letters to nature

z 0s is the spread of the ground-state wavefunction for one ion in the stretch mode and f L is ..............................................................
the phase of the driving field. Evaluation of the integrals in equation (8) yields:
a#" ðtÞ ¼iQD =d½expð2idtÞ 2 1Š expðifL Þ An electronic Mach–Zehnder
ð12Þ
F#" ðtÞ ¼ðQD =dÞ2 ½sinðdtÞ 2 dtŠ interferometer
For the p-phase gate the maximum excursion in phase space is ja"# max j ¼ 1: The
normalized fluorescence is equal to: Yang Ji, Yunchul Chung, D. Sprinzak, M. Heiblum, D. Mahalu
& Hadas Shtrikman
Snorm ðtÞ ¼ P## ðtÞ þ 1/2½P#" ðtÞ þ P"# ðtފ ¼ 1/2½1 þ expð2ja"# ðtÞj2 /2Þ cosðF"# ðtÞފ ð13Þ
For the fit in Fig. 3 we allowed for an additional exponential decay: Braun Center for Submicron Research, Department of Condensed Matter Physics,
2
Sf it ðtÞ ¼ 1/2½1 þ expð2t=t0 Þ expð2ja"# ðtÞj /2Þ cosðF"# ðtÞފ ð14Þ Weizmann Institute of Science, Rehovot 76100, Israel
.............................................................................................................................................................................
where t 0 is a phenomenological decay constant mimicking decoherence effects. Double-slit electron interferometers fabricated in high mobility
two-dimensional electron gases are powerful tools for studying
Gate speed considerations coherent wave-like phenomena in mesoscopic systems1–6. How-
The gate speed in our experiments was limited by the intensity of the displacement beams.
The observed ratio of gate time 2p/d to stretch oscillation period 2pq S was 238. The largest
ever, they suffer from low visibility of the interference patterns
contribution to gate infidelity from off-resonant excitations is that due to excitation of the due to the many channels present in each slit, and from poor
centre-of-mass (COM) mode (as opposed to excitation of the internal-state carrier sensitivity to small currents due to their open geometry3–5,7.
transition in the Cirac–Zoller and Sørensen–Mølmer gates). The ratio of excitation Moreover, these interferometers do not function in high mag-
amplitude of the stretch mode to that of the COM mode scales as d COM/d where
d COM / q S is the detuning from the COM mode. Therefore if the required fidelity
netic fields—such as those required to enter the quantum Hall
restricts the COM mode excitation to be below a certain size, then the gate rate (/ d) will effect regime8—as the field destroys the symmetry between left
scale linearly with q S. For our experimental parameters, we estimate the contribution to and right slits. Here we report the fabrication and operation of a
infidelity from off-resonant excitation of the COM mode to be below 1024. In a more single-channel, two-path electron interferometer that functions
refined scenario, we could intentionally excite both the COM and the stretch mode. For
in a high magnetic field. This device is the first electronic
suitable parameters the COM and stretch-mode amplitudes can return to their initial
values even if the gate rate exceeds the trap frequency. analogue of the optical Mach–Zehnder interferometer9, and
opens the way to measuring interference of quasiparticles with
Received 18 December 2002; accepted 7 February 2003; doi:10.1038/nature01492.
fractional charges. On the basis of measurements of single edge
1. Cirac, J. I. & Zoller, P. Quantum computations with cold trapped ions. Phys. Rev. Lett. 74, 4091–4094
(1995).
state and closed geometry transport in the quantum Hall effect
2. Sørensen, A. & Mølmer, K. Quantum computation with ions in thermal motion. Phys. Rev. Lett. 82, regime, we find that the interferometer is highly sensitive and
1971–1974 (1999). exhibits very high visibility (62%). However, the interference
3. Sackett, C. A. et al. Experimental entanglement of four particles. Nature 404, 256–259 (2000). pattern decays precipitously with increasing electron tempera-
4. Wineland, D. J. et al. Experimental issues in coherent quantum-state manipulation of trapped atomic
ions. J. Res. Natl Inst. Stand. Technol. 103, 259–328 (1998).
ture or energy. Although the origin of this dephasing is unclear,
5. Kielpinski, D., Monroe, C. & Wineland, D. J. Architecture for a large-scale ion-trap quantum we show, via shot-noise measurements, that it is not a decoher-
computer. Nature 417, 709–711 (2002). ence process that results from inelastic scattering events.
