Anal. Chem.

2007, 79, 2850-2858

Enantiomer Separation of Amino Acids by

Complexation with Chiral Reference Compounds
and High-Field Asymmetric Waveform Ion Mobility
Spectrometry: Preliminary Results and Possible
Limitations
Axel Mie,*, Magnus Jo
1 rnten-Karlsson, Bengt-Olof Axelsson, Andrew Ray, and Curt T. Reimann

Department of Analytical Chemistry, Chemical Center, Lund University, P.O. Box 124, SE-221 00 Lund, Sweden, Analytical
Development, AstraZeneca R&D, Lund, Sweden, and Analytical Development, AstraZeneca R&D, Loughborough, United
Kingdom

We present a new method for separation of enantiomers

with high-field asymmetric waveform ion mobility spectrometry (FAIMS), coupled to mass spectrometric detection. Upon addition of an appropriate chiral reference
compound to the analyte solution and subsequent ionization of the solution by electrospray ionization, analyte
enantiomers formed diastereomeric complexes, which
were potentially separable by FAIMS. The methodology
being developed is intended to be general, but here amino
acid analytes are specifically considered. In the examples
presented herein, six pairs of amino acid enantiomers
were successfully separated as metal-bound trimeric
complexes of the form [MII(L-Ref)2(D/L-A)-H]+, where MII
is a divalent metal ion, L-Ref is an amino acid in its L form
acting as chiral reference compound, and A is the amino
acid analyte. For example, D- and L-tryptophan were
separated in FAIMS as [NiII(L-Asn)2(D-Trp)-H]+ and [NiII(L-Asn)2(L-Trp)-H]+. As FAIMS separation typically takes
place over a time scale of only a few hundred milliseconds,
the presented separation method opens new possibilities
for rapid analysis of one analyte enantiomer in the
presence of the other enantiomer. Preliminary quantification results are presented, which suggest that fast and
sensitive quantitative chiral analyses can be performed
with FAIMS. Method limitations are discussed in terms
of diverse phenomena, which are not yet understood.
Different structural isomers and stereoisomers of molecules
often feature different physical, chemical, and biological properties.
One group of stereoisomers, the enantiomers, is of specific interest
because two enantiomers have different chemical properties only
in asymmetric environments. Life itself offers one of the key
examples of asymmetric environments, and many cases are known
for which two enantiomers have very different effects on biological
systems. One example is drug substances, one enantiomer of
* Corresponding author. E-mail: Axel.Mie@analykem.lu.se.
Lund University.
AstraZeneca R&D, Lund, Sweden.
AstraZeneca R&D, Loughborough, UK.

2850 Analytical Chemistry, Vol. 79, No. 7, April 1, 2007

which has a therapeutic effect (eutomer) while the other enantiomer has no therapeutic effect, or an adverse effect (distomer).
It is a challenge for analytical chemists to develop techniques that
can either separate enantiomers or else reliably detect one such
enantiomer in the presence of the other, even when there is a
large concentration difference.
A number of techniques have been developed with the aim of
quantifying enantiomers. All of them have in common that they
provide some kind of local chiral environment that offers stereoselective interaction with the enantiomers, in order to distinguish
between them. Probably most widespread today are liquid
chromatographic (LC) methods,1 either employing chiral stationary phases or using derivatization steps in order to transform the
pair of enantiomers into a pair of more easily separable diastereomers. Another chromatographic technique used is supercritical
fluid chromatography1 (SFC). Another common technique is
capillary electrophoresis (CE) with chiral selectors in the mobile
phase.1 Also, NMR with chiral shift reagents is commonly
employed.2
The kinetic method3 is a tandem mass spectrometric (MS/
MS) technique in which diastereomeric complex ions are formed
from the enantiomers and chiral enantiopure reference compounds, these complexes then being subjected to fragmentation.
Differences in the structures and dissociation energies of these
complexes or their dissociation products result in different
fragment branching ratios (i.e., different relative intensities of the
fragments), making it possible to measure the amounts of the
two enantiomers. A number of different mass spectrometric
approaches toward chiral recognition have been reviewed.4
Unlike LC, SFC, and CE, neither NMR nor the kinetic MS/
MS method involves physical separation of the enantiomers;
rather, the two enantiomers are measured simultaneously. LC,
SFC, CE, and NMR are rather time-consuming; NMR and the
kinetic method normally have limited sensitivity; and none of the
(1) Gubitz, G.; Schmid, M. G. Biopharm. Drug Dispos. 2001, 22, 291-336.
(2) Wenzel, T. J.; Wilcox, J. D. Chirality 2003, 15, 256-270.
(3) Tao, W. A.; Zhang, D.; Wang, F.; Thomas, P. D.; Cooks, R. G. Anal. Chem.
1999, 71, 4427-4429.
(4) Schug, K. A.; Lindner, W. J. Sep. Sci. 2005, 28, 1932-1955.
10.1021/ac0618627 CCC: $37.00

