrickson, Proc. 115 (1992) 1–19.
A.G. Marshall, in: Proceedings of the 50th Pittsburgh [86] E.N. Nikolaev, M.V. Gorshkov, A.V. Mordehai, V.L. Talrose,
Conference on Analytical Chemistry and Applied USSR Patent (1987).
Spectroscopy, Orlando, FL, 1999 (Abstract #2065P). [87] Y.P. Pan, D.P. Ridge, J. Wronka, A.L. Rockwood, Rapid
[57] W. Shockley, J. Appl. Phys. 9 (1938) 635. Commun. Mass Spectrom. 1 (1987) 121.
[58] L. Schweikhard, A.G. Marshall, J. Am. Soc. Mass Spectrom. [88] Y.P. Pan, D.P. Ridge, A.L. Rockwood, Int. J. Mass Spectrom.
4 (1993) 433–452. Ion Proc. 84 (1988) 293.
[89] P. Kofel, M. Allemann, H. Kellerhals, K.P. Wanczek, in: P.
[59] A.G. Marshall, M.B. Comisarow, G. Parisod, J. Chem. Phys.
Longevialle (Ed.), Advances in Mass Spectrometry, Heyden
71 (1979) 4434–4444.
and Son., Ltd., London, 1989, 676–677.
[60] S. Guan, G.-Z. Li, A.G. Marshall, Int. J. Mass Spectrom.
[90] P.B. Grosshans, A.G. Marshall, Int. J. Mass Spectrom. Ion
Ion Proc. 167/168 (1998) 185–194.
Proc. 107 (1991) 49–81.
[61] L. Chen, C.E. Cottrell, A.G. Marshall, Chemometr. Intell.
[91] P.A. Limbach, P.B. Grosshans, A.G. Marshall, Int. J. Mass
Lab. Syst. 1 (1986) 51–58.
Spectrom. Ion Proc. 123 (1993) 41–47.
[62] M. Wang, A.G. Marshall, Anal. Chem. 60 (1988) 341–344.
[92] M. Knobeler, K.P. Wanczek, Int. J. Mass Spectrom. Ion
[63] J.W. Cooley, J.W. Tukey, Math. Comput. 19 (1965) 297.
Proc. 125 (1993) 127.
[64] M.B. Comisarow, J. Melka, Anal. Chem. 51 (1979) 2198–
[93] E.N. Nikolaev, M.V. Gorshkov, A.V. Mordehai, V.L. Talrose,
2203.
Rapid Commun. Mass Spectrom. 4 (1990) 144–146.
[65] A.G. Marshall, Chem. Phys. Lett. 63 (1979) 515–518. [94] L. Schweikhard, M. Lindinger, H.-J. Kluge, Int. J. Mass
[66] A.G. Marshall, D.C. Roe, J. Chem. Phys. 73 (1980) 1581– Spectrom. Ion Proc. 98 (1990) 25–33.
1590. [95] K.D. Henry, F.W. McLafferty, Org. Mass Spectrom. 25
[67] Z. Zhang, S. Guan, A.G. Marshall, J. Am. Soc. Mass (1990) 490–492.
Spectrom. 8 (1997) 659–670. [96] P.A. Limbach, L. Schweikhard, K.A. Cowen, M.T.
[68] G.M. Alber, A.G. Marshall, Appl. Spectrosc. 44 (1990) McDermott, A.G. Marshall, J.V. Coe, J. Am. Chem. Soc.
1111–1116. 113 (1991) 6795–6798.
[69] E.R. Williams, S.Y. Loh, F.W. McLafferty, R.B. Cody, Anal. [97] M.V. Gorshkov, L. Pasa-Tolic, H.R. Udseth, G.A. Anderson,
Chem. 62 (1990) 698–703. B.M. Huang, J.E. Bruce, P.D.C., S.A. Hofstadler, L. Tang,
[70] C.W. Ross III, S. Guan, P.B. Grosshans, T.L. Ricca, A.G. L.-Z. Chen, J.A. Willett, A.L. Rockwood, M.S. Sherman,
Marshall, J. Am. Chem. Soc. 115 (1993) 7854–7861. R.D. Smith, J. Am. Soc. Mass Spectrom. 9 (1998) 692–700.