6. Monroe, C., Meekhof, D. M., King, B. E., Itano, W. M. & Wineland, D. J. Demonstration of a
Direct phase measurements of electrons, customarily done in
fundamental quantum logic gate. Phys. Rev. Lett. 75, 4714–4717 (1995).
7. Roos, Ch. et al. Quantum state engineering on an optical transition and decoherence in a Paul trap. double-slit interferometers1–4, are difficult to perform under strong
Phys. Rev. Lett. 83, 4713–4716 (1999). magnetic fields. Electrons are diverted by the Lorentz force, perform
8. Sørensen, A. & Mølmer, K. Entanglement and quantum computation with ions in thermal motion. chiral skipping orbits, and prefer one slit to the other—thus break-
Phys. Rev. A 62, 02231 (2000).
9. DeVoe, R. G. Elliptical ion traps and trap arrays for quantum computation. Phys. Rev. A 58, 910–914
ing the symmetry of the interferometer. At the extreme quantum
(1998). limit (that is, in the quantum Hall effect, QHE, regime), the
10. Cirac, J. I. & Zoller, P. A scalable quantum computer with ions in an array of microtraps. Nature 404, skipping orbits quantize to quasi-one-dimensional-like states,
579–581 (2000). named chiral edge states. We have exploited the chiral motion of
11. Rowe, M. A. et al. Transport of quantum states and separation of ions in a dual rf ion trap. Quant. Inf.
Comput. 4, 257–271 (2002).
the electrons, and constructed an electronic analogue of the ubi-
12. Carruthers, P. & Nieto, M. M. Coherent states and the forced quantum oscillator. Am. J. Phys. 7, quitous optical Mach–Zehnder interferometer9 (Fig. 1a). A beam
537–544 (1965). splitter BS1 splits an incoming monochromatic light beam from
13. Walls, D. F. & Milburn, G. J. Quantum Optics (Springer, Berlin, 1994).
source S into two beams, which, after reflection by mirrors M1 and
14. Monroe, C., Meekhof, D. M., King, B. E. & Wineland, D. J. A “Schrödinger Cat” superposition state of
an atom. Science 272, 1131–1136 (1996). M2, recombine and interfere at BS2 to result in two outgoing beams
15. Myatt, C. J. et al. Decoherence of quantum superpositions through coupling to engineered reservoirs. (collected by detectors D1 and D2). When the phase along one of the
Nature 403, 269–273 (2000). paths varies, signals in both D1 and D2 oscillate out of phase, and as
16. Milburn, G. J., Schneider, S. & James, D. F. Ion trap quantum computing with warm ions. Fortschr.
Physik 48, 801–810 (2000).
no photons are being lost, the sum of both signals stays always equal
17. Wang, X., Sørensen, A. & Mølmer, K. Multibit gates for quantum computing. Phys. Rev. Lett. 86, to the input, S. In the electronic counterpart (Fig. 1b), quantum
3907–3910 (2001). point contacts (QPCs) function as beam splitters, and ohmic
18. Wineland, D. J. et al. Quantum information processing with trapped ions. Preprint quant-ph/0212079 contacts serve as detectors. A QPC is formed in the two-dimensional
available at khttp://arXiv.orgl (2002).
19. King, B. E. et al. Cooling the collective motion of trapped ions to initialize a quantum register. Phys.
electron gas (2DEG) by depositing a split metallic gate on the
Rev. Lett. 81, 1525–1528 (1998). surface of the semiconductor and biasing it negatively with respect
20. Rowe, M. A. et al. Experimental violation of a Bell’s inequality with efficient detection. Nature 409, to the 2DEG. The induced potential in the 2DEG creates a barrier
791–794 (2001).
under the gate bringing the two oppositely propagating edge
21. Steane, A. et al. Speed of ion trap quantum information processors. Phys. Rev. A 62, 042305
(2000). currents to the small opening in the barrier, thus allowing back-
22. Steane, A. Overhead and noise threshold of fault-tolerant quantum error correction. Preprint quant- scattering. As shown schematically in Fig. 1b, QPC1 splits the
ph/0207119 available at khttp://arXiv.orgl (2002). incoming edge current from S to two paths, a transmitted outer
path and a reflected inner path; both later recombine and interfere
Acknowledgements We thank J. Chiaverini, T. Schätz and A. Steane for comments on the
manuscript. This work was supported by the US National Security Agency (NSA), the Advanced
in QPC2, resulting in two edge currents (collected by D1 and D2).