2007 American Chemical Society

Published on Web 02/28/2007

mentioned techniques provides a general solution for quantifying

enantiomers in the presence of each other. Thus, the need to
develop alternative methodologies for the fast and sensitive
analysis of enantiomers remains.
A way to separate ions in the gas phase is ion mobility
spectrometry (IMS),5 which has long been employed for discrimination of different kinds of isomers. In IMS, ions subjected to an
electric field E drift through a bath gas with a velocity v ) KE,
where K is the ion mobility. K is dependent on several factors,
including ion structure.5 IMS has been employed for showing that
C59+ and C61+ each appear in dramatically different structural
forms, while C60+ forms the famed fullerene (magic number 60
and compact spherical structure) virtually exclusively.6 Leucine
and isoleucine, structural isomers with identical mass, have been
distinguished with IMS.7 It has been shown that tryptic peptides
containing one or more proline residues exhibit traces of cistrans forms in IMS experiments.8 Diastereomeric ions are generally separable by IMS as well: Zn-ligand-hexose diastereomers
have been separated with IMS aiming ultimately at a structural
characterization that could explain fragmentation spectra observed
in tandem MS of these ions.9 A pair of olefin-linked bisparacyclophane diastereomers has also been separated by IMS.10 For the
most part, separation of enantiomers may be considered as
impossible by IMS (a typical gas-phase environment does not offer
the asymmetry needed to sense chiral intermolecular contacts).
However, by providing chiral molecules to the drift gas, enantiomers can undergo formation of short-lived clusters with the chiral
gas molecules in a stereoselective way and can thus be separated
by IMS. It has been proposed to provide selectively interactive
gaseous particles in the gas phase of the IMS, and thereby a
rough separation of chiral components of fluoxetine has been
demonstrated using a chiral collision gas, 2-butanol.11 Very
recently, this concept has been further investigated and extended
to the separation of amino acids, carbohydrates, and drug
compounds.12
Recently, the capabilities of IMS were extended with the
development of high-field asymmetric waveform ion mobility
spectrometry (FAIMS).13,14 FAIMS exploits the fact that, at
sufficiently high electric fields, most ion species display a
dependence of their ion mobility on the electric field strengths
the mobility is no longer electric-field-independent, K ) K(E). If
we express K(E) as K0[1 + h(E)], the function h(E) will typically
vary depending on the structure and identity of the ion, while K0
constitutes the ion mobility at the low-field limit. FAIMS separates
ions characterized by differing ratio of high-field mobility to lowfield mobilitysi.e., ions are separated according to the quantity 1
+ h(E). This is done by alternating a high electric field of one
(5) Eiceman, G. A.; Karpas, Z. Ion Mobility Spectrometry, 2nd ed.; Taylor &
Francis/CRC Press: Boca Raton, FL, 2005.
(6) von Helden, G.; Hsu, M. T.; Gotts, N.; Bowers, M. T. J. Phys. Chem. 1993,
97, 8182-8192.
(7) Asbury, G. R.; Hill, H. H., Jr. J. Microcolumn Sep. 2000, 12, 172-178.
(8) Counterman, A. E.; Clemmer, D. E. Anal. Chem. 2002, 74, 1946-1951.
(9) Leavell, M. D.; Gaucher, S. P.; Leary, J. A.; Taraszka, J. A.; Clemmer, D. E.
J. Am. Soc. Mass Spectrom. 2002, 13, 284-293.
(10) Baker, E. S.; Hong, J. W.; Gidden, J.; Bartholomew, G. P.; Bazan, G. C.;
Bowers, M. T. J. Am. Chem. Soc. 2004, 126, 6255-6257.
(11) Karas, M. PCT Int. Appl 2002096805, 2002.
(12) Dwivedi, P.; Wu, C.; Matz, L. M.; Clowers, B. H.; Seims, W. F.; Hill, H. H.,
Jr. Anal. Chem. 2006, 78, 8200-8206.
(13) Guevremont, R.; Purves, R. W. Rev. Sci. Instrum. 1999, 70, 1370-1383.
(14) Guevremont, R. J. Chromatogr., A 2004, 1058, 3-19.

Figure 1. Cylindrical FAIMS interface.

polarity during a short time with a low electric field of the opposite
polarity during a longer time. In such a scheme, ions would
oscillate and return to their starting positions after each cycle, if
it were not for the dependence of the ion mobility on the electric
field strength. This dependence gives the ions a net nonzero
displacement during each cycle. Typically, FAIMS separation takes
place between two parallel plates or two concentric cylinders with
a spacing of a few millimeters. A longitudinal (or axial) gas flow
sweeps the ions through the plate system, while the electric field
applied between the plates causes transverse ion motion perpendicular to the gas flow. When cylindrical electrodes are employed
(as shown in Figure 1), FAIMS features an additional ion focusing
effect because the electric field between the cylinders possesses
a gradient.13 Unfortunately, the functions h(E) are not known in
advance for analytes of interest, so that FAIMS separation needs
to be determined empirically.
FAIMS separation is driven by the dispersion voltage (DV), a
high-voltage asymmetric waveform that is applied between the
two electrodes. By applying an additional constant but adjustable
compensation voltage (CV), the net displacement of a certain ion
species during each DV waveform cycle can be compensated,
resulting in those ions being transmitted through the FAIMS while
noncompensated ions are lost in collisions with the electrode
surfaces. By scanning CV and measuring the transmitted ions at
the ion outlet, data can be obtained in the form of a CV scan or
CV spectrum. Alternatively, the CV can be kept constant or can
be cycled over a few predetermined values, so as to transmit only
ions of interest.
FAIMS, a relatively new technique, has not yet been applied
in many analytical areas, but the method shows considerable
promise in enhancing our ability to resolve various isomers in
critical applications. The structural isomers o-, m-, and p-phthalic
acid have been baseline-separated using FAIMS.15 Anomers,
linkage, and position isomers of disaccharides have also been
(15) Barnett, D. A.; Purves, R. W.; Ells, B.; Guevremont, R. J. Mass Spectrom.
2000, 35, 976-980.