[71] C.P. Williams, A.G. Marshall, Anal. Chem. 61 (1989) 428– [98] C.L. Hendrickson, J.J. Drader, D.A. Laude Jr., S. Guan,
431. A.G. Marshall, Rapid Commun. Mass Spectrom. 10 (1996)
[72] C.P. Williams, A.G. Marshall, Anal. Chem. 64 (1992) 916– 1829–1832.
923. [99] S.D.-H. Shi, J.J. Drader, C.L. Hendrickson, A.G. Marshall,
[73] J.E. Meier, A.G. Marshall, Anal. Chem. 62 (1990) 201–208. J. Am. Soc. Mass Spectrom. 10 (1999) 265–268.
[74] J.E. Meier, A.G. Marshall, Anal. Chem. 63 (1991) 551–560. [100] M.B. Comisarow, A.G. Marshall, Can. J. Chem. 52 (1974)
[75] J.F. Loo, M.D. Krahling, T.C. Farrar, Rapid Commun. Mass 1997–1999.
Spectrom. 4 (1990) 297–299. [101] B.A. Vining, R.E. Bossio, A.G. Marshall, Anal. Chem. 71
[76] S. Guan, A.G. Marshall, Anal. Chem. 69 (1997) 1156–1162. (1999) 460–467.
 A.G. Marshall, C.L. Hendrickson / International Journal of Mass Spectrometry 215 (2002) 59–75 75

[102] E.C. Craig, I. Santos, A.G. Marshall, Rapid Commun. Mass [111] B.L. Tomlinson, H.D.W. Hill, J. Chem. Phys. 59 (1973)
Spectrom. 1 (1987) 33–37. 1775–1784.
[103] P.P. Genequand, Z. Angew. Math. Phys. 22 (1971) 951– [112] W.T. Scott, Am. J. Phys. 27 (1959) 418–419.
975. [113] M.B. Comisarow, Adv. Mass Spectrom. 8 (1980) 1698–1706.
[104] R.T. Hunter, R.T. McIver Jr., Chem. Phys. Lett. 49 (1977) [114] S.H. Lee, K.-P. Wanczek, H. Hartmann, Adv. Mass
577. Spectrom. 8B (1980) 1645–1649.
[105] J. Dadok, R.F. Sprecher, J. Magn. Reson. 13 (1974) 243. [115] P. Kofel, M. Allemann, H. Kellerhals, K.-P. Wanczek, Int.
[106] S.C. Beu, in: Proceedings of the 46th American Society J. Mass Spectrom. Ion Proc. 74 (1986) 1–12.
of Mass Spectrometry Conference on Mass Spectrometry [116] J.L. Elkind, F.D. Weiss, J.M. Alford, R.T. Laaksonen, R.E.
and Allied Topics, Am. Soc. Mass Spectrom., Orlando, FL, Smalley, J. Chem. Phys. 88 (1988) 5215–5224.
1998, p. 502. [117] P. Caravatti, M. Allemann, Org. Mass Spectrom. 26 (1991)
[107] A.G. Marshall, Acc. Chem. Res. 18 (1985) 316–322. 514–518.
[108] I. Solomon, Compt. Rend. Acad. Sci. Paris 248 (1959) 92– [118] G. Gabrielse, L. Haarsma, S.L. Rolston, Int. J. Mass
94. Spectrom. Ion Proc. 88 (1989) 319–332.
[109] R. Chen, A.G. Marshall, M. Wang, Chem. Phys. Lett. 181 [119] S.C. Beu, D.A. Laude Jr., Anal. Chem. 64 (1992) 177–180.
(1991) 168–174. [120] S.C. Beu, D.A. Laude Jr., Int. J. Mass Spectrom. Ion Proc.
[110] G. Bodenhausen, R.R. Ernst, J. Am. Chem. Soc. 104 (1982) 112 (1992) 215–230.
1304–1309. [121] P. Langevin, Ann. Chim. Phys. 5 (1905) 245.