Research and Development Activity (ARDA). This is a publication of a US government agency. The actual device (Fig. 1c) was fabricated in a high-mobility
2DEG embedded in a GaAs–AlGaAs heterojunction. A ring-shaped
Competing interests statement The authors declare that they have no competing financial mesa, 3 mm in width, was defined by plasma etching with ohmic
interests.
contacts (for S, D1 and D2) connected to the inner and outer edges
Correspondence and requests for materials should be addressed to D.J.W. of the ring. The inner contact, D2, and the two QPCs are connected
(e-mail: [email protected]). to outside sources via metallic films that hover above the surface of

NATURE | VOL 422 | 27 MARCH 2003 | www.nature.com/nature © 2003 Nature Publishing Group 415
letters to nature
the mesa, called ‘air bridges’. A phase difference J between the two
paths is introduced via the Aharonov–Bohm (AB) effect10,11, J ¼
2pBA=f0 ; with B the magnetic field, A the area enclosed by the two
paths (,45 mm2), and f0 ¼ 4:14 £ 10215 T m2 the flux quantum. A
few modulation gates, MG, are added above the outer path in order
to tune the phase J by changing the area A.
We briefly review the operation of the interferometer. At filling
factor 1 in the QHE regime, a single chiral edge state carries the
current. The interfering current, in turn, is proportional to the
transmission probability from source to drain, TSD. Neglecting
dephasing processes and having the transmission (reflection) ampli-
2 2
tude t i (r i) of the ith QPC fulfilling jr i j þ jt i j ¼ 1; then7 I D1 /
iJ 2 2 2
T SD1 ¼ j t 1 t 2 þ r 1 r 2 e j ¼ j t 1 t 2 j þ jr1 r 2 j þ 2jt 1 t 2 r 1 r 2 jcosJ and
2 2 2
I D2 / T SD2 ¼ jt 1 r 2 þ r 1 t 2 eiJ j ¼ jt 1 r 2 j þ jr 1 t 2 j 2 2jt 1 t 2 r 1 r 2 jcosJ;
where I D1 and I D2 are the currents in detectors D1 and D2,
respectively. Note that ideally the two currents oscillate out of
phase as function of J while T SD1 þ T SD2 ¼ 1: The visibility of the
oscillation is defined as: n ¼ ðI max 2 I min Þ=ðI max þ I min Þ, where I max
and I min are the maximum and minimum currents in one of the
detectors. For example,
p when QPC2 is tuned so that T 2 ¼ 0.5, the
2
Figure 1 The configuration and operation of an optical Mach–Zehnder interferometer, visibility is n ¼ 2 T 1 ð1 2 T 1 Þ, where jt i j ¼ T i .
and its realization with electrons. a, An optical Mach–Zehnder interferometer. D1 and D2 Measurements were done at filling factor 1 (magnetic field
are detectors, BS1 and BS2 are beam splitters, and M1 and M2 are mirrors. With 0 (p) ,5.5 T) and also at filling factor 2 with similar results. With a
phase difference between the two paths, D1 measures maximum (zero) signal and D2 refrigerator temperature of ,6 mK, the electron temperature was
zero (maximum) signal. The sum of the signals in both detectors is constant and equal determined by measuring the equilibrium noise12 to be ,20 mK.