Analytical Chemistry, Vol. 79, No. 7, April 1, 2007

2851

shown to be separable in FAIMS.16 Three pairs of ephedrinerelated small-molecule diastereomers have been successfully
separated.17 Recently, the partial separation in FAIMS of the
drinking water contaminants D- and L-lactic acid by forming
complexes with L-tryptophan18 was demonstrated. In the work
presented below, we have investigated the separability by FAIMS
of diastereomeric complex ions composed of chiral amino acids
and reference compounds, where the ultimate goal is unequivocal
determination of chirality. The complexes have the form [MII(LRef)2(D/L-A)-H]+, where MII is a divalent metal ion, L-Ref is a chiral
reference compound in its L form, and A is the analyte. These
analytes were chosen because they are readily available in
enantiopure form. Moreover, the chirality of the analyte for such
complexes was previously detected indirectly by Tao and Cooks
using the kinetic MS/MS method,19 giving a good comparison
point for our experiments. Here the focus is on development and
optimization of a qualitative gas-phase FAIMS-based separation
methodology for such complexes. The separation is confirmed
through both use of enantiomerically pure amino acids as test
compounds and by applying the kinetic MS/MS method in
conjunction with FAIMS. Preliminary quantitative data are also
presented.
EXPERIMENTAL SECTION
All experiments were performed using a beta unit of a
Selectra FAIMS system (Ionalytics Corp., Ottawa, Canada;
Thermo Finnigan, San Jose, CA) coupled with a custom-built
PEEK support to an Agilent 1100 series SL MSD ion trap mass
spectrometer. Samples were supplied using either an Agilent 1100
series autosampler, a syringe pump, or the autosampler together
with a syringe pump, both coupled together via a T-junction. All
samples were supplied to the ion source through direct infusion,
without prior liquid chromatographic or other separation. The ion
source was a custom-built microelectrospray (ESI) ion source
using stainless steel TaperTip needles (New Objective, Woburn
MA) with 50-m inner diameter as electrospray needles.
Operating conditions of the ESI, FAIMS, and MS were as
follows: The spray voltage applied to the microelectrospray needle
was +3400 V; the FAIMS front plate voltage was +900 V; the
FAIMS outer electrode was set to ground potential; and the
FAIMS inner electrode DV was -4000 V (zero-to-peak voltage of
the asymmetric waveform) following the equation U ) [2 sin(2ft)
+ sin(4ft - )]DV/3, f ) 750 kHz, ) /2. The CV scan speed
was 2 V/min unless otherwise specified. The CV was varied within
an overall range of 0 to -16 V.
The MS inlet capillary tip was set to ground potential. As the
focus of this work was on the FAIMS separation, generic values
were used as MS and MS/MS tune parameters. The MS capillary
exit was set to 128.5 V; the skimmer was kept at 40 V; lens 1 was
set to -5 V; the octopole 1 dc offset was set to 12 V; the octopole
2 dc offset was kept at 1.7 V; and the octopole rf amplitude was
set to 178.1 Vpp. The partition was kept at 6.8 V; lens 2 was kept
at -60 V; and the trap drive was 46.7 V. The following MS/MS
(16) Gabryelski, W.; Froese, K. L. J. Am. Soc. Mass Spectrom. 2003, 14, 265277.
(17) McCooeye, M.; Ding, L.; Gardner, G. J.; Fraser, C. A.; Lam, J.; Sturgeon, R.
E.; Mester, Z. Anal. Chem. 2003, 75, 2538-2542.
(18) Sultan, J.; Gabryelski, W. Anal. Chem. 2006, 78, 2905-2917.
(19) Tao, W. A.; Cooks, R. G. Anal. Chem. 2003, 75, 25A-31A.

2852 Analytical Chemistry, Vol. 79, No. 7, April 1, 2007

Figure 2. Spectral features determining peak resolution in FAIMS.

For discussion see the text.

parameters were used throughout all experiments if not otherwise

specified: the isolation width was 4 m/z and the fragmentation
amplitude was 0.8 V.
The actual FAIMS interface (see Figure 1) was a domed-type
FAIMS interface20 (ion inlet lateral, ion outlet axial) with 10.0mm diameter of the inner electrode, 14.0-mm inner diameter of
the outer electrode, and thus with a 2.0-mm electrode spacing in
the cylindrical section, and adjustable axial electrode position (tip
distance).
Optically pure D- and L-amino acids, tryptophan (Trp), arginine
(Arg), phenylalanine (Phe), methionine (Met), glutamine (Gln),
asparagine (Asn), isoleucine (Ile), valine (Val), proline (Pro), and
lysine (Lys), were obtained from SigmaAldrich. NiCl26H2O,
CuCl25H2O, MgAc24H2O, and ZnAc22H2O were obtained from
SigmaAldrich. Methanol (LC-grade) was obtained from VWR Intl.
(Poole, England). Water was taken from a Millipore water supply
(Millipore, Billerica MA; Solna, Sweden).
Stock solutions of the amino acids (10 mM) were prepared in
50:50 methanol/water. Stock solutions of the metal salts (25 mM)
were prepared in water.
For the optimization experiments, samples were prepared
containing 0.033 mM metal ions, 0.133 mM reference amino acid,
and 0.067 mM concentrations each of the D and L forms of the
analyte. Samples were infused at 0.5 L/min with a syringe pump
into the electrospray ion source.
For the screening experiments, samples were prepared containing 0.133 mM reference amino acid and 0.067 mM concentrations of each of the D and L forms of the analyte. Samples were
supplied by the autosampler at 10 L/min. Metal solutions at 2.5
mM were delivered by the syringe pump at 8 L/h. The sample
and metal solution flows were combined in a T-junction, resulting
in the same molar ratio of metal, analyte, and reference compound
as for the optimization experiments. The combined flow was split
1:10 to result in a flow rate of 1 L/min at the electrospray ion
source.
For the confirmation experiments, samples were prepared
containing 0.033 mM metal ions, 0.133 mM reference amino acid,
(20) Guevremont, R.; Thekkadath, G.; Hilton, C. K. J. Am. Soc. Mass Spectrom.
2005, 16, 948-956.

and 0.067 mM of either the D or L form of the analyte, or both.

Samples were delivered at 10 L/min from the autosampler, and
the flow was split 1:10 to result in a flow rate of 1 L/min at the
electrospray ion source.
In electrospray ionization, a variety of cluster ions are formed
from these solutions. After ions have passed the FAIMS interface,
clusters with m/z corresponding to the general formula [MII(LRef)2(D/L-A)-H]+ (where MII is a divalent metal ion, L-Ref is a chiral
reference compound in its L form, and A is the analyte) were
subjected to tandem mass spectrometry.
Although the method presented herein is not dependent on
the use of tandem mass spectrometry, the mass spectrometer was
operated in MS/MS mode unless otherwise specified, in order to
confirm the molecular identity of peaks in the CV spectrum. In
certain cases, the branching ratio of fragmentation by loss of either
analyte or reference was used in order to confirm enantiomeric
identity according to the kinetic method.19
We evaluated peak resolution in FAIMS with two different
measures (Figure 2; see also Results and Discussion below). The
discrimination factor Rdf is defined by

Rdf ) 100% (hpeak - hvalley)/hpeak

(1)

where hpeak is the signal intensity of the smaller peak and hvalley is
the signal intensity at the position of the valley between the two
peaks.21 Thus, Rdf ) 100% means perfect (baseline) separation.
The peak overlap factor Rof of a peak 1 is defined as

Rof,1 ) Itail,2/Ipeak,1

(2)

where the Ipeak,1 is the peak signal intensity of peak 1 and Itail,2 is
the signal intensity of the tail of peak 2 at the CV position of peak
1. Two peaks, peak 1 and peak 2, have two overlap factors; we
define the minimum overlap factor Rmof as being the smaller value
of Rof,1 and Rof,2:

Rmof ) min(Rof,1, Rof,2)

(3)

We express Rof and Rmof as percent values. Thus, Rmof ) 0%

indicates perfect separation.
Throughout this work, all Rdf and Rmof values calculated refer
to cases where equal concentrations of both analyte enantiomers
were used.
For all calculations of resolution of peaks appearing in the CV
spectrum, the intensities of all fragments appearing at significant
intensities, which could be assigned to the loss of one analyte
molecule or one reference molecule from the parent cluster ion
of interest, were added. In certain cases where the collision energy
was too low to induce fragmentation, the intensity of the parent
ion was used instead.
RESULTS AND DISCUSSION
The aim of this study was to investigate whether enantiomers
can be physically separated by FAIMS, when they are ionized as
metal-bound trimeric diastereomeric clusters with enantiopure
chiral reference compounds. Six pairs of amino acid enantiomers
(21) El Fallah, M. Z.; Martin, M. Chromatographia 1987, 24, 115-122.

Figure 3. ESI-MS/MS measurements of (a) [NiII(L-Asn)2(L-Trp)-H]+

and (b) [NiII(L-Asn)2(D-Trp)-H]+ showing different branching ratios.

were chosen as model compounds to address this question. We

now present our results in three stages: results of optimization,
demonstration of separation in several cases, and finally discussion.
I. Optimization. As a first step, the relevant experimental
parameters for the separation were identified and optimized. In
order to generally optimize the FAIMS experimental parameters,
one pair of cluster ions was chosen for comparative purposes:
[NiII(L-Asn)2(L-Trp)-H]+ and [NiII(L-Asn)2(D-Trp)-H]+. In addition
to being separable by FAIMS, these clusters are known to display
different branching ratios (i.e., different relative intensities of the
fragments) in MS/MS experiments22 (Figure 3). The experimentally determined branching ratios R were RL ) 0.136 and RD )
0.905, resulting in a chiral selectivity Rchiral ) RD/RL ) 6.65. This
is in good agreement with previously published results,22 where
the respective values were RL ) 0.140 and RD ) 1.10, with Rchiral
) 7.86. Thus, the FAIMS separation can be confirmed by MS/
MS studies with the kinetic method.
FAIMS factors that were investigated for optimization of
performance in terms of signal intensity and resolution along the
CV axis included carrier gas composition, longitudinal position
of the inner electrode relative to the output orifice (referred to as
the tip distance), and makeup gas flow. The choice of drift gas
composition is important for obtaining a successful separation of
ions in many cases.15,23 Based on previous work,24 the evaluated
gas compositions were nitrogen/helium mixtures containing
0-50% helium. Recent work20 has investigated the dependence
of the resolving power of domed FAIMS devices on the tip
distance, indicating that this is an important factor in optimization.
Finally, the makeup gas decreases the gas flow rate through the
FAIMS interface; prolonged ion residence time of ions in FAIMS
can in some cases lead to improved resolution, but too high
makeup gas flow can hinder entrance of ions into the FAIMS
interface. Other factors, such as ambient pressure, FAIMS
interface and carrier gas temperatures, and interface design
features, have been disregarded in the experiments presented
herein, because they were not accessible as experimental variables
due to hardware limitations.
The influences of tip distance and gas composition on ion
throughput and separation have been evaluated in a multivariate
(22) Zhang, D.; Tao, W. A.; Cooks, R. G. Int. J. Mass Spectrom. 2001, 204, 159169.
(23) Ells, B.; Barnett, D. A.; Purves, R. W.; Guevremont, R. Anal. Chem. 2000,
72, 4555-4559.
(24) Barnett, D. A.; Ells, B.; Guevremont, R.; Purves, R. W. J. Am. Soc. Mass
Spectrom. 2002, 13, 1282-1291.

Analytical Chemistry, Vol. 79, No. 7, April 1, 2007

2853

Figure 5. Response surface of peak resolution Rdf of [NiII(L-Asn)2(L-Trp)-H]+ and [NiII(L-Asn)2(D-Trp)-H]+ as a function of FAIMS tip
distance and carrier gas composition.