to the input signal. b, The electronic Mach–Zehnder interferometer and the measurement High-sensitivity measurements of the interference pattern were
system. Edge states are formed in a high, perpendicular, magnetic field. The incoming conducted at ,1.4 MHz with a spectrum analyser. Current at
edge state from S is split by QPC1 (quantum point contact 1) to two paths; one moves D1 (or D2) was filtered and amplified in situ by an LC
along the inner edge, and the other along the outer edge, of the device. The two (inductance þ capacitance) circuit and a low-noise, purpose-built
paths meet again at QPC2, interfere, and result in two complementary currents in D1 and pre-amplifier, both placed near the sample and cooled to 1.5 K. A
in D2. By changing the contours of the outer edge state and thus the enclosed area standard lock-in technique, with a low-frequency signal (7 Hz,
between the two paths, the modulation gates (MGs) tune the phase difference between 10 mV r.m.s.), gave similar results, but the measurement lasted
the two paths via the Aharonov–Bohm effect. A high signal-to-noise-ratio measurement of much longer and was affected by the instability of the sample. At
the current in D1 is performed at 1.4 MHz with a cold LC resonant circuit as a band-pass 5.5 T, each flux quantum occupies an area of some 10215 m2 (some
filter followed by a cold, low-noise, preamplifier. c, Scanning electron micrograph of the 60,000 flux quanta thread the area A), so a minute fluctuation in the
device. A centrally located small ohmic contact (3 £ 3 mm2), serving as D2, is connected superconducting magnet’s current or in the area would smear the
to the outside circuit by a long, metallic, air bridge. Two smaller metallic air bridges interference signal. Two measurement methods were used. The first
bring the voltage to the inner gates of QPC1 and QPC2—both serve as beam splitters for relied on the unavoidable decay of the short-circuited current that
edge states. The five metallic gates (at the lower part of the figure) are MGs. circulates in the superconducting magnet, which is in the so-called

Figure 2 Interference pattern of electrons in a Mach–Zehnder interferometer and the voltage on a modulation gate, V MG (red plot), and as function of the magnetic field, B (blue
dependence on transmission. a, Two-dimensional colour plot of the current collected by plot)—along the cuts shown in a. The visibility of the interference is 0.62. c, The visibility
D1 as function of magnetic field and gate voltage at an electron temperature of ,20 mK. of the interference pattern (data points) as a function of the transmission probability T1 of
The magnet was set in its persistent current mode (B < 5.5 T at filling factor 1 in the bulk) QPC1 when QPC2 is set to T2 ¼ 0.5. Red dashed line is a fit to the experimental data with
p
with a decay rate of some 0.12 mT h21, hence time appears on the abscissa. The two visibility 2h T 1 ð1 2 T 1 Þ: The normalization coefficient h ¼ 0.6 accounts for possible
QPCs were both set to transmission T 1 ¼ T 2 ¼ 0:5: Red (blue) stands for high (low) decoherence and/or phase averaging.
current. b, The current (a.u., arbitrary units) collected by D1 plotted as function of the

416 © 2003 Nature Publishing Group NATURE | VOL 422 | 27 MARCH 2003 | www.nature.com/nature
 letters to nature
persistent current mode. In this mode, the magnetic field decays p 2c). It agrees well with the expected expression, n ¼
(Fig.
smoothly at a rate of ,0.12 mT h21 (,1 flux quantum every 2h T 1 ð1 2 T 1 Þ; with h < 0.6 a normalization factor that accounts
50 min). The second was via scanning the voltage on a modulation for dephasing (either due to phase averaging in the energy window
gate at a rate much faster than the decay rate of the magnetic field, of the electrons, or due to inelastic scattering processes). Moreover,
thus changing the area A, the enclosed flux, and consequently the the period of the oscillations, in time and in MG voltage, agrees well
AB phase. with one flux quantum being added (or subtracted) in the ring’s area.
We first test the ideality of the ohmic contacts and the validity of The time period is ,50 min, which is the time needed for one flux
the edge-states picture. For both QPCs open, a nearly ideal Hall quantum decay in the superconducting magnet, while the voltage
plateau was observed in I D1 while no current was measured in D2 period agrees approximately with that needed to deplete one electron
ðI D2 ¼ 0Þ: This confirms that current was confined to the outer edge (hence, one flux quantum for filling factor 1) under the MG gate.
with no backscattering across the 3-mm-wide mesa. We then Although the visibility is very high, it is still smaller than unity
pinched off QPC1 or QPC2, and found again a Hall plateau in (n < 0.6). An obvious reason for this is the finite energy spread of
I D2 with zero current in D1 (I D1 ¼ 0). This proved that the small the electrons at the edge (due to their finite temperature), and the
ohmic contact of D2 was ideal and fully absorbed the current. unavoidable dependence of the AB area on the energy (hence, the
Setting then both QPCs to T 1 ¼ T 2 < 1=2 and varying the magnetic AB phase)—leading to phase averaging (thermal smearing). Indeed,
field B (actually the time) or the area A (the voltage on a MG) led to the visibility was found to drop precipitously with increasing
a pronounced interference signal in D1 (or in D2) with visibility as temperature or applied voltage at S, as seen in Fig. 3. In this
high as 0.62 (Fig. 2). As the field decays linearly with time, and the example, a mere increase of the temperature to 100 mK (some
area (or electron density) varies in proportion to the gate voltage, ,9 meV) reduced the visibility from n < 0.53 to n < 0.01 (plotted
changing these parameters leads to the diagonal, straight, lines of in red in Fig. 3a). If indeed phase averaging is the cause of the
colour (of constant phase) seen in Fig. 2a. Figure 2b shows similar dephasing, it could, in principle, be eliminated with monoenergetic
data taken along two cuts (the dotted lines shown in Fig. 2a)—one electrons. A minute a.c. signal (,0.5 mV) at 1.4 MHz was added to a
for constant B and one for constant A. The cleanliness of the variable d.c. voltage Vdc and the synchronous a.c. part of the
interference pattern and the high visibility prove the nearly ideal interfering signal was measured at 20 mK. This signal leads to a
nature of the interferometer. differential visibility nd, resulting only from the electrons in an
In order to verify further the two-path nature of the interference, energy window ,0.5 meV around an energy eVdc. Surprisingly, as
the visibility was measured as function of T 1 for a constant T 2 ¼ 0.5 seen in Fig. 3a (plotted in blue), the energy-dependent differential
visibility at T ¼ 20 mK is similar to the temperature-dependent
visibility, with a relation between the scales eV dc < 4kB T: The
visibility (in colour scale) is plotted as function of both T and V dc
in Fig. 3b. The clear symmetry across the diagonal suggests that the
dephasing processes due to temperature and voltage are similar.
Unfortunately, this contradicts our previous assertion of phase
averaging taking place in a wide window of energy, and points at
decoherence, induced by inelastic scattering events, as the main
source of dephasing. In other words, for an increased temperature
or for high-energy monoenergetic electrons, empty states are being
created, allowing energy loss via scattering.
In order to test this hypothesis, current shot noise was measured.
Its spectral density is defined as the averaged square of the current

Figure 4 Shot-noise measurement (at filling factor 2) as function of T2 when the

transmission of QPC1 was set to T1 ¼ 0.5. A 30-mV d.c. voltage (under which the AB
interference pattern was quenched) was used to measure shot noise. The shot noise of
Figure 3 The dependence of the visibility of the interference pattern on temperature and the current collected by D1 is shown by the black dots (normalized to a maximum), while
applied voltage. a, Visibility as function of temperature at small excitation voltage for the two solid lines are the expected noise for k ¼ 0 and k ¼ 0.9, respectively, according
Vdc ¼ 0 (red plot), and as function of Vdc with a small a.c. voltage Vac superimposed on it to the simple model described in the text. The agreement with the simple model indicates
at electron temperature 20 mK (blue plot). Both QPCs were set to T 1 ¼ T 2 ¼ 0:5: b, A that the electrons are coherent even at the d.c. voltage where the interference pattern
two-dimensional colour plot of the visibility as function of temperature and applied d.c. fades away. Inset, the current at D1 as function of T2 at V dc ¼ 30 mV: As expected from
voltage. Red (blue) stands for high (low) visibility. the lack of the interference pattern, the current is independent of T2 (see text).

NATURE | VOL 422 | 27 MARCH 2003 | www.nature.com/nature © 2003 Nature Publishing Group 417
letters to nature
fluctuations per unit of frequency, S ¼ kði2 Þl=Df ; for stochastic ..............................................................
partitioning at zero temperature, S / eV dc T SD ð1 2 T SD Þ (ref. 13).