Figure 4. Dependence of peak resolution and intensities of [NiII(LAsn)2(L-Trp)-H]+ (left peak) and [NiII(L-Asn)2(D-Trp)-H]+ (right peak)
on FAIMS tip distance and carrier gas composition. Visually, acceptable signal intensities and peak separations are achieved in the area
around 40% helium/2.4 and 2.6-mm tip distance.

way in the form of a full factorial design. Finally, the makeup gas
flow rate was optimized manually in order to find out whether
additional improvements in resolution could be obtained without
loss of ion transmission.
Figure 4 shows results of studying FAIMS resolution as a
function of tip distance and gas composition, using a solution of
D- and L-Trp, L-Asn, and Ni, varying the tip distance between 2.0
and 3.0 mm in increments of 0.2 mm, and varying the gas
composition between 0 and 50% helium in nitrogen in increments
of 10%. In those cases where two peaks appeared in the CV
spectra, the left peak was associated with the presence of [NiII(L-Asn)2(L-Trp)-H]+, while the right peak was associated with the
presence of [NiII(L-Asn)2(D-Trp)-H]+.
Of note is the observation that the relative intensities of the
peaks associated with D-Trp and L-Trp in Figure 4 are strongly
dependent on the carrier gas composition. With the data available
from the experiments presented herein, it is not possible to
understand this phenomenon. However, unlike in, for example,
liquid chromatographic techniques in which normally nothing is
lost and everything sooner or later elutes from a column, the
effectiveness of ion transmission in FAIMS is dependent on
favorable ion focusing conditions, which in turn depend on the
high-to-low-field mobility ratio function. FAIMS can thus favor a
certain ion species transmission over that of another ion species,
by providing different favorable focusing conditions for certain
ions under certain experimental conditions.
There is still no widely accepted measure for the resolution of
two peaks in a FAIMS CV scan, largely due to the fact that the
2854 Analytical Chemistry, Vol. 79, No. 7, April 1, 2007

peaks, unlike in chromatography, do not generally have Gaussian

shape.20,25 Many peaks observed in the CV scans displayed shapes
that visually could be regarded as being roughly Gaussian;
however, fitting of the experimental data to Gaussian curves
showed that the wings (tails) of these peaks were consistently of
higher intensity than expected for a Gaussian model. As the tails
are important for the calculation of peak overlap, we decided not
to use chromatographic models for peak resolution. Instead, we
chose the discrimination factor21 (Rdf) to evaluate peak resolution.
The discrimination factor does not make any assumptions on the
peak shapes, and it is readily accessible from the experimental
data. However, all information on the peak separation that may
be associated with the peak shape is disregarded; the discrimination factor must therefore be regarded as a rather crude measure
of peak resolution.
Figure 5 shows the response surface of the peak resolution as
a function of tip distance and gas composition. The response
surface is somewhat noisy, due to the fact that only a few data
points comprise the CV spectra of Figure 4. Nevertheless, it is
clearly visible that, within the experimental domain, good separation is achieved at tip distance values around 2.6 mm and 40% He
in N2 carrier gas composition. Ion transmission has been disregarded here, although it was affected as well, as can be seen from
the peak heights in Figure 4.
The optimization has been continued on a finer level in much
the same way, varying the tip distance between 2.5 and 2.7 mm
at 0.1-mm increments and varying the carrier gas composition
between 20 and 50% helium in nitrogen. The results, presented
in Figure 6, allowed selection of tip distance 2.5 mm and carrier
gas composition 40% He/60% N2 as the optimum; separation as
well as ion transmission have been weighed in making this
decision.
Figure 7 shows the effect of makeup gas. A small flow of
makeup gas (0.2 L/min) resulted in a slightly enhanced resolution,
while a stronger makeup gas flow had a strong detrimental effect
on the ion transmission.
CV spectra of the separation of the two complexes, as well as
CV spectra of the individual complexes, using the parameters
determined above, are presented in Figure 8. Though lacking
(25) Guevremont, R.; Purves, R. J. Am. Soc. Mass Spectrom. 2005, 16, 349362.

Figure 6. Dependence of peak resolution and intensities of [NiII(LAsn)2(L-Trp)-H]+ (left peak) and [NiII(L-Asn)2(D-Trp)-H]+ (right peak)
on FAIMS tip distance and carrier gas composition, optimizing at a
finer level than in Figure 4.
Figure 8. Optimized FAIMS separation of L-Trp and D-Trp as
trimeric cluster ions [NiII(L-Asn)2(D/L-Trp)-H]+, with DV ) -4000 V,
carrier gas composition 60% N2/40% He, FAIMS tip distance 2.5 mm,
and makeup gas 0.2 L/min N2. (a) Mixture of L-and D-Trp; (b) L-Trp
individually; (c) D-Trp individually.
Table 1. Successful Separationsa
Figure 7. Dependence of peak resolution and intensities of [NiII(LAsn)2(L-Trp)-H]+ (left peak) and [NiII(L-Asn)2(D-Trp)-H]+ (right peak)
on makeup gas flow.