Introducing a phenomenological parameter k that accounts for African vegetation controlled by
decoherence ffi the interferometer with T 1 ¼ 1/2 and T SD ¼ 0:5 þ
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiin tropical sea surface temperatures
k T 2 ð1 2 T 2 Þ cosJ; we find that for complete phase averaging or
for a complete decoherence T SD ¼ 0.5, namely, a constant. On the
other hand, shot noise in D1 is SD1 / T SD1 ð1 2 T SD1 Þ ¼ 1=4 2
in the mid-Pleistocene period
k2 T 2 ð1 2 T 2 Þ cos2 J; with S D1 ¼ const. for k ¼ 0, but SD1 ¼ 1=4 2 Enno Schefuß*, Stefan Schouten, J. H. Fred Jansen
k2 T 2 ð1 2 T 2 Þ=2 for complete phase averaging (resulting from an & Jaap S. Sinninghe Damsté
integration of cos2 J in the range J ¼ 0……2p). Hence, noise is
expected to exhibit a parabolic dependence on T2 in a coherent Royal Netherlands Institute for Sea Research (NIOZ), PO Box 59, 1790 AB Den
system. Shot noise was measured12,13 with a relatively large Vdc Burg, The Netherlands
applied at S so that interference signal was quenched (negligible .............................................................................................................................................................................
visibility). The dependence of S on T2, shown in Fig. 4, followed the The dominant forcing factors for past large-scale changes in
above expression with k < 0.9, proving that phase averaging is vegetation are widely debated. Changes in the distribution of C4
indeed dominant while decoherence is negligibly small. plants—adapted to warm, dry conditions and low atmospheric
A single-particle model (that is, a non-interacting model) would CO2 concentrations1—have been attributed to marked changes in
lead to the following dependences of the visibility on energy: for environmental conditions, but the relative impacts of changes in
V ¼ 0 and finite T; n / bT=sinhðbTÞ; with b a constant; for finite V aridity, temperature2,3 and CO2 concentration4,5 are not well
but T ¼ 0, n / sin½ðe=2pÞVŠ=½ðe=2pÞVŠ; while the differential visi- understood. Here, we present a record of African C4 plant
bility at T ¼ 0 is expected to be voltage independent. Because the abundance between 1.2 and 0.45 million years ago, derived
experimental results contradict these predictions, we propose (with from compound-specific carbon isotope analyses of wind-trans-
no proof yet) two possible reasons for the dephasing. One might be ported terrigenous plant waxes. We find that large-scale changes
low-frequency noise (of, say, the 1/f type due to moving impurities), in African vegetation are linked closely to sea surface tempera-
which might be induced by a higher current, leading to fluctuation tures in the tropical Atlantic Ocean. We conclude that, in the mid-
in the area and consequently, phase smearing. The other could be Pleistocene, changes in atmospheric moisture content—driven by
related to the self-consistent potential contour at the edge. As it tropical sea surface temperature changes and the strength of the
depends on the local density of the electrons in the edge state14, African monsoon—controlled aridity on the African continent,
fluctuation in the density due to partitioning are expected to lead to and hence large-scale vegetation changes.
fluctuation in the AB area enclosed by the two paths and hence to Two main carbon fixation pathways of higher plant photosyn-
phase randomization. For example, for B < 5.5 T, a shift of the edge thesis, the Calvin–Benson (C3) and the Hatch–Slack (C4) cycles,
of only 1–2 Å suffices to add one flux quantum into the enclosed area. occur in natural ecosystems6. Nearly all trees, cold-season grasses
We believe that this electron interferometer might prove useful in and sedges use the C3 pathway, whereas C4 photosynthesis is found
future work on the interference of electrons. One possible area of in warm-season grasses and sedges7. Thus, C4 plants are found
research is the coherence and phase of fractionally charged quasi- predominantly in tropical savannahs, temperate grasslands and
particles in the fractional QHE regime15. A semideserts7. Most African grasslands are dominated currently by
C4 plant vegetation1. C4 plants use a CO2-concentrating mecha-
Received 1 October 2002; accepted 12 February 2003; doi:10.1038/nature01503.
nism, thereby outcompeting C3 plants at low atmospheric p CO2
1. Yacoby, A., Heiblum, M., Umansky, V., Shtrikman, H. & Mahalu, D. Unexpected periodicity in an
electronic double slit interference experiment. Phys. Rev. Lett. 73, 3149–3152 (1994).
(ref. 8) and causing them to be isotopically enriched in 13C (ref. 4).
2. Yacoby, A., Heiblum, M., Mahalu, D. & Shtrikman, H. Coherency and phase sensitive measurements At high p CO2, however, C3 plants will outcompete C4 plants owing to
in a quantum dot. Phys. Rev. Lett. 74, 4047–4050 (1994). the higher energy need of C4 plants during photosynthesis8. The
3. Schuster, R. et al. Phase measurement in a quantum dot via a double-slit interference measurement. crossover between C3 and C4 plants depends also on the daytime
Nature 385, 417–420 (1997).
4. Ji, Y., Heiblum, M., Sprinzak, D., Mahalu, D. & Shtrikman, H. Phase evolution in a Kondo-correlated
growing-season temperature, with higher temperatures favouring
system. Science 290, 779–783 (2000). C4 plants7,8. Other factors, however, may also affect the occurrence
5. Buks, E., Schuster, R., Heiblum, M., Mahalu, D. & Shtrikman, H. Dephasing in electron interference of C3 and C4 plants. Where precipitation and nutrient availability
by a ‘which-path’ detector. Nature 391, 871–820 (1998).
permit trees to grow, C3 trees will outcompete C4 grasses, such as in
6. van der Viel, W. G. et al. The Kondo effect at unitary limit. Science 289, 2105–2108 (2000).
7. Buttiker, M. Four terminal phase coherent conductance. Phys. Rev. Lett. 57, 1761–1764 (1986).
the tropical forests1. The significance of the climatic factors determin-
8. Prange, R. E. & Girvin, S. M. (eds) The Quantum Hall Effect (Springer, New York, 1987). ing the large-scale C4 plant abundance is still not well understood, but
9. Born, M. & Wolf, E. Principles of Optics 348–352, 7th edn (Cambridge Univ. Press, Cambridge, UK, insights can be gained from the analysis of past vegetation changes.
1999).
10. Aharonov, Y. & Bohm, D. Significance of electromagnetic potentials in the quantum theory. Phys. Rev.
Long-chain, odd-numbered C25 to C35 n-alkanes are major lipid
115, 485–491 (1959). constituents of the epicuticular wax layer of terrestrial plants9. These
11. Aronov, A. G. & Sharvin, Yu. V. Magnetic flux effects in disordered conductors. Rev. Mod. Phys. 59, plant waxes are easily removed from the leaf surface by rain or wind,
755–779 (1987). especially by sandblasting during dust storms. They are, therefore,
12. de-Picciotto, R. et al. Direct observation of a fractional charge. Nature 389, 162–165 (1997).
13. Reznikov, M. et al. Quantum shot noise. Superlattice Microstruct. 23, 901–915 (1998).
common organic components of eolian dust10. In surface sediments
14. Chklovskii, D. B., Shklovskii, B. I. & Glazman, L. I. Electrostatics of edge channels. Phys. Rev. B 46, of the eastern South Atlantic, the plant wax n-alkanes exhibit a
4026–4034 (1992). moderate to high odd versus even carbon-number predominance
15. Kane, C. L. Telegraph noise and fractional statistics in the quantum Hall effect. Preprint cond-mat/
(carbon preference index (CPI) of 2.3–6.4), and thus predominantly
0210621 at khttp://xxx.lanl.govl (2002).
represent leaf waxes of terrestrial higher plants9. Their plume-like
Acknowledgements We thank Y. Levinson for clarifying the issue of phase averaging, and C. Kane distribution in surface sediments (Fig. 1a) indicates that they are
for comments on the manuscript. The work was partly supported by the MINERVA Foundation, primarily transported by southeasterly winds from the dry areas in
the Israeli Academy of Science, the German Israeli Project Cooperation (DIP), the German Israeli southern Africa, the Kalahari savannah and Namib Desert. The
Foundation (GIF), and the EU QUACS network.
discharge of the Congo River is apparently of minor importance for
the supply of n-alkane leaf waxes. During the austral winter (June to
Competing interests statement The authors declare that they have no competing financial
interests. August) the strong Southern Hemisphere trade winds transport

Correspondence and requests for materials should be addressed to M.H. * Present address: Research Center Ocean Margins, University of Bremen, PO Box 330440, 28334 Bremen,
(e-mail: [email protected]). Germany.

418 © 2003 Nature Publishing Group NATURE | VOL 422 | 27 MARCH 2003 | www.nature.com/nature