predictive knowledge about the behavior of these complexes in

FAIMS, it appears reasonable to employ the experimental settings
determined above in other cases as well.
II. Separation of Various Enantiomeric Amino Acids. For
all six pairs of enantiomeric amino acids, all combinations of nine
reference compounds and four metals were screened for their
ability to form pairs of [MII(ref)2(D/L-analyte)-H]+ complexes that
are separable in FAIMS, using the above optimized FAIMS
parameters. In case the screening suggested separation, by
resulting in two distinct peaks in the CV spectrum that have a
discrimination factor Rdf of more than 50%, confirmation experiments were performed by infusing each of the enantiomers
separately with the respective combination of metal ion and
reference substance. Successful separations according to this
screening and confirmation procedure are presented in Table 1.
We observed that in about half of the cases in which two or more
peaks appeared in a CV scan of a solution containing both of the
enantiomers, it turned out that all peaks had contributions from
both enantiomers. In these cases, not included in Table 1, no chiral
separation occurred. It should also be mentioned that, of the 23
successful FAIMS separations, we found in only 4 cases a
substantial distinction between D- and L-analyte using the kinetic
method. These four cases were [(NiII(L-Asn)2(D/L-Trp)-H]+ Rchiral
) 6.65, compared to Rchiral ) 7.86;22 [CuII(L-Pro)2(D/L-Phe)-H]+
Rchiral ) 6.43, compared to 7.4;26 [(ZnII(L-Gln)2(D/L-Trp)-H]+ Rchiral
) 4.05; and [(CuII(L-Gln)2(D/L-Trp)-H]+ Rchiral ) 6.21, compared
to Rchiral ) 5.7 or 6.8, depending on collision energy.26
(26) Tao, W. A.; Zhang, D.; Nikolaev, E. N.; Cooks, R. G. J. Am. Chem. Soc. 2000,
122, 10598-10609.

m/z
fragments used
reference
metal
analyte
(D/L)- substance (L-) (M2+) complex ion for evaluation
Trp

Pro
Phe

Val
Arg

Lys

Gln
Pro
Arg
Asn
Gln
Lys
Asn
Gln
Val
Pro
Pro
Lys
Pro
Gln
Ile
Trp
Trp
Met
Val
Gln
Lys
Ile
Val

Zn
Zn
Mg
Ni
Cu
Cu
Zn
Ni
Mg
Mg
Ni
Ni
Cu
Cu
Mg
Cu
Mg
Ni
Ni
Cu
Cu
Cu
Cu

559
497
575
525
558
558
443
446
422
418
452
514
457
519
402
587
605
529
465
528
528
470
443

355 + 413
382
371
321 + 393
354 + 412
354
310 + 327
349
422
418
287 + 337
349
292 + 342
354
402
470
401 + 464
380
348
382
382
339
325

Rmof
(%)b
0.04*
0.91
0.24*
0.37
0.73
0.91*
0.68
1.51
2.16*
2.24*
0.85
3.08*
0.44
0.58*
1.29*
0.19
3.23
1.39
4.65*
4.53*
0.92*
5.24
3.88

a Note that FAIMS settings were optimized for the case [NiII(LAsn)2(D/L-Trp)-H]+. The factor Rmof characterizing peak overlap can
thus probably be improved (decreased) in all other cases by optimizing
the individual separations with respect to tip distance, gas composition,
and makeup gas. b These / cases involve multiple peaks, in such a
way that some peaks appear at the same CV for both enantiomers and
others do not. See the examples in Figures 10 and 11.

One of the more interesting applications of the method

described herein is the sensitive analysis of one enantiomer in
the presence of a large excess of the other enantiomer. After
having developed a successful separation method, it would not
be necessary to perform CV scans in order to do such analyses;
rather, it would be sufficient to monitor the signal intensities at
Analytical Chemistry, Vol. 79, No. 7, April 1, 2007

2855

two distinct CVs only, namely, the two CVs corresponding to the
peak maximums of the two chirality-linked diastereomers. It is
therefore of interest to evaluate what contribution the tail of one
peak makes to the intensity of the main portion of the other peak
(Figure 2). If this contribution is negligible, then the separation
may be considered as perfect for the purposes of measuring
enantiomeric excess.
Obviously, the simplest and ideal case is a full separation
between two peaks in the CV spectrum (Rof,1 ) Rof,2 ) 0%), each
one originating from a complex ion containing one and only one
analyte enantiomer. While we found some cases that are quite
close to this situation, e.g., the separation of D/L-valine with copper
and tryptophan as reference compound, there are other cases that
are not as favorable. One of the best cases we found is shown in
Figure 8b and c. If a sensitive analysis of an L-Trp contamination
in a sample of D-Trp is required, it is beneficial if Itail,D, representing
the signal intensity of the tail of the peak related to D-Trp at the
CV of the peak related to L-Trp, be as small as possible, in order
not to obscure very small signals that could arise from a minority
component L-Trp. In fact, a ratio Rof,L ) Itail,D/Ipeak,L, calculated from
separate measurements of L-Trp and D-Trp as in Figure 8b and c,
could serve as a measure for sensitivity of a specific separation
method. In this case, the ratio 0.37% is quite low, making a
sensitive analysis possible.
In the opposite case, when a sensitive analysis of a D-Trp
contamination in a sample of L-Trp is required, the separation is
somewhat different: The ratio Rof,D ) Itail,L/Ipeak,D is quite large,
because the main peak in Figure 8b has a smaller satellite peak
at CV ) -7.8V, which is close to the CV ) -7.6V of the main
peak in Figure 8c (see also the discussion below). A small amount
of D-Trp in the presence of a large excess of L-Trp would thus
probably be more difficult to measure using this method. However,
if the reference compound L-Asn is exchanged for its enantiomer
D-Asn, the peaks of Figure 8 should swap their positions, so that
a sensitive analysis of D-Trp would then be possible. This peak
position interchange is expected, because, for example, the
complexes [NiII(L-Asn)2(L-Trp)-H]+ and [NiII(D-Asn)2(D-Trp)-H]+
are themselves enantiomers and thus should behave identically
in FAIMS. This has been confirmed experimentally (data not
shown).
As becomes clear from these considerations, it is the smaller
of the ratios Rof,L and Rof,D that determines the quality of this
separation. We report this value as Rmof in Table 1. Compared to
Rdf, Rmof is more suitable for characterizing a separation, because
the broad flanks of each peak (peak tailing) and the potential
contribution to the other peak are considered. Compared to
chromatographic measures of resolution, Rmof has the advantage
that calculating it does involve making any assumptions about the
peak shape. Rmof should be regarded as a purity of separation
rather than as a statement about our ability to quantitate. It implies
the existence of a background or baseline for quantification; the
noise with which this baseline is afflicted is really what will
determine the ultimate sensitivity of a quantification method.
However, Rmof is not as easily accessible as either Rdf or
chromatographic resolution, because it cannot be determined from
a single experiment. Also, it is susceptible to contaminations of
each pure analyte enantiomer with the other enantiomer and thus
requires very clean substances.
2856 Analytical Chemistry, Vol. 79, No. 7, April 1, 2007

Figure 9. Separation of D/L-valine as [CuII(L-Trp)2(D/L-Val)-H]+. (a)

D-Val; (b) L-Val.

Figure 10. Separation of D/L-arginine as [CuII(L-Gln)2(D/L-Arg)-H]+.

(a) D-Arg; (b) L-Arg.

Figure 9 shows a case where only one peak related to the

presence of each enantiomer appears in the CV spectrum. Here,
in principle, it is directly possible to sensitively measure either
analyte enantiomer in the presence of a large excess of the other.
As has already been shown in Figure 8, it is possible that a
single type of cluster gives rise to more than one peak in the CV
spectrum. Another such example is shown in Figure 10. The
complex [CuII(L-Gln)2(D-Arg)-H]+ seems to appear in two different
forms that are separable by FAIMS. This phenomenon is referred
to as multiple peaks throughout this work. Similar behavior was
observed in more than 10 cases (*s in Table 1). In this case, we
could measure a small concentration of D-Arg in a large excess
of L-Arg; measuring a small concentration of L-Arg in a large excess
of D-Arg would require exchanging the reference compound for
its enantiomer.
An even more complex situation can be seen in Figure 11.
Here, each enantiomer gives rise to two distinct peaks in the CV
spectrum, one pair appearing at the same CV for the two
enantiomers and the other pair being clearly separated. The
arrangement of the peaks indicates that, despite the complexity
of the CV spectra, each enantiomer could be determined in the
presence of a large amount of the other.
Figure 12 shows results from a quantification of L-Trp in the
presence of a large excess of D-Trp. The measurements for the
calibration curve of Figure 12 have been performed by monitoring
the CV values -5.9 and -7.6 V, corresponding to peak maximums

Figure 11. Separation of D/L-valine as [MgII(L-Ile)2(D/L-Val)-H]+. (a)

D-Val; (b) L-Val.

Figure 12. Calibration curve for L-Trp in presence of a large excess

of D-Trp. Separation as [NiII(L-Asn)2(D/L-Trp)-H]+ (see Figure 8). r2 )
0.9989. Error bars show ( estimated standard deviation of two
experiments with one replicate each, performed on different days,
with samples freshly prepared from stock solutions each day. ICV,1
and ICV,2 denote signal intensities at CV ) -5.9 and -7.6 V,
respectively. These CV values correspond to peak maximums in
Figure 8.

of [NiII(L-Asn)2(L-Trp)-H]+ and [NiII(L-Asn)2(D-Trp)-H]+ (Figure 8),

rather than by performing CV scans. The response for each
sample is the ratio of the signal intensities at -5.9 and -7.6 V.
The offset of the calibration curve is due to imperfect separation
of the two complexes. It should be noted that these quantitative
results are of preliminary nature, as the effect of factors such as
sample pH, solvent content, ionic strength, or changed relative
concentrations of analyte, metal, and reference compound have
not yet been addressed. Also, repeatability has not yet been
investigated in a rigorous way. Still, these results suggest that
our approach to separate enantiomers has potential for the
determination of enantiomeric excess, even in cases where there
is a large excess of one enantiomer present. In this specific case,
it appears that less than 0.1% L-Trp of total Trp could be detected.
III. Discussion. All six pairs of amino acid enantiomers have
been separated successfully using the experimental settings
determined above (see Table 1). Of a total of 216 different
combinations of analyte, reference compound, and metal ion, 23
have been successfully separated, averaging just over 10% successful separation attempts. Our current experimental setup allows
for four CV scans 0 to -16 V, each taking 8 min. The remaining
time is needed for liquid transport, loading MS method, and
instrument synchronization. Although we do not want to generalize our findings on the basis of our limited number of analytes

from only one compound group, this would suggest that on

average a pair of unknown enantiomers could be separated after
10 attempts using different combinations of metal and reference
compound, taking 2.5 h. In an instrumental system more
integrated than ours, probably this time could be substantially
shortened. This compares favorably to widespread screening
approaches for suitable HPLC columns and conditions, typically
taking 24 h. Also worth noting is the fact that the number of
possible combinations of chiral reference compounds and metals
is very high and is by no means limited to the metals and reference
compounds used in the present work.
Though our data set is limited, it appears that analytes that
comprise an aromatic system (tryptophan, phenylalanine) were
more easily separated than aliphatic analytes. However, the same
aromatic compounds were hardly represented among the reference compounds of successful separations. Here, L-Gln and L-Pro
were represented 5 and 4 times out of 23 separations, respectively.
Copper appeared to be the most successful metal ion with 9
separations, compared to nickel (6), magnesium (5), and zinc (3).
It can also be seen that, in the few cases zinc was successful, it
yielded very good separations. The only combinations of metal
and reference compounds that appeared several times as successful were Cu and L-Gln (three times) and Cu and L-Lys (twice).
We would like to stress that these observations are as yet of only
empirical nature and do not provide rules of general validity.
We found that only about one-third of cases where we obtained
good separation by FAIMS were also amenable to treatment by
the kinetic method, owing to lack of suitable fragment branching
in about two-thirds of the cases. This result suggests that these
two methods generally probe different complex ion properties
(which was expected) and that these two methods are complementary. It should be noted that both methods share the need
for well-defined standard solutions of the relevant enantiomers,
which in some cases could be seen as disadvantageous.
For this and other cluster types,26,27 results of the kinetic
method fragmentation experiments are interpreted in terms of
each diastereomer having one structure of partly but not completely known character. From the results presented herein, it
has become clear that many clusters of the form [MII(L-Ref)2(A)H]+, where A is present in only one enantiomeric form, are
resolved into at least two species transmitted at different CVs
(Figures 8, 10, and 11). It is tempting to attribute these features
to the presence of different conformers or constitutional isomers
of the complex ions. This could possibly include complex ions
with (a) different binding sites to the metal ion, (b) different
binding sites between analyte and reference molecule or between
reference molecules, (c) different sites of deprotonation, or (d)
identical constitution but sterically hindered intramolecular rotation (conformers). While FAIMS separation of different conformations of protein ions has been reported before,28-30 this would be,
to our knowledge, the first time that FAIMS separation of different
(27) Augusti, D. V.; Carazza, F.; Augusti, R.; Tao, W. A.; Cooks, R. G. Anal. Chem.
2002, 74, 3458-3462.
(28) Borysik, A. J. H.; Read, P.; Little, D. R.; Bateman, R. H.; Radford, S. E.;
Ashcroft, A. E. Rapid Commun. Mass Spectrom. 2004, 18, 2229-2234.
(29) Purves, R. W.; Barnett, D. A.; Ells, B.; Guevremont, R. J. Am. Soc. Mass
Spectrom. 2000, 11, 738-745.
(30) Robinson, E. W.; Williams, E. R. J. Am. Soc. Mass Spectrom. 2005, 16, 14271437.

Analytical Chemistry, Vol. 79, No. 7, April 1, 2007

2857

conformers or constitutional isomers of such small ions could be

suggested.
On the other hand, FAIMS and the kinetic method are carried
out at completely different parts of the experimental chain. FAIMS
is carried out at atmospheric pressure just after electrospray
ionization, whereas the kinetic method is carried out after ions
have passed through the collisional desolvation region of the MS
and into the mass analyzer. This implies the possibility that
multiple peaks in the CV spectrum may represent different
molecular species, such as complexes with solvent adducts or
complex dimers. In this context, it may be worth pointing out
that our electrospray-FAIMS interface has no provision for heated
dry gas. After transmission through FAIMS, these species would
be subjected to declustering or other gas-phase reactions and then
detected as identical ions by the mass spectrometer. One
experimental observation that would point toward this interpretation is the fact that, out of 19 cases where multiple peaks were
observed and where the kinetic method was useable, none showed
a distinction of these multiple peaks by the kinetic method. But
from the experimental data available, the interesting question of
what different FAIMS-separated species actually represent
remains unsolved. Use of enantiomerically pure standard substances still allows the method to be used for qualitative and
quantitative purposes, however.
It appears that several issues have to be addressed before this
method can be routinely used for chiral separation and quantification of enantiomers. However, the preliminary quantitation data
we show in Figure 12 seem to be encouraging. The calibration
was performed by mixing the pure enantiomers of tryptophan in
known concentrations. This approach requires that pure enantiomers, or at least one pure enantiomer and the racemate, have
to be available; this could in some cases be regarded as
disadvantageous and is a limitation of our method. However, in
principle, the calibration curve can be calculated from the relative
heights (or response factor) of the two peaks of a racemate. This
approach would require that each of the two peaks of a mixture of enantiomers be associated with the presence of one and

only one enantiomer. We thus believe that it is of relevance for

the ultimate value of our method, if the detailed identity of the
multiple peaks in the CV spectrum and their origin can be
revealed, and preferably if the formation of multiple peaks could
be avoided.

(31) Eiceman, G. A.; Krylov, E.; Krylova, N.; Douglas, K. M.; Porter, L. L.;
Nazarov, E. G.; Miller, R. A. Int. J. Ion Mobility Spectrom. 2002, 5, 1-6.
(32) Krylov, E.; Nazarov, E. G.; Miller, R. A.; Tadjikov, B.; Eiceman, G. A. J.
Phys. Chem. A 2002, 106, 5437-5444.

Received for review October 4, 2006. Accepted January 5,

2007.

2858

Analytical Chemistry, Vol. 79, No. 7, April 1, 2007

CONCLUSIONS
We have successfully demonstrated the ability to separate six
pairs of amino acid enantiomer complexes by FAIMS-MS.
It seems to be difficult to draw any conclusions as to which
combinations of metal and reference compound should be tested
first in case an unknown analyte should be separated, from the
available experimental data. It can be concluded, though, that the
screening approach described herein is a promising one even for
other analytes. While computational tools exist for calculating the
gas-phase structure of diastereomeric complexes of the kind we
used in our experiments, it is still unclear how these structures
correlate to high electric field mobility. Although progress has
been made in this subject,31,32 the prediction of FAIMS behavior
from the ionic structure or other characteristics is still not
generally possible.
Upcoming improvements in FAIMS instrumentation, such as
temperature control and increased dispersion voltage, would make
it even easier to find good separations or else improving existent
separations.
We believe it is also of interest to apply the method described
herein for analytes other than amino acids. Work on separation
and quantitation of drug compound enantiomers is in progress.
ACKNOWLEDGMENT
The authors acknowledge the loan of a FAIMS beta unit from
Ionalytics Corporation, now part of Thermo Electron Corporation.
Acknowledged are also Crafoordska Stiftelse, Carl Tryggers
Stiftelse. and Stiftelsen J. Gust. Richerts Minne, and AstraZeneca,
Lund/Sweden, for providing funding for this work, the Faculty of
Sciences at Lund University for providing funding for a Ph.D.
position. We also thank Ulrika Nilsson at the Department of
Analytical Chemistry at Stockholm University for providing the
mass spectrometric instrumentation and laboratory facilities for
this work.

AC0